eHam

eHam Forums => Amplifiers => Topic started by: W1BR on August 15, 2015, 08:46:03 AM



Title: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on August 15, 2015, 08:46:03 AM
Tom

Here is a poser for you... a lot of hams have expressed opinions that a linear's IMD performance somehow affects key clicks. Since IMD is products caused by undesirable mixing of multiple signals, it would seem that a single tone CW signal would NOT require an linear amplifier.

On the other hand, rise and decay time of the waveform might be adversely affected at the extremes between Class C and true linear service.  That may cause key click artifacts when a properly shaped CW waveform is amplified in a non linear stage?

The question remains:  How much affect does the linearity of an external amplifier have on a CW signal?

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SWL2002 on August 16, 2015, 05:32:04 AM
Tom

Here is a poser for you... a lot of hams have expressed opinions that a linear's IMD performance somehow affects key clicks. Since IMD is products caused by undesirable mixing of multiple signals, it would seem that a single tone CW signal would NOT require an linear amplifier.

On the other hand, rise and decay time of the waveform might be adversely affected at the extremes between Class C and true linear service.  That may cause key click artifacts when a properly shaped CW waveform is amplified in a non linear stage?

The question remains:  How much affect does the linearity of an external amplifier have on a CW signal?

Pete

If you are talking about what hams call "CW" (morse code) what you are calling CW or continuous wave is not CW or continuous wave at all.  You are turning a carrier on and off with some finite rise and fall time.  This is an amplitude modulated signal.  It also has a bandwidth defined by the speed and also the rise/fall times of the signal, among other things.

In the simplest form, "CW" consists of a pure sine wave multiplied with a square wave that's either 0 or 1, corresponding to the keying of the carrier. As with a mixer, or amplitude modulation, multiplying two signals generates frequency components that are the sum and difference of each frequency component of the multiplicands. In the simple form described above, where the switching waveform is a square wave, the bandwidth can be very large, because a square wave consists of an infinite series of odd harmonics.  Real transmitters filter this square wave to some extent, and perfect square waves can't exist in practice anyway. The slower the transition from "on" to "off" is made, the less bandwidth is required. If you send faster, then there are more transitions per second. Since each transition requires a certain amount of energy away from the carrier frequency, faster speed means more sideband power, that is, more bandwidth.

It is amazing that many (most?) Hams do not know or are not taught this.  I guess it should not be surprising since most Hams nowadays are appliance operators.

As to the effect of using a Linear or class C amplifier on the "CW" signal, I will leave that up to you to deduce.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: N3QE on August 16, 2015, 06:30:27 AM
I'm not Tom, but I know for a lot of half-century in old ham transmitters, the keyed class C stages often went through a region of instability every keying transition and the resulting parasitic squawks were just everywhere - clicks not just kHz away but up into VHF.

The handbooks always wrote this up as if it were the sparks from the key contacts, recommending filters on the key line, and I think they got this wrong. I'm sure the sparks on the key contacts caused local very much local RFI but the key filters recommended by the handbooks, made the instabilities on keying transitions last longer.

I don't think this is applicable to any modern equipment but I could be wrong. Maybe Tom has some thoughts on neutralization instabilities that happen as grid current starts to get drawn with capacitive divider neutralizing circuits.

Incidentally, my experiences with parasitics in ham equipment cause my company's lawyers to get the heebie-jeebies when it comes to parasitics in other equipment. They claim that the half century of postcards I've gotten from OO's, in no way make me qualified to testify on the subject in court :-).


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on August 16, 2015, 06:47:52 AM
Firstly, to be really pedantic about that which we usually refer to simply as CW, let's clarify it as having the ITU-R designation 100HA1AAN.

A Class C stage can 'sharpen' the rise and fall times depending on the circuit. If, for example, a 'clamp' tube is used without a VR tube in series with the screen, it will be in Class A at the start of the pulse and the effects will be minimal. If it is biased just to cut off with grid leak bias making up the rest of the necessary Class C bias, again, there won't be much sharpening. If the bias is derived solely from a bias supply, then you will see sharpening, the amount depending how far back the tube is biased. So a true linear amplifier should not make any difference, although the presence of an ALC spike could have an effect.

In theory, any amplifier COULD also add a chirp to a keyed signal by virtue of AM to PM conversion: in reality, it will be so small as to be negligible, and only noticeable with special test gear, but I mention it for the sake of completeness, since pedantry is the order of the day. That's with it acting purely as an amplifier, without N3QE's spurious oscillations (they can do - oh, yes they CAN - but not inevitably!) taking place during the power ramp up/ramp down.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on August 16, 2015, 06:58:45 AM

 
If you are talking about what hams call "CW" (morse code) what you are calling CW or continuous wave is not CW or continuous wave at all.  You are turning a carrier on and off with some finite rise and fall time.  This is an amplitude modulated signal.  It also has a bandwidth defined by the speed and also the rise/fall times of the signal, among other things.
 

As to the effect of using a Linear or class C amplifier on the "CW" signal, I will leave that up to you to deduce.

I clearly stated rise and fall time, so what does infer?  Continuous carrier? 

So, you don't know the answer. The question was simple--how much effect does using a Class Amp have on the rise and fall time of a properly shaped CW waveform?  Is there enough distortion to affect the signal's BW?  It has been argued on here before that certain "CB" amps would not even be suitable for use on CW because of their poor two tone IMD performance.

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on August 16, 2015, 07:08:01 AM
Firstly, to be really pedantic about that which we usually refer to simply as CW, let's clarify it as having the ITU-R designation 100HA1AAN.

A Class C stage can 'sharpen' the rise and fall times depending on the circuit. If, for example, a 'clamp' tube is used without a VR tube in series with the screen, it will be in Class A at the start of the pulse and the effects will be minimal. If it is biased just to cut off with grid leak bias making up the rest of the necessary Class C bias, again, there won't be much sharpening. If the bias is derived solely from a bias supply, then you will see sharpening, the amount depending how far back the tube is biased. So a true linear amplifier should not make any difference, although the presence of an ALC spike could have an effect.

In theory, any amplifier COULD also add a chirp to a keyed signal by virtue of AM to PM conversion: in reality, it will be so small as to be negligible, and only noticeable with special test gear, but I mention it for the sake of completeness, since pedantry is the order of the day. That's with it acting purely as an amplifier, without N3QE's spurious oscillations (they can do - oh, yes they CAN - but not inevitably!) taking place during the power ramp up/ramp down.

Hi Peter

I should have been a bit more precise:  tube finals are bit more involved, since there must be some sort of protection when drive is removed; most likely grid block or cathode keying, or a clamp tube (which I haven't seen used for CW.)  There was usually some method of providing an RC time constant, either on the grid bias or in the keyed cathode to help control the rise and fall time. Fancier rigs had sequential keying to control chirp problems, etc.

I was more concerned about statements that were made in this forum in the past concerning the use "CB" amplifiers that are being sold on the ham market.  Their two tone IMD characteristics aren't that great, and it was claimed that they were entirely unsuitable for CW because they weren't linear enough for that service.  I've never seen any actual measurements that prove or disprove that theory. I suspect the waveform rise and fall times will be affected to some degree, but how much does that increase the key clicks to a point that is noticeable?

Pete



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SWL2002 on August 16, 2015, 07:14:02 AM

 
If you are talking about what hams call "CW" (morse code) what you are calling CW or continuous wave is not CW or continuous wave at all.  You are turning a carrier on and off with some finite rise and fall time.  This is an amplitude modulated signal.  It also has a bandwidth defined by the speed and also the rise/fall times of the signal, among other things.
 

As to the effect of using a Linear or class C amplifier on the "CW" signal, I will leave that up to you to deduce.

I clearly stated rise and fall time, so what does infer?  Continuous carrier?  

So, you don't know the answer. The question was simple--how much effect does using a Class Amp have on the rise and fall time of a properly shaped CW waveform?  Is there enough distortion to affect the signal's BW?  It has been argued on here before that certain "CB" amps would not even be suitable for use on CW because of their poor two tone IMD performance.

Pete

Obviously this went over your head or you are having reading comprehension problems again.  I thought it was so clear that it was not worth stating.  Unless you are keying the POWER SUPPLY of a class A,B,C,D,E,F, or whatever amp, the amplifiers linearity will affect the "CW" signal.  The more IMD the amp generates because of non-linearity, the wider the "CW" signal becomes (beyond the bandwidth attributed to rise/fall times, keying speed, carrier phase noise, etc...).  You are putting an AMPLITUDE MODULATED signal into the "linear" amplifier in question.  What do you think happens when an AM signal is put into a non-linear amplifier?  The same thing happens when a keyed "CW" signal is put into a non-linear amplifier.

The answer was laid out in front of you, but you still could not see it.  Pitiful...


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on August 16, 2015, 07:15:54 AM
Pete,

Depends on the bias. Because a bipolar will start turning on at a low voltage, I suspect the sharpening won't be noticeable - especially if driven with a rig with lousy key clicks like so many! MOS without bias, quite possibly. I'd worry more about harmonic attenuation to Part 97.....

See your messages....

Peter G3RZP


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on August 16, 2015, 07:17:56 AM
Indeed, harmonics are an issue, but that can be taken care with proper output filtering.

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on August 16, 2015, 08:18:14 AM
Quote
The more IMD the amp generates because of non-linearity, the wider the "CW" signal becomes (beyond the bandwidth attributed to rise/fall times, keying speed, carrier phase noise, etc...).

The amount is so negligible as to be hardly worth considering. Refer to ITU-R Rec SM.1138, where the necessary bandwidth for a 25 wpm signal in fading is considered to be 100Hz. 25 wpm means with the standard word, dots 100ms long:  the 10-90% rise and fall time specified in ITU-R Rec. SM328-11 requires an occupied bandwidth (99%) of 66.4Hz. So take a 10 Hz modulating frequency: there is a carrier with two 10Hz sidebands each 6dB down, and drive the amplifier to clipping. Then the worst case 3rd order IMP each side will be -15dB rel carrier at 20Hz: the worst case 5th order will be -29dB and the 7th at 40Hz away will be around -38dB: so all still within both the necessary bandwidth and indeed the occupied bandwidth and lower than modulation components themselves.

A reference on IMD in limiting amplifiers  may be found in  S.F. George and J.W.Wood, 'Ideal Limiting' Part 1, U.S. Naval Research Laboratory, AD266069, October 2nd, 1961.

Phase noise would not intermodulate, although the 3dB lower AM component could. However,
since at a 10Hz offset, one would expect the AM component to be at least 20dBc/Hz down, the IMPs will be much lower than those for the modulation sidebands.

Thus to all intents and purposes, unless the CB amplifier has a special noise generator built in, it is OK for CW. Or all those transmitters built over the years with acceptable performance by people like Collins,  Marconi, STC , KW, Labgear etc all got it wrong - as did the ITU.

harmonics - yes. Just needs filtering but the CB amps don't bother! One point that needs thinking about with those amps is stability when feeding a load which is not 50+j0 at the operating frequency. There was an article in Electronics Design magazine on that back in the mid 1970s.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SWL2002 on August 16, 2015, 08:45:26 AM
Quote
The more IMD the amp generates because of non-linearity, the wider the "CW" signal becomes (beyond the bandwidth attributed to rise/fall times, keying speed, carrier phase noise, etc...).

The amount is so negligible as to be hardly worth considering. Refer to ITU-R Rec SM.1138, where the necessary bandwidth for a 25 wpm signal in fading is considered to be 100Hz. 25 wpm means with the standard word, dots 100ms long:  the 10-90% rise and fall time specified in ITU-R Rec. SM328-11 requires an occupied bandwidth (99%) of 66.4Hz. So take a 10 Hz modulating frequency: there is a carrier with two 10Hz sidebands each 6dB down, and drive the amplifier to clipping. Then the worst case 3rd order IMP each side will be -15dB rel carrier at 20Hz: the worst case 5th order will be -29dB and the 7th at 40Hz away will be around -38dB: so all still within both the necessary bandwidth and indeed the occupied bandwidth and lower than modulation components themselves.

A reference on IMD in limiting amplifiers  may be found in  S.F. George and J.W.Wood, 'Ideal Limiting' Part 1, U.S. Naval Research Laboratory, AD266069, October 2nd, 1961.


No where did I make a judgement on whether the levels were worth considering nor did I make any comments on whether a CB amp would be acceptable.  I was talking about the result, not whether the result was negligible of not.  Those are two different things.  

Quote

Phase noise would not intermodulate, although the 3dB lower AM component could. However,
since at a 10Hz offset, one would expect the AM component to be at least 20dBc/Hz down, the IMPs will be much lower than those for the modulation sidebands.


I did not say that phase noise would inter-modulate.  I said that the CW bandwidth is affected by multiple things one of which is carrier phase noise.

Quote

Thus to all intents and purposes, unless the CB amplifier has a special noise generator built in, it is OK for CW. Or all those transmitters built over the years with acceptable performance by people like Collins,  Marconi, STC , KW, Labgear etc all got it wrong - as did the ITU.

harmonics - yes. Just needs filtering but the CB amps don't bother! One point that needs thinking about with those amps is stability when feeding a load which is not 50+j0 at the operating frequency. There was an article in Electronics Design magazine on that back in the mid 1970s.

Yes, they might be acceptable for CW, but many use them for SSB service also, and that is not OK if the CB amp is biased class C or whatever.  Some CB amps are so poorly built, I would not even use them to amplify a true continuous wave signal (not what Hams call "CW").

  


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on August 16, 2015, 09:01:40 AM
So Pope is proved right again......


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: DL8OV on August 16, 2015, 10:19:51 AM
There is one thing I am not sure of. If I generate a morse code signal with a 5mS rise and fall time then feed this into a Class A amplifier I will see a larger output signal which still has a 5mS rise and fall time, a faithful reproduction. If I do the same exercise using a Class C amplifier will I still see that 5mS rise and fall time on the output? A little voice in the back of my head keeps on telling me that the output will have crisp sharp edges and be a source of key clicks because a Class C amplifier can be either 'on' or 'off'.

It's tempting to take my homebrew rig and rebias the P.A. into Class C just to see what happens.

Peter DL8OV


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on August 16, 2015, 10:56:12 AM
it's going to depend on how Class C it is. Let's say you are driving a silicon bipolar. Then there's  going to be conduction starting at around 0.4 to 0.5 volts vbe. Now if the 5mS is 10 - 90% (which the ITU feels a bit short unless you are talking of 50wpm), then the transistor will be starting to conduct at 2 to 3ms. Of course, it won't be an exponential rise and fall.....

Now take a tube biased to say -100 volts and starting to conduct at say -20 and running to +20 volts. Your 5ms is down to about 1.5ms. Duty cycle is around 80 degrees and efficiency up there with the Gods. Well, not quite.

The imperfections in all this mean, as usual, that a few rough calculations can give you an idea and then empiricism has advantages. Simulation works well too, PROVIDED you have an accurate model, and much of the time, you don't really have that. Especially in IC design, tying the CAD guys down to what the devices REALLY do at RF and what all the process corners really are is a bit like picking up liquid mercury from a Teflon crucible while wearing boxing gloves with your hands tied behind you!

Cynical? Who, me?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: DL8OV on August 16, 2015, 01:03:16 PM
Well, now I have something to do on a wet Monday, tweak the bias on my P.A. Driver until I see < half an AC cycle on the output then try sending some CW into a dummy load***. Another local ham will be listening for my signal (and any keyclicks) using an FT-817 and a mobile whip.

Much more fun than feeding the lot into SPICE and seeing what the computer spits out the other end.

Peter DL8OV

*** This dummy load leaks, on purpose.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on August 16, 2015, 01:05:43 PM
Good idea, but he should add an attenuator equal to the amplifier gain to make a realistic guess.

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W8JI on August 16, 2015, 02:08:13 PM

I clearly stated rise and fall time, so what does infer?  Continuous carrier? 

So, you don't know the answer. The question was simple--how much effect does using a Class Amp have on the rise and fall time of a properly shaped CW waveform?  Is there enough distortion to affect the signal's BW?  It has been argued on here before that certain "CB" amps would not even be suitable for use on CW because of their poor two tone IMD performance.

Pete

Any signal changing amplitude must have sidebands. This means the CW signal has sidebands when the rising and falling edges occur. It is obviously not a single frequency except during the time when level is not changing.

When level is changing, it is actually an AM signal.

Bandwidth, or number of sidebands and frequency spread of the sidebands during the rise and fall, is determined by the slope and time of carrier power change while rising and falling.

If we ran that signal into a non-linear amplifier and changed the rise and fall shape, we would alter the bandwidth.

A class C amplifier will distort the rise and fall, usually flattening it off or squaring the edges. The square edges would greatly increase keying sideband bandwidth over something like a raised sine shape.

I think people are probably getting a bit too fussy if they think normal IMD of a linear amplifier can significantly increase bandwidth, but clipping of the leading edge (even at low power) or sharpening of roll over at the top edge can add significant bandwidth. A grossly non-linear amp, one that increases the change in slope of the envelope, can make bandwidth go way up.

The average power in the sidebands goes up with keying rate, but the signal sideband width stays the same (since it is set by the rise and fall shapes). Faster CW with clicks is more annoying because the clicks come more often, but the distance we tune off and are bothered by the clicks remains the same as long as the shape and duration of the rise and fall stay the same.

ALC induced clicks are reduced with higher speed, which is another irony. This is because the ALC stays more constant with higher speeds. At slow speeds, the ALC can drop to zero at times, and the next element can have an exaggerated rise with overshoot.

73 Tom


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on August 16, 2015, 10:47:43 PM
Many of the QRQ ops prefer a signal with 'hard' i.e. clicky keying, as being easier to read at speeds of 40 wpm plus. With the advantage that they can the n argue that as they run at QRQ in a contest and need hard keying, the fact that it keeps people away from 'their' frequency is a bonus!


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: JS6TMW on August 16, 2015, 11:57:01 PM
This is important to me because I plan to use a cheapy "CB amp" on the higher HF bands on CW with my QRP CW rig when needed. (I'm making a good lowpass filter for it so harmonics should not be an issue). It's impossible to know what is really going on inside without opening it and taking voltage readings while it's operating, but as far as I can tell from the schematic, it runs exactly the same for SSB and CW. In fact the only difference between the "SSB/CW" and "AM/FM" switch settings seems to be a time constant (delay?) for Transmit-Receive.

Steve AI6KX waiting for my Japanese license


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: ZS6DX on August 17, 2015, 12:39:55 AM
So Pope is proved right again......

Seems like the Pope has issues......  :(

Interesting discussion! I have played around quite a bit with the rise and fall times of my CW keying, adjustable on many modern radios. I found that for "slow" CW, up to about 25wpm more gentle rise and fall times make for a very pleasant signal and I have had people comment on it.

When you go up in speed readability becomes an issue, dos/dashes seems to flow together and it becomes a strain to read. I sometimes run at 40-65wpm and then I prefer very sharp rise/fall times, sounds more harsh and the signal does get wider but it is crisp and readable.

I have not tried using CW on a non linear amplifier but I think it will mostly be ok if the radio/amplifier is clean.

Thanks or you informative posts, I have learned a lot from you!

73, Rudi de ZS6DX/V51VE


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SWL2002 on August 17, 2015, 03:27:05 AM

When level is changing, it is actually an AM signal.

73 Tom

Interesting, I wonder where we heard that before?

Quote from: SWL2002

If you are talking about what hams call "CW" (morse code) what you are calling CW or continuous wave is not CW or continuous wave at all.  You are turning a carrier on and off with some finite rise and fall time.  This is an amplitude modulated signal.  It also has a bandwidth defined by the speed and also the rise/fall times of the signal, among other things.




Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: JS6TMW on August 17, 2015, 03:49:36 AM
I think just now the Pope is trying to decide how many angels can dance on the head of a pin.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W8JI on August 17, 2015, 05:05:25 AM
A CB amplifier without bias will square the envelope rise at the bottom. That is the main damage class C does.

If any amplifier saturates hard, it will square the top edge.

Generally if those two points are taken care of, everything else will be acceptable. The low end change is the difficult one to see on a scope.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on August 17, 2015, 05:08:56 AM
However, even when limiting, the numbers showed that the bandwidth increase caused by intermodulation was minimal compared with the bandwidth caused by the modulation. The sharpened rise and fall times are not a characteristic of intermodulation but the non-linearity.

The Pope in question is Alexander......


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W8JI on August 17, 2015, 05:31:03 AM
Right. IMD of the amplifier is not such a concern as any abrupt change in transfer function as level changes. You don't want to hose up the shape with something sharp.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on August 17, 2015, 05:45:54 AM
Now that I think about it a bit more... a dual trace scope could be used to sample and compare the waveform shaping before and after the amplifier.  But, I don't have a cheap SS amp so it is a moot point.

I'd also suspect that if the amp is using RF sensing the resultant hot keying and lead to problems.

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: JS6TMW on August 17, 2015, 06:03:19 AM


I'd also suspect that if the amp is using RF sensing the resultant hot keying and lead to problems.

Pete

Yes, I am worried about that. I have no experience with this kind of amp yet. I might have to slow it down. Not sure how the QRP rig  (Weber Tribander) works as far as keying goes either since it is now under construction.

I don't understand why W8JI assumes CB amplifiers are Class C, since they are mainly intended to boost AM signals (is there much SSB on the CB band???)

There seems to be a common opinion that just because these CB amps are illegal for CB use, they are also by nature producers of dirty signals. Oh, and I HAVE read W8JI's test of the HLA-150 using a two-tone input signal. Not very flattering. The RM KL300 I bought might have more headroom. Then again, it might be the same transistors with a new nameplate.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on August 17, 2015, 06:54:00 AM
Quote
I don't understand why W8JI assumes CB amplifiers are Class C, since they are mainly intended to boost AM signals (is there much SSB on the CB band???)

Surprisingly, very heavily clipped AM through a Class C stage is still understandable. But you don't need Class C for poor IMD. There was an experimental RAF VHF TX in WW2 that never got anywhere but used a Class C 8018 (selected 807 for full output at 120MHz) fed with the modulated signal. It's probably that legacy that leaves aircraft VHF sounding to me like 90% distortion and 10% unmentionable!

Even 'Linear amplifiers' are not necessarily so. G3SJX reviewed the Expert 1.3K-FA amplifier in July RadCom. He found the 3rd order IMD at outputs of 1.3 to 1.5kW to be  -20 to -26dB relative to PEP - that's -13 to -20dB relative to one tone of a 2 tone signal. At 1kW, it was -30dB  rel PEP  i.e. -24dB relative to tone.

Not exactly brilliant


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on August 17, 2015, 08:40:36 AM


I'd also suspect that if the amp is using RF sensing the resultant hot keying and lead to problems.

Pete

Yes, I am worried about that. I have no experience with this kind of amp yet. I might have to slow it down. Not sure how the QRP rig  (Weber Tribander) works as far as keying goes either since it is now under construction.

I don't understand why W8JI assumes CB amplifiers are Class C, since they are mainly intended to boost AM signals (is there much SSB on the CB band???)

There seems to be a common opinion that just because these CB amps are illegal for CB use, they are also by nature producers of dirty signals. Oh, and I HAVE read W8JI's test of the HLA-150 using a two-tone input signal. Not very flattering. The RM KL300 I bought might have more headroom. Then again, it might be the same transistors with a new nameplate.

Well,  many of the cheaper CB amps are running with no forward bias on the devices, and are designed for AM amplification. I believe Tom had written up the "advantages" for doing so (as in the reasoning why it "worked" for CB lids, and not that it was technically desirable!!,)  at the expense of destroying adjacent channels, etc.

Also, I am not sure that a lack of forward bias is actually true Class C for transistor devices; by definition they'd need to have some reverse bias voltage placing them well into the cut off region. Non linear for sure, but I am positive that the devices are fully into the Class region.

Gain compression for CW signals may slightly affect the waveform, but again, how much of a problem does it cause unless the waveform rise or fall times become to close to being a square wave?  Most CB amps have some forward bias, albeit with poor regulation, but is that more of a concern for SSB waveforms vs. CW waveforms?

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: DL8OV on August 17, 2015, 09:47:53 AM
Gentlemen, I have some test results for you:

Transmitter - Homebrew rig with RD16HHF1 final producing about 5W into a leaky dummy load, variable output power and variable bias.

Receiver - FT817 with untuned mobile whip at about 200m range. Transmit power was adjusted to give an S9 signal on the receiver to avoid overload. For the purposes of the test 'Signal Width' is assumed to be the two points either side of my signal where I measure S5 or four S points down on the carrier.

1) Biased for Class A and transmitting a clean sine wave the signal width was reported as 400 Hz and no key clicks were heard.

2) Biased so that clipping could just be seen on the output waveform signal width was reported as 400 Hz with very slight keyclicks, maybe too small to notice unless you were listening out for them.

3) Biased for Class B with 50% of the waveform on the scope signal width was reported as 700 Hz with heavy key clicks. They could also copy my 14 MHz signal on 28 MHz despite using a low-pass filter on the TX.

4) Biased for Class C with 40 % of the waveform on the scope signal signal width was reported as 1,3 KHz and the key clicks were described as 'terrifying'. Signal level was confirmed as S9 so this was not due to receiver overload.

Peter DL8OV





Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on August 17, 2015, 09:53:22 AM
What were the actual rise and fall time variations?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: DL8OV on August 17, 2015, 10:00:40 AM
The rig normally transmits using a 4,5mS rise and fall time. I must confess that I forgot to measure it at the clipping and Class B tests but on Class C it was under 2mS so it looks like a Class C amplifier halves the times.

Peter DL8OV


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K7KBN on August 17, 2015, 10:01:47 AM
The Pope in question is Alexander......

And there were more or less eight Popes named Alexander, although #5 is often listed as an antipope.  I guess he erred ....


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: DL8OV on August 17, 2015, 10:08:04 AM
Maybe so but Alexander VI was an interesting fellow, also known as Rodrigo Borgia. An interesting time in European history.

Peter DL8OV


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on August 17, 2015, 10:09:59 AM
Clicks from later stages isn't new (in the 1948 ARRL Handbook), although I'm a little surprised that they become so bad by only halving the rise time.

Pat, the relevant Alexander Pope lived from 1688 to 1744.....


















Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: N3QE on August 17, 2015, 10:18:30 AM
4) Biased for Class C with 40 % of the waveform on the scope signal signal width was reported as 1,3 KHz and the key clicks were described as 'terrifying'. Signal level was confirmed as S9 so this was not due to receiver overload.

Were there any signs of parasitics during the key-on/key-off transitions? These are best seen by triggering the scope from the key transitions and turning the scope timebase up and down by a few decades around 1 us/division. When you find the fuzzies, use the trigger delay to zoom in.

A "hot device" like a 6CL6 or RD16HHF1 can simultenously be squegging at LF, HF, and UHF frequencies!!!

A recent article about squegging in RF MOSFET's: http://www.highfrequencyelectronics.com/Archives/Nov13/1311_HFE_squegging.pdf

Tim N3QE


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on August 17, 2015, 10:30:05 AM

Even 'Linear amplifiers' are not necessarily so. G3SJX reviewed the Expert 1.3K-FA amplifier in July RadCom. He found the 3rd order IMD at outputs of 1.3 to 1.5kW to be  -20 to -26dB relative to PEP - that's -13 to -20dB relative to one tone of a 2 tone signal. At 1kW, it was -30dB  rel PEP  i.e. -24dB relative to tone.

Not exactly brilliant

Reading this, and realising that the stock 1.3K-FA has sufficient gain to be driven by the FT-817, I become slightly worried.
G3SJX measures, if I remember correctly, amplifier IMD using a very low IMD drive source.
If the drive source is quite bad from the beginning, due to i.a. ALC modulation, how would the output spectrum look?

73/
Karl-Arne
SM0AOM


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: DL8OV on August 17, 2015, 10:37:52 AM
There were no parasitics that I know of but I can check again tomorrow if you want. The RD16HHF stage that I used is intended to be the driver of a 150W-200W PA and as such was constructed to be as clean and stable as possible. The usual things were done as a matter of course including lots of groundplane, decoupling using multiple capacitors and a resistor in the gate connection. The device was also under run, normally it produces between 1W and 2W but for this test I was running about 5W from a 16W device.

As for the risetime changes, normal operation is a risetime of 4,5mS using a raised cosine profile, under Class C bias this dropped to under 2mS and I am quite sure that the careful cosine shape had been destroyed. I forget what the unmodified FT-1000 produced but I seem to remember that it was around this 2mS figure (if I am wrong then please could someone correct me).

Peter DL8OV


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: JS6TMW on August 17, 2015, 03:23:52 PM

Well,  many of the cheaper CB amps are running with no forward bias on the devices, and are designed for AM amplification. I believe Tom had written up the "advantages" for doing so (as in the reasoning why it "worked" for CB lids, and not that it was technically desirable!!,)  at the expense of destroying adjacent channels, etc.

Also, I am not sure that a lack of forward bias is actually true Class C for transistor devices; by definition they'd need to have some reverse bias voltage placing them well into the cut off region. Non linear for sure, but I am positive that the devices are fully into the Class region.


I only have a schematic without any component values indicated, but it looks like there is no bias on the finals. The bases go to an input xfmr whose center tap is at ground, as are the emitters. I suppose that is to keep no-signal current drain low without having to switch off the B+.  I need to read up on biasing in case it looks necessary. I'm using a power supply that is easily switched remotely.

Steve


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K7KBN on August 17, 2015, 03:31:02 PM
Clicks from later stages isn't new (in the 1948 ARRL Handbook), although I'm a little surprised that they become so bad by only halving the rise time.

Pat, the relevant Alexander Pope lived from 1688 to 1744.....

Ah!  He of the Pierian Spring caveat.  Might give Perrier a bit of a challenge - spelling and all!



















Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K7KBN on August 17, 2015, 03:37:36 PM
Maybe so but Alexander VI was an interesting fellow, also known as Rodrigo Borgia. An interesting time in European history.

Peter DL8OV

"What would you do if you were up a dark alley with Cesare Borgia
And he was coming torgia?"    ...Ogden Nash

Might have worked with Rodrigo as well - or Lucrezia


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W8JI on August 17, 2015, 05:52:50 PM
The rig normally transmits using a 4,5mS rise and fall time. I must confess that I forgot to measure it at the clipping and Class B tests but on Class C it was under 2mS so it looks like a Class C amplifier halves the times.

Peter DL8OV

It isn't just time. The slope at every point sets the ultimate bandwidth.

Picture the slope as the envelope of an AM transmitter. If there is an area where it is rising like a 2kHz sine wave, that will be the minimum bandwidth of the "click".

The total rise or fall times sets the BEST POSSIBLE bandwidth, but it can always be any amount worse than that best possible width.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on August 18, 2015, 12:55:35 AM

Well,  many of the cheaper CB amps are running with no forward bias on the devices, and are designed for AM amplification. I believe Tom had written up the "advantages" for doing so (as in the reasoning why it "worked" for CB lids, and not that it was technically desirable!!,)  at the expense of destroying adjacent channels, etc.

Also, I am not sure that a lack of forward bias is actually true Class C for transistor devices; by definition they'd need to have some reverse bias voltage placing them well into the cut off region. Non linear for sure, but I am positive that the devices are fully into the Class region.


I only have a schematic without any component values indicated, but it looks like there is no bias on the finals. The bases go to an input xfmr whose center tap is at ground, as are the emitters. I suppose that is to keep no-signal current drain low without having to switch off the B+.  I need to read up on biasing in case it looks necessary. I'm using a power supply that is easily switched remotely.

Steve

I would say that connecting the base returns directly to ground constitutes Class C in BJT work, as the forward voltage drop in the base-emitter junction is about 0.6V before collector current
starts to flow. This makes the conduction angle less than 180 degrees which is one of the definitions of Class C operation.
The no-signal collector current should be only the leakage current which for modern transistors is very near zero.

Regarding amplifier non-linearity creating or worsening key-clicks, any departure from a perfect straight-line transfer function in an amplifier will affect the spectrum of the modulated output signal.

An A1A or "CW" signal is a carrier wave 100% amplitude modulated with a keying envelope which creates the output spectrum.
It has be shown (by i.a. W9CF) that a keying envelope that has the shape of the Gauss error function erfc(t) provides the narrowest spectrum for given rise or fall times and keying rate.
Any departure from this shape will create a broader spectrum.

A perfect linear amplifier has a transfer function which is a straight line or constant K, which only multiplies the input signal.
The Gaussian shaped input keying envelope will come out of the amplifier still Gaussian shaped and with the same spectral properties.

However, any realisable amplifier will have a transfer function which deviates from this ideal straight line.
Depending on the nature of the deviation, the effects of the non-linearity will be different;
a dead-band or cross-over distorsion discontinuity usually has worse effects than more gradual gain compression on the keying envelope.

It would probably take IM distorsion which makes an amplifier more or less unusable for serious SSB use to create an objectionable amount of key-clicks,
but the shortening of the rise/fall-time and the discontinuities at low drive levels that a Class C-biased amplifier provides are more than sufficient to create adjacent channel problems.

In the "old days" when marine radio transmitters were using Class C amplifiers and long multiplier chains, the keying envelope requirements were taken care of by
quite complex differential keying schemes, where suitably shaped grid-block keying was used in the final stage, and cathode or oscillator keying in the exciter.
This enabled the transmitters to comply with the output spectrum requirements of the 1947 Atlantic City regulations, which narrowed the channel spacing in the radiotelegraphy bands
considerably compared to the 1938 Cairo regulations.

Finally, the channel spacing for A1A emissions was set to 500 Hz in the 1959 Geneva regulations and their follow-ups in 1967.
Transmitters that were specificed to comply with these requirements had a very tight keying envelope, which was hard or almost impossible to create using Class C exciter and amplifier chains.
Fortunately, the state-of-the-art had progressed to use SSB exciters and linear amplifiers where control of the keying envelope shape became much easier.

The ITU regulatory requirements were repeated or even extended in many national type-acceptance documents,
there was a British MPT specification for the allowable keying spectrum mask that was considered very difficult to fulfill.
These 1960's requirements were for all practical purposes in force until the end of maritime Morse on February 1 1999.

All commercial, military and shipboard Morse transmitters of recent designs were compliant with the ITU/CCIR and MPT keying shaping requirements,
which is one of the reasons why objectionable key-clicks were so seldom encountered from commercial A1A transmitters.

The coast radio Morse transmitters that I supervised work on up to the closure of HF Morse services in 1991 used CCIR
compliant keying shaping in the exciters. It was a regular maintenance procedure to check the rise and fall times of the output RF envelope using
"1:1 reversals keying" to ensure that component ageing or amplifier adjustments had not introduced any discontinuities.

73/
Karl-Arne
SM0AOM


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: JS6TMW on August 18, 2015, 03:08:21 AM
Thanks Karl-Arne. Lots of good information and history!

And thanks to YO9IRF I now know how to bias the RM amp into linear modes if necessary for clean CW.

73,

Steve


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on August 18, 2015, 08:29:39 AM
Thanks Karl-Arne. Lots of good information and history!

And thanks to YO9IRF I now know how to bias the RM amp into linear modes if necessary for clean CW.

73,

Steve

The RMI amps are  biased for linear operation if rated for SSB. The  power ratings are greatly exaggerated for most of their HF amps. But, I'd wager they are good enough for CW service. 

Many of the really cheap CB amps were designed for AM, and may NOT have any biasing. Those amps would require considerable modification to be usable on the ham bands... stiff biasing, good harmonic filters, etc. And that assumes the ferrite transformer cores aren't going into saturation. 

I'd bet my UHF and VHF ham bricks probably have horrendous IMD figures as well.. there have been articles in VHF journals that show how to modify the VHF/UHF bricks to a more desirable active biasing circuit.

Pete



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: JS6TMW on August 18, 2015, 02:58:36 PM
Yes, the RMI HLA series and others sold for ham use do have real bias circuits. The "CB" models like my KL300 run Class C. I have seen some Youtubes of scope displays from these on SSB and they look good, somehow.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W8JI on August 21, 2015, 11:00:56 AM
Yes, the RMI HLA series and others sold for ham use do have real bias circuits. The "CB" models like my KL300 run Class C. I have seen some Youtubes of scope displays from these on SSB and they look good, somehow.

This is a great example of the problem caused by using a scope to measure bandwidth. A scope is the wrong tool.

Lack of conduction bias causes distortion near zero envelope power. It does not affect the envelope once the envelope has the device turned on. Lack of quiescent current sharpens the envelope rise at the low level points of the envelope. This is almost impossible to see on a scope, even if the rise is very rapid.

A scope has extremely limited usefulness in indicating bandwidth, despite the fact many people rely on them for ensuring a clean signal.

A properly used communications receiver is actually far better than a scope.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on August 21, 2015, 02:39:02 PM
To get good linearity from a  bipolar PA, you would ideally have a constant current bias source that sagged with increasing temperature. Except that would be a TOTAL DISASTER when the base rectified the signal and drove the base negative. It is possible - at least in an IC and was done for cellular back in the mid 1990s - to produce a constant current bias source that has a almost zero reverse impedance, so if the transistor tries to rectify the signal, it see an almost zero impedance load. At the same time, if the forward conduction raises the DC bias, that gets cancelled, too.

The guy who designed this has retired from what was GEC-Plessey Semiconductors in the UK (later Zarlink and what is now left is part of Microsemi)and gone back to Russia - he's exactly a year older than I am, so he's 69 - it was never published or patented, and the ICs have doubtless disappeared and the documentation certainly has - it was for a mid 1990s device for Motorola for CDMA 'phones.

But he got over 50dB 3rd order IMD easily from Class B stages.......


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 22, 2015, 02:48:17 AM
Yes, the RMI HLA series and others sold for ham use do have real bias circuits. The "CB" models like my KL300 run Class C. I have seen some Youtubes of scope displays from these on SSB and they look good, somehow.

This is a great example of the problem caused by using a scope to measure bandwidth. A scope is the wrong tool.

Lack of conduction bias causes distortion near zero envelope power. It does not affect the envelope once the envelope has the device turned on. Lack of quiescent current sharpens the envelope rise at the low level points of the envelope. This is almost impossible to see on a scope, even if the rise is very rapid.

A scope has extremely limited usefulness in indicating bandwidth, despite the fact many people rely on them for ensuring a clean signal.

A properly used communications receiver is actually far better than a scope.

Skillful use of the maths functions in a modern digital scope (eg a low cost 8 bit scope made in the last 15 years or so) can give lots of info about SSB bandwidth and also distortion terms over about a 60dB range.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: DL8OV on August 22, 2015, 01:11:01 PM
"But he got over 50dB 3rd order IMD easily from Class B stages......."

I know of some people on this board who would sell their mothers into slavery to achieve something like that on an HF solid state amplifier.

Peter DL8OV


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on August 22, 2015, 02:01:54 PM
That was at 400mW PEP from a surface mount  (albeit a BIG one!) IC with a glued on heatsink!

Sad that the knowledge has been lost. At the time, it wasn't considered that it was worth patenting, especially as MOS was on the horizon.....


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: DL8OV on August 23, 2015, 02:01:45 AM
Would -46 dBm 3rd Order IMD at 1W 10 MHz be close enough?

http://www.qsl.net/in3otd/ham_radio/PD85004_PP_PA/PD85004_PP_PA.html#ver3

Scroll down to just above the half way point on the page and read about the "With center-tapped drive transformer and resistive feedback" amplifier. With performance like this I think this website is the Internet's best kept secret.

http://www.qsl.net/in3otd/ham_radio/PD85004_PP_PA/PD85004_PP_fbk_CT_PA_IM_1W_PEP.png

Peter DL8OV


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on August 23, 2015, 02:37:49 AM
That 400mW PEP was at 1800 MHz....


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W8JI on August 23, 2015, 05:33:10 AM
Yes, the RMI HLA series and others sold for ham use do have real bias circuits. The "CB" models like my KL300 run Class C. I have seen some Youtubes of scope displays from these on SSB and they look good, somehow.

This is a great example of the problem caused by using a scope to measure bandwidth. A scope is the wrong tool.

Lack of conduction bias causes distortion near zero envelope power. It does not affect the envelope once the envelope has the device turned on. Lack of quiescent current sharpens the envelope rise at the low level points of the envelope. This is almost impossible to see on a scope, even if the rise is very rapid.

A scope has extremely limited usefulness in indicating bandwidth, despite the fact many people rely on them for ensuring a clean signal.

A properly used communications receiver is actually far better than a scope.

Skillful use of the maths functions in a modern digital scope (eg a low cost 8 bit scope made in the last 15 years or so) can give lots of info about SSB bandwidth and also distortion terms over about a 60dB range.



Most Hams look for peak clipping, and assume if the peaks are not flat the bandwidth is good. The test of this is if you can explain how a typical Ham can use his scope to accurately measure transmitter bandwidth. If you can explain it, and most people can do it, you are correct. Otherwise, a good receiver is a better test.



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W0BTU on August 23, 2015, 06:20:15 AM
Most Hams look for peak clipping, and assume if the peaks are not flat the bandwidth is good. The test of this is if you can explain how a typical Ham can use his scope to accurately measure transmitter bandwidth. If you can explain it, and most people can do it, you are correct. Otherwise, a good receiver is a better test.

(http://www.w0btu.com/images/share/facebook_like_stamp.jpg)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 23, 2015, 07:20:52 AM
Just set the scope up to look at the 2 tone waveform in the time domain (i.e. as you would with an analogue scope) and then hit the MATH>>FFT function and set the sample rate to undersample the signal to show a low frequency alias.

You will then get a spectrum analyser display showing a low frequency bandpass sample (Alias) of the RF signal.

So you should see a similar plot to the classic ARRL spectrum analyser plots of transmitter IMD. My creaky old TDS scope is based on a 20 year old design and it can do this and show IMD to about 60dB below PEP on a 12.5kHz span for example. It's not as good as a regular spectrum analyser but still very useful indeed. Some skill is needed to ensure the image isn't an inverted alias but that probably doesn't matter much anyway.
Not all DSOs can do this very well but mine is adequate. The newer scopes are much better in this respect.

It's possible to make a reasonable measurement of an amplifier IP3 just using a DSO and a two tone test signal using this technique. Obviously, there are better ways to do it but even a basic entry level Tek DSO like mine can do a few useful tricks :)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W8JI on August 23, 2015, 08:41:36 AM

This is all a huge diversion from the fact a regular scope is almost useless for indicating cleanliness of a signal, especially when it is a quiescent current bias issue or amplifier class issue caused by bias.

But I think it is an interesting topic we need to think about.....

Just set the scope up to look at the 2 tone waveform in the time domain (i.e. as you would with an analogue scope) and then hit the MATH>>FFT function and set the sample rate to undersample the signal to show a low frequency alias.

You will then get a spectrum analyser display showing a low frequency bandpass sample (Alias) of the RF signal.

So you should see a similar plot to the classic ARRL spectrum analyser plots of transmitter IMD. My creaky old TDS scope is based on a 20 year old design and it can do this and show IMD to about 60dB below PEP on a 12.5kHz span for example. It's not as good as a regular spectrum analyser but still very useful indeed. Some skill is needed to ensure the image isn't an inverted alias but that probably doesn't matter much anyway.
Not all DSOs can do this very well but mine is adequate. The newer scopes are much better in this respect.

It's possible to make a reasonable measurement of an amplifier IP3 just using a DSO and a two tone test signal using this technique. Obviously, there are better ways to do it but even a basic entry level Tek DSO like mine can do a few useful tricks :)

That probably will not work accurately for real world because:

1.) Almost no Ham has a scope that can do that accurately. The scope would have to resolve a bandwidth of a few Hz that is 30-50 dB down from fundamental, while looking at few kHz snapshot of a 1.8-30 MHz waveform.

2.) Speech is dynamic, changing all the time. A two tone test is static. The class C bias will only cause a very slight knock in the waveform when the two tones are out-of-phase and virtually equal. The two tones must be EQUAL, or they won't cross zero and the bias error will not show!

Unless your scope can do multiple sweeps and be able to learn and display the peak and average values of narrow band slices in a bandwidth around the operating frequency while the signal has dynamic changes similar to voice, or better an actual voice, the data doesn't mean much.

When I look at a rig with a spectrum analyzer, I have to use a 50-100 sweep average and peak, a really slow sweep, a few hundred Hz or less bandwidth, and do it with voice or a three tone test that roughly simulates voice. Any test has to have a syllabic modulation component or it does not check regulation or bias dynamics.

It actually is far easier and much more accurate to just look at it with a very good selective level meter with voice or three tone, if we want to know the truth. If I want to make someone happy, I can print a plot from the spectrum analyzer the way they want, but it generally does not actually honestly indicate performance.

Even a notched noise test will not simulate voice performance, unless the noise is turned off and on at a syllabic rate.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 23, 2015, 09:04:35 AM
Quote
1.)  The scope would have to resolve a bandwidth of a few Hz that is 30-50 dB down from fundamental, while looking at few kHz snapshot of a 1.8-30 MHz waveform.
Look up bandpass sampling. The waveform is only a few kHz wide at RF so only needs to be sampled at a rate of maybe 50kHz to capture all of the info. So even a cheap/common DSO can do this as long as it has a reasonable MATH/FFT function.

Quote
Almost no Ham has a scope that can do that accurately.
Define what you mean by accurately. The DSO I have here is old and has little value today but it can refresh at maybe 2-5 sweeps a second and show a decent spectrum of a two or n tone signal.

Alternatively I could mix the signal down to LF and do it all with a PC based scope/analyser and sample the data in real time using its onboard soundcard. How cheap does the solution have to be?

Quote
When I look at a rig with a spectrum analyzer, I have to use a 50-100 sweep average and peak, a really slow sweep, a few hundred Hz or less bandwidth, and do it with voice or a three tone test that roughly simulates voice. Any test has to have a syllabic modulation component or it does not check regulation or bias dynamics.

Then (in your case) buy a decent/modern spectrum analyser or a modern scope that can capture this in real time. This isn't the 1970s any more and test gear that can do this is very cheap nowadays :)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 23, 2015, 09:45:09 AM
My old DSO is a very common entry level scope and it represents the bottom of the barrel today because these scopes are outclassed today. But even it can make a reasonable stab at a spectrum plot of a small test amplifier driven by two test tones up at 28MHz. The scope is set to undersample the signal. It can't measure the frequency of the signal very well in this mode but this signal is made up of two test tones spaced about 500Hz apart at about 28.5MHz.

The display is scaled at 10dB/div and the span is 12.5kHz. At this narrow span it does several screenupdates per second. Not quite fluid but still very good :)

I also compared the IMD against an old HP/Agilent spectrum analyser and it agreed quite well. Certainly close enough for ham stuff.

I'm guessing my scope is worth maybe $200 today although some people would argue it's worth a lot less because it is so old and slow with limited sample memory etc.


(http://i1227.photobucket.com/albums/ee439/G0HZU/dso_fft_ssb_28.jpg) (http://s1227.photobucket.com/user/G0HZU/media/dso_fft_ssb_28.jpg.html)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W0BTU on August 23, 2015, 11:22:40 AM
We can talk about the right way to use good test equipment for this purpose all we want to. The advice given here by several technically knowledgeable hams is all well and good and much appreciated. However, the stark reality is that the vast majority of hams are either unable or disinclined to do this.

What W8JI said here about using a good receiver to check bandwidth is absolutely correct. It's THE choice piece of test equipment at any price for hams like those.

Case in point:
I've been unsuccessfully trying to explain to an Extra class fellow --via many e-mail exchanges-- how to to use an ohmmeter to check for a short or open in the coax coming from his Beverage antenna. He's a nice fellow and has other talents, but he needs someone to come to his house and help him. He has absolutely no concept of what the reading on his ohmmeter means. There are many others like him. That's just a fact. And there are people with scopes and other test equipment that are every bit as clueless. That's just a fact, as well.

I don't know how else to put it. Take a good hard look at https://en.wikipedia.org/wiki/Dunning%E2%80%93Kruger_effect#Original_study because they're still splattering, my friends.  ;D


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 23, 2015, 11:49:55 AM
Quote
However, the stark reality is that the vast majority of hams are either unable or disinclined to do this.

To dismiss my input because a few hams don't care or can't learn this stuff is missing the point. Some hams will want to be able to do this and these hams are getting advice from people who are out of touch.

The whole point of ham radio is for self learning and self improvement (for all ages) and the stuff I'm posting up here could be replicated for very little money for anyone keen enough to try.

The cheapest is a homebrew downconverter to a PC soundcard. The soundcard technology for this PC based system has been around for decades and it could be built for little more than the cost of a beer assuming a PC/laptop with a soundcard is available.

A keen operator could sample and capture the spectrum and play it back and pause it looking at key clicks or SSB modulation splatter from a two tone test or from speech. They could even email it to someone else to look at.

It's 2015 not 1975. Not every ham is a chequebook CB type operator. Many will want to learn new ways to test what they are designing or using. Not every ham will be using an old analogue scope. Some (many?) will have a scope better than my cheap and old/obsolete Tek TDS. Although W8JI seems to think otherwise.






Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SWL2002 on August 23, 2015, 11:54:44 AM
Quote
However, the stark reality is that the vast majority of hams are either unable or disinclined to do this.

To dismiss my input because a few hams don't care or can't learn this stuff is missing the point. Some hams will want to be able to do this and these hams are getting advice from people who are out of touch.

The whole point of ham radio is for self learning and self improvement (for all ages) and the stuff I'm posting up here could be replicated for very little money for anyone keen enough to try.

The cheapest is a homebrew downconverter to a PC soundcard. The soundcard technology for this PC based system has been around for decades and it could be built for little more than the cost of a beer assuming a PC/laptop with a soundcard is available.

A keen operator could sample and capture the spectrum and play it back and pause it looking at key clicks or SSB modulation splatter from a two tone test or from speech. They could even email it to someone else to look at.

It's 2015 not 1975. Not every ham is a chequebook CB type operator. Many will want to learn new ways to test what they are designing or using. Not every ham will be using an old analogue scope. Some (many?) will have a scope better than my cheap and old/obsolete Tek TDS. Although W8JI seems to think otherwise.






Well stated... AMEN!

http://www.rigolna.com/products/digital-oscilloscopes/ds1000Z/ds1054z/ (http://www.rigolna.com/products/digital-oscilloscopes/ds1000Z/ds1054z/)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on August 23, 2015, 12:06:10 PM
The whole point of ham radio is for self learning and self improvement (for all ages) and the stuff I'm posting up here could be replicated for very little money for anyone keen enough to try.

This also happens to be one of the reasons that the ITU still allocates spectrum for amateur radio.

However, when the full truth about thë ongoing transformation of amateur radio into CB
due to deregulation and rapidly declining technical and operating standards becomes known
to the ITU and the national Administrations they may become less inclined to do so.

73/
Karl-Arne
SM0AOM


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 23, 2015, 01:52:14 PM
The whole point of ham radio is for self learning and self improvement (for all ages) and the stuff I'm posting up here could be replicated for very little money for anyone keen enough to try.

This also happens to be one of the reasons that the ITU still allocates spectrum for amateur radio.

However, when the full truth about thë ongoing transformation of amateur radio into CB
due to deregulation and rapidly declining technical and operating standards becomes known
to the ITU and the national Administrations they may become less inclined to do so.

73/
Karl-Arne
SM0AOM


Yes, it's all a bit sad really :(

This forum and QRZ.com often appear to be stuck in a 70s time capsule. New hams need to be encouraged to exploit modern and cheap and readily available technologies to advance the hobby into the future :)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W0BTU on August 23, 2015, 01:58:13 PM
Quote
However, the stark reality is that the vast majority of hams are either unable or disinclined to do this.

To dismiss my input because a few hams don't care or can't learn this stuff is missing the point.

I am not dismissing your input. As I said, "The advice given here by several technically knowledgeable hams [you included] is all well and good and much appreciated. "

Things might be different in the USA vs. the rest of the world. I wrote what I did based on my observations here in the central USA, and I stand by my statements.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KM1H on August 23, 2015, 02:59:19 PM
Quote
We can talk about the right way to use good test equipment for this purpose all we want to. The advice given here by several technically knowledgeable hams is all well and good and much appreciated. However, the stark reality is that the vast majority of hams are either unable or disinclined to do this.

The true stark reality is your strong biases for one person Mike. EHAM doesnt revolve around him for technical prowess; those days are over.

Quote
What W8JI said here about using a good receiver to check bandwidth is absolutely correct. It's THE choice piece of test equipment at any price for hams like those.

Lots of things mentioned on this thread are correct; OTOH expecting a transceiver bound ham these days to invest in a separate receiver capable of doing the job (which includes an AM filter bandwidth at a minimum) is ludicrous when a well under $50 scope does the job with visual proof and not just an uncalibrated S meter. Even 1980-90's SA's have come way down in price as digital has replaced them in ever increasing numbers. Analog is dead in the corporate engineering world, I bought my HP 8568B at a company auction for $525 which included various options and a complete manual set, recent overhaul and current cal sticker. Big and heavy but it does the job and I also run it off a bucking transformer at 110 VAC to minimize heat and maximize life.

Im real glad to see G0HZU, SM0AOM, G3RZP and others speak up in response to JI's rather self centered view of things and a few knee pad wearing followers.

Carl


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W0BTU on August 23, 2015, 03:06:06 PM
Carl,

Don't say that I have a bias for anyone. W8JI has said things that I absolutely disagreed with. If Tom Rauch had never even been born, I would still have believed what I just did.

There's nothing wrong with you, me, or the others in this thread suggesting using other methods beside a receiver, but I maintain that's a better method for the majority than the way most hams use their scopes.

Have a good afternoon, Carl.  :)

73 Mike

Quote
We can talk about the right way to use good test equipment for this purpose all we want to. The advice given here by several technically knowledgeable hams is all well and good and much appreciated. However, the stark reality is that the vast majority of hams are either unable or disinclined to do this.

The true stark reality is your strong biases for one person Mike. EHAM doesnt revolve around him for technical prowess; those days are over.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KM1H on August 23, 2015, 03:23:53 PM
Quote
Don't say that I have a bias for anyone. W8JI has said things that I absolutely disagreed with. If Tom Rauch had never even been born, I would still have believed what I just did.

Your bias has been obvious for years Mike even when others have stated otherwise in a tech forum on several sites.

Quote
There's nothing wrong with you, me, or the others in this thread suggesting using other methods beside a receiver, but I maintain that's a better method for the majority than the way most hams use their scopes.

You skated over that real fast. How about a nice list of suitable receivers for starters...with AM at an absolute minimum or wider filters included to get well beyond IMD3 ?

As far as scopes how do you know what others are using? Most of what I see mentioned are Kenwood and Heath monitors and ancient scopes which we all know are useless for IMD measurements and as soon as G0HZU suggested an affordable modern scope your knees hit the floor in support of your hero without having a clue to the scopes capability.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W0BTU on August 23, 2015, 03:28:05 PM
Respectfully, I have never seen anyone who has so much hate. That's not the way I was brought up.

I am done discussing this.

(http://www.eham.net/html/ehamforum/smf/Themes/eham/images/english/ignore.gif)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W8JI on August 23, 2015, 03:49:34 PM
This went off the rails pretty quick. Let's try to bring it back on track.

Quote
When I look at a rig with a spectrum analyzer, I have to use a 50-100 sweep average and peak, a really slow sweep, a few hundred Hz or less bandwidth, and do it with voice or a three tone test that roughly simulates voice. Any test has to have a syllabic modulation component or it does not check regulation or bias dynamics.

Then (in your case) buy a decent/modern spectrum analyser or a modern scope that can capture this in real time. This isn't the 1970s any more and test gear that can do this is very cheap nowadays :)

This is where I think we are not communicating. I must not be making my point well.  let's see if we can stay on point.

I have relatively modern test equipment. The issue is in the signal and nature of the splatter, not in the equipment. First, the topic was bias on a solid state amp. The linearity we are concerned with is a transfer function issue, and the lack of bias causes a linearity or transfer function issue when the envelope crosses zero. If the test envelope does not cross zero, perhaps because the tones are not equal, there won't be significant distortion.

A two tone test, or a notched noise test, also loads the bias or other supply circuits pretty steady with the tone difference frequency. This isn't anything like a voice does.

This is why I generally use a voice (or three tone at a syllabic rate plus two tone at normal pitch differences) and have to do a lot of sweeps. I have to catch the distortion.

I'm not saying the FFT system won't work. What I am saying is the two tone test isn't a very good test when done correctly, and most people could never do it correctly with the equipment they have, although they almost always think no flat topping means no splatter.

By the way, what is your old DSO? I wonder what the dynamic range is. I think you said 60 dB DR, but that seems pretty good for cheap gear. I was not able to do that well with some new Agilent stuff I tried out. Admittedly, it was Chinese manufactured.

73 Tom


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SWL2002 on August 23, 2015, 04:31:20 PM
Respectfully, I have never seen anyone who has so much hate. That's not the way I was brought up.

I am done discussing this.

(http://www.eham.net/html/ehamforum/smf/Themes/eham/images/english/ignore.gif)

What is wrong with big baby's like the one above? They put you on ignore if you dare disagree with them or you contradict something they have said.  Argument is good - more things are can be learned from heated disagreement than everyone huddled together in agreement with flowers in their hair singing Kumbaya and wearing rose colored glasses. 


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 23, 2015, 04:37:51 PM
This went off the rails pretty quick. Let's try to bring it back on track.

Quote
When I look at a rig with a spectrum analyzer, I have to use a 50-100 sweep average and peak, a really slow sweep, a few hundred Hz or less bandwidth, and do it with voice or a three tone test that roughly simulates voice. Any test has to have a syllabic modulation component or it does not check regulation or bias dynamics.

Then (in your case) buy a decent/modern spectrum analyser or a modern scope that can capture this in real time. This isn't the 1970s any more and test gear that can do this is very cheap nowadays :)

This is where I think we are not communicating. I must not be making my point well.  let's see if we can stay on point.

I have relatively modern test equipment. The issue is in the signal and nature of the splatter, not in the equipment. First, the topic was bias on a solid state amp. The linearity we are concerned with is a transfer function issue, and the lack of bias causes a linearity or transfer function issue when the envelope crosses zero. If the test envelope does not cross zero, perhaps because the tones are not equal, there won't be significant distortion.

A two tone test, or a notched noise test, also loads the bias or other supply circuits pretty steady with the tone difference frequency. This isn't anything like a voice does.

This is why I generally use a voice (or three tone at a syllabic rate plus two tone at normal pitch differences) and have to do a lot of sweeps. I have to catch the distortion.

I'm not saying the FFT system won't work. What I am saying is the two tone test isn't a very good test when done correctly, and most people could never do it correctly with the equipment they have, although they almost always think no flat topping means no splatter.

By the way, what is your old DSO? I wonder what the dynamic range is. I think you said 60 dB DR, but that seems pretty good for cheap gear. I was not able to do that well with some new Agilent stuff I tried out. Admittedly, it was Chinese manufactured.

73 Tom

OK but I'm still confused why you think no one will have a scope that could do a simple FFT analysis. My old scope is pretty flaky by today's standards but it does OK with human speech as you can see below. It's an 8bit Tek TDS2012 100MHz scope. It's spurious free dynamic range is quite limited but it is still useful for making casual measurements and its display refresh rate is much faster than an old school swept analyser on spans like this giving it a useful edge.


I'm afraid I'm not a very good speaker/presenter (and it's nearly bedtime) but I did a quick youtube video showing a downconverter + laptop vs my old scope vs a fairly old Agilent signal analyser and a Yaesu FT290R. Obviously the scope is only a 100MHz model so it is a bit deaf by 144MHz but it still does OK. It also has a display hold feature to capture and hold the peak response in infinite persistence mode (not shown in the video).

You can see how effective the soundcard option is. In my opinion the scope provides useful info even though it isn't quite as fluid as the other two. The camera will probably add some jerkiness to the laptop and the Agilent display so they will be a bit more fluid than they appear in the video. But I think it demonstrates that a cheap DSO can provide useful info about signal bandwidth.

https://www.youtube.com/watch?v=hdtQL1MrhHo&feature=youtu.be



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SWL2002 on August 23, 2015, 04:49:15 PM

https://www.youtube.com/watch?v=hdtQL1MrhHo&feature=youtu.be



Very interesting.  Thanks for posting this.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KB4QAA on August 23, 2015, 04:52:00 PM
Respectfully, I have never seen anyone who has so much hate. That's not the way I was brought up.

I am done discussing this.

(http://www.eham.net/html/ehamforum/smf/Themes/eham/images/english/ignore.gif)

What is wrong with big baby's like the one above? They put you on ignore if you dare disagree with them or you contradict something they have said.  Argument is good - more things are can be learned from heated disagreement than everyone huddled together in agreement with flowers in their hair singing Kumbaya and wearing rose colored glasses. 
What is wrong is the unwarrranted hate and venom being spewed here.   Technical remarks and long user experienced are welcomed and valued.

Vile and baseless personal attacks are not.   Keep it professional and courteous folks!


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SWL2002 on August 23, 2015, 04:54:05 PM
Respectfully, I have never seen anyone who has so much hate. That's not the way I was brought up.

I am done discussing this.

(http://www.eham.net/html/ehamforum/smf/Themes/eham/images/english/ignore.gif)

What is wrong with big baby's like the one above? They put you on ignore if you dare disagree with them or you contradict something they have said.  Argument is good - more things are can be learned from heated disagreement than everyone huddled together in agreement with flowers in their hair singing Kumbaya and wearing rose colored glasses. 
What is wrong is the unwarrranted hate and venom being spewed here.   Technical remarks and long user experienced are welcomed and valued.

Vile and baseless personal attacks are not.   Keep it professional and courteous folks!

So sorry there are so many shrinking violets here on eHam.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 23, 2015, 05:09:59 PM
I also did a very brief video showing a comparison on a two tone test of a little amplifier module.

The scope FFT is compared to the HP/Agilent signal analyser and it does OK here in my opinion if you look at the tone levels on both instruments. Some of the subtle differences could be down to the quality of the 50R termination used on the scope. I don't use the scope like this as often as I used to but I've had good service from this old scope for over 10 years now. It was secondhand when I bought it so I don't know how old it is.

I can also generate more than two tones here (one sig gen can generate up to 64 tones/carriers on its own) and the scope does OK with 3 or 4 or 8 test tones.

https://www.youtube.com/watch?v=kksRWwY9v6E

Also note that the spectrum analyser software used on that laptop is over 15 years old. It's an ancient trial program I dug out and installed on the laptop for the video. There are newer/better version of this SW available today or you could write your own analyser software for the soundcard using readily available DSP libraries etc. I was tempted to use my own software for the laptop analyser demo but it's not as pretty or as versatile as the demo SW used in the video. The trial SW can capture and record/play back the samples in slow or real time and you can save the captured samples as a wav file etc and do any post processing/analysis you want. This would be most useful for looking at key clicks because this software can show dual displays for the time domain and also frequency domain as well as doing waterfall displays.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KM1H on August 23, 2015, 05:46:57 PM
Quote
Respectfully, I have never seen anyone who has so much hate. That's not the way I was brought up.

You actually believe anyone believes that "respectfully" ploy?

Quote
I am done discussing this.

That is the way your type try to get out of a bucket of s... they stepped into. I save my hate for those who deserve it and are in the news often. On here it is mostly pity at the weak knee shakers.

I was brought up in the Brooklyn-NYC-LI area where discussions got heated even in school. When I moved to MA and NH and watched local politics then I saw the bar raised even more. I guess I have contempt for those without a backbone to fight for what they believe in even if I disagree with them.

Im still waiting for that receiver list.




Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KB4QAA on August 23, 2015, 05:49:52 PM
Quote
Respectfully, I have never seen anyone who has so much hate. That's not the way I was brought up.

You actually believe anyone believes that "respectfully" ploy?

Quote
I am done discussing this.

That is the way your type try to get out of a bucket of s... they stepped into. I save my hate for those who deserve it and are in the news often. On here it is mostly pity at the weak knee shakers.

I was brought up in the Brooklyn-NYC-LI area where discussions got heated even in school. When I moved to MA and NH and watched local politics then I saw the bar raised even more. I guess I have contempt for those without a backbone to fight for what they believe in even if I disagree with them.
Karl,
Name calling and insults are not dicussions nor even arguing.  Chill out.  Put on some manners.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KM1H on August 23, 2015, 05:53:02 PM
Quote
Karl,
Name calling and insults are not dicussions nor even arguing.  Chill out.  Put on some manners.

If you cant contribute to the discussions just mind your own business instead of a grandstanding broadcast to the full forum.

And the name is CARL as you should know after all these years.......my grandfather was Karl.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W8JI on August 24, 2015, 05:55:22 AM
OK but I'm still confused why you think no one will have a scope that could do a simple FFT analysis. My old scope is pretty flaky by today's standards but it does OK with human speech as you can see below. It's an 8bit Tek TDS2012 100MHz scope. It's spurious free dynamic range is quite limited but it is still useful for making casual measurements and its display refresh rate is much faster than an old school swept analyser on spans like this giving it a useful edge.

Based on what I see dealing with Hams, only a few have scopes. I think we've been through this a few times, but it might have been buried in the sideshow noise (which tends to take all useful value out of discussions).

Anyway, those that do have scopes look at waveform. They generally focus on the tops, but the tops are only one part of the issue. A class C amplifier can look good at the top, which is what started my comments.

Quote
I'm afraid I'm not a very good speaker/presenter (and it's nearly bedtime) but I did a quick youtube video showing a downconverter + laptop vs my old scope vs a fairly old Agilent signal analyser and a Yaesu FT290R. Obviously the scope is only a 100MHz model so it is a bit deaf by 144MHz but it still does OK. It also has a display hold feature to capture and hold the peak response in infinite persistence mode (not shown in the video).

Regular scopes are useless for determining bandwidth. All of the DSO's I've seen are limited to roughly about 40 dB dynamic range, with older units much worse. I certainly agree they are much better than a regular scope (which is useless), but someone can generally do better for this application in this thread with a typical fairly good receiver in a proper setup.
Quote
You can see how effective the soundcard option is. In my opinion the scope provides useful info even though it isn't quite as fluid as the other two. The camera will probably add some jerkiness to the laptop and the Agilent display so they will be a bit more fluid than they appear in the video. But I think it demonstrates that a cheap DSO can provide useful info about signal bandwidth.

Thanks for the video. I was trying to offer technical suggestions to the person looking at his amplifier with equipment he might have, before this thread derailed into the Brooklyn Syndrome show. It's a phenomena explained here:

http://content.time.com/time/magazine/article/0,9171,852006,00.html

or

https://news.google.com/newspapers?nid=1964&dat=19430912&id=k1kyAAAAIBAJ&sjid=27YFAAAAIBAJ&pg=954,4855471&hl=en

I've tried a few DSO's (several years ago) but:

1.) They don't change the basic problems with testing

2.) They had pretty limited dynamic range

It was relatively easy to build amplifiers beyond the ability of the DSO's measurement dynamic range.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: JS6TMW on August 24, 2015, 08:02:14 AM
All this makes me want to just go on the air and SPLATTER!


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on August 24, 2015, 08:12:18 AM
Good grief. There is nothing wrong with using a cheap monitor scope to ensure that flat topping is not occurring!

Even the best or worst amplifiers will be infinitely worst when driven into compression.  The monitor scope assumes that the amplifier is properly biased, and properly loaded... that is the responsibility of the amplifier's owner, not the scope display. As is over driving the amplifier. The scope is a tool, not a guarantee that all other "systems are a go..."

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 24, 2015, 09:33:21 AM
Quote
Good grief. There is nothing wrong with using a cheap monitor scope to ensure that flat topping is not occurring!

I agree that this is often a useful indicator of problems. When I was student I used to repair and service SSB ham/CB radios to supplement my income and I used an old Tek 585 scope and a homebrew two tone generator.
A badly set up amplifier was easy to spot even using this basic setup if looking for flat topping or bias/crossover issues as this scope had very good optics. The Tek DSO below improves on this a lot and allows the two tone (or n tone) waveform to be analysed in the frequency domain for issues and will give a useful indicator of bandwidth with speech on both FM and SSB.


To show this I had another go at setting up the camera and the scope to show how fluid the scope can be when showing a test signal. This time I've used a signal within the scopes range at 10MHz and optimised the camera as much as I can.

This signal is a clean two tone signal direct from a pair of sig gens to show the residual distortion levels of the test system. I think this is a useful system for checking an amplifier for classic IMD terms.
Note again that my scope is OLD and was a cheap entry level scope in its day. Modern scopes will show a much lower noise floor than this slow old clunker of mine.
This crude setup is capable of displaying similar two tone plots as found in the ARRL lab tests and it can display voice patterns quite well too. Sometimes the scope can pause or lag briefly and miss some data but if viewed for a few seconds it can show a lot of useful info. it can also show the bandwidth of an FM signal very well, especially if looking for Bessel nulls if you want to calibrate a deviation meter using this scope :)

The display looks very fluid in this video and is closer to what is seen with the naked eye. it's just as fluid with extra tones added but you have to allow extra headroom for the PEP. eg 18dB headroom for 8 tones.
Note that I've selected the Hanning window in this video rather than the preferred 'flat top' but it doesn't make much visible difference in this test so I'll leave it as it is rather than reshoot the video.

Newer scopes will perform much better than my old scope.

In the hobby electronics world my scope is classed as ancient junk. Slow, limited memory depth, poor display and no data capture features etc. But it still does OK in the video below.

https://www.youtube.com/watch?v=TpHbF6O8T7Y&feature=youtu.be


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 24, 2015, 01:20:49 PM
Quote
W8JI  1.) Almost no Ham has a scope that can do that accurately. The scope would have to resolve a bandwidth of a few Hz that is 30-50 dB down from fundamental, while looking at few kHz snapshot of a 1.8-30 MHz waveform.

Hopefully you can now see that even an incredibly common and cheap/basic/old DSO can do just this if you exploit the MATH/FFT feature and set up the instrument to bandpass sample the signal at maybe 100ksps. Note that if the transmitter has out of band spurious or harmonic content above -60dBc then these can appear on the screen at the same time as alias/ghost terms but they can usually be managed/pushed off screen once the user has a basic understanding of how to use the scope in this mode. It really is quite easy to spot and manage any unwanted alias terms that may creep in.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W8JI on August 24, 2015, 05:49:33 PM
Good grief. There is nothing wrong with using a cheap monitor scope to ensure that flat topping is not occurring!

Even the best or worst amplifiers will be infinitely worst when driven into compression.  The monitor scope assumes that the amplifier is properly biased, and properly loaded... that is the responsibility of the amplifier's owner, not the scope display. As is over driving the amplifier. The scope is a tool, not a guarantee that all other "systems are a go..."

Pete

Once again, the point of all this that was lost in the noise is the fellow tried to see problems with a scope in class C. This is the problem when a thread gets personal, and people start arguing about unrelated personal crap.

What the original person (who is probably long gone) said was he looked at a class C amp on a scope and it looked good.  Lack of quiescence current does not cause flat topping, it causes distortion where down near zero on the envelope where it is almost impossible to see, but it can cause very wide band splatter.

All that being said, a scope like that is a poor way to ensure good bandwidth. It doesn't measure bandwidth. It can look flat, and not be wide, it can look peaky, and be wide.

A good receiver properly used is a better way to determine bandwidth issues than a regular scope. Even a DSO with FFT is not that great. They generally have very limited dynamic range, not many people have them, and the nature of splatter or excessive bandwidth is that a two-tone will not show many problems.

I don't think this is all that complicated. I am puzzled why rational point discussion is so difficult.

73 Tom


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KM1H on August 24, 2015, 06:08:12 PM
Quote
I don't think this is all that complicated. I am puzzled why rational point discussion is so difficult.

It becomes that way when you cant even remember what the opening post was about and have spun it every way possible since. At least in Brooklyn I was taught reading comprehension and did extremely well later in the USN.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 24, 2015, 07:13:39 PM
Quote
I am puzzled why rational point discussion is so difficult.

You kicked this off by saying that a scope can only provide extremely limited info about bandwidth. That sounds like a remark from the 1970s or 1980s :)

So I gave you an example of how I've been using my cheap old DSO to measure signal bandwidths for over a decade here at home.

So you then asked me this:
Quote
If you can explain it, and most people can do it, you are correct

So what am I expected to do here apart from try and demo the concept?

I then tried to explain it in a way I'd expect a typical DSO user to understand but you replied saying this:

Quote
1.) Almost no Ham has a scope that can do that accurately. The scope would have to resolve a bandwidth of a few Hz that is 30-50 dB down from fundamental, while looking at few kHz snapshot of a 1.8-30 MHz waveform.

I interpreted from this that you didn't understand my explanation or maybe you don't know the capabilities of a DSO or maybe you thought I was referring to some kind of $$$ super scope. So I then showed you how a common and cheap DSO can be exploited to do this over the 30-50dB range that you asked for using the bandwidth resolution that you asked for with a 'few kHz' span that you asked for with a test signal in the 1.8-30MHz signal range that you asked for.

You also asked for this:

Quote
Unless your scope can do multiple sweeps and be able to learn and display the peak and average values of narrow band slices in a bandwidth around the operating frequency while the signal has dynamic changes similar to voice, or better an actual voice, the data doesn't mean much.

I showed you it can display a fairly fluid spectrum of human speech in a video. It can also store the peaks in its infinite persistence mode which is a kind of 'max hold' feature. Sure, it can't match the Agilent signal analyser for speed or dynamic range but it can produce two tone plots and IMD bandwidth info that are similar to the plots produced by the ARRL and it can do it in real time allowing one to tweak the PA and observe the two tone IMD terms change in real time. You can't do this with a conventional swept spectrum analyser because it sweeps too slowly with a narrow RBW.

Just to repeat. a degree of caution is needed with this method because the display will also show out of band spurious or harmonics on the display as an onscreen alias unless the user manages the scope etc to push these (potentially confusing) terms off the scope display. But this isn't difficult to do once the user understands the limitations of the scope.



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: JS6TMW on August 25, 2015, 02:21:10 AM
Quote
I don't think this is all that complicated. I am puzzled why rational point discussion is so difficult.

It becomes that way when you cant even remember what the opening post was about and have spun it every way possible since. At least in Brooklyn I was taught reading comprehension and did extremely well later in the USN.

The opening post was about key clicks, not SSB splatter. (I also learned to read in Brooklyn - PS138 to be precise). DL8OV posted actual on-air observations that were scary but useful to me.

About halfway through this pileon, I mentioned that I had seen a Youtube of a Class C amp used on SSB monitored with a simple scope and the envelope looked OK to the operator.  And yes, I am still around. The debate over instrumentation that followed was not in the least interesting to me because I will never own any gear like that.

I am going to use the same Class C amp on CW only, but if it seems like I generate keyclicks, I know how to bias it into linear modes.

Over and out.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SWL2002 on August 25, 2015, 03:53:57 AM

You kicked this off by saying that a scope can only provide extremely limited info about bandwidth. That sounds like a remark from the 1970s or 1980s :)


Some "experts" are stuck back in the past and poo poo such things as digital storage oscilloscopes and software defined radios.  Unfortunately, they also tender their "expert" opinions on things they are not "experts" on at all.

I, for one, appreciate your efforts to demonstrate your point as well as your Youtube videos. 


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W8JI on August 25, 2015, 08:31:14 AM
The opening post was about key clicks, not SSB splatter. (I also learned to read in Brooklyn - PS138 to be precise). DL8OV posted actual on-air observations that were scary but useful to me.

About halfway through this pileon, I mentioned that I had seen a Youtube of a Class C amp used on SSB monitored with a simple scope and the envelope looked OK to the operator.  And yes, I am still around. The debate over instrumentation that followed was not in the least interesting to me because I will never own any gear like that.

I am going to use the same Class C amp on CW only, but if it seems like I generate keyclicks, I know how to bias it into linear modes.

Over and out.

That's what I thought. I thought the debate about instrumentation took the thread totally off track. I'm not saying a DSO with FFT or SDR won't work, although some people seem obviously to enjoy that concept as means of making this a little personal discussion about people, rather than useful technical info.

What I am saying, and this is important, is that a standard scope pattern will not tell you much, if anything, about bandwidth. The problem is wide bandwidth is not indicated or caused by being somewhat flat across the top, which is what most people look for.

Wide bandwidth is caused when the envelope rises very fast, or rolls over very fast, not when it is flat horizontally on the scope. Understanding most people are not going to run out and buy a FFT capable scope and do an almost worthless two-tone test, we have to look carefully for a very sharp rise or roll-over in the envelope waveform. How rapidly the envelope changes slope at any point is tied to the ultimate bandwidth.  How tall that rapid slope change area is determines the level, but even a very short height (voltage change) that is difficult to see can be significant sideband power.

If you have a transistor stage with zero bias, it makes the sharp rise near zero level. This is because the transistor is off until the RF level gets to a certain point, and then abruptly turns on. This decreases the zero to full power rise time and full power to zero fall time by changing the very bottom, and that is where the clicks (or splatter) is formed.

It is almost always much better to just listen with a receiver. A reasonably good receiver with narrow filter, that is not overloaded, should not hear clicks further away than you can hear the tone. The tone should not be "clicky".

I hope this is useful.

As a final point,  most Hams think if the waveform is not flat at the top, there won't be splatter and everything is good. That isn't true. The flat part doesn't cause or indicate bandwidth. The sharpening of the envelope rise and fall transitions are what actually cause and indicate bandwidth issues.

Since the issues are in rising and falling areas of the envelope, that is where we have to look.

73 Tom   


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on August 25, 2015, 10:23:23 AM

Once again, the point of all this that was lost in the noise is the fellow tried to see problems with a scope in class C. This is the problem when a thread gets personal, and people start arguing about unrelated personal crap.

What the original person (who is probably long gone) said was he looked at a class C amp on a scope and it looked good.  Lack of quiescence current does not cause flat topping, it causes distortion where down near zero on the envelope where it is almost impossible to see, but it can cause very wide band splatter.

  is all that complicated. I am puzzled why rational point discussion is so difficult.

73 Tom

This why it is wise to do a quote before posting a response that may be off topic, or directed to a response that doesn't follow the original intent of the thread. 

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 25, 2015, 02:13:56 PM
Quote
This why it is wise to do a quote before posting a response that may be off topic, or directed to a response that doesn't
follow the original intent of the thread.  


I think you should be listening to G3RZP's earlier posts because he answered your query quite early in the thread IMO.

However, my input to this thread was intended to show how a common hobby DSO scope can be used to actually analyse waveforms like this for bandwidth. i.e. you don't need an expensive modern analyser or a slow old swept analyser like W8JI was using for his SSB analysis.

If it wasn't for the fact I had to spoonfeed W8JI (across loads of posts) about the math capabilities of an ancient and very common hobby DSO then we could have fast forwarded to using the DSO to look at the response of a typical keyed cw signal after the SSB stuff.

The scope can show the close in modulation sidebands and the bandwidth of the keyed cw signal in FFT mode. i.e. it can zoom in to the RF waveform using bandpass sampling and view it at 25Hz per division or a 250Hz span. it ideally requires the cw to be sent electronically at a precise rate but this is quite valid for the analysis.

It could also show the effect of altering the pulse rise/fall time/shape away from a pure OOK format and presumably show what happens through a well (or poorly) biased amplifier as this is done. The resolution bandwidth of the scope is just a few Hz so it can show the tiny occupied BW of the keyed cw signal really well and show the relative levels of each of the cw sidebands in dB with a useful refresh rate of 1-2 a second.

There's one sideband every few Hz so there loads of them to look at :)

If you don't have access to a scope like this then try using the mixer + PC/soundcard method because this type of PC based scope/analyser can be really cheap to set up and use and it will be faster and will have more dynamic range than an older scope like mine :)

I thought this was what your topic question related to?








Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 25, 2015, 02:38:53 PM
I, for one, appreciate your efforts to demonstrate your point as well as your Youtube videos. 

Thanks for that :)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SWL2002 on August 25, 2015, 02:50:50 PM

However, my input to this thread was intended to show how a common hobby DSO scope can be used to actually analyse waveforms like this for bandwidth. i.e. you don't need an expensive modern analyser or a slow old swept analyser like W8JI was using for his SSB analysis.

If it wasn't for the fact I had to spoonfeed W8JI (across loads of posts) about the math capabilities of an ancient and very common hobby DSO then we could have fast forwarded to using the DSO to look at the response of a typical keyed cw signal after the SSB stuff.


I don't get why W8JI keeps talking about seeing (or not seeing) flat topping on a scope, and how useless that is, when you are talking about using the FFT functions of a DSO, not looking at the time domain waveform. 


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 25, 2015, 04:48:54 PM
TBH I'm not sure it was worth me trying because I don't think many other people appreciated my input...

However, the scope or the soundcard method is really powerful and I've been using it for many years to look at LO phase noise or SSB or narrowband FM.

It can also be extended up into the UHF and microwave bands with a suitable mixer and local oscillator. The LO doesn't have to be a commercial sig gen either. For FM or SSB use it could be a simple/cheap/homebrew free running oscillator because a bit of drift wouldn't really matter than much if the user was able to tolerate this visually after a suitable warmup time. Obviously a PLL based LO would be a lot better but that would cost more.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: JS6TMW on August 25, 2015, 06:44:13 PM
I don't have any desktop PCs anymore. Is there a USB soundcard that will meet the requirements of spectrum-analysis software?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W8JI on August 25, 2015, 07:49:36 PM
I don't get why W8JI keeps talking about seeing (or not seeing) flat topping on a scope, and how useless that is, when you are talking about using the FFT functions of a DSO, not looking at the time domain waveform. 

Well, what happened here is a mix of things. First, there is personal ranting about people.  Not technical facts or disagreements, just name calling or insults. That just wastes space, and makes any technical exchange difficult.

For myself, I don't care about that. It is so old and tiring to me, I just ignore it. The problem is when I ignore it I might miss an actual technical point, but I don't really care other than I don't it belongs in a technical forum for the people interested in learning.

Someone brought up a scope and SSB with their example. When it is SSB, the issue for measurements becomes a time problem and a problem simulating voice in a test. This is because speech is not even remotely close to being two steady equal tones, just as CW is not two tones.

For SSB, you can't measure bandwidth with a system that looks at a snapshot of time. A two-tone is bad enough with any system.

A time domain display of voltage with a scope is a pretty poor way to measure clicks, especially clicks caused by not having quiescent current, because the problem is right down near the zero line. I'm not saying it never tells someone when something is bad, just that it won't always be obvious.

I also tried to point out most people look at the wrong thing. They look at the top to see of it is flat, but the flat part has no bandwidth (ignoring frequency drift or things like that). The only part of the envelope producing or requiring bandwidth are the changes in level.

Now I certainly agree that something that captures the envelope waveform, analyzing it and displaying it as a frequency domain is useful within the dynamic range of that system, but when I have tried to use those systems they have been generally poor. I can't take enough sweeps to see ALC issues, and everything I've tried out is not nearly as good as my 8593E.

I can see some cases where a DSO might be better, say capturing one instance of an envelope change with close spaced sidebands (slow rise and fall). A spectrum analyzer has to use very narrow selectivity. Unless it is a repeating waveform and multiple sweeps are used with peak storage, it will miss the sidebands with narrow selectivity. This is because narrow selectivity dictates slow sweep. In a case like this a waveform snapshot with FFT can convert the time domain voltage snapshot into frequency display.

However, for the typical Ham trying to learn WHY a class C amplifier can make clicks (which was the topic of this thread) all the ranting was useless. He isn't going to buy a digital scope with FFT, and he doesn't need to buy one. He can hear what he needs to hear with a receiver, or he could even find it by doing a drive power vs. RF output plot.

We could not really discuss how he could solve his problem, because it was more fun saying I had antique equipment because my 8593E has to be on multiple sweep peak power storage to resolve close spaced sidebands.

Frankly, I think the whole thing became pretty silly. Rather than leaning anything ()which to me is the general idea of a forum), it was like being in bar listening to a bunch of drunks. 

73 Tom


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K6UJ on August 25, 2015, 09:58:12 PM
Respectfully, I have never seen anyone who has so much hate. That's not the way I was brought up.

I am done discussing this.

(http://www.eham.net/html/ehamforum/smf/Themes/eham/images/english/ignore.gif)


Our resident troll SWL2002 scores  again  ;D

Bob
K6UJ





Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SWL2002 on August 26, 2015, 03:23:09 AM
Respectfully, I have never seen anyone who has so much hate. That's not the way I was brought up.

I am done discussing this.

(http://www.eham.net/html/ehamforum/smf/Themes/eham/images/english/ignore.gif)


Our resident troll SWL2002 scores  again  ;D

Bob
K6UJ





Sorry no cigar.  That comment made by W0BTU was directed at Carl, K1MH, not me.  

P.S. please learn to quote correctly.  You need to match each [quote.] with  [/quote.].  Otherwise, it just makes you look dumb.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SWL2002 on August 26, 2015, 03:27:51 AM

However, for the typical Ham trying to learn WHY a class C amplifier can make clicks (which was the topic of this thread) all the ranting was useless. He isn't going to buy a digital scope with FFT, and he doesn't need to buy one. He can hear what he needs to hear with a receiver, or he could even find it by doing a drive power vs. RF output plot.

73 Tom

Got it. Thanks.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: JS6TMW on August 26, 2015, 08:34:39 AM
BUMP

I don't have any desktop PCs anymore. Is there a USB soundcard that will meet the requirements of spectrum-analysis software?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 26, 2015, 09:29:21 AM
BUMP

I don't have any desktop PCs anymore. Is there a USB soundcard that will meet the requirements of spectrum-analysis software?

I can't be sure. The software I'm using is 15 years old and it may not support an external soundcard. But there are plenty of other analysis programs that are available as freeware. Some are written by hams and some may support external USB.

A basic external USB soundcard only cost peanuts on ebay so it may be worth the gamble that it will work. I'm just using the internal soundcard in my laptop. It isn't a good quality example but it's good enough for things like this :)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 26, 2015, 10:06:13 AM
Quote
We could not really discuss how he could solve his problem, because it was more fun saying I had antique equipment because my 8593E has to be on multiple sweep peak power storage to resolve close spaced sidebands.
Frankly, I think the whole thing became pretty silly. Rather than leaning anything ()which to me is the general idea of a forum), it was like being in bar listening to a bunch of drunks.  

I take it from the above that my comments include me as a 'drunk'?

Look, I've worked in the electronics industry all my working life and I'm just offering people the benefit of my professional experience on cheap/common ways to analyse the spectrum bandwidth using modern technology that is relatively cheap. A modern sampling system is going to perform very well for stuff like this. This could be a decent DSO (ideally better than my old DSO) or a simple mixer and soundcard or some other form of PC/USB based scope.

I suspect that you don't know how powerful such a modern and simple system can be? Especially compared to your old HP8593E? I'm trying to help you (and hopefully other people too) but you seem to want to push people like me away?







Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on August 26, 2015, 11:19:33 AM
As the OP who started this thread, I appreciate all input.  And I don't understand the chip-on-the-shoulder attitude in a few recent posts?  I am interested in the DSO scope approach, but it must be remembered that not all of us have similar instrumentation available to us.  I have three HP-141T analyzers that are fully operational with all of the accessory TG options and plugins.. and so far the few 100 MHz analog scopes that I own are up to the task. Not all hams can afford to have one of everything at their disposal.  I would consider buying a good DSO scope, provided I had some assurance that it would be useful to me in my old age!!! But, you have to understand that your technique would be a learning experience for many of us, and not something that is first hand to old hands as us who are still old school and still using a SA with two or three tone analysis for IMD evaluation.   Some of us older dogs, are stuck back in the 60's, 70's and 80's.

 Heck, I recently ended up with several HP-8407A (with the 9601A and Normalizers) NA mainframes and accessories, and for the life of me I will be hard pressed to find a use for them, especially for simple scalar analysis needs.  I'm always willing to learn a few tricks, even at my age. And, the few savant, fawning lapdogs who really bring nothing to the conversation have earned their "ignore" status by many of us users.

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 26, 2015, 04:17:08 PM
Quote
I would consider buying a good DSO scope, provided I had some assurance that it would be useful to me in my old age!!!

Well, I hope nobody reading this thread rushes out and buys a scope like mine because my old TDS2012 scope is poor in lots of ways :) This range of scopes was poor in lots of ways maybe 20 years ago when they were first introduced. We have loads and loads of them at work and they are really good for doing very basic stuff but the screen resolution and quality is pretty woeful and the scope limits the FFT size to a mere 2K samples. Also you can't grab raw sample data from it for post processing and analysis on a PC.

So in many ways its technical spec is dire by today's standards. Most people would be very disappointed with it when using it as a sampling scope.

If you need any more hints that this is not a modern scope then watch Dave/EEVblog  give his opinion on a TDS2024C model that is two generations newer than mine with better sample rate specs etc.

https://www.youtube.com/watch?v=3K04bHJJQQA.

However, even my old TDS2012 can have a reasonable stab at looking at cw clicks in MATH mode and it can show the spectrum being reduced with some basic waveshaping. Better/newer scopes will be much better here.

The soundcard based sampling scope really scores here and that's why I suggested using a PC based scope to grab and save raw data (looking at key clicks) and then play it back in slow motion or just play with the FFT settings etc on the recording to display the captured data in a way that works for you.

Buying a (suitable) DSO for your needs can be a minefield. On some forums people passionately argue over scopes and their makes/models/specs like they are a religion! The PC/soundcard alternative can be had for virtually no cost if you make the mixer and LO yourself and it can look very, very impressive as a narrowband signal analyser :)




Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 26, 2015, 05:51:31 PM
Here's a video showing my old scope trying to look at the bandwidth of an on/off keyed 28.5xxMHz signal that is being keyed at about a 10Hz rate. So this is a bit like a morse code signal.

At the start of the video there is very little pulse shaping but during the video you can see the bandwidth go down in stages as I adjust the rise and fall times of the RF waveform. By the end of the video the bandwidth goes down quite a bit because I've got rise/fall times of 3 or 4 milliseconds.

The span is 5kHz and the scale is 10dB/div.

At the end of the video I zoom in to look at a much narrower span of 500Hz. The scope hits a RBW limit wrt its 2K FFT size here so there isn't much detail. I can improve on this and zoom in further to get finer RBW resolution but I need to set the scope up for a different sample rate to do this. See the image in my next post for this.

But you can see that even my old 8 bit dog scope isn't limited to displaying just 40dB of range. There's much more range than this. Newer and better scopes improve on this. I'll see if I can borrow a better scope from work but I'm not sure this will be allowed (new rules at work)

In my opinion there is useful bandwidth information here even if you just look at the trend as the rise/fall time increases. The PC/Soundcard option is vastly better than this though :)

https://www.youtube.com/watch?v=kQT5U46ge9M&feature=youtu.be



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 26, 2015, 06:13:27 PM
In the still image below you can see that the scope can show the 10Hz modulation sidebands of the keyed 28.5xxMHz input waveform and the other sidebands are at multiples of 20Hz away from this as expected.

It is possible to predict the levels of these using a simulator or some equations for an OOK signal and the levels look to be correct for the basic theory of the 28.515MHz signal that is being keyed at 10Hz.

(http://i1227.photobucket.com/albums/ee439/G0HZU/OOK.jpg) (http://s1227.photobucket.com/user/G0HZU/media/OOK.jpg.html)

Again... the PC based scope + mixer will be much better than this. Faster refresh rate and more range :)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on August 26, 2015, 06:57:01 PM
One of the nice features of newer equipment is no more cameras.  :)

(http://www.w6rz.net/alias28.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on August 26, 2015, 10:47:10 PM
In the still image below you can see that the scope can show the 10Hz modulation sidebands of the keyed 28.5xxMHz input waveform and the other sidebands are at multiples of 20Hz away from this as expected.

It is possible to predict the levels of these using a simulator or some equations for an OOK signal and the levels look to be correct for the basic theory of the 28.515MHz signal that is being keyed at 10Hz.

(http://i1227.photobucket.com/albums/ee439/G0HZU/OOK.jpg) (http://s1227.photobucket.com/user/G0HZU/media/OOK.jpg.html)

Again... the PC based scope + mixer will be much better than this. Faster refresh rate and more range :)


Here we see a textbook example of the spectrum associated with OOK or ASK.

First, we have the carrier in the center, the first-order AM sidebands 6 dB down spaced from the carrier
at half the baudrate (10 Hz = 20 Baud = 24 WPM or a signal element length of 50 ms assuming an 1:1 mark/space ratio), and the higher-order odd-harmonic sidebands extending further out spaced with the baudrate.

It doesn't show what the rise and fall times are used in this waveform, but as the harmonic spectral lines extend quite far away outside the necessary A1A bandwidth, which according to the table of emissions in the ITU Radio Regulations is B=3K for "non-fading circuits" and B=5K for "fading circuits" where more distinct keying pulses are required, turns out to be 5*20 = 100 Hz or +/- 50 Hz at the -20 dB points, the rise and fall times are probably in the 2 ms region and thus shorter than optimum.

The ITU calculations assume a close to Gaussian shape of the leading and trailing edges, and the optimum rise and fall times of about 10% of the element length.

73/
Karl-Arne
SM0AOM



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 27, 2015, 11:43:54 AM
Hi Karl-Arne

The close in image showing the mod sidebands was taken with a fairly fast rise and fall time so the roll off is minimal and was taken before I slowed the rise/fall time. Much faster rise/fall than a few ms. Sorry for any confusion :)


DEMO OF MIXER + LAPTOP + OOK waveform wrt rise and fall time


For anyone interested in a more detailed analysis, here's a video of the mixer/PC/soundcard system in operation and you can see the effect of changing rise and fall times on a 10Hz OOK signal. This is like sending dits at this rate in morse code. The laptop is showing a 5kHz span with a 100dB display range in almost real time.

I'm using my Tek scope to show the keying waveform on the yellow trace and also the resulting RF envelope on the blue trace. Note that I've set the scope to peak detect mode to improve the definition of the RF envelope. This gives it a strange cross hatched appearance but this is normal for this mode. The scope is set at 10ms/div so the rise/fall time gets adjusted by several milliseconds in the demo.

To generate the adjustable rise/fall time I'm using an old HP pulse gen that has variable rise/fall time and I feed this pulse to one of the mod inputs of an old HP/Agilent 4000B vector sig gen which is set to output a signal around 28MHz.

https://www.youtube.com/watch?v=VnkhJvwFNqM&feature=youtu.be

I could repeat the test with it zoomed right in if anyone would like to see it? The other bonus with this method is you can save the captured data as a file on the PC and play back the same spectrum and pause it or look at the time domain in another window at your leisure.

I hope it's useful or interesting to some people :)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on August 27, 2015, 01:38:04 PM
I could repeat the test with it zoomed right in if anyone would like to see it? The other bonus with this method is you can save the captured data as a file on the PC and play back the same spectrum and pause it or look at the time domain in another windowccr at your leisure.

If you could take two still pictures or screen dumps, one using fast rise time, less than 0,5 ms and one using the proper shaping for 24 WPM which is about 5 ms symmetrical rise and fall times, using a span of +/- 125 Hz and the narrowest possible RBW so the individual keying spectral lines can be identified, it would be most instructive.

A visual comparison between the occupied and necessary bandwidths for a given keying rate and different rise/fall times can then be made.

73/
Karl-Arne
SM0AOM
 


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 27, 2015, 02:11:58 PM
Ok, I'll do this soon.

There is a slight problem though... I'd really want to do that on my Agilent signal analyser because it is a decent bit of kit and should be very accurate. But it isn't here at the moment. It's currently located a mile away in a big test rack connected to loads of other test gear and it's being used for some very important research work. I don't think I'll have it here again for maybe a week.

As you probably already know, it's possible to predict the spectrum of an ideal OOK with superfast edges using a few basic sums. The spectrum bandwidth/response is then just a function of the OOK keying rate and it's easy to predict the relative levels of each of the many sidebands to <1dB. So the first plot could be drawn by hand without needing any test gear. This could be done for offsets out to several kHz if needed.

I would expect that the second plot could be computed fairly easily too but I'll have a go at measuring it :)

If I get time I could also do it using a Excel worksheet using the FFT feature on Excel as a crosscheck. Or I could do it on an RF simulator.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 28, 2015, 06:19:56 AM
Is it OK to just sent continuous dits in this test? It's easier for me this way. At 24wpm I think the dit length is 50ms so this would be OOK at 10Hz.

Also, note that my ancient HP pulse gen does rise and fall time as a variable linear ramp using rotary controls.

Do you want me to measure rise time as 10 to 90% of this ramp on the scope or the whole time for the ramp from bottom to top? It would be nice to set this up as accurately as you want it.

I really wish I still had the signal analyser here so this will have to be done on the PC/soundcard instead :(







Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on August 28, 2015, 10:58:17 AM
For checking keying continuous dits, a square wave, or "1:1 reversals" as we old-timers with telegraph engineering background say,
provides a steady-state spectrum.
 
20 Baud or 10 Hz corresponds to 24 WPM which is a representative keying rate for aural reception Morse.
Many coast stations ran their CQ slips and traffic lists at 20 or 24 WPM.

ITU practice is to use the 10 and 90% points on the leading and trailing edges for expressing rise and fall, and to recommend a rise/fall time of between 10 and 20% of the signalling element duration.
It would be interesting to compare the keying spectra for the same keying rate for unshaped (shortest possible rise time) medium shaped at 10% or 5 ms and finally very soft at 20% or 10 ms.
My recollection from commercial Morse is that the 20% rise time resulted in a signal that sounded "nice" but could be difficult to read in fading conditons at 25 WPM.

Looking forward to your results.

73/
Karl-Arne
SM0AOM



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 28, 2015, 12:37:20 PM
I've had a first attempt at this but there were a few camera flicker issues and the image quality isn't that great either.

But see below for a 10Hz OOK waveform with rise/fall times of 0.1, 2ms, 3ms, 4ms and 5ms.

Note that it's best to select 1080P playback so you can read the rise and fall time data on the scope display.

The rise and fall time is shown on the right of the scope's display in yellow screen text. So you can see I pause at each of the required settings.

The vertical scale is 100dB at 10dB/div and the span is 500Hz. Note that I do zoom out a bit in the last few seconds of the video to wider spans.

https://www.youtube.com/watch?v=OGHoKS1t88w


I'd really like to repeat this with the signal analyser because it can do a few tricks wrt measuring power wrt bandwidth etc.

I hope this is useful and hopefully W8JI will comment because I think even he will find this interesting and useful because I doubt he will have seen this level of detail in real time before.






Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on August 28, 2015, 01:15:57 PM
Most interesting.

It appears that discontinuities in the keying envelope are dominant when it comes to creating
high-order keying sidebands.

Close to the center frequency, the expected smooth fall-off of the
sidebands with increasing rise and fall times takes place, but at larger spacings the sin(x)/x envelope
appears. Sharp edges or overshoots on the envelope create high-order sidebands in the same way as too short rise/fall times do.

The importance of proper shaping of the keying envelope so it corresponds as closely to the Gaussian erfc(t) shape as possible
if the occupied and the necessary bandwidths are not to be widely different is demonstrated in this video.
It is no coincidence that the CCIR and MPT keying shapes got the forms they have.

Thanks again.

73/
Karl-Arne
SM0AOM



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 28, 2015, 01:39:20 PM
Yes, it does look interesting.

Realistically, though the whole thing can be done much better on a PC using something like Matlab or even an RF simulator.

At the start of the video you can see a little row of low level tones sitting in between the main sidebands. They kind of bounce up and down as if they were at sea.

These are there because I can't generate a precise 50:50 mark:space ratio with the old HP pulse gen. It's an ancient thing and the controls really need cleaning as you can see it is a bit jerky in places.

So this imperfect signal will contain an extra series of continuous tones at these frequencies along with the expected suite of continuous tones (sidebands) that make up the OOK waveform.

You can see a similar effect if I imbalance the rise and fall times. The signal grows lots of extra (continuous) tones.

If I could generate the initial (fast edge) waveform perfectly then it would just contain a set of continuous tones that make up the carrier and the fanout of sidebands at +/-10Hz, +/-30Hz, +/-50Hz, +/-70Hz etc etc. There would  be nothing at +/-20Hz, +/-40Hz, +/-60Hz etc etc.



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 29, 2015, 06:25:12 PM
I got bored by the lack of activity on here and reread the thread and spotted this bit of pseudoscience from W8JI.

Quote
Any signal changing amplitude must have sidebands. This means the CW signal has sidebands when the rising and falling edges occur. It is obviously not a single frequency except during the time when level is not changing.


Isn't he making the classic error that all students make when considering the nature of an OOK signal?

Maybe I'm misinterpreting what he means here but basic comms theory suggests that a 28MHz signal sent for 10 minutes with 10Hz OOK consists of a series of tones that are there for all of the 10 minutes. The bandwidth of the signal is the same throughout the 10 minute transmission.

The sidebands and 'bandwidth' don't just appear during parts of the 10Hz modulation. They are there all of the time. The carrier tone frequency is there all of the time and so are all the sideband tones. They never disappear even though it looks like they are all gone in the OFF time of the OOK signal if you look at it in the time domain on a scope. It's a classic trap for students to fall into to think that there is no bandwidth in either the ON or OFF states of an OOK transmission :)



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on August 29, 2015, 06:43:29 PM
So you are saying key clicks exist even when the transmitter is not keyed.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 29, 2015, 06:51:00 PM
So you are saying key clicks exist even when the transmitter is not keyed.

I'm a bit unsure what the exact definition of a key click is because it is open to interpretation. I'd much rather just stick to the basic physics of OOK signals and the sideband tones that are associated with it.

If you were to send OOK slowly at 1Hz for 10minutes then a student would argue that there can't be bandwidth in the keyed state (or the unkeyed state) because a second is such a long time. But in reality, the signal does have bandwidth all through the 10 minutes because OOK is being sent at 1Hz for the whole 10 minutes. So there will be (continuous) modulation tones/sidebands for all of the 10 minutes.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on August 29, 2015, 06:54:51 PM
There is no time axis on a spectrum plot. Now a waterfall is a different matter.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 29, 2015, 07:05:53 PM
Quote
There is no time axis on a spectrum plot. Now a waterfall is a different matter.
I'd expect to see continuous spectral lines on the waterfall at the carrier frequency and the mod sidebands at +/-1Hz and also all the other sidebands at +/- 3Hz, +/- 5Hz, +/-7Hz etc etc. So it would look like a load of vertical streamer lines with no gaps in the streamers for any of the tones on the spectrum waterfall.

You can synthesise an OOK waveform on a computer either using equations or with a waveform simulator and there has to be bandwidth there all of the time. Or you could do an FFT on the waveform and the FFT will show the continuous tones that make up the waveform.

Or I guess you could synthesise it for real using a load of signal tones all set to the correct amplitude and starting phase and frequency offset in order to generate the OOK waveform. There wouldn't be any keying required but you would see the same OOK waveform on a scope for as long as you kept all of the continuous tones running.

But why bother with hardware when you can do it on a computer to prove it? It's just a classic OOK waveform to analyse?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on August 29, 2015, 07:18:32 PM
Try it. When the signal is continuous there are no sidebands. When key gets stuck "on" I hear no sidebands.

We're talking about a spectrum display that does not show time.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 29, 2015, 07:31:31 PM
It would be much easier* to just post me a link to a (credible) source that backs up what W8JI says below:

Quote
W8JI:  This means the CW signal has sidebands when the rising and falling edges occur. It is obviously not a single frequency except during the time when level is not changing.

* easier for me but harder for you because I don't think the credible link exists :)



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on August 29, 2015, 07:32:47 PM
On rethink, the sharp rise time of a loooong dash may not spread out in frequency like a normal dash. If that's the case the clicks are only Hz away from center, very narrow bandwidth.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on August 29, 2015, 07:34:29 PM
I'm trying but may fall on my face. ;)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 29, 2015, 07:37:16 PM
Quote
I'm trying but may fall on my face. Wink
By credible I mean formally accepted literature and not someone's homebrewed pseudoscience on their web page  :)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on August 29, 2015, 08:02:53 PM
This is a frequency vs. time ambiguity issue regarding what a spectrum of a signal represents. Since no time is displayed on the spectrum chart there is no way to determine when the signals are present. The viewing time has to at least cover the signal time period for a FFT. So you are saying that sidebands are present and detectable during the say 9 minute period after the keying and before unkeying?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on August 29, 2015, 10:28:37 PM
I think this is the waterfall you're looking for. Simulated with GNU Radio.

(http://www.w6rz.net/ook3.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 30, 2015, 04:10:27 AM
There isn't enough info there for me in the GNU plot. If it is being sent at 1Hz OOK then you would need a monumentally huge FFT size to resolve the individual spectral tones on such a huge span.

I'm describing the behaviour of an ideal 50:50 OOK test waveform here and I'm not trying to blaze a trail or be clever here. This stuff dates back >200 years to when Joseph Fourier was alive. I'm just a 'small' engineer stood on the shoulders of giants like Fourier and I'm trying to describe the nature of the OOK test waveform based on general theory :)

I suspect I've just opened a huge can of worms here... Maybe this wasn't a good idea.




Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 30, 2015, 04:26:08 AM
This is a frequency vs. time ambiguity issue regarding what a spectrum of a signal represents. Since no time is displayed on the spectrum chart there is no way to determine when the signals are present. The viewing time has to at least cover the signal time period for a FFT. So you are saying that sidebands are present and detectable during the say 9 minute period after the keying and before unkeying?

I'm not sure what you mean. But in the case of a perfectly sent 1Hz OOK test waveform there will be (in theory) an infinite number of test tones being sent continuously for the duration of the test transmission. It is a bit weird I know but theory tells us this and there are things you can do to prove it with real hardware. But theory should be enough in the case of a synthetic test signal like this :)

However, in reality morse code isn't sent like this. So it gets much muddier if you want to analyse a real cw signal consisting of manually sent dits and dahs with imperfect sending. There are also so many other real world factors that can cause clicks or chirps and this includes the transmitter synthesiser and the transmit path hardware and the comms receiver used to receive the signal.





Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on August 30, 2015, 05:01:35 AM
I wasn't trying to show the spectral tones. I was trying to show that they exist only during the rise and fall portion of the CW element. During the steady state period of the CW element, the bandwidth is very small. This agrees with W8JI's statement.

In your setup, consider delaying the trigger of the oscilloscope. Instead of triggering on the rising or falling edge, trigger on the steady state portion of the CW pulse.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on August 30, 2015, 05:09:07 AM
BTW, that was 24 WPM CW, but slowed down in playback 10X to exaggerate the rise and fall portions


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on August 30, 2015, 05:18:00 AM
I'm a bit unsure what the exact definition of a key click is because it is open to interpretation. I'd much rather just stick to the basic physics of OOK signals and the sideband tones that are associated with it.

I have never seen a formal definitíon of "key clicks", but using the language of the ITU Radio Regulations and the CCIR Recommendations, a proper description could be:
 
"Higher order keying sidebands, created by deviations from the optimum shaping of the keying envelope,  causing the occupied bandwidth of an A1A or A1B radiotelegraph emission to become much greater than its necessary bandwidth as established by the relevant relations between signalling rate and bandwidth set out in ITU-R SM.328"

The rationale for the relationships given in the ITU literature between keying rate, keying shape, necessary and finally occupied bandwidths are understood by anyone having a proper engineering background.

73/
Karl-Arne
SM0AOM


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 30, 2015, 05:31:46 AM
I'm a bit unsure what the exact definition of a key click is because it is open to interpretation. I'd much rather just stick to the basic physics of OOK signals and the sideband tones that are associated with it.

I have never seen a formal definitíon of "key clicks", but using the language of the ITU Radio Regulations and the CCIR Recommendations, a proper description could be:
 
"Higher order keying sidebands, created by deviations from the optimum shaping of the keying envelope,  causing the occupied bandwidth of an A1A or A1B radiotelegraph emission to become much greater than its necessary bandwidth as established by the relevant relations between signalling rate and bandwidth set out in ITU-R SM.328"

73/
Karl-Arne
SM0AOM

Ok thanks. I think there could be other interpretations as well but that would probably be OK for many people :)



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 30, 2015, 05:38:43 AM
I wasn't trying to show the spectral tones. I was trying to show that they exist only during the rise and fall portion of the CW element. During the steady state period of the CW element, the bandwidth is very small. This agrees with W8JI's statement.

In your setup, consider delaying the trigger of the oscilloscope. Instead of triggering on the rising or falling edge, trigger on the steady state portion of the CW pulse.
I think your setup is flawed as I said in my earlier posts.

My scope is just sampling at a low sample rate and it does an FFT on the waveform. My old scope can only do a 2K FFT so in order to resolve down to a few Hz across a narrow span I have to use bandpass sampling at a very low sample rate. I can get away with this because the 28MHz waveform is quite narrow wrt the Nyquist frequency. So aliasing isn't a significant issue.



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on August 30, 2015, 05:49:13 AM
It's not flawed, you're just not able to wrap your head around it. The key clicks only exist during the rise and fall of the CW element. If the sidebands existed for the entire time of the CW element, you wouldn't hear clicks off frequency, you'd hear the CW, but at a reduced level.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on August 30, 2015, 05:50:05 AM
I'm a bit unsure what the exact definition of a key click is because it is open to interpretation. I'd much rather just stick to the basic physics of OOK signals and the sideband tones that are associated with it.

Ok thanks. I think there could be other interpretations as well but that would probably be OK for many people :)
Having worked with technical and spectrum matters in the ITU Study Groups probably makes you somewhat partial in favour of the "ITU-speak"  :)...

73/
Karl-Arne
SM0AOM


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on August 30, 2015, 07:04:19 AM
I'm very late to this debate, but .......

Try abruptly switching on a carrier at t=0; switch it off about an hour later.

At t=1 minute, examine the carrier with your spectrum analyser; keep doing it for 5 minutes. At no time will you see any sidebands.

But examine the signal during the t=0 switch on and you will see sidebands whose extent depends on the rise time of the carrier.

I think that was what W8JI was saying.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 30, 2015, 08:35:19 AM
I'm very late to this debate, but .......

Try abruptly switching on a carrier at t=0; switch it off about an hour later.

At t=1 minute, examine the carrier with your spectrum analyser; keep doing it for 5 minutes. At no time will you see any sidebands.

But examine the signal during the t=0 switch on and you will see sidebands whose extent depends on the rise time of the carrier.

I think that was what W8JI was saying.

Steve G3TXQ

I'm not sure your example helps because it just lowers the OOK rate from 1 a second to 1 an hour.

Or are you suggesting that the physics and maths changes if you slow everything down? It's hard enough (in human terms) to cope with this concept at 1Hz let alone 1 per hour but I don't think the theory changes at all. If you send OOK at the hourly rate (say for 3 weeks) the infinite series of tones that make up this are still there all the time for all of the 3 weeks.

You can scale it in time any way you like but the theory stays the same? However, I don't think it's wise to explore extreme timescales of OOK as you may start attracting other arguments due to various physical limits. It's much more sensible to stick to 1Hz OOK or 10Hz OOK or even OOK into the MHz range.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on August 30, 2015, 09:26:06 AM
It's not flawed, you're just not able to wrap your head around it. The key clicks only exist during the rise and fall of the CW element. If the sidebands existed for the entire time of the CW element, you wouldn't hear clicks off frequency, you'd hear the CW, but at a reduced level.

How could it be any other way?

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on August 30, 2015, 09:50:49 AM
I'm not sure your example helps because it just lowers the OOK rate from 1 a second to 1 an hour.
I do find it helpful to think what happens if we lower the rate - that way you can begin to unravel the separate effects that the signal periodicity and the keying rise-time have on the sidebands.

In fact I find it even more helpful to avoid periodic signals altogether and just look at the Fourier Transform of a baseband Unit Step function.

All that said, I suspect W8JI was thinking purely in practical terms. The common experience of most Hams is that if you tune a receiver off to one side of a carrier that is "hard" keyed at a slow rate, you will hear sideband energy at the time of the key closures and openings, and not in between; in fact you can often read the Morse from the timing of the "clicks". Clearly the observed spectral density of the signal changes at the key closures and openings.

Theoretically you are right - if we analyse any periodic signal over a long enough time with infinite resolution, we will have a time-invariant frequency transform. But in the real world of Ham radio Morse, the signal isn't periodic, we don't have infinite frequency resolution and our observation times are relatively short.

Steve G3TXQ



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on August 31, 2015, 01:01:20 AM
Having thought about this some more, let me give my qualitative explanation of what is happening.

Let's take G0HZU's example of a simple periodic OOK square wave with a frequency of, say, 1Hz. Theory tells us that the spectrum of this signal will be stable, and will have sidebands spaced by 2Hz (only the odd order components are present).

But experience tells us that if we tune off to one side of this signal we will hear a series of "clicks" - one every 0.5 seconds - coincident in time with when the square wave changes state. How can that be if the spectrum of the signal is not varying?

The answer is that our receiver is encompassing more than just one sideband component. In a typical bandwidth of, say 100Hz, we will encompass 50 components, all spaced by 2Hz. The phase relationship of those closely-spaced components is such that for long periods of time they virtually cancel to zero; but for short periods of time - coincident with the changes of state of the square wave - they come into phase and add to produce a narrow, high amplitude, pulse; that's what we hear as the "click".

It has some similarity to passing a square wave through a high-pass filter: excluding the low order frequency components leaves us with a series of short pulses at the zero crossings.

I think the following statements are probably true:

1) The "clicks" are coincident with the key closures
2) In typical ham situations the net sideband energy is almost zero other than at the key closures
3) Individual sideband components are present all the time.

Steve G3TXQ



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 31, 2015, 06:04:14 AM
Yes, I agree that the 'click' phenomenon is a complete system phenomenon including the quality of the Tx and also the receiver hardware in terms of AGC, linearity and bandwidth and I suppose you have to include the 'human hearing' factor too.

That's why I suggested earlier that the definition of a click is open to debate. I've tried to just look at the actual sidebands in the transmission of an OOK signal and how to measure then using an analyser :)




Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on August 31, 2015, 06:13:18 AM
Yes, I agree that the 'click' phenomenon is a complete system phenomenon including the quality of the Tx and also the receiver hardware in terms of AGC, linearity and bandwidth and I suppose you have to include the 'human hearing' factor too.

I wasn't really including those practical aspects. I was simply explaining how taking a subset of higher-order components from the theoretical spectrum of a perfect square wave would result in a time-domain signal which was impulsive in nature, the "pulses" being coincident with the square-wave zero-crossings.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on August 31, 2015, 06:25:31 AM
Maybe this simulation will help. It's a set of triggered FFT plots. The top trace is the IQ values being sent with I = green and Q = violet. The bottom trace is a 2048 bin FFT. The sample rate is 48 kHz, so the length of the FFT is 2048/48000 = 42.666 milliseconds. I've adjusted the time scale of the IQ trace to be close to 42.666 milliseconds to show what's in the FFT.

The first trace is during the space between elements. The transmitter is off and spectrum is zero (limited to -200 dB).

(http://www.w6rz.net/a0.png)

Second trace is right in the middle of a hard rising edge. Spectrum is wide and energy is high off frequency.

(http://www.w6rz.net/a1.png)

Third trace is in the middle of an element. Spectrum is narrow.

(http://www.w6rz.net/a2.png)

The spectrum of a CW signal is clearly dynamic.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on August 31, 2015, 06:47:38 AM
It's a wet Bank Holiday afternoon in UK, so here's something to occupy the brain:

Say we send a series of Dashes for 10 seconds followed by a series of Dots for 10 seconds. If we analyse the signal in the middle of a single Dash (avoiding the edges) we will get a single spectral line - call it S1.

If we analyse the signal over the first 10 seconds - a period which includes just the Dashes and nothing else - we will get a more complex, wider, spectrum; call it S2.

If we analyse the signal over the last 10 seconds - a period which includes the just Dots - we will get a different spectrum again; call it S3.

And if we analyse the signal over the full 20 seconds we will get yet another Spectrum because it contains both Dashes and Dots; call it S4

Questions:
Which of the 4 spectra actually existed during that first Dash?
If it was S4, how could it have been determined by something which had not yet happened.
If it was S1, how could it be retrospectively changed when the Dots came along?

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on August 31, 2015, 06:54:45 AM
While I'm at it, here's the 2K FFT spectrum of a filtered element.

(http://www.w6rz.net/a4.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: AC7CW on August 31, 2015, 07:04:49 AM
I got bored by the lack of activity on here and reread the thread and spotted this bit of pseudoscience from W8JI.

Quote
Any signal changing amplitude must have sidebands. This means the CW signal has sidebands when the rising and falling edges occur. It is obviously not a single frequency except during the time when level is not changing.


Isn't he making the classic error that all students make when considering the nature of an OOK signal?

Maybe I'm misinterpreting what he means here but basic comms theory suggests that a 28MHz signal sent for 10 minutes with 10Hz OOK consists of a series of tones that are there for all of the 10 minutes. The bandwidth of the signal is the same throughout the 10 minute transmission.

The sidebands and 'bandwidth' don't just appear during parts of the 10Hz modulation. They are there all of the time. The carrier tone frequency is there all of the time and so are all the sideband tones. They never disappear even though it looks like they are all gone in the OFF time of the OOK signal if you look at it in the time domain on a scope. It's a classic trap for students to fall into to think that there is no bandwidth in either the ON or OFF states of an OOK transmission :)




gadz... I had a physics professor, PhD no less... the class was discussing the trajectory of a cannon projectile... he was certain that the forward momentum of the shell actually was zero, actually was stopped, at apogee... I tried to tell him that the vertical component of the shell's motion went to zero for an infinitely short period of time but the forward velocity remained largely unchanged... he just had the PhD [and the introvert personality that can JUST NOT BE WRONG] and I was ignored... LOL


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: N3QE on August 31, 2015, 07:20:46 AM
If it was S4, how could it have been determined by something which had not yet happened.
If it was S1, how could it be retrospectively changed when the Dots came along?

Strictly speaking the Fourier transform has to not go back just over the 10 seconds you talk about, but it has to backwards to infinitely before the big bang and forwards infinitely past the heat death of the universe!

Obviously most of our QSO's don't last quite that long (although some of the guys on AM... you might think they're not gonna stop before the heat death of the universe.)

Broadly there is the question about what to do when doing a Fourier transform on a finite data set. The choices available are described as the different "window" possibilities. https://en.wikipedia.org/wiki/Window_function

Most audio-based waterfall analyzers seem to use a transform window that is about 0.010 to 0.025 of a second or so and the "shmear" on the waterfall graph around each on/off transition seems to match closely how we usually think about key clicks. Take some example off-the-air CW audio (I like to use the LZ Open 40M wav's available at http://www.lzopen.com/audio/index.htm ), open in audacity, select a 20 or so second patch that is a QSO, then view the "spectrogram". The transform window used in audacity seems to match pretty well how humans think about sounds - and you can clearly see all signals have keyclick at both "on" and "off" transitions, the "soft" sounding ones have less keyclick, that some are worse than others, that some have asymmetric clicks, etc.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on August 31, 2015, 10:26:43 AM
I think the first thing to do is try to clear up the idea that the sidebands are present all the time.

Would someone show me the graph/plot which lead to this idea? Certainly not a spectrum plot since it has no time axis. What lead to that statement?







Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 31, 2015, 10:38:32 AM
From my point of view you can synthesise the 50:50 OOK waveform by summing together a whole load of continuous tones.

My guess is it goes something like this:

You start off with a continuous tone at the centre carrier frequency. Let's asssume it has an average power of 1W.

What you can then do is start summing in other continuous tones at the right amplitude and phase and offset frequency such that you produce (or synthesise) an OOK signal of 2W average power.

Obviously, you need a lot of tones :)

So to get that 2W average power you have to sum in another 1W worth of average power in the form of continuous tones (or sidebands). Isn't that just basic Fourier synthesis?

If this is all wrong then I'd definitely be interested to know why. Like I said, I'm just a small engineer stood on the shoulders of giants like Fourier :)

All of the above can be done in theory rather than in hardware. eg on a simulator or in something like Matlab.




Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on August 31, 2015, 10:51:33 AM
The fact that a waveform can be can be synthesized from continuous tones does not suggest or imply the time of the tones.

I just can't seem to get it into words that there is no time shown in the frequency domain.

One has to use a waterfall to have any hopes of determining time.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on August 31, 2015, 11:11:56 AM
G0HZU obviously has a proper engineering background and correctly applies the "steady state" concept of alternating currents.

This was taught with exemplary clarity during my third semester at Chalmers Univ. of Tech. by senior lecturer Charles Davidson who then added that the observation window theoretically has to extend in time from minus infinity to infinity, "which may be perfectly acceptable to mathematicans and physicists, but we engineers seldom have the patience to wait that long"

73/
Karl-Arne
SM0AOM


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on August 31, 2015, 12:11:47 PM
I think the first thing to do is try to clear up the idea that the sidebands are present all the time.

Would someone show me the graph/plot which lead to this idea? Certainly not a spectrum plot since it has no time axis. What lead to that statement?


I regret bringing up this topic.

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W5WSS on August 31, 2015, 12:31:07 PM
Sorry you regret bringing up this topic I have been enjoying it ;D


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on August 31, 2015, 12:44:31 PM
Sorry you regret bringing up this topic I have been enjoying it ;D

i see no reasons for regretting bringing up a topic that needs to be discussed far more among radio amateurs than it usually has been.

Methinks that we now have a through understanding of the key-click phenomenon, both in the time and frequency domains, mainly thanks to G0HZU and G3TXQ.

It is now time to direct our attention to what can be done to mitigate this menace to Morse amateur radio operation  which is an art, and should be equally enjoyable.

However, one does not have to operate for long on the amateur bands to run across signals which have outright anti-social spectral properties.

The current discrepancies between necessary and occupied bandwidths shown by amateur gear are hard to understand for us that have professional radiotelegraphy and engineering backgrounds.
How can it be that the equipment designers in Japan feel that they must design their keying envelopes with both improper shapes and rise/fall times of the leading and trailing edges?
Are there customer requirements to design in this way? And if so, which category of customers have had their say in this process? 

Sometimes opinions are voiced that a "wide signal" is beneficial in the competitive environment, which leads to the question of what happened to:
´"The Amateur is Gentlemanly . . . He never knowingly uses the air for his own amusement is such a way as to lessen the pleasure of others"

Also, as licensed radio amateurs operating in an ITU recognised service we are bound by the general provisions in the Radio Regulations about bandwidth conservation and signal quality. It is no excuse that we "only are amateurs", because due to our perceived technical and operational competence, we are supposed to know better.

The "bottom line" is that if we want better spectral masks for our Morse code transmitters we need to get the manufacturers to listen to other opinions than the contesters.
Otherwise it may be the "homebrew" or professional gear route that needs to be taken.

73/
Karl-Arne
SM0AOM
 


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on August 31, 2015, 12:49:33 PM
Aw, this is fun. Now only if I could remember stuff I knew a loooong time ago. Laplace or Fourier, one sided or two to infinity integral. Heaviside and unit step function.

Maybe with enough thought something insightful will come out.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 31, 2015, 01:26:11 PM
The fact that a waveform can be can be synthesized from continuous tones does not suggest or imply the time of the tones.

I just can't seem to get it into words that there is no time shown in the frequency domain.

One has to use a waterfall to have any hopes of determining time.


What do you mean by the time of the tones?

I'm suggesting you set off all of the tones at the same time and then let them run freely from their original preprogrammed starting phase at time T0.

You don't have to do anything else. Just sit back and look at the OOK waveform on a scope as if it had been keyed? i.e. you will see a self keying waveform on the scope that will run and repeat for as long as you want :)





Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on August 31, 2015, 01:55:45 PM
Time means time; duration, absolute, or relative. Given a spectral plot and not knowing the original waveform it is impossible to state the duration or absolute time of a spectral line. There is no time base on a spectrum plot.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 31, 2015, 02:02:27 PM
Given a spectral plot and not knowing the original waveform it is impossible to state the duration or absolute time of a spectral line.

Are you saying I can't just sit here with a piece of paper and predict how to construct a reasonable OOK waveform using lots of continuous tones?

I think I can do it with just a basic calculator. i.e. predict the required (starting) phase of every tone and its required amplitude.

Actually, I could do it with just a paper and pencil. That isn't because I'm good at this. I think the calculations could be done by a schoolkid if you gave them the instructions and a piece of paper and a pencil.

Is there something else I'm missing here? If I sat down and did the above sums and it produced an OOK waveform would you still be saying that something else is missing or wrong or undefined?





Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on August 31, 2015, 02:32:15 PM
I think the difference is the confusion between a know periodic/symmetric waveforms and unknown ones. Only waterfall displays show when the key clicks occur and only then can we associate the rise/fall time of the signal to the generation of clicks.

The spectrum plot missed the "quiet period" between the clicks.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on August 31, 2015, 02:43:34 PM
I explained in my Post #142 that the "silence" between key clicks does not necessarily mean there are no sidebands during those periods.

The sidebands can be present, but have a phase relationship that causes cancellation during those "quiet" periods. The "click" occurs when they momentarily all drift into phase, and that produces a narrow pulse.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on August 31, 2015, 02:49:57 PM
So an impulse/dirac delta function creates an infinite bandwidth and lasts for zero seconds?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on August 31, 2015, 03:09:19 PM
Correct - and it contains unit energy. But I didn't say it was a Dirac Delta function, I said it was a narrow pulse.

G0HZU is correct when he says that a regular OOK signal has a stable frequency spectrum. If the OOKeying was a continuous 10Hz square wave you would have a wide, time-invariant, spectrum of sidebands each spaced 20Hz from each other. If you had a receiver with a narrow enough bandwidth you could tune away from the carrier and resolve each separate sideband - they would each sound like a continuous sine wave with no trace of the keying changes of state.

However, with the receiver tuned away from the carrier, if you widen its bandwidth you begin to encompass more of those individual sidebands and their net envelope starts to become impulsive for the reasons I've explained already. As you widen the receiver bandwidth, the hearing perception changes from one of individual tones to impulses; the sidebands haven't changed - they are still present as individual tones 20Hz apart - but we now perceive them as "clicks" that are coincident with the keying changes of state.

The more we open up the receiver bandwidth, the more sidebands we encompass, and the narrower those pulses become.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on August 31, 2015, 03:21:05 PM
What formula or plot shows or even suggests that there are continuous spectral lines generated during the entire time period of the OOK?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on August 31, 2015, 03:28:02 PM
A simple Fourier transform of the square wave!

That's the whole basis of Fourier transformation - that you can synthesise a complex, periodic, time-domain signal with a set of steady sinusoids.

The spectral components are continuous for as long as the OOK continues.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on August 31, 2015, 03:41:22 PM
OK, now lets try a step function representing an "on" state with no "off". Will there be a click?  If so what is it's duration?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on August 31, 2015, 03:45:26 PM
Much more complex, because it's not a periodic signal!

Do I take it you now accept that the sidebands are present continuously if the signal is periodic?

PS: now off to bed!

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: JS6TMW on August 31, 2015, 03:50:33 PM

Actually, I could do it with just a paper and pencil. That isn't because I'm good at this. I think the calculations could be done by a schoolkid if you gave them the instructions and a piece of paper and a pencil.


Not by an American schoolkid.

And SM0AOM said, "How can it be that the equipment designers in Japan feel that they must design their keying envelopes with both improper shapes and rise/fall times of the leading and trailing edges?"

I suppose they are restricted by cost and use only simple RC circuits. But it would be interesting to know what shapes and rise/fall times would result in minimal keyclicks and still maintain "readibility", which admittedly is subjective.

73,

Steve AI6KX in Okinawa (hoping to minimize keyclicks by adjusting the bias on my solid-state amplifier..... still here!)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on August 31, 2015, 03:55:48 PM
I'll accept that the inverse Fourier will provide the time domain signal. Still mulling whether the spectral components need to physically exist over the entire period. Seems to suggest that energy is constantly being provided over the entire time which makes no sense.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 31, 2015, 04:09:46 PM
I've got a couple of oldish HP/Agilent vector sig gens here that I have set up to try and show a crude OOK waveform.

I think I could also generate a file containing the tone data and send it as an ARB but that might not work very well. Even if it did it wouldn't be very convincing. But that could do it with just one generator. The lazier option is to create the sidebands in one signal generator and then sum in the carrier from another generator at the right amplitude and phase to cause an OOK waveform to appear. I've also locked both generators to a common reference.

I tried it with 6 upper and 6 lower sidebands plus the manually inserted carrier and it looks OK on my old Tek scope. I think the sig gen maxes out at 64 tones but I'm much too lazy to manually enter the start phase and amplitude settings this many times :)

Note that the OOK is being sent much faster than 1Hz because 100Hz is the lowest frequency spacing the multitone generator will accept. Also I have to turn off the ALC in the top generator to preserve the relative amplitude accuracy of all of the sidebands. Sorry that it's not 1Hz OOK but I'll have a think about a way to do this using something else.

The top generator produces a continuous stream of sideband tones and the bottom one sums in the continuous carrier at the required phase and ampltude in the little combiner block. This composite signal then gets fed to the scope.

Does this prove anything? Is there something I'm missing that means the continuous tones aren't 'valid' in some way?

I have to say this again. I'm not trying to be clever here . This is where my experience ends in terms of Fourier synthesis. i.e. the stuff I learned years ago and am trying to remember.

If there is some other aspect to all this that I'm missing then it would be nice to know what it is :)

oops.. there seems to be an issue with the photo upload. Photobucket is down for maintenance and I tried IMGUR but it doesn't seem to work either. How annoying!

Oh no.. now the image has  reappeared and it's enormous. (sorry)

(http://i1227.photobucket.com/albums/ee439/G0HZU/multitone_OOK.jpg) (http://s1227.photobucket.com/user/G0HZU/media/multitone_OOK.jpg.html)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 31, 2015, 04:37:07 PM
If anyone else wants to try this then probably the easiest way to synthesise it is to generate the tones like this:

Assume a carrier frequency F.

The sidebands on the high side will need to start 180deg out of phase with the sidebands on the low side. The carrier F needs to start in quadrature with both of them.

So generate a suite of tones thus.

Set the main carrier to F and give it a starting phase of +90 degrees. Give it an amplitude of about 1V

The next few tones all need to be starting at +180deg phase
Then add a tone at +1Hz at about 0.64V and call this tone S1.
Then add a tone +3Hz from the carrier that is a third of the voltage at S1
Then add a tone +5Hz from the carrier that is a fifth of the voltage at S1
Then add a tone +7Hz from the carrier that is a seventh of the voltage at S1
Then add a tone +9Hz from the carrier that is a ninth of the voltage at S1

Keep adding these + tones until you get bored. But I would suggest you add quite a few.


The next few tones all need to be starting at 0deg phase

Then add a tone at -1Hz at the same level as S1
Then add a tone -3Hz from the carrier that is a third of the voltage at S1
Then add a tone -5Hz from the carrier that is a fifth of the voltage at S1
Then add a tone -7Hz from the carrier that is a seventh of the voltage at S1
Then add a tone -9Hz from the carrier that is a ninth of the voltage at S1
Keep adding these + tones until you get bored but add as many as you did for the +xHz tones


That should produce a waveform that looks like OOK. It won't be perfect because you can't add an infinite set of sidebands.

It's probably easiest to do the above in Matlab. Or do it in a SPICE simulator if it allows this many signal sources. If you use a SPICE simulator then I'd suggest you use a low carrier frequency and sum all the tones into the inverting input of a precision summing opamp. Each one via a 1k resistor. But your SPICE simulator may complain about the number of signal sources.

You might also want to scale the sideband spacings in frequency to make the FFT analysys easier and quicker but the OOK rate will go up accordingly.

Hope this is interesting and/or useful (and relevant?)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 31, 2015, 05:04:08 PM
For completeness I wheeled in an old school swept analyser and looked at the composite signal on a 2kHz span.

The analyser wasn't fully settled so the signal is a bit off frequency (too lazy to wait) but you can see the makeup of this signal on the analyser. I can also display a crude post detection FFT using this analyser but it doesn't look as impressive as the signal analyser.

I should have the signal analyser back here in a few days. Possible tomorrow. If there is enough interest I could see if I can redo this with more tones. But I think I'll upload the amplitude and phase info via GPIB. It takes ages to do it vis the LCD display on the generator.


(http://i1227.photobucket.com/albums/ee439/G0HZU/Multi_OOK_8566.jpg) (http://s1227.photobucket.com/user/G0HZU/media/Multi_OOK_8566.jpg.html)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on August 31, 2015, 05:08:27 PM
Nice demonstration of adding sinusoidal waves to create a time domain OOK waveform.

I guess it would be important to know whether the key clicks are heard twice on both on and off transitions or as one click.

The other question would be as to whether the reconstructed OOK waveform contains key clicks.



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on August 31, 2015, 05:25:52 PM
Quote
Nice demonstration of adding sinusoidal waves to create a time domain OOK waveform.


Thanks :) I think this is all OK but it would be nice to have this stuff peer reviewed by a few people. Maybe G3TXQ can have a look tomorrow and maybe a few others.

I worked out the phases and amplitudes quite easily and it seemed to work first time although I have to set the phase of the lower generator to account for the coax cable feed etc.

It is quite interesting to watch the waveform change on the scope  if I mess with the phase of the carrier and also its amplitude. i.e. to watch the null zone appear and disappear.

But I'm nervous that some very clever physics guru may pop up and say that it doesn't actually work like this and my conclusions are all wrong  ;D



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on August 31, 2015, 06:50:32 PM
Everyone here is more or less correct. It all depends on how long the duration (in time) of the FFT is. The integration time (or whatever you want to call it) of human hearing is very short. You have to match the FFT duration to something that makes sense for human hearing and the duration of the pulses you're measuring.

In my last set of simulations, I used a 42.6 millisecond duration. That's probably much longer than human hearing, but short enough for 24 WPM CW elements to show different spectra at different points in the waveform.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on August 31, 2015, 11:15:39 PM


I suppose they are restricted by cost and use only simple RC circuits. But it would be interesting to know what shapes and rise/fall times would result in minimal keyclicks and still maintain "readibility", which admittedly is subjective.

73,

Steve AI6KX in Okinawa (hoping to minimize keyclicks by adjusting the bias on my solid-state amplifier..... still here!)


Actually this is the easiest to answer. Modern equipment designs are not restricted to simple RC circuits, because nowadays CW signal generation is made in DSP.
This permits the synthesis of an optimum keying shape with minimal effort.

However, even in the "old days" designers could create "decent" keying shapes in grid-block keyed transmitters,
by careful control of time constants in the circuits. In the early-70's I designed and built a tubed Morse transmitter
using differential grid-block keying which after some "tweaking" had very good element shapes, maybe a little on the soft side.

For the designer that takes Morse adjacent-channel performance seriously, there is a lot of guidance material in the ITU Radio Regulations and in the CCIR Recommendations, with the SM.329 in particular. From these are spectrum masks derived that were included in the various type-acceptance specifications that the more serious Administrations applied when approving Morse transmitters for fixed and mobile use.

One widely used was the late-60's specification from the British Ministry of Post and Telecommunication (MPT) which most European Administrations applied. To fulfill the spectral mask outlined in this specification was considered difficult, but not impossible. All ship's Morse transmitters of reasonable age that I have run across were compliant with this specification.

In my former field, coast radio systems engineering, there were no prescribed "spectrum masks", but the ITU Regulations and the MPT Specifications were applied when specifying and adjusting equipment. Older exciters, as the Telefunken early 60's units had selectable keying rise/fall times,in the 1 to 8 ms range which usually were set at 6 ms as a suitable compromise for the most commonly used telegraph keying rates around 20-25 WPM. Newer exciters, as the ITT-SRT TD90 used a fixed rise/fall time of around 6 ms.

When comparing keying spectra between amateur and commercial equipment, the peak levels of "key-click energy"
in a 100 Hz bandwidth at the nominal channel spacing of 500 Hz may differ by over 40 dB (!). The amateur equipment used in this comparison was not one of the "worst offenders".

The "bottom line" is that if the designers wanted, and the customers demanded, all amateur equipment could be as "clickless" as their professional counterparts.

73/
Karl-Arne
SM0AOM





Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: JS6TMW on September 01, 2015, 02:07:04 AM

The "bottom line" is that if the designers wanted, and the customers demanded, all amateur equipment could be as "clickless" as their professional counterparts.


I have never seen any specification other than 'rise-time' and 'fall-time', but how is this defined? Rise to 90% as in a simple passive circuit..., or a linear rise to 100%, or... there are a lot of possibilities. Or does specifying a spectrum mask uniquely define the OOK waveform?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: DL8OV on September 01, 2015, 02:29:08 AM
I suspect that the keyclicks problem will be easier to solve than the IMD issue as (as has been pointed out) generation of the CW waveform is now done in DSP software. If the specification demands a raised cosine rise and fall waveshape with 5mS between the 10% and 90% points we will be almost there. The issue used to be restricted by cost but there have also been some very strange design decisions, I did some reading on the issue and one manufacturer even installed extra circuitry to route the CW transmit signal through the SSB filter when a path through the CW filter would have resulted in a cleaner waveform.

Can anyone remember which rig had that strange design in its CW transmit path?

As for the rest of the technical discussion with Dirac Delta functions and Fourier transformations , this beyond the limits of my technical knowledge but I am still having fun reading it all.

Peter DL8OV


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 01, 2015, 02:30:50 AM
I guess it would be important to know whether the key clicks are heard twice on both on and off transitions or as one click.
Yes - they would be heard at each transition, positive-going or negative-going

The other question would be as to whether the reconstructed OOK waveform contains key clicks.

Yes - because it was synthesised from frequency components which are responsible for the "click".

It's very important in this discussion to note the big difference between analysis (or synthesis) of a periodic signal - that's one where f(t)=f(t+T)=f(t+2T) etc - and one that's non-periodic. Imagine trying to synthesise a unit step function: f(t)=0 for t<=0 and =+1 for t>0. What sine wave generators, at what frequencies , what phases and what amplitudes, would you need? How could you stop the synthesised waveform from being periodic?

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: N3QE on September 01, 2015, 03:11:37 AM
I suspect that the keyclicks problem will be easier to solve than the IMD issue as (as has been pointed out) generation of the CW waveform is now done in DSP software. If the specification demands a raised cosine rise and fall waveshape with 5mS between the 10% and 90% points we will be almost there. The issue used to be restricted by cost but there have also been some very strange design decisions, I did some reading on the issue and one manufacturer even installed extra circuitry to route the CW transmit signal through the SSB filter when a path through the CW filter would have resulted in a cleaner waveform.

Can anyone remember which rig had that strange design in its CW transmit path?

I don't think it's all that unusual for rigs of a certain vintage and mixing scheme. For CW, with the IF transmit frequency shifted 800 Hz from the receive IF frequency, going to a wide SSB filter instead of the narrow CW filter during transmit means you have fewer oscillators to shift on each key-up/key-down.

On a slightly different tack but using the "why not transmit through the narrow filter" note, note that phase-continuous FSK used in RTTY by default has bad keyclick (worse than we talk about for "bad CW keyclick"). This guy hacks a K3 to route the FSK through the narrow CW filter to clean it up in his figure 14: http://www.frontiernet.net/~aflowers/k3rtty/k3rtty.html (Note that later revisions of K3 clean up FSK considerably - but there are still lots of non-K3 bad-FSK-RTTY-keyclick transmitters out there.)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on September 01, 2015, 11:09:37 AM
What I am puzzling over is that the Fourier transform has no information on the clicks existing at the rise and fall times of the waveform. This is understandable since no time axis exists for the frequency components, the spectral components are different at those times than those between. 

Currently I suggest that the transform, being an integration, depicts the spectral components arising from the waveform but does not imply they are continuous throughout.

There seems to be a difference between the information contained in the transform and FFT waterfall and it is all about time.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 01, 2015, 11:15:58 AM
The spectral components exist for as long as the periodic time waveform persists.

Think about how you would synthesise a periodic square wave from a set of sine waves - the sine waves would be continuous. Look at the synthesised square wave in the charts on this page:
http://www.allaboutcircuits.com/textbook/alternating-current/chpt-7/square-wave-signals/

The component sine waves are continuous and of constant amplitude - they don't do anything different at the square wave transitions.

The timing of the "clicks" is contained within the phase relationship of the spectral components: when they all add in phase you get the "click".

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on September 01, 2015, 11:36:08 AM
Agree, but is this a mathematical construct or do the spectral in reality lines exist throughout? Again thinking of a 10 min pulse, the waveform is a clean sine wave between the keying, there cannot be additional sideband energy generated during this time.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 01, 2015, 11:56:30 AM
Agree, but is this a mathematical construct or do the spectral in reality lines exist throughout? Again thinking of a 10 min pulse, the waveform is a clean sine wave between the keying, there cannot be additional sideband energy generated during this time.
I keep stressing the difference between periodic and non-periodic functions. If your short pulse is periodic and is repeated every 10 minutes there will indeed be constant-amplitude sideband sine-waves present continuously. But with a repetition period of 10 minutes they will be spaced 0.00167Hz apart. You'll struggle to resolve them individually with a receiver or spectrum analyser, but if you could, they would be there.

Asking if they "really exist" gets us into philosophical discussions. However, they are real insofar as a device providing sufficient discrimination in the frequency domain - such as a bank of incredibly narrow filters - will identify them and the energy they contain.

I could equally well take your 10 minute periodic time waveform and transform it into another domain - say, the sequency domain. The resulting sequency components would be "real" insofar as they could be individually detected by the appropriate sequency discriminators.

Representations of a signal can be in the time domain, the frequency domain, the sequency domain, maybe others - they are simply different ways of viewing or representing the same thing. It's tempting to ask " ... but which is the real one?" to which, of course, there is no real answer  ;)

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 01, 2015, 12:21:23 PM
If it helps at all, I can tell you how I solved the multitone Fourier synthesis? i.e. how I was easily able to work out the phases and amplitudes?

I basically constructed the spectrum from an unbalanced switching mixer.

Would it help if you viewed the 'keyer' as the switching LO?  That's why I was able to predict that for a 2W carrier going into the 'mixer' only 1W comes out (50% duty cycle)

0.5W of that 1W is the carrier and the other 0.5W is in all of the sidebands. It was easy to construct the amplitudes of the sidebands for the Fourier synthesis because they are derived from a square wave. So to solve it all you just have to make sure that it all adds up to 1W. Put the carrier in quadrature and make sure the upper and lower sidebands are in anti phase to each other.

Does that help at all?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on September 01, 2015, 12:33:18 PM
Next project: Sequency Receiver for eliminating key clicks. :)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 01, 2015, 01:11:46 PM
Next project: Sequency Receiver for eliminating key clicks. :)
Then you'll be interested in the schematics here:
http://www.vtvt.ece.vt.edu/research/references/uwb/signalings/WALSHFCT.pdf
;)

Humour aside, it turns out that transforming time-domain signals into a set of orthogonal Walsh functions (the Sequency domain) can have real practical benefit. For example, if you convert a cardiogram from the time domain into the sequency domain, some heart conditions show up very distinctly as an abrupt change in particular sequency components; if you try to spot the same change in the time domain or the frequency domain it becomes much less distinct.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on September 01, 2015, 01:37:00 PM
Lots can be done with orthogonal functions, met my match with wavelets. Never came across Welsh.

Yes the whole idea is to better discriminate and/or find traits for useful purposes.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 01, 2015, 02:46:51 PM
I've set up a demo to show the 'existence' of the sidebands and to prove the multi tone Fourier synthesis is realistic to what really happens.

Some people might question the use of a conventional envelope detector in a spectrum analyser so the demo below uses an average power meter as a detector

My old HP8566B analyser has a 21.4MHz IF output at the rear. If I set the analyser to linear mode then it becomes a tunable downconverter receiver with variable bandwidth if I select zero span. i.e. in this mode it is no longer a sweeping receiver.

I can then connect an Anritsu thermocouple based power meter to the 21.4MHz IF port. This meter responds to average power and isn't just a peak reading or envelope detector. Because there is a ~20dB conversion loss in this path to the IF I have to fit a small booster amplifier/filter at the IF port to get the signal level into the range of the power meter.

So, if you allow me to increase the OOK rate to 1kHz I can set one of the Agilent generators into internal pulse/OOK mode and send out a 70MHz carrier with 1kHz square OOK.

If tune the analyser to 70MHz CF and set the receiver bandwidth fairly wide, eg 100kHz then this lets in the carrier and all of the (significant) sidebands to the power meter at the 21.4MHz IF once the OOK is turned on. I then set the signal power from the generator to give 0dBm (1mW) on the power meter with the modulation OFF (continuous unmodulated carrier).

I then turned on the modulation and the power meter dropped by very close to 3dB to 500uW as expected because we now have a 50:50 OOK signal and the carrier and sidebands are all 'hitting' the power meter.

I can then change the analyser/receiver BW to 30Hz in zero span mode and then go fishing in 1kHz tuning step sizes to explore the average power levels of each product in the spectrum.

The carrier at 70.000MHz reads about -6dBm or 250uW

The first upper sideband at 70.001MHz reads about -9.9dBm or just over 100uW

The next upper sideband at 70.003MHz reads  about -19.5dBm or about 11.5uW

The first lower sideband at 69.999MHz reads about -9.9dBm or just over 100uW

The next lower sideband at 69.997MHz reads about -19.5dBm or about 11.5uW

You can see that the total power in the sidebands is just under 250uW. You can also see that the total power in the carrier is about 250uW

This accounts for nearly all of the 500uW average power in the OOK signal. Now the power meter reads average power so this shows that even though the OOK signal on the scope looks like it is just a 70MHz signal appearing and disappearing it actually consists of a wide spectrum of signals that are really there.

If the sidebands only exist in the tiny rise/fall time then how come they account for 50% of the signal power? Their share of the duty cycle would be so tiny they would have a tiny share of the average power if they only resided in the tiny risetime period.

But instead they have 50% of the signal power as conventional theory predicts.

Are you bored yet? :)




Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 01, 2015, 05:34:59 PM
Here's an actual live demo. It's two SDR's hooked up back to back (through a 30 dB attenuator). The transmitter is an Ettus B210 and the receiver is a Nuand bladeRF. The steady carrier in the middle is the not perfectly corrected DC offset from the B210 transmitter. There's also a little IQ imbalance on the other side. The key clicks are easily controlled with proper DSP filtering.

Here's what the waterfall looks like with the root raised cosine filters turned off.

(http://www.w6rz.net/nofilt.png)

With the RRC filters on.

(http://www.w6rz.net/filt.png)

The GNU Radio flow graph.

(http://www.w6rz.net/cwfilt.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 01, 2015, 06:42:36 PM
Quote
W6RZ: I wasn't trying to show the spectral tones. I was trying to show that they exist only during the rise and fall portion of the CW element. During the steady state period of the CW element, the bandwidth is very small. This agrees with W8JI's statement.

I think you are still misusing and/or misinterpreting what your tools are telling you.

I don't think there's much more I can do except set you a challenge. If what you say is true then no switching mixer could have any useful purpose or efficiency.
If what you say is true the mixer products would have such a low average power wrt the input signal to the mixer that all of our receivers and transmitters would need huge amplifiers to make up for the losses in the mixers. You need to prove to me and the rest of the engineering world that the efficiency of a switching mixer has somehow been achieved through voodoo because the IF products only exist very briefly according to you.

Also you need to tell them that the faster the edges of the switching the less efficient the mixer becomes because the wanted mixer terms appear for even less time. So you need to prove that the designer of the switching mixer was a fool.

Best of luck :)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 01, 2015, 07:12:56 PM
I'm sorry that the techniques I'm using do not fit into your world view. The algorithm I'm using is no different than what you would find in a modern DSP based amateur radio transceiver.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 02, 2015, 12:35:23 AM
"I shall say this only once ......"

One of you is using a periodic signal, the other is using a non-periodic signal !!!!!!

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 02, 2015, 01:14:58 AM
"I shall say this only once ......"

One of you is using a periodic signal, the other is using a non-periodic signal !!!!!!

Steve G3TXQ

I can send a periodic signal. It doesn't matter. It's the keying speed that matters. If I raise the keying rate to something that's very short compared to the FFT duration, I get a spectrum with side bands.

(http://www.w6rz.net/fastcw.png)

But at human readable keying rates, it's the same as before.

(http://www.w6rz.net/periodic.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 02, 2015, 01:41:14 AM
What's the resolution of spectrogram? It needs to be small compared to the OOK rate. If not, you are seeing more than one sideband, which as I explained several times already will result in a spectrogram that shows key clicks.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 02, 2015, 03:12:42 AM
What's the resolution of spectrogram?
I can't perform high resolution spectrograms in GNU Radio. That was some ridiculously high rate just to show the concept of FFT duration versus keying speed.

But high resolution spectrograms are part of the problem. You can't have a high resolution spectrogram with a short integration time. On my spectrum analyser, an 8 kHz span requires 800 milliseconds. At 24 WPM, that's 8 dits. Way too long.

But the FFT length I've been showing in my waterfalls is 8192 points at 160000 samples per second. That's 8192/160000 = 51.2 milliseconds. At the 5 wpm rate I was using, the dits are 240 milliseconds long and the dashes are 720 milliseconds. So I get to do around 5 FFT's for each dit and around 14 for each dash. In other words, I'm doing a complete FFT at different parts of the waveform (rising and falling edges and the steady state on and off periods).

We know folks can copy CW at 75 WPM, so the integration time (or FFT length if you will) of human hearing has to be less than 16 milliseconds. That's why we can hear key clicks. We have high resolution hearing.

So it's all about sampling at an appropriate rate to show what a key click really is in terms of human perception.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 02, 2015, 06:02:10 AM
I can't perform high resolution spectrograms in GNU Radio. That was some ridiculously high rate just to show the concept of FFT duration versus keying speed.
And therein lies the problem! Unless you have high enough resolution to discriminate the individual sidebands, it's impossible to resolve the question of whether or not they exist continuously.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 02, 2015, 08:11:39 AM
I suppose if the dits where short enough, and sent fast enough, the RF waveform would be triangular, and the artifacts would be continuous.

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 02, 2015, 08:44:40 AM
Pete,

The sidebands are continuous, however long the Dits, if the signal is periodic.

Say we send continuous Dits at 20wpm with hard On-Off keying; that's equivalent to a square wave with a frequency of about 8Hz. That signal will have discrete spectral lines spaced every 16Hz; they are constant amplitude throughout the time that the Dits are sent - including the silences in-between the Dits.

If we "tune in" to some of those sidebands with a receiver which has sufficiently narrow bandwidth that it can resolve a single spectral line (in other words a bandwidth less than 16Hz) we will "hear" a steady tone.

However, if the receiver has a bandwidth which encompasses several of the spectral lines, they combine to produce an impulsive waveform which has high amplitude at the key openings and closures. That doesn't mean the sidebands aren't there and continuous, it simply means we can't resolve them. That's why W6RZ's spectrogram doesn't prove the absence of continuous sidebands.

Reducing the time of the Dits (increasing the Morse speed) doesn't change the theory, but it would make resolution of the individual spectral lines easier because they would be further apart in the frequency domain.

If the signal is non-periodic, things get much more complex!

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 02, 2015, 12:33:46 PM
That's why W6RZ's spectrogram doesn't prove the absence of continuous sidebands.

I'm not trying to prove anything here. I'm just presenting information from my toolbox for folks to consider. At this point, it's become tedious for me and I'm going to bail out. There's more fun stuff to experiment with like OFDM spectrum painting. You can download the code that produced this waterfall here.

https://github.com/drmpeg/gr-paint

(http://www.w6rz.net/paintgqrx2.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 02, 2015, 05:40:24 PM
I've got my signal analyser back at last and I thought it might be interesting to see the spectrum of a 'square' 1Hz OOK waveform. I think this is the dit rate for 2.4 WPM. The signal is being generated in an Agilent vector signal generator set to square wave pulse or OOK mode.

This has fast keying edges but you can see that the 99% OBW bandwidth is very small.

I did a quick excel spreadsheet to predict the 99% occupied BW for 1Hz OOK and it comes out at about 42Hz or +/- 21Hz. The marker in the image is at +21Hz.

The excel spreadsheet is for a perfect OOK signal and it predicts that the sideband at the marker frequency of +21Hz should be about 30dB below the carrier level response at the centre frequency.
The measured plot agrees with this very closely. So the occupied BW will be very close to theory at 42Hz.

Hope this is interesting. The RBW needs to be very low to show the definition of the sidebands at this low sending rate so I've turned it down to 0.2Hz resolution to get this plot across a 100Hz span.

(http://i1227.photobucket.com/albums/ee439/G0HZU/OOK_1Hz.gif) (http://s1227.photobucket.com/user/G0HZU/media/OOK_1Hz.gif.html)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on September 02, 2015, 11:09:15 PM

This has fast keying edges but you can see that the 99% OBW bandwidth is very small.

I did a quick excel spreadsheet to predict the 99% occupied BW for 1Hz OOK and it comes out at about 42Hz or +/- 21Hz. The marker in the image is at +21Hz.

The excel spreadsheet is for a perfect OOK signal and it predicts that the sideband at the marker frequency of +21Hz should be about 30dB below the carrier level response at the centre frequency.
The measured plot agrees with this very closely. So the occupied BW will be very close to theory at 42Hz.

This plot also proves the cause for adequate shaping of the keying envelope in a quite convincing way.
The "occupied bandwidth" calculates and shows to be 42 Hz, with higher-order sidebands extending very far outside this value,
which causes key-click interference far from the center frequency.
For a symmetric keying rate of 1 Hz or 2 Baud, the "necessary bandwidth" using erfc(t) shaped signalling elements would have been only 6 or 10 Hz depending on the
operational need for distinct edges.

In this example, the occupied bandwidth is a factor 7 or more larger than the frequency range actually needed by properly shaped keying.
This is reflected in the ITU necessary bandwidth formulas which depend on using the proper shaping criteria.
A fact that was throughly understood many decades ago by the professional world.

I find it somewhat strange and distressing to see that radio amateurs, that presumably have Extra Class licences and are supposed to know at least the very basics of
modulation and information theory, are disbelieving the underlying fundamentals of modulation,
which furthermore have been formulated by peer review and are accepted by the whole scientific and engineering world.

Understanding modulation spectra is one of cornerstones of "scientific spectrum management", the application of which is one of the reasons that we still have amateur frequency bands.

73/
Karl-Arne
SM0AOM





Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 03, 2015, 01:38:15 AM
Here are a couple of graphs which illustrate how we might hear "clicks", even though individual spectral components are continuous.

(http://www.karinya.net/g3txq/temp/sqwave1.png)

This first graph shows the synthesis of a square wave (red) using 9 odd-order spectral components at frequencies f, 3f, 5f etc up to 17f . Notice that the individual spectral components are continuous sine waves - they don't "do anything different" at the square-wave zero crossings.

Now let's see what happens if we "tune off" to one side of the signal and lose some of the low order spectral components but still encompass more than one of the higher order components. This next graph is the resulting waveform from adding the 9f, 11f, 13f and 15f components:

(http://www.karinya.net/g3txq/temp/sqwave2.png)

Notice that the signal has become impulsive with peaks coincident with the square-wave zero crossings. It is still formed from continuous individual sine waves even though we perceive it as "clicks".

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 03, 2015, 11:09:06 AM
Pete,

The sidebands are continuous, however long the Dits, if the signal is periodic.

 
Steve G3TXQ

Steve, do the side bands EXIST beyond the transitional rise and fall time periods??  That is question. 

Separating fly shit from pepper may be fun, but let's cut to the chase. I still say W8JI has it right.

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 03, 2015, 11:39:06 AM
Steve, do the side bands EXIST beyond the transitional rise and fall time periods??  That is question.

If the signal is periodic, yes the sidebands exist for as long as the signal exists.

Look at the first graph in my post #204. The individual sine waves - the "sidebands" - making up the square-wave exist throughout the period of the square-wave - not just at the transitions. To repeat what I said there: "Notice that the individual spectral components are continuous sine waves - they don't "do anything different" at the square-wave zero crossings."

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 03, 2015, 12:21:23 PM
Pepper.... fly shit... what ever.... then the bands should be full of key click artifacts going back for eons.... 


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 03, 2015, 12:30:10 PM
Why? I said "... the sidebands exist for as long as the signal exists"; I didn't say they existed for ever!

What I'm saying is nothing unusual - it's the whole basis of Fourier analysis.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 03, 2015, 12:37:51 PM
And what I am saying is that Tom is absolutely correct, and I quote:

Any signal changing amplitude must have sidebands. This means the CW signal has sidebands when the rising and falling edges occur. It is obviously not a single frequency except during the time when level is not changing.

When level is changing, it is actually an AM signal.

Bandwidth, or number of sidebands and frequency spread of the sidebands during the rise and fall, is determined by the slope and time of carrier power change while rising and falling.


You guys can dance around the subject, post any number or graphs, but yet no one has disproved the fact that sidebands are only present during a "level change" the carrier, albeit modulation or other artifacts.  Modulation and sidebands go hand in hand. You are still trying to make pepper out of fly shit.

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 03, 2015, 12:38:54 PM
For W8JI to be correct then we wouldn't be able to make receivers or transmitters using mixers. Perhaps the most elegant example of this is if we examine the single balanced switching mixer.

(http://i.stack.imgur.com/YfBrV.gif)

Hopefully I don't have to explain in too much detail about how this device works but at the output of the mixer you will get a timeslice where it appears that you just get the carrier 'in phase' leaving the mixer. Then you get a time period when the carrier leaves the mixer 'out of phase'
The time period is set by the LO.

According to the logic of W8JI and K1ZJH, this mixer can't work the way the engineering world claims because there is zero bandwidth for all of the time except the switch rise/fall time. This would even be true if the rise/fall time was in nanoseconds if they are correct.

But what they have to explain to the world of conventional science is why (in the real worls) we see VERY LITTLE power leaving this mixer at the carrier frequency and almost all of the power is contained in the frequencies that make up the mixer products or 'sidebands'.


Hopefully, this will end this debate :)
 


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 03, 2015, 12:44:25 PM
Products exist for as long as at least two of two signals exist. Otherwise, you are still looking for fly shit . 



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 03, 2015, 12:48:09 PM
 
Products exist for as only AS long as at least two signals exist. Otherwise, you are still looking for fly shit .  Spectral and time domain are two different things.


 


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 03, 2015, 12:52:54 PM
@ K1ZJH: I think you would find it helpful to read some basic texts on Fourier analysis. You'll find they nearly all begin with a statement to the effect that: "any periodic time varying signal can be represented by a series of sinusoids".

Notice that it doesn't say "by bits of sinusoids that only exist for some of the time."

I tried my best, but I give up now  ::) I'll refer other readers to the perceptive comments made by Karl-Arne at the end of his post #203.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 03, 2015, 12:55:58 PM
Simple question... do the key click artifacts cease to exits once the keying waveform is ended? Let's stop the bullshit and get down reality.  Once the keying transitions, stop, do the unwanted artifacts also cease?  Stop being an ass and stick to what Tom was trying to say?  If you can't understand that, I give up.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on September 03, 2015, 12:56:20 PM
 The problem is that this FT model requires a periodic signal, and for all sidebands to be active during the entire period. The idea that periodic waveforms can be represented as a sum of sines and cosines is a very clever and powerful concept as is the transform to a frequency domain in describing bandwidth and signal signatures. Unfortunately the model does not neatly describe non-periodic signals and being a model it may not describe/reflect the real world.

 I think we can prove the idea that clicks are generated during the on/off transitions by blanking out the transitions of a 10 minute pulse and listening for the clicks.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 03, 2015, 01:03:01 PM
The single balanced mixer example is intended as a real world example that flies in the face of W8JI's claims.

Set the LO switch rate to 1Hz and the RF to 1MHz and according to W8JI there will be zero bandwidth for 1 second with the switch set to the in phase signal.

Then zero bandwidth for a second when it is set to the other.


(http://i.stack.imgur.com/YfBrV.gif)

Don't be confused by the triangle symbol. This is normally just a transformer and the 180deg and 0deg taps from the transformer go to the two arms of the switch.

For 1 second you 'send' the in phase output of the transformer and then for 1 second you send the out of phase arm.

But at the output there will be signals at the sidebands frequencies but virtually nothing at the carrier frequency even though it looks like the carrier was 'there' for a whole second with 'zero bandwidth'.

If W8JI is correct then why is there virtually no signal leaving the mixer at the carrier frequency? What happened to it if we think we can see it on a scope for a whole second? Where did the power come from at the sideband frequencies if they only 'exist' for the switching time of the switch (could be as fast as a few nanoseconds)

So how come we have a 'zero bandwidth carrier' for one second and only a 10ns in 1 second  'duty cycle' for the sidebands to exist.

W8JI would have to prove that all of the sideband power occured in that tiny duty cycle and he would have to explain where all of that power that was there at zero bandwidth for 1 second has gone at the carrier frequency?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 03, 2015, 01:06:36 PM
Time domain spectral analysis.  One of two signals disappears... what products remain?

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on September 03, 2015, 01:11:20 PM
But this is NOT OOK. It's a totally different animal - BPSK.#

You cannot compare grapes with turnips......


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 03, 2015, 01:15:37 PM
But this is NOT OOK. It's a totally different animal - BPSK.

OK - completely remove the inverter that provides the 180 degree signal, making it an unbalanced mixer. Do the sidebands now only exist during the transitions of the LO, or are they there all the time?

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on September 03, 2015, 01:23:38 PM
I would argue that without the LO, the sidebands can't exist: thus they must only exist during transitions. But if the LO transitions once per hour and you look for one minute 10 minutes after the transition, your 'window' isn't wide enough to see the sidebands. So depending on the observation window, the sidebands may or not be seen to be present. So 'present' depends on the observation window width....

I think this got raised some inputs back.....


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 03, 2015, 01:28:59 PM
I would argue that without the LO, the sidebands can't exist: thus they must only exist during transitions.

I don't understand that leap?

But if the LO transitions once per hour and you look for one minute 10 minutes after the transition, your 'window' isn't wide enough to see the sidebands. So depending on the observation window, the sidebands may or not be seen to be present. So 'present' depends on the observation window width....

I think this got raised some inputs back.....

Agreed, but does an inability to observe something - because you don't look long enough - mean that the "thing" doesn't exist?

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 03, 2015, 01:36:46 PM
But this is NOT OOK. It's a totally different animal - BPSK.#

You cannot compare grapes with turnips......

Peter

I think this thread was based on OFF ON KEYING.  And, I suspect Tom's observations are close enough to be practical.

So, again, I ask, if Tom was wrong, for how long after the keying waveform transition do unwanted artifacts exist? Nanoseconds, milliseconds? Seconds, hours, days?  Lets get down to earth, and deal with practical matters concerning key clicks and what may or not occur for normal keying rates and waveform shapes.

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K6JH on September 03, 2015, 01:38:07 PM
You guys are all arguing about a pretty narrow point. Of course clicks are generated during a transition. And the signal does not have to be periodic.

How about a single dit, or heck, put a brick on the key and just enable the carrier as long as the electric bill is paid. There will still be a click at the start. How long with the spectra last before they die out below the noise floor of the receiver and be undetectable? Probably about as long as the rise time. Does it really matter if the math says they go on forever, if no one can hear them anymore?

I don't really care to bother doing a fourier analysis on it - I haven't had to do that for 35 years. I'm not going to start in again now.  :D

Where's my popcorn???   ;)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 03, 2015, 01:40:57 PM
You guys are all arguing about a pretty narrow point. Of course clicks are generated during a transition. And the signal does not have to be periodic.

How about a single dit, or heck, put a brick on the key and just enable the carrier as long as the electric bill is paid. There will still be a click at the start. How long with the spectra last before they die out below the noise floor of the receiver and be undetectable? Probably about as long as the rise time. Does it really matter if the math says they go on forever, if no one can hear them anymore?

I don't really care to bother doing a fourier analysis on it - I haven't had to do that for 35 years. I'm not going to start in again now.  :D

Where's my popcorn???   ;)

My hero. You get a bag of popcorn, and a beer. And something more substantial should we me ever have an eyeball in a pub. And I think Tom is invited as well.

Really, isn't this akin to saying that an exited LC circuit will ring beyond the last pulse? The question is what effect in the real world it exhibits.

pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 03, 2015, 01:58:15 PM
Lets get down to earth, and deal with practical matters concerning key clicks and what may or not occur for normal keying rates and waveform shapes.

OK I'll try one more time!

Set your keyer for 20wpm continuous Dots and leave it going for 1 minute. That will generate sidebands - individual spectral components - that are time-continuous and spaced by 16Hz. They will disappear at the end of the 1 minute when the transmitter shuts off.

Whilst the transmitter is on, if you tune away from the carrier with a receiver having a bandwidth significantly greater than 16Hz you will perceive a series of "clicks" coincident with the key closures and openings - not because the individual spectral components are appearing and disappearing, but because you are observing several spectral components which sum to become an impulsive time waveform.

Those are the facts!

Arguing that practical bandwidths don't allow the individual spectral lines to be resolved, and that in practice all we'll ever hear is "clicks" is fair comment; but the enquiring mind will want to understand what the underpinning mechanisms are, rather than deny them.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 03, 2015, 02:01:47 PM
No one is denying anything.... again, using your example, what is the PRACTICAL concern of those artifacts, beyond the last key up????

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 03, 2015, 02:05:41 PM
No one is denying anything.... again, using your example, what is the PRACTICAL concern of those artifacts, beyond the last key up????

Nothing - as I explained, they don't exist in any meaningful way after the transmitter shuts off.

Great about no one denying anything! So now can I assume no-one is denying that the sidebands - the individual spectral components - are present continuously throughout the 1 minute transmission.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 03, 2015, 02:15:59 PM
But this is NOT OOK. It's a totally different animal - BPSK.#

You cannot compare grapes with turnips......

Sure, but it can be interpreted as a different form of on/off keying because you (digitally) turn each phase on and off at the switching rate. Just because you replace the OFF period with something else doesn't mean that the 0deg phase signal hasn't been turned OFF and will be turned ON again maybe a second later.

You can look at that circuit as BPSK or you can look at it as a conventional single balanced switching mixer.

You need to explain to me why the output of that mixer produces the wanted mixer products so efficiently even though you think there is zero bandwidth for maybe 99.99999% of the time according to you.

If the mixer products only exist for a few nanoseconds and the carrier is there with zero bandwidth for a relatively HUGE amount of time  for a 1Hz LO then why is their power share at the output the other way around? i.e. almost ALL of the power is in the mixer products (AKA sidebands)?

BPSK is the equivalent of double sideband suppressed carrier DSBSC when you look at the spectrum. (emphasis on Suppressed Carrier)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 03, 2015, 02:58:09 PM
G0HZU's mixer example is a revealing one, but I would simplify it further to better represent the OOK Morse situation:

(http://www.karinya.net/g3txq/temp/mixer1.jpg)

If the LO signal fo was 1MHz and the input signal f1 was 9MHz, we'd get something like this:

(http://www.karinya.net/g3txq/temp/mixer2.jpg)

The key question then is this: if I filter the output signal with a narrowband filter centred on 10MHz to observe the first-order 10MHz  sideband, do I see a steady 10MHz sine wave or is it only present for brief periods around the transitions of fo?

[Images from http://users.tpg.com.au/users/ldbutler/MixerTheory.htm]

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 03, 2015, 10:35:24 PM
Those are the facts!

Those are most definitely NOT the facts.

Since you are proclaiming factuality, the onus is on you to PROVE it. Of course, the tool of proof here is mathematics, specifically Fourier analysis (in the form of the FFT).

First of all, you need to analyse the correct signal. As G3RZP correctly points out, a square wave is NOT OOK. OOK is a modulated sine wave as pictured below.

Get yourself an FFT analysis tool and come back to us with the results. Be sure to take FFT's with appropriate time intervals (10 to 20 milliseconds, the granularity of human hearing) on a slowly keyed waveform (5 WPM).

If you can't provide these results, then you're just "talking the talk" and not "walking the walk".

Also, I don't appreciate SM0AOM's comment that was obviously targeted at myself and K1ZJH. What are you, the poster child for passive aggressiveness? I've been a ham for 33 years and an engineer for longer. Over the past 3 years I've deeply explored Software Defined Radio and developed DSP code for DVB-S2 and DVB-T2 transmitters (arguably some of the most advanced modulation schemes on the planet). Don't suggest, even for a second, that I don't understand modulation.

https://github.com/drmpeg/gr-dvbs2

https://github.com/drmpeg/gr-dvbt2

(http://www.w6rz.net/ookwaveform.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 04, 2015, 02:23:32 AM
First of all, you need to analyse the correct signal. As G3RZP correctly points out, a square wave is NOT OOK. OOK is a modulated sine wave as pictured below.

G3RZP is correct, but what is true of the spectral components of a baseband square-wave is equally true when it modulates a sine-wave carrier. The spectral components simply move up in frequency around the carrier.

Be sure to take FFT's with appropriate time intervals (10 to 20 milliseconds, the granularity of human hearing) on a slowly keyed waveform (5 WPM).

If I'm trying to prove the existence (or otherwise) of the individual sidebands associated with the signal you show - which basic modulation theory tells us will be spaced about 10Hz apart - I'd be foolish to choose a sampling regime with a resolution bandwith of  100Hz or 50Hz.

Using the granularity that you suggest will simply reveal an impulsive component formed from the summation of at least 5 of the spectral components. It will tell us nothing about the separate components.

I see no-one has yet answered the question I asked in #229

Steve G3TXQ



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 04, 2015, 06:08:10 AM
What is the granularity of human hearing? We know for a fact that humans can copy CW at 36 WPM. At that rate, the dit on time is 33 milliseconds and the dit off time is 33 milliseconds. So there's no doubt that human hearing can discern whether a tone is present or not at that rate. Agreed?

If we can discern whether a tone is present or not every 33 milliseconds, we know that human hearing perceives the world at at least that rate.

In other words, your hearing system is taking a slice of time (no longer than 33 milliseconds) and processing it such that it can discern frequency components. This is exactly what an FFT does. It takes a slice of time and displays the frequency components that occurred during that time.

Now let's slow the rate down to 5 WPM. At that rate, the dit on time is 240 milliseconds and the dit off time is 240 milliseconds.

But we know that the human hearing is taking at most a 33 millisecond slice of time for processing. The tone on time is 240 milliseconds, so we can easily hear a long period of time while the amplitude of the tone is constant. In other words, we're hearing at least seven 33 millisecond slices of tone where the amplitude is constant.

What is the FFT of a constant amplitude sine wave over the selected period (33 milliseconds)? It's a single frequency.

What is the FFT of a zero amplitude sine wave over 33 milliseconds.  It's nothing.

What is the FFT of a sharp rising or falling edge over 33 milliseconds? It's a complex spectrum of tones. A key click.

You can't ignore the granularity of the FFT. It's crucial to the understanding of how things actually work. For example, G0HZU provided us with a wonderful spectrum of a 1 Hz signal. How long did the signal analyser take to make that measurement? It's has to be very long, especially at 25 averages, probably in the minutes. So it's taking a very long duration FFT that contains many many rising and falling edges. It has to take a long time, an FFT can't have both spectral and temporal resolution at the same time.

https://en.wikipedia.org/wiki/Short-time_Fourier_transform#Resolution_issues

As it turns out, I was mistaken before. I can take medium resolution spectrograms with GNU Radio. Here's my calibration image. The sample rate is 210000 Msps. Since it's complex sampling, it's looking at 210 kHz of spectrum. There are 14 horizontal ticks and the zoom is set to 50X. That means each horizontal tick mark is 210000 / 14 / 50 = 300 Hz.

Here's what 360 WPM dits looks like. Oh my goodness, it's the exact spectrum that all of us expect. Tones spaced every 300 Hz.

(http://www.w6rz.net/calibration.png)

Here's 72 WPM.

(http://www.w6rz.net/72wpm.png)

And 36 WPM. Good so far, right?

(http://www.w6rz.net/36wpm.png)

Now here's 2.4 WPM, or 1Hz. I swear on a stack of bibles that the only thing I changed was the keying rate. Same spectrogram as I've been showing right along.

(http://www.w6rz.net/1Hz.png)

In the spectrograms that nicely show spectral tones, we're taking FFT's over many rising and falling edges (just like G0HZU's long FFT duration 1Hz spectrum). I'm taking a 32768 point FFT at 210000 Msps, so the FFT interval is 156 milliseconds. A bit long compared to human hearing, but still in the ballpark.

At 360 WPM, that's an FFT over 46 dits. At 72 WPM, 5 dits and at 36 WPM, 2.5 dits. As long as the FFT time is longer than the dit time, we'll see tones at the correct spacing.

But when the FFT duration is much shorter than the occurrences of rising and falling edges (like human hearing is at up to 36 WPM), the individual on, off and rise/fall time periods are distinctly visible.

If you can't wrap your head around these timing constraints, you'll make the mistake of thinking that side bands exist while the transmitter off.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 04, 2015, 06:25:00 AM
Here are the results from some On-Off keying measurements I did this morning.

I built this circuit to generate the keying:

(http://www.karinya.net/g3txq/temp/mixer1.jpg)

Input f1 was connected to a signal generator and fo to a square wave generator. The output was connected to a spectrum analyser, an oscilloscope and the input of a narrow-band filter. Frequencies were chosen to place a predicted sideband spectral component in the passband of the filter, and to replicate W6RZ's keying waveform which has approximately 50 carrier cycles during the On period.

Here's the keyed waveform as viewed on the oscilloscope connected to the output:

(http://www.karinya.net/g3txq/temp/sqwave5.jpg)

And here's its spectrum as viewed on the spectrum analyser:

(http://www.karinya.net/g3txq/temp/sqwave4.jpg)

The spectrum is centred on the carrier frequency. Notice that there is a carrier component because this is essentially an AM signal with 100% modulation. Notice too that the  even-order spectral components are not fully suppressed - that's a result of the square-wave not being exactly 50/50 On-Off.

The question being debated is whether or not those spectral components are present throughout the keying period. So, here's the output of the filter which is centred on the first sideband (lower trace) compared to the keyed waveform (upper trace).

(http://www.karinya.net/g3txq/temp/sqwave3.jpg)

Notice that it is present continuously whether the keyed waveform is On or Off. This is nothing new - it's basic modulation theory.

Frequencies used were: carrier=1.29MHz, keying square-wave=12.9kHz; those values put the first sideband (1.3029MHz) in the passband of my narrowband filter and emulated W6RZ's keying waveform.

If this were all scaled down in frequency by a factor 1290, to a carrier of 1kHz and a keying waveform of 10Hz - exactly replicating W6RZ's waveform - nothing would change, providing you have the ability to separately resolve the spectral components which will be spaced by 10Hz.

Arguments about "granularity" and what the human ear perceives are obscuration. The fact that the human sensory system can't resolve the sidebands doesn't mean they don't exist; by that logic an audio tone of 25kHz can't exist because I can't hear it!

We can debate this ad-infinitum, but without a common definition of the term "sidebands" we will never have a meeting of minds. G0HZU and I have taken it to mean "individual spectral components", and I've demonstrated that they are present continuously for the waveforms we've considered; I think W8JI ( and perhaps W6RZ) take it to mean something different: the net effect of a number of spectral components defined by human perception/practical receiver bandwidths.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: N3QE on September 04, 2015, 06:30:19 AM
You guys can dance around the subject, post any number or graphs, but yet no one has disproved the fact that sidebands are only present during a "level change" the carrier, albeit modulation or other artifacts.  Modulation and sidebands go hand in hand. You are still trying to make pepper out of fly shit.

If you do not like the results of the Fourier transform (which by definition extends backwards and forwards infinitely in time), you may be happier with a different transform.

Fourier transform converts between a pure 1D time-series, to a 1D pure-spectrum, and back, with no time locality in the spectrum. Yes, if it's a Fourier transform, by definition it goes back before the beginning of time and extends past the heath death of the universe.

"Short-time Fourier transform" is a little closer to how humans think about time and frequency. Wavelet transforms are localized in both time and frequency and used for a lot of "lossy compression" algorithms today because they seem to match human concepts pretty well. These sorts of transforms are used to make "waterfall" displays as well, that show 2D graphs as a function of both frequency and time.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 04, 2015, 06:46:44 AM
@W6RZ: I fully follow your post #232 and agree with most of it; where we differ is in our understanding of when the sidebands exist!

In your final waterfall display they cease to exist because you've deliberately limited the resolution bandwidth to one that you judge is consistent with human perception. However, we both know that they would have existed in that display had you narrowed the resolution bandwidth of the FFT.

I'm pretty uncomfortable with any definition of "existence" which depends on human perception. What if the keying rate were right on the cusp of Fred's "granularity" but not Bill's. Are we saying that sidebands exist for Bill but not for Fred? What sort of science is it that depends on human subjectivity?

Steve G3TXQ



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 04, 2015, 07:03:15 AM
@W6RZ: I fully follow your post #232 and agree with most of it; where we differ is in our understanding of when the sidebands exist!

In your final waterfall display they cease to exist because you've deliberately limited the resolution bandwidth to one that you judge is consistent with human perception. However, we both know that they would have existed in that display had you narrowed the resolution bandwidth of the FFT.

I'm pretty uncomfortable with any definition of "existence" which depends on human perception. What if the keying rate were right on the cusp of Fred's "granularity" but not Bill's. Are we saying that sidebands exist for Bill but not for Fred? What sort of science is it that depends on human subjectivity?

Steve G3TXQ


No, the resolution bandwidth is the same in all of the images (4200 Hz). The frequency scale in the middle doesn't change. As I said, all I changed was the keying rate.

And yes, limiting things to what humans can perceive un-augmented is not ideal. However, we use tools to augment our perception all the time. Telescopes, microscopes, oscilloscopes and high speed computers.

All you have to do is throw more computing power at it (along with higher sampling rates), and you'll "see" distinct on/off periods up to very high rates.

Check out this video.

https://www.youtube.com/watch?v=mjD-l3GAghU


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 04, 2015, 07:44:34 AM
If you do not like the results of the Fourier transform (which by definition extends backwards and forwards infinitely in time), you may be happier with a different transform.

Fourier transform converts between a pure 1D time-series, to a 1D pure-spectrum, and back, with no time locality in the spectrum. Yes, if it's a Fourier transform, by definition it goes back before the beginning of time and extends past the heath death of the universe.

"Short-time Fourier transform" is a little closer to how humans think about time and frequency. Wavelet transforms are localized in both time and frequency and used for a lot of "lossy compression" algorithms today because they seem to match human concepts pretty well. These sorts of transforms are used to make "waterfall" displays as well, that show 2D graphs as a function of both frequency and time.

Good point. I was being lazy and using the term "FFT" to really mean "DFT" (Discrete Fourier Transform).


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 04, 2015, 07:57:35 AM
No, the resolution bandwidth is the same in all of the images (4200 Hz). The frequency scale in the middle doesn't change. As I said, all I changed was the keying rate.

Yes - I expressed that badly. But the significant change was that the reduction in keying rate narrowed the sideband spectral-line spacing to below that of the resolution bandwidth.

And yes, limiting things to what humans can perceive un-augmented is not ideal. However, we use tools to augment our perception all the time. Telescopes, microscopes, oscilloscopes and high speed computers.

In that case, if I'm checking for the existence of narrow-spaced spectral components, I'll choose to augment my human perception with tools that have the required resolution.

But if others choose to say that the sidebands don't exist simply because humans don't perceive them, I guess we'll just have to agree to differ. I prefer my science to be a little more objective that that.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 04, 2015, 08:29:38 AM
Yes - I expressed that badly. But the significant change was that the reduction in keying rate narrowed the sideband spectral-line spacing to below that of the resolution bandwidth.

So what? You see a muddy representation of the sidebands only during the start and end of the dit. The key clicks themselves. Between dits, you don't see anything because there's nothing to see. The sidebands only exist during the rise and fall time of the dit.

In that case, if I'm checking for the existence of narrow-spaced spectral components, I'll choose to augment my human perception with tools that have the required resolution.

But if others choose to say that the sidebands don't exist simply because humans don't perceive them, I guess we'll just have to agree to differ. I prefer my science to be a little more objective that that.

Steve G3TXQ

Don't get hung up on the human perception thing. I was just using that to relate the timings to human readable CW rates. The author of the high speed spectrogram in the video just told be that distinct OOK on/off transitions should be able to be seen at 50 kHZ rates.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 04, 2015, 08:41:27 AM
Between dits, you don't see anything because there's nothing to see.

Sure there is - keep the keying rate the same and run your waterfall again with a narrower resolution bandwidth and you'll see them magically appear!

Your last waterfall shows a periodic signal that repeats every 1 second. So if you want to understand what its spectral components are you need a total sampling time of at least 1 second.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 04, 2015, 08:58:18 AM
The existence of side band energy may be continuous for a sinusoidal (or similar) modulating signal, but it doesn't hold true for a waveform that starts and ends abruptly.

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 04, 2015, 09:10:24 AM
The existence of side band energy may be continuous for a sinusoidal (or similar) modulating signal, but it doesn't hold true for a waveform that starts and ends abruptly.

See my #233
See W6RZ's #232 (First 3 waterfall displays)

Those are both On-Off "hard-keyed" signals and both show continuous sideband components.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 04, 2015, 09:12:42 AM
Between dits, you don't see anything because there's nothing to see.

Sure there is - keep the keying rate the same and run your waterfall again with a narrower resolution bandwidth and you'll see them magically appear!

No. Nothing will appear. The only thing that will happen is the key clicks at the beginning and end of the dit will be more clearly resolved in frequency.

Your last waterfall shows a periodic signal that repeats every 1 second. So if you want to understand what its spectral components are you need a total sampling time of at least 1 second.

Steve G3TXQ

No. All you need is high spectral resolution during the rise and fall times.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 04, 2015, 09:47:03 AM
No. Nothing will appear. The only thing that will happen is the key clicks at the beginning and end of the dit will be more clearly resolved in frequency.

You demonstrated very convincingly in the first three spectrograms of your #232 that the spectral components are continuous provided the keying rate is high in relation to the resolution bandwidth.

You then reduced the keying rate by 15 times from 36wpm to 2.4wpm - without changing the resolution bandwidth - and the continuous spectral components "disappeared". But if you now improve your resolution bandwidth commensurate with the drop in keying rate, you'll see continuous spectral components again. Try it - you can't expect to represent a 1Hz signal properly without sampling it over at least 1 second.

But even so, your third spectrogram is very telling. It shows that on OOK signal, at a perfectly practical keying rate of 36wpm, has continuous spectral components.

Steve G3TXQ




Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on September 04, 2015, 09:51:49 AM
This discussion does seem to be trending towards something like the Council of Trent (1545 - 63) with the discussion of 'how many angels could dance on the head of a pin'.

If you consider a carrier with a very short rise time switched on and left on for say, one minute, there will be a click when it switched on and a click when it is switched off. Those clicks will be seen during the amplitude transition as sidebands, but while there is no carrier or while it is not changing in amplitude, there won't be any sidebands. I think we all agree on that, at least.

The original question was on the lines of 'does non-linearity make the sidebands worse because of intermodulation?' I suspect that it will depend on the effect of the non-linearity on the rise and fall times and to quantify it by saying that X dB down of third order product will degrade the transient by YdB isn't possible because the non-linearity in the transfer coefficient that produces XdB down third order product is not a fixed quantity - it's made up from terms in both the third and fifth order coefficients of the transfer characteristic.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 04, 2015, 09:55:49 AM
Those clicks will be seen during the amplitude transition as sidebands, but while there is no carrier or while it is not changing in amplitude, there won't be any sidebands. I think we all agree on that, at least.

 

No,  that is what W8JI said--the sidebands only exist when an AC component (or composite waveform)  is modulating a carrier, and he was immediately attacked and ridiculed for stating the obvious. It went to hell afterwards.  None the less, this thread remains amusing.  Sure wish Tom would defend his position. 

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 04, 2015, 09:58:37 AM
You demonstrated very convincingly in the first three spectrograms of your #232 that the spectral components are continuous provided the keying rate is high in relation to the resolution bandwidth.

You then reduced the keying rate by 15 times from 36wpm to 2.4wpm - without changing the resolution bandwidth - and the continuous spectral components "disappeared". But if you now improve your resolution bandwidth commensurate with the drop in keying rate, you'll see continuous spectral components again. Try it - you can't expect to represent a 1Hz signal properly without sampling it over at least 1 second.

But even so, your third spectrogram is very telling. It shows that on OOK signal, at a perfectly practical keying rate of 36wpm, has continuous spectral components.

Steve G3TXQ




All we have to do is look at something in between. Here's 12 WPM.  Spectral components in the upper portion and visible on/off periods in the waterfall. If I could increase the waterfall rate to something higher than 60 fps, the on/off periods would stand out more.

(http://www.w6rz.net/12wpm.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 04, 2015, 09:58:53 AM
OK, once more I will try and end this debacle...

If keyclicks have a wide spectrum only at very brief intervals as W6RZ claims then I can replicate this waveform using one of my Agilent E4433 sig gens. I can turn on the arb waveform generator and tell it to generate a wideband noise jammer across the whole of the 2metre band.

If I do this then I see a 2MHz wide wall of noise across the whole band and it it wipes out everything below S9. i.e. the whole band is just a constant S9 of wideband mush noise.

I can then tell it to only pulse this noise ON at a 5Hz rate and only enable it for maybe 500 microseconds at this 5Hz PRF.  So this is approximately what W6RZ claims the spectrum of an unfiltered CW signal looks like with a shortish envelope risetime. But there is obviously no carrier in this case. Just the 'clicks'.

So I'm now generating the sideband/click spectrum that W6RZ claims exists. The whole band shows very familiar sounding clicks at a 5Hz rate no matter where I tune my FT290R in CW mode.

But my FT290R doesn't agree with W6RZ's theory because it's classic (gating) noise blanker can wipe away these brief wideband gated noise 'clicks' very easily. They completely disappear even if they are huge in level and very loud when the noise blanker is turned off. The noise blanker works by blanking the receiver at each click event for slightly longer than the duration of the impulse noise to remove these pulses.

i.e. it receives each 500us slice of noise and detects its duration and then generates a slightly wider blanking pulse that straddles the original pulse of noise in time. so the noise pulse gets blanked very efficiently. It is extremely effective :)

But the FT290R can't deal with the (weak or strong) 'clicks' from a real 5Hz OOK signal from the same type of generator if I tune up and down across 20kHz either side of the OOK signal.

This is because these clicks are different even though they 'sound' exactly the same via the speaker. They consist of a constant wide spectrum of tones at 10Hz intervals and dozens of them are able to get through the SSB filter in real time (rather than in short bursts)

So this classic type of noise blanker can't blank them out because they don't produce the same brief impulse (at the 10.7MHz IF) that it is designed to detect and blank.

Does this help you at all W6RZ? If not then you need to explain to me why the noise blanker is brilliant, absolutely BRILLIANT at removing the brief impulse noise you claim exists with a CW key. But it is HOPELESS at dealing with 'clicks' from a real 5Hz OOK signal at 144.1MHz from the same generator. They don't go away in the slightest!

This is a really good test. But then I think all the tests I've done have been valid.






Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 04, 2015, 10:11:50 AM
OK, once more I will try and end this debacle...

If keyclicks have a wide spectrum only at very brief intervals as W6RZ claims then I can replicate this waveform using one of my Agilent E4433 sig gens. I can turn on the arb waveform generator and tell it to generate a wideband noise jammer across the whole of the 2metre band.

If I do this then I see a 2MHz wide wall of noise across the whole band and it it wipes out everything below S9. i.e. the whole band is just a constant S9 of wideband mush noise.

I can then tell it to only pulse this noise ON at a 5Hz rate and only enable it for maybe 500 microseconds at this 5Hz PRF.  So this is approximately what W6RZ claims the spectrum of an unfiltered CW signal looks like with a shortish envelope risetime. But there is obviously no carrier in this case. Just the 'clicks'.

So I'm now generating the sideband/click spectrum that W6RZ claims exists. The whole band shows very familiar sounding clicks at a 5Hz rate no matter where I tune my FT290R in CW mode.

But my FT290R doesn't agree with W6RZ's theory because it's classic (gating) noise blanker can wipe away these brief wideband gated noise 'clicks' very easily. They completely disappear even if they are huge in level and very loud when the noise blanker is turned off. The noise blanker works by blanking the receiver at each click event for slightly longer than the duration of the impulse noise to remove these pulses.

i.e. it receives each 500us slice of noise and detects its duration and then generates a slightly wider blanking pulse that straddles the original pulse of noise in time. so the noise pulse gets blanked very efficiently. It is extremely effective :)

But the FT290R can't deal with the (weak or strong) 'clicks' from a real 5Hz OOK signal from the same type of generator if I tune up and down across 20kHz either side of the OOK signal.

This is because these clicks are different even though they 'sound' exactly the same via the speaker. They consist of a constant wide spectrum of tones at 10Hz intervals and dozens of them are able to get through the SSB filter in real time (rather than in short bursts)

So this classic type of noise blanker can't blank them out because they don't produce the same brief impulse (at the 10.7MHz IF) that it is designed to detect and blank.

Does this help you at all W6RZ? If not then you need to explain to me why the noise blanker is brilliant, absolutely BRILLIANT at removing the brief impulse noise you claim exists with a CW key. But it is HOPELESS at dealing with 'clicks' from a real 5Hz OOK signal at 144.1MHz from the same generator. They don't go away in the slightest!

This is a really good test. But then I think all the tests I've done have been valid.






500 microseconds is too long. Key clicks are much shorter in duration.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 04, 2015, 10:24:57 AM
If you consider a carrier with a very short rise time switched on and left on for say, one minute, there will be a click when it switched on and a click when it is switched off. Those clicks will be seen during the amplitude transition as sidebands, but while there is no carrier or while it is not changing in amplitude, there won't be any sidebands. I think we all agree on that, at least.
No - I don't agree! If it's a signal that repeats every one minute it will have spectral components that are spaced every 0.017Hz from the carrier - again, basic modulation theory. If you want to know whether or not they exist and are continuous, measure them with something that has a bandwidth which can resolve those separate components.

The "clicks" you hear are a result of not resolving them separately. When you sum a large subset of those individual components, their envelope becomes impulsive and that's what you hear.

If by "sidebands" you mean a summation of several individual spectral components, then I agree they are not continuous; but if we are talking about separate spectral components, then they are continuous.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 04, 2015, 10:31:31 AM
If you consider a carrier with a very short rise time switched on and left on for say, one minute, there will be a click when it switched on and a click when it is switched off. Those clicks will be seen during the amplitude transition as sidebands, but while there is no carrier or while it is not changing in amplitude, there won't be any sidebands. I think we all agree on that, at least.
No - I don't agree! If it's a signal that repeats every one minute it will have spectral components that are spaced every 0.017Hz from the carrier - again, basic modulation theory. If you want to know whether or not they exist and are continuous, measure them with something that has a bandwidth which can resolve those separate components.

The "clicks" you hear are a result of not resolving them separately. When you sum a large subset of those individual components, their envelope becomes impulsive and that's what you hear.

If by "sidebands" you mean a summation of several individual spectral components, then I agree they are not continuous; but if we are talking about separate spectral components, then they are continuous.

Steve G3TXQ

You know the saying, "pictures or it didn't happen". Show us a mathematical tool that proves your point. Words aren't enough for this discussion.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 04, 2015, 10:42:55 AM
All we have to do is look at something in between. Here's 12 WPM.  
Sorry - that's not enough to resolve the spectral components!

Keep the keying at 12wpm, lengthen the total period over which you are taking samples - it needs to be at least 0.2s,  and reduce your sampling rate to compensate if you need to. Then show us the results.

Quote
You know the saying, "pictures or it didn't happen"

See  #204 for the pictures

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 04, 2015, 10:43:53 AM
Quote
500 microseconds is too long. Key clicks are much shorter in duration.

It does the same thing at 100us pulse width.

How quickly do you think a ham transmitter can key up a carrier? 100us risetime is going to be 50 times quicker than the recommended risetime.

Why did they choose 5ms when you could have told them they only need 100us to avoid clicks?
 


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 04, 2015, 10:46:09 AM
Quote
500 microseconds is too long. Key clicks are much shorter in duration.

It does the same thing at 100us pulse width.

How quickly do you think a ham transmitter can key up a carrier? 100us risetime is going to be  50 times quicker than the recommended risetime.
 

What does that have to do with the discussion at hand?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 04, 2015, 10:53:44 AM
Quote
You know the saying, "pictures or it didn't happen". Show us a mathematical tool that proves your point. Words aren't enough for this discussion.

Just look up the equations for a single unbalanced switching mixer.


If you on/off key a carrier by hand at 10 dits per second they you are the LO of a switching mixer. Look up the equations of a switching mixer to find what you are looking for.

You seem to want to argue that mixers don't work as (efficient) frequency conversion devices because the wanted mixer conversion products (AKA sidebands) only exist for a tiny moment according to you.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 04, 2015, 10:56:29 AM
Your are right... I just listened on twenty meters, and the  key clicks from the last CQ Contest are still there!!! 

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 04, 2015, 11:00:56 AM
Quote
What does that have to do with the discussion at hand?

I want W6RZ to tell me how long he/she thinks the click spectrum lasts for :)



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 04, 2015, 11:41:00 AM
Just to recap, I'm now down to a 52us pulse width for the click duration. i.e. I'm sending a 52us wide slice of wideband noise to the FT290R at a 10Hz rate. See the image below.

So ten times a second I send a very brief burst (or click?) of wideband noise to the FT290R.

Sure enough the FT290R shows loud clicks at a 10Hz rate and they sound just like CW keyclicks. But because they are so brief in existence the noise blanker can totally remove this click even if it registers S9 on the meter.

But if I send genuine OOK at 10Hz then the FT290R noise blanker can't remove the clicks. It can't reduce them at all. This is because the sidebands from OOK that cause the clicks are NOT appearing in tiny timeslices like the waveform below. So W6RZ needs to explain to me why this is so :)

Note that 52us is 100 times shorter than the 5ms rise/fall duration recommended for CW :)


(http://i1227.photobucket.com/albums/ee439/G0HZU/52us_burst.jpg) (http://s1227.photobucket.com/user/G0HZU/media/52us_burst.jpg.html)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 04, 2015, 12:28:54 PM
@G0HZU: W6RZ has already demonstrated that the sidebands are continuous.

He has shown us in #232 Spectrogram 1 that a carrier keyed on and off at 360wpm has continuous spectral components. In #232 Spectrogram 2 he demonstrates that reducing the keying rate to 72wpm still generates continuous spectral components; and in #232 Spectrogram 3 he demonstrates that they are still continuous at 36wpm.

Clearly, nothing abruptly changes as you reduce the keying rate still further - why would it? The sidebands simply get closer together. His latest spectrogram at 12wpm appears to show a change, but that's because he's not yet analysing the signal over a full Dot cycle.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 04, 2015, 02:02:30 PM
Agreed.

Going back about 10 pages I did try and tell everyone what I expected a waterfall to display.
Quote
G0HZU: I'd expect to see continuous spectral lines on the waterfall at the carrier frequency and the mod sidebands at +/-1Hz and also all the other sidebands at +/- 3Hz, +/- 5Hz, +/-7Hz etc etc. So it would look like a load of vertical streamer lines with no gaps in the streamers for any of the tones on the spectrum waterfall.

If anyone else is still in doubt then look up the equations for a simple mixer based on a series ON/OFF switch. A CW key 'IS' the LO of a switching mixer.

These equations are based on trig identities that date back a very long time and these identities are even taught at school level. In the above case you are mixing the RF signal with the LO and all of the odd order harmonics of the LO to produce a suite of frequencies.
That's what mixers do :)

If you depart from a 50:50 duty for the OOK then you have to expand the equations to include even order LO terms in the multiplication. It makes it more complicated to work out a Fourier synthesis but it can still be done :)



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 04, 2015, 07:23:03 PM
Your are right... I just listened on twenty meters, and the  key clicks from the last CQ Contest are still there!!!  

Pete

Although K1ZJH is making a joke here, it certainly pertains to the matter at hand. The "sidebands exist between dits" crowd are essentially saying that energy is being transmitted while the transmitter is physically OFF. It doesn't take much common sense to realize that that idea is FALSE.

So G3TXQ and G0HZU, solve the insurmountable paradox you've created in your theory. Where does the energy in the sidebands come from? It can't the transmitter, it's OFF. Is it the luminiferous aether ringing?

And G3TXQ, you can't point to the waterfalls I posted that appear continuous. If I could have used a waterfall update rate higher than 60 fps, they wouldn't look continuous. Please think about the things I'm showing you, instead of reacting to them to support your incredibly flawed premise.



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 05, 2015, 12:50:27 AM
So G3TXQ and G0HZU, solve the insurmountable paradox you've created in your theory. Where does the energy in the sidebands come from? It can't the transmitter, it's OFF.

It comes from the transmitter - there's nowhere else for it to come from. The total energy released during the ON period is redistributed in the frequency domain amongst the multitude of continuous spectral components. If you need to know more, see Parseval's Theorem - it addresses this very issue of energy conservation.

The mixer example I described in #233 demonstrates the process well. To remind you, here are the waveforms:

(http://www.karinya.net/g3txq/temp/sqwave3sm.jpg)

The top trace is the output of the switching mixer - the carrier is ON for only 50% of the time, but the energy in the sideband shown in the lower trace is present continuously. This is the real world, not some mathematical abstraction - you can extract energy from that sideband even during the periods that the carrier is OFF!

That should surprise no-one - it's our familiar experience and pervades the design of our radios. But some folk then seem to find it difficult to accept when the ON-OFF keying rate becomes slow, although of course the physics remains exactly the same.

The apparent paradox is no different than the equivalent synthesis one: we know we can sum the output from a large number of sine-wave generators to produce a periodic impulse - a short ON period followed by a long OFF period. Energy is being supplied continuously by the generators, but energy only exists in the pulse during the ON period; so where does the energy go that was supplied by the generators during the OFF period?

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 05, 2015, 04:06:18 AM
This discussion seems to be going nowhere, so this will be my last posting. I'll try to summarise the points I've been making over these many pages.

Read a textbook on Fourier Analysis and you'll learn that any signal can equally well be represented in the frequency domain as in the time domain. If the time domain representation is periodic with interval T, the frequency domain equivalent will be a set of sinusoids; they will be spaced in frequency by 1/T and will be constant amplitude - they exist for as long as the signal exists.

We readily accept this is true when applied to something like the switching mixer example I showed in #233. It's obvious that there is energy in the sideband even during the OFF period of the mixer. It only seems to become a paradox for some folk when the switching rates become low.

There's an obvious explanation: When dealing with RF mixer circuitry, the bandwidths of the filters involved allow us easily to select an individual spectral component which, as the Fourier Theorem tells us, must be continuous. But we're much less familiar with viewing individual components at much slower switching rates. When we analyse a signal with a keying rate equivalent to say, 12wpm, the individual spectral components are only 5Hz apart, and with the typical bandwidths available to us we never view a single component.

When we sum several spectral components the resulting waveform is impulsive as I showed in #204 and as W6RZ's spectrograms also illustrate. The perception then is one of impulsive "clicks", even though they result from an aggregation of continuous spectral components.

The amplitudes of the individual components in the frequency domain satisfy Parseval's Theorem - that the total energy in the individual sinusoids equals the total energy in the signal, provided you integrate the energy over time T. Conservation of energy between the time domain and frequency domain does not apply if you integrate over "time slices" less than T.

We're familiar with this temporal dispersion of energy when moving from time-domain to frequency-domain, even though we might not realise it. It's inherent in the mixer example, where energy is available from the sideband even when none is being supplied from the mixer. We experience it, too, when we feed a narrow pulse into a high-pass filter - we continue to see energy at the output of the filter long after the pulse has finished. In either of those examples it would be foolish to try to apply conservation of energy to a narrow time slice of the input and output waveforms.

I've enjoyed the discussion - thanks to everyone who has contributed.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 05, 2015, 06:49:25 AM
So power can exist when no power is being input. So you're describing some kind of storage media. Where is the RF power being stored between dits?

Is it in some quantum based alternate universe?

Is it in pixie dust?

Is it in 3 day old donuts?

It seems like you're willing to argue all of the above answers to support your logically inconsistent theory.

You guys are wasting your time with ham radio. You should be building perpetual motion machines.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 05, 2015, 06:56:08 AM
Once the AC component is gone, the side bands also cease  to exist.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on September 05, 2015, 07:00:05 AM
I have trouble with this concept of power when not transmitting. I turn on the carrier - it's a sharp rise so there's a transient and sidebands. The power reaches a steady state - are there sidebands now? There's no change in amplitude, so where would they come from? I turn off the carrier, there's a transient and the power decays and there are sidebands decaying. Now the transmitter is off, and there are no sidebands.

So my argument is that sidebands - clicks - can only exist during the period when the amplitude is changing. When it isn't changing, there aren't any sidebands, which is why the observation window width is important.

I think I'm saying much the same as K1ZJH and W6RZ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 05, 2015, 07:09:01 AM
I have trouble with this concept of power when not transmitting. I turn on the carrier - it's a sharp rise so there's a transient and sidebands. The power reaches a steady state - are there sidebands now? There's no change in amplitude, so where would they come from? I turn off the carrier, there's a transient and the power decays and there are sidebands decaying. Now the transmitter is off, and there are no sidebands.

So my argument is that sidebands - clicks - can only exist during the period when the amplitude is changing. When it isn't changing, there aren't any sidebands, which is why the observation window width is important.

I think I'm saying much the same as K1ZJH and W6RZ

Yes, we're all on the same page. The logically consistent one.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 05, 2015, 08:36:25 AM
I turn on the carrier - it's a sharp rise so there's a transient and sidebands. The power reaches a steady state - are there sidebands now? There's no change in amplitude, so where would they come from? I turn off the carrier, there's a transient and the power decays and there are sidebands decaying. Now the transmitter is off, and there are no sidebands.

So my argument is that sidebands - clicks - can only exist during the period when the amplitude is changing. When it isn't changing, there aren't any sidebands, which is why the observation window width is important.

If a fresh faced student listened to what you say above then I think they would conclude this:

They would conclude that if they monitored a stream of (100dps) OOK dits sent at 7MHz (with average power = 1W) then nearly all of this 1W average power would be at 7.00000MHz i.e. maybe 99%.
This is because the rise/fall time is a tiny percentage of the on/off period and there's effectively a 7MHz signal present nearly half of the time.

They would conclude that the average power in the sidebands would be tiny in comparision.

Do you agree with the student?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 05, 2015, 08:40:41 AM
The apparent paradox is no different than the equivalent synthesis one: we know we can sum the output from a large number of sine-wave generators to produce a periodic impulse - a short ON period followed by a long OFF period. Energy is being supplied continuously by the generators, but energy only exists in the pulse during the ON period; so where does the energy go that was supplied by the generators during the OFF period?

Steve G3TXQ


And therein lies the fallacy of your argument.  There are no "generators" feeding a mixer.  Only the carrier, and the AC modulating component during the fall and rise times.  Once the key is open, BOTH signals are gone once the falling edge decays to zero reference.  Nothing goes on forever, since all energy has been dissipated.

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 05, 2015, 08:58:25 AM
Quote
And therein lies the fallacy of your argument.  There are no "generators" feeding a mixer.  Only the carrier, and the AC modulating component during the fall and rise times.  Once the key is open, BOTH signals are gone once the falling edge decays to zero reference.  Nothing goes on forever, since all energy has been dissipated.

Quote
Your are right... I just listened on twenty meters, and the  key clicks from the last CQ Contest are still there!!!

What are you saying here? No one is saying the OOK signal lasts forever. Keeping it 'real', the spectral tones exist for as long as the pattern of dits is sent. That's how a mixer works.
If you want to argue against this by claiming that the mixer doesn't work like this because you eventually have to turn the carrier or LO signal off (permanently) then this is just going to get silly.






Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 05, 2015, 09:01:26 AM
I am saying that the sidebands exist for only as long as the RF signal is varying in amplitude.  Pretty simple. Think about it.



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 05, 2015, 09:09:39 AM
So do you agree with the 'student' in the example below:

Quote
If a fresh faced student listened to what you say above then I think they would conclude this:

They would conclude that if they monitored a stream of (100dps) OOK dits sent at 7MHz (with average power = 1W) then nearly all of this 1W average power would be at 7.00000MHz i.e. maybe 99%.
This is because the rise/fall time is a tiny percentage of the on/off period and there's effectively a 7MHz signal present nearly half of the time.

They would conclude that the average power in the sidebands would be tiny in comparision.

Please give me an answer to this rather than just dodge it :)






Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 05, 2015, 09:26:37 AM
The argument is how long the side bands persist. Stop with your nonsense.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 05, 2015, 09:51:16 AM
I can only conclude from your aggressive reply that you either want to dodge the question or you simply don't know the answer :)

It's a basic question about AM modulation with a simple answer.




Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 05, 2015, 10:10:19 AM
You were given answers. 


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 05, 2015, 10:17:32 AM

It comes from the transmitter - there's nowhere else for it to come from. The total energy released during the ON period is redistributed in the frequency domain amongst the multitude of continuous spectral components. If you need to know more, see Parseval's Theorem - it addresses this very issue of energy conservation.

The mixer example I described in #233 demonstrates the process well. To remind you, here are the waveforms:

(http://www.karinya.net/g3txq/temp/sqwave3sm.jpg)

 
Steve G3TXQ

The top waveform clearly shows that the RF waveform is free of RF energy when the key is released. We are not taking about mixers, we are talking about a keyed Class Amplifier that has all RF drive being removed.

You can have two signals into a mixer, and produce the waveform you are showing by removing one, while showing the CW carrier on the other sweep, but to compare apples and oranges is pure nonsense!!  There is only one RF signal into the amplifier. It IS NOT a mixer, and neither the RF carrier or AC components exist beyond the rise and fall times!  You guys can continue this nonsense forever, but your views defy physics.  And while I agree that mixers are indeed modulators, you also need to realize that in the examples that were being discussed, ALL inputs to BOTH mixer ports were removed once the key was released.  To have an output, a mixer requires TWO inputs. Once the RF carrier, and once the rise and fall times have ended, BOTH requirements have ceased to exist.

You are confusing time and frequency domains, and using superfluous arguments to prove pseudo science. Besides my own modest opinions, several others have also tried to explain matters.  You can continue to bait, I am done with this nonsense.

To be fair, there may be some ringing in the RF tank, pursuant to the Q of tank, that may allow RF to persist for a few cycles due to "ringing," or flyback energy in the plate tank circuit.

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 05, 2015, 10:37:05 AM
 

It comes from the transmitter - there's nowhere else for it to come from. The total energy released during the ON period is redistributed in the frequency domain amongst the multitude of continuous spectral components. If you need to know more, see Parseval's Theorem - it addresses this very issue of energy conservation.

The mixer example I described in #233 demonstrates the process well. To remind you, here are the waveforms:

(http://www.karinya.net/g3txq/temp/sqwave3sm.jpg)

 
Steve G3TXQ

The top waveform clearly shows that the RF waveform is free of RF energy when the key is released. We are not taking about mixers, we are talking about a keyed Class Amplifier that has all RF drive being removed.

You can have two signals into a mixer, and produce the waveform you are showing by removing one, while showing the CW carrier on the other sweep, but to compare apples and oranges is pure nonsense!!  There is only one RF signal into the amplifier. It IS NOT a mixer, and neither the RF carrier or AC components exist beyond the rise and fall times!  You guys can continue this nonsense forever, but your views defy physics.  And while I agree that mixers are indeed modulators, you also need to realize that in the examples that were being discussed, ALL inputs to BOTH mixer ports were removed once the key was released.  To have an output, a mixer requires TWO inputs. Once the RF carrier, and once the rise and fall times have ended, BOTH requirements have ceased to exist.

You are confusing time and frequency domains, and using superfluous arguments to prove pseudo science. Besides my own modest opinions, several others have also tried to explain matters.  You can continue to bait, I am done with this nonsense.

To be fair, there may be some ringing in the RF tank, pursuant to the Q of tank, that may allow RF to persist for a few cycles due to "ringing," or flyback energy in the plate tank circuit.

A few years ago a PE used a page of math to "prove" that a diode in series with a AC supply produced !/2 voltage instead of I/2 power (.707 x 1/2 RMS)... eventually Fluke set him straight, but he had produced a webpage full of equations that "proved" he was correct.

Pete
  


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 05, 2015, 11:13:48 AM
Quote
ALL inputs to BOTH mixer ports were removed once the key was released.  To have an output, a mixer requires TWO inputs. Once the RF carrier, and once the rise and fall times have ended, BOTH requirements have ceased to exist.

I think the disconnect in thinking is that in my example I keep keying on and off at the LO rate. You seem to want to just look at a single step event.

If I am allowed to keep keying the transmitter then I am mixing the carrier with the frequency content of a square wave because a repeated ON/OFF sequence is a square wave input to the mixer LO port.
This square wave will contain frequencies at 100Hz, 300Hz, 500Hz, 700Hz.....etc etc.

My argument applies for as long as I keep sending the pattern of dits.

So in my example I am 'mixing' a 7MHz carrier with a 100Hz square wave to provide a spectrum that contains 1W average power. 0.5W of that average power is the carrier and the other 0.5W is the total power in the sidebands that exist at 100Hz, 300Hz, 500Hz, 700Hz.... offsets etc.

When you look at this waveform on a scope it will look like OOK at a 100Hz rate even though it actually consists of a wide spectrum of signals.

Does this help at all?

If this makes sense then we can all go home :)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 05, 2015, 11:25:24 AM
Unfortunately, the carrier is also keyed. it also ceases to exist.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 05, 2015, 12:49:16 PM
Keeping it 'real', the spectral tones exist for as long as the pattern of dits is sent.

It's logically inconsistent.

So we have a magical receiver that can detect the spectral tones between dits. If we detect tones between dits, we know that another dit is coming along.

In other words, we can predict the future.

I take it back, you guys shouldn't be building perpetual motion machines, you should be investing in the stock market.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K5WLR on September 05, 2015, 01:15:02 PM
 :o ??? ::) :P


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 05, 2015, 01:30:34 PM
Keeping it 'real', the spectral tones exist for as long as the pattern of dits is sent.

It's logically inconsistent.

So we have a magical receiver that can detect the spectral tones between dits. If we detect tones between dits, we know that another dit is coming along.

In other words, we can predict the future.

I take it back, you guys shouldn't be building perpetual motion machines, you should be investing in the stock market.


It's not logically inconsistent.

The mixer inputs are a 100Hz squarewave and a 7MHz carrier frequency. The output looks like an OOK signal in the time domain.

So I'm arguing that the output OOK waveform is a function of these frequency inputs for as long as you keep sending these frequencies to the mixer.

If you want to prove that 'output' is not a function of 'input' then I think you are being logically inconsistent and not me.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 05, 2015, 02:18:06 PM
Let me see if I can take you through the thought experiment.

First of all, we're NOT talking about mixers. We are talking about OOK transmission. Specifically, a CW transmitter.

We will key this transmitter on for 1/2 second and key it off for 1/2 second. In other words, 1 Hz.

We have a magical receiver that we bought at the Mordor flea market from Gandalf the wizard. It can detect the spectral tones that you claim exist when the CW transmitter OFF. It has a speaker like any radio, but it also has an indicator that tells us if spectral tones are present.

From the speaker of our special receiver, we hear 1/2 second of audio and then 1/2 second of silence, just like in the real world.

From the spectral tone indicator, we see that it's indicating that tones are present continuously.

During the 1/2 second when the CW transmitter is OFF, we can have two cases.

1) The spectral tone indicator is indicating that spectral tones are present. 1/2 second is a long time. A normal human being can easily perceive that there is silence from the speaker and that the spectral tone indicator is on. From this information, we know that another 1/2 second dit is coming along. We can predict the future.

2) The spectral tone indicator is indicating that spectral tones are NOT present. Same case as above. We can predict that there will be no following dit.

It's logically inconsistent. We cannot predict the future.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 05, 2015, 02:40:55 PM
Quote
First of all, we're NOT talking about mixers. We are talking about OOK transmission. Specifically, a CW transmitter.

But we are talking about a mixer. That's why this debate is never going to end. A series switch that repeatedly gates on and off a carrier is a simple form of an unbalanced switching mixer.

There will be pictures and equations in theory books that show this including a series make/break switch that behaves like a cw key.

They will show the equations for the switch based mixer. That's ALL I'm trying to provide here. This isn't my theory. I'm not trying to be clever or novel.

The equations for the mixer don't include risetime or falltime. But they still accurately predict the output spectrum and the conversion efficiency of the mixer in terms of the frequency and power levels of all the mixer sidebands.

Your argument revolves around human perception. I think G3TXQ made the same observation about your arguments. I don't know how to debate human perception because it's not an exact science. I prefer to look at the design equations. I'm just an EE not an expert in human science :)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 05, 2015, 02:47:01 PM
So you're saying that you can build a receiver in the real world that can detect spectral tones between dits? Yes or no.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 05, 2015, 03:15:28 PM
Quote
So you're saying that you can build a receiver in the real world that can detect spectral tones between dits? Yes or no.
No, I'm not saying that. I'm just presenting you with basic mixer theory that's all. Trying to prove that mixer theory is wrong just because I can't build you a receiver isn't very scientific.

I'll see if I can surf for some established textbooks with equations for the series switch type mixer. This type of mixer isn't a classic but its output can be represented by a series of tones at various frequencies. It's up to you if you consider that these tones don't exist in terms of human perception time.

As a start you could try looking at page 7 of this pdf file where it covers a simple series switch based mixer

http://www.ece.ucsb.edu/Faculty/rodwell/Classes/ece218b/notes/Mixer1.pdf

There is an equation for the output and it is a series of continuous tones.

Does this help?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 05, 2015, 07:36:06 PM
Does this help?

It does. In fact, I believe I can show the solution that satisfies both camps.

I went ahead and simulated the switching mixer in GNU Radio. The first plot is a 100 kHz sine wave switched with a 50 kHz square wave. Be aware that the frequency plot is not implemented properly. Since I simulated with real numbers, the negative frequency side of the plot should not be shown. So you just have to ignore it.

Anyway, we get the expected spectrum. The RF feedthrough at 100 kHz, the 2nd order products at 50 and 150 kHz and higher order products further out.

(http://www.w6rz.net/mixer5k10k.png)

Then I slowed the switching rate down to 800 Hz. Back to a single frequency! Note that the DFT length is 1024 points and the sample rate is 1 Msps. So the DFT duration is 1024 microseconds, which is just encompassing the stable amplitude sine wave portion of the waveform.

(http://www.w6rz.net/mixer800.png)

The key is to look at the time domain waveforms. At the 50 kHz switching rate, the time domain waveform is continuous. At the 800 Hz switching rate, the time domain waveform is not continuous (it has zero amplitude for long periods).

So the solution that satisfies both camps is simple. Continuous time domain waveforms create continuous spectra. Non-continuous time domain waveforms create non-continuous spectra.

Is everyone happy?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 05, 2015, 07:39:40 PM
Here's the GNU Radio simulation.

(http://www.w6rz.net/gnumixer.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: N3QE on September 06, 2015, 02:59:00 AM
So you're saying that you can build a receiver in the real world that can detect spectral tones between dits? Yes or no.

Yes!!!

Have you ever used a super-narrow filter with poor phase response? They're still ringing in between dits at the exact same frequency of the CW tone!!


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 06, 2015, 05:45:36 AM
I've received several eMails encouraging me to contribute again, so I'll give it another try!

Let's begin with an example that is simple to analyse and where we might find some common ground - a carrier AM modulated by a sine wave. You'll find an equation for that signal in almost any text book on modulation; here it is:

(http://www.karinya.net/g3txq/temp/sidebands/am2.png)

That equation tells us that the carrier and the two sidebands are all constant amplitude sinusoids.

The same textbook I took that from has a helpful diagram illustrating the carrier, sidebands, and the composite AM signal:

(http://www.karinya.net/g3txq/temp/sidebands/am1.png)

Notice again that the carrier and the two sidebands are constant amplitude sine waves even though the composite AM signal appears to be a carrier which varies in amplitude. The explanation is that the sidebands and the carrier are on 3 slightly different frequencies: at times they all drift into phase, giving us the peak in the AM signal envelope; at other times they drift out of phase to give us the troughs.

Nothing in the theory constrains the frequency fm of the modulating signal. It could be as low as 0.1Hz and the waveforms would still be valid; even though the envelope would be high for 5 seconds and then low for 5 seconds, the carrier and the sidebands would still be constant amplitude throughout. All that has changed is that the sidebands have moved closer to the carrier - in fact to within 0.1Hz of it.

There is absolutely nothing novel here - it's standard modulation theory which I hope we would all agree on. And yet even this simple example contains the apparent paradoxes that seemed to worry folk when talking about the OOK example:

1) How can the carrier be constant at all times, when it apparently goes to zero at some points in the AM wave?

2) There appears to be much more energy transmitted around the high amplitude peak of the envelope compare to the trough; so how can the energy in the carrier and sidebands be constant with time?

Once folk can understand these simple principles they are well on their way to understanding the OOK example, which is simply an extension of the same process but with a different modulating signal containing more sidebands.

[Equations and diagram taken from "Amplitude Modulation Fundamentals http://www.pa2old.nl/files/am_fundamentals.pdf]

Steve G3TXQ



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 06, 2015, 08:00:33 AM
Part II

Now let's extend these basic principles to the case of On-Off keying. From that same text book:

Quote
When complex signals such as pulses or rectangular waves modulate a carrier, a broad
spectrum of sidebands are produced. According to Fourier theory, complex signals such
as square waves, triangular waves, sawtooth waves, and distorted sine waves are simply
made up of a fundamental sine wave and numerous harmonic signals at different amplitudes
.

So let's take a more complex modulating signal; this one is made up of a constant amplitude carrier and 4 constant amplitude sidebands: fc, fc+fm, fc-fm, fc+3fm, and fc-3fm

(http://www.karinya.net/g3txq/temp/sidebands/am5.png)

Notice how - because of their complex and changing phase relationships - the 5 individual, constant-amplitude sinusoids have "interfered" constructively during the first half of the modulating waveform, and destructively during the second half.

Now take an even more complex modulating signal; this one is made up of a constant carrier and 9 constant amplitude sidebands each side of the carrier:

(http://www.karinya.net/g3txq/temp/sidebands/am4.png)

Amazing, isn't it - the simple addition of a constant-amplitude carrier and 18 constant-amplitude sidebands has produced a signal with apparently no carrier for half of the period. My goodness, it almost looks like On-Off keying  :)

Please understand this: the carrier and the sidebands exist at constant amplitude throughout the period of that graph, not just during the first half or just at switching times. If we examine the carrier or any sideband in isolation, we see a constant amplitude sinusoid. It's only when we view the carrier plus one or more sidebands that the envelope begins to vary.

Note that the On period of the modulated signal can be as short as you like - all those same principles apply. The only constraint is that the signal be periodic (i.e. it repeats regularly like continuous Morse Dots or continuous Morse Dashes) - a point I made about 1,000,000 postings ago  :)

Note also that everything W6RZ has shown is consistent with these basic principles - you just need to understand what you are looking at; in particular you need to understand whether the discrimination bandwidth is allowing you to see just one spectral component or the complex envelope formed from several.

Finally, notice that the analysis involves just simple mathematics - there's no need to conjure up thoughts of filters ringing, or artefacts resulting from DFT etc. The simple equations tell us that the On-Off keying comprises a continuous carrier plus a lot of continuous sidebands.

Hope that helps.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 06, 2015, 09:42:05 AM
Well folks, that's all I got. I tried to show things as they are. My results don't have huge matzo ball logical inconsistencies, like power getting through the switching mixer when the switch is open for long periods of time. So all the folks reading this thread have to decide for themselves which camp is correct.

If you were a fan of this thread (and it would seem with so many views there were at least some), chime in with your opinion. The choices are:

1) Sidebands do NOT exist between dits. They only exist during the time when the amplitude is changing at the beginning and end of a dit.

2) Sidebands exist between dits, no matter how long the interval between dits is.

Of course, my choice is number 1. So the current tally is:

Sidebands do NOT exist between dits. = 1

Sidebands do exist between dits. = 0


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on September 06, 2015, 10:22:53 AM
Where I feel Steve's analogy fails is that when applied to a long dash. During the dash, there is no modulation so there are no sidebands. Now you can extend that to dots IF the period of the dot exceeds the time that it is being modulated i.e. its amplitude is changing.

If rise time plus fall time are less than the dot length, you have a partial response system and there will be sidebands all the time.

Otherwise you must have the ridiculous situation that no matter how long the dash is, the clicks are there all the time....


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 06, 2015, 10:25:08 AM
Sidebands do NOT exist between dits. = 2

Sidebands do exist between dits. = 0


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 06, 2015, 10:49:25 AM
Where I feel Steve's analogy fails is that when applied to a long dash. During the dash, there is no modulation so there are no sidebands.

Peter,

Let's go back to the simple sine-wave modulation case where we may have common ground.

Q1. Do you agree that the carrier and sidebands are continuous as shown in the diagram from the text book?

If your answer is "no", there is no point continuing because that is just elementary modulation theory. If your answer is "yes":

Q2.  Do you think the carrier and sidebands are still continuous if the modulating frequency drops to 0.1Hz?

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 06, 2015, 11:36:18 AM
Quote
Well folks, that's all I got. I tried to show things as they are. My results don't have huge matzo ball logical inconsistencies, like power getting through the switching mixer when the switch is open for long periods of time.

It isn't a case of 'power getting through an open switch"

It's a case of analysing the spectrum of signals that EXIT the mixer. OUTPUT is a function of INPUT.

The output consists of a multiplication of the carrier INPUT and the INPUT of the keying LO at 100Hz and all of its odd harmonics. This produces a massively wide spectrum of signals.

If you had a 1W carrier and the pulsed it (or OOK'ed it) at say 100Hz then you have only 0.5W average power EXITING the mixer because the switch is open half of the time.

Of this 0.5W, about half is in the carrier and about half is in the sidebands. Switching mixer theory dating back many decades proves this.

If you think this is all wrong then PLEASE post up why it is wrong because the same paper I posted a link to is also published by Agilent below. Do you think Professor Long is a mad professor or something?

http://cp.literature.agilent.com/litweb/pdf/5989-9102EN.pdf



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 06, 2015, 11:48:24 AM
Mixer:  Nothing in equals Nothing out.  No rise/fall time, no carrier, no side bands. 


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 06, 2015, 11:57:59 AM
Sidebands do NOT exist between dits. = 3
W6RZ, G3RZP, K1ZJH

Sidebands do exist between dits. = 2
G3TXQ, G0HZU


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 06, 2015, 12:00:17 PM
I realise that I'm just wasting my time because W6RZ seems to think he can rewrite mixer theory 'by vote' on a ham forum. The scary thing is I think this guy actually believes his voting results will prove something.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 06, 2015, 12:01:32 PM
If you think this is all wrong then PLEASE post up why it is wrong because the same paper I posted a link to is also published by Agilent below.
No more arguments from me. Protons will decay before you and G3TXQ will change your position. What's the point of continuing?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 06, 2015, 12:08:00 PM
I realise that I'm just wasting my time because W6RZ seems to think he can rewrite mixer theory 'by ballot' on a ham forum. The scary thing is I think this guy actually believes his ballot proves something.
I'm not trying to use the ballot to prove things. I'm interested in the social aspects.

1) How many people even care?

2) How many people are biased by their perception of the participants?

3) Assuming people even care, are they willing to expose their opinion?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 06, 2015, 12:08:18 PM
As we can't agree, why don't we just do the experiment and measure it:

(http://www.karinya.net/g3txq/temp/sqwave8_5khz.jpg)

Top trace is the On/Off keyed signal; bottom trace is the filtered carrier. Does anyone see any carrier between the Dits?

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 06, 2015, 12:41:10 PM
Quote
No more arguments from me. Protons will decay before you and G3TXQ will change your position. What's the point of continuing?

I can set you another conundrum if you still don't believe the mixer theory :)

If I send OOK dits at, say 10kHz rate on a 7MHz carrier from my Agilent sig gen and look at this on a scope I see the carrier on for half of the time.

If the average power on a thermocouple power meter reads 1mW with the modulation ON then according to your theory I should see almost the full 1mW on a spectrum analyser at the carrier frequency if the rise/fall times are really, really fast.

If the rise/fall times are maybe 100ns then there is almost no time for any power in the sidebands. The rise/fall time is an insignificant time contained to the ON time of the carrier.

So why does the analyser only show about 0.5mW average power at the carrier frequency? where has the other 0.5mW gone? Why does it show about 0.2mW at each of the first two sidebands if I tune away from the carrier by 10kHz? So that totals 0.4mW in average sideband power already.

So you need to tell me where half of the power you 'think' you see on the scope at the carrier frequency has 'gone' and how are you able to hide about 0.5mW of sideband power in the tiny timeslot of the rise and fall time?

This is a fundamental question that challenges your position very strongly.



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 06, 2015, 01:14:14 PM
As I said, no more arguments from me. I did the best I could with the tools I had. At this point, I'm hoping that maybe one or two people thought the GNU Radio stuff was cool and might become interested enough to experiment with SDR technology.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 06, 2015, 01:46:42 PM
Show me a time domain display of the keyed waveform, show all of us where RF exists in the time domain once the unkeyed CW carrier reaches zero power.  We are not talking high level AM modulation where the carrier "appears to go away."  In that instance the PA stage is being excited. Big difference when you remove the drive energy for CW.  The only thing I can think of is that you are visualizing the keying square wave as being a modulating component and thus appears at the output of singly balanced mixer. That is correct. But, again, the clicks in the RF spectrum are gone once the PA is unkeyed.  The RF input from the driver is also gone.  That leaves the modulating waveform, which is probably sub-audible, and won't pass through the HF tank regardless. And it certainly can't cause RF interference at that point.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 06, 2015, 01:50:34 PM
Show me a time domain display of the keyed waveform, show all of us where RF exists in the time domain once the unkeyed CW carrier reaches zero power.

(http://www.karinya.net/g3txq/temp/sqwave8_5khz.jpg)

Top trace is the keyed carrier; bottom trace is the same signal filtered to extract the carrier component.

Steve G3TXQ





Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 06, 2015, 02:03:00 PM
There in lies the BS.

The lower scope trace? Where does that come from???  The exciter and PA are BOTH keyed. The upper waveform is probably the best representation of the RF envelope going into the transmitter PA.  There is NO RF drive when the rig is unkeyed. Mixers need two inputs.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 06, 2015, 02:06:04 PM
The upper trace is the signal from a 1.4MHz signal generator passed through a diode switch to turn it on and off. The lower trace is that keyed signal passed through a 1.4MHz filter to extract the carrier component.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 06, 2015, 02:58:43 PM
It's getting late here, but for any readers out there who still have open minds on this issue here is the time domain expression for an On-Off keyed signal:

fo(t) = Ac.sin(ωct) x [0.5 + 2/π [sin(ωst) + 1/3 sin(3ωst) + 1/5 sin(5ωst) + ......  etc  ........]]

where Ac is the amplitude of the carrier, ωc is its angular frequency, and ωs is the angular frequency at which it is being keyed.

Ac.sin(ωct) is the carrier, which is being gated on and off by the stuff in the square brackets which toggles between +1 and zero. Check it in any modulation theory text book.

It's an infinite series because a perfectly keyed signal has sidebands extending to infinity. But we only need to look at the first term in the expansion, 0.5.Ac.sin(ωct), to see that there is a steady carrier at all times.

The maths simply confirms what I showed in the experiment.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 06, 2015, 03:13:28 PM
This just isn't going to end...

We can all agree that the composite waveform of all of the 'stuff' from the mixer output shows a time when there appears to be s flat line on the scope and a time where there just appears to be a signal waveform at the carrier frequency.

However, we know there are sidebands with significant power because we can see them on a spectrum analyser. So the signal has bandwidth.

Switching mixer theory tells us that there is a wide spectrum of signals leaving the mixer and the average power in this composite signal is shared between the carrier and the sidebands at about 50:50. About half of the power is in the sidebands and half is in the carrier.

What we see on the scope is the result of all of these tones summing and/or cancelling over time on the scope. This results in the OOK waveform with a flat time and a busy time on the scope.

If the average power of the OOK signal is 0dBm or 1mW then about 500uW of this is in the carrier and about 500uW of this is in the sidebands.

Can't we just leave it at that?






Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 06, 2015, 03:50:09 PM
If anyone is interested I just did the analyser and scope test at 7MHz with 2kHz OOK.

The scope is set to 0.2V/div and it has a 50R feedthru termination on its input. Note that the analyser is set to 5dB/div and not the usual 10dB/div.

The (recalibration due in 2016) Agilent sig gen is set to a +3dBm or 2mW carrier (unmodulated) and then set to gate this on and off at 2kHz.

The scope shows the modulation ON and also modulation OFF states of the sig gen and so does the spectrum analyser. On the scope the peak voltage is about 0.45V so the PEP is about (0.45*0.45)/100 = 2mW which agrees with the sig gen setting of +3dBm or 2mW.

Mixer theory tells us that the power on the analyser at the carrier frequency will be 6dB down on the unmodulated carrier at +3dBm on the analyser.

This is what we see on the analyser with the mod on. We see -3dBm which is 500uW. It also tells us that the first sidebands will each be just under 4dB below this carrier at about -7dBm or 200uW each. So 400uW total for the pair of sidebands.

Again, this is what we see on the analyser. If you added up all the other sideband powers you would end up with about half of the 1mW average power in the sidebands and half in the carrier.

The composite OOK waveform we see on the scope is made up of all of these tones beating together over time.

(http://i1227.photobucket.com/albums/ee439/G0HZU/OOK7_SG.jpg) (http://s1227.photobucket.com/user/G0HZU/media/OOK7_SG.jpg.html)


(http://i1227.photobucket.com/albums/ee439/G0HZU/OOK7_8566_OFF.jpg) (http://s1227.photobucket.com/user/G0HZU/media/OOK7_8566_OFF.jpg.html)

(http://i1227.photobucket.com/albums/ee439/G0HZU/OOK7_8566_ON.jpg) (http://s1227.photobucket.com/user/G0HZU/media/OOK7_8566_ON.jpg.html)


(http://i1227.photobucket.com/albums/ee439/G0HZU/OOK7_8566_ONSB.jpg) (http://s1227.photobucket.com/user/G0HZU/media/OOK7_8566_ONSB.jpg.html)


(http://i1227.photobucket.com/albums/ee439/G0HZU/OOK7_scopeOFF.jpg) (http://s1227.photobucket.com/user/G0HZU/media/OOK7_scopeOFF.jpg.html)

(http://i1227.photobucket.com/albums/ee439/G0HZU/OOK7_scope_ON.jpg) (http://s1227.photobucket.com/user/G0HZU/media/OOK7_scope_ON.jpg.html)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 06, 2015, 04:20:16 PM
Note that if anyone claims that the analyser is being 'spoofed' to show the wrong power levels because it has an envelope detector that responds to peaks  (rather than displaying as a true average reading power) then I can remind everyone that I already did this this test way back in post#190.

I placed an average reading thermocouple power meter at the analyser IF output with the analyser set to linear mode and did a similar test using zero span to stare at the average power of each spectral component in the OOK signal. Although I did this test at 70MHz and with a different OOK rate and at a lower test power level, the results were the same in terms of relative power levels for each spectral component in the OOK signal.  :)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 06, 2015, 04:41:45 PM


(http://i1227.photobucket.com/albums/ee439/G0HZU/QQQ.jpg) (http://s1227.photobucket.com/user/G0HZU/media/QQQ.jpg.html)



So I guess the conundrum is... where is the 500uW (average) heating power of all of these mixer sidebands hiding in the time domain?

According to W6RZ etc this can't be in the blue regions on the scope because there's either 'nothing' or 'carrier' apart from the tiny, tiny timeslice of the rise/falltime.

This tiny timeslice must be about 1% of the duty time.

So why isn't nearly ALL of the average power just at the carrier frequency? Why is only half of it at the carrier frequency?

I think the answer is that the OOK waveform consists of a wide spectrum of tones that beat together over time to produce the OOK waveform on the scope.

Just like unbalanced switching mixer theory (that has been proven for decades) tells us :)



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: AC7ZN on September 06, 2015, 06:53:41 PM
It is dangerous for a newcomer to the conversation to pop in here as some opinions have already petrified, so things probably can't be resolved easily.  Additionally, I consider the participants to this conversation to be very competant, who I have learned much from, and appreciate their contributions.  I use a homebrew hexbeam at my home and Steve's work is exemplary.

Isn't it interesting that even something as simple as OOK and AM modulation seems to generate these dilemmas?  And it is not just here...the pages of the ARRL handbook's explanation of AM states that carrier amplitude is unchanged with AM modulation, then states that CW (OOK) is a form of AM. Just a little thought will tell you that CW does NOT have a constant carrier amplitude.  If our cat falls asleep on the key we have a constant amplitude output, but when it gets up we have nothing.  Our common sense tells us a string of dits from our paddle will give an average carrier of about 50% full strength since the duty cycle of this signal is 50%.  And indeed, those who disagree with this are saddled with how to explain how our carrier maintains full amplitude when our power supply consumption has dropped in half... the extra sideband energy from the keying only subtracts energy from this half amplitude carrier...it does not add anything, so our actual carrier energy is less than half.

But the math doesn't lie either.  A sinewave modulated carrier is a full strength carrier..the math tells us so, and so does the ARRL handbook.  The handbook also says AM modulation is not as simple as just modulating the amplitude of a carrier...yet anyone who can read a schematic knows AM modulation does EXACTLY that. No more, no less.

How do we resolve this apparent paradox?  The trick is in understanding how we calculate and measure the spectrum of our signals. When we see the equation showing the result of sinewave modulation, we see three sinewaves...an upper sideband sinewave, a lower and the carrier...and the carrier is indeed full amplitude.  The handbook states 'we have proven AM modulation does not change the carrier amplitude'.  Unfortunately, the proof given is only a for very special case: sinewave modulation that has, and will exist for all time.  As soon as we interrupt the sinewave modulation the proof no longer holds.

A constant amplitude sinewave gives a spectrum of a single, constant amplitude frequency (it is actually positive and negative frequency but we can ignore the negative frequency stuff here).  But the assumption is that the sinewave is periodic...that is, its maximum/minimum amplitude never varies over all time.  When we make this assumption we are able to use the fairly simple Fourier transfer to calculate the spectrum of this signal.  The math in the Fourier transfer assumes a periodic signal for all time, and for many things it gives a useful result because things change slowly enough that we can assume the signals are periodic.  Our spectrum analyzers also do not respond to instant spectral changes (because of the energy storage in the filters) but rather averages the spectral energy over time (the time is dependent on our resolution bandwidth setting).

Now, if our AM modulation is very slow (I think Steve had an example of 5 second period) and we look at it on our spectrum analyzer, we find the carrier shown by the analyzer is not constant at all...it goes up and down at at 5 second rate, just as our common sense dictates.  But if we were to set our resolution bandwidth to something really low (say, a half a day...most analyzers can't do this, but I think some some software-based ones could...) we would have to wait a day or so for the result but we would find that carrier is shown as a constant full amplitude carrier. 

The point is that we must take the time and periodicity of modulation into account both when calculating and when measuring a spectrum.  If the modulating signal is not periodic (and Morse usually isn't), the Fourier transfer, the sinewave equations, and standard spectrum analyzer no longer give an accurate result of what is happening. We must resort of more powerful, complicated tools such as the Fourier integral (The Fourier integral is the non-periodic equivalent of the Fourier transfer, and is usually more difficult to calculate) when calculating, and advanced spectrum analysis techniques that can accurately process signals of this nature.

When we do this, the Fourier integral will tell us that we have a nearly full carrier single frequency when our key is down, and nearly nothing when it is up.  And an appropriately set up spectrum analyzer will tell us the same thing.  But the interesting part, as you might expect , comes when we apply the Fourier integral to the transitions between the mark and space.  Here is where the clicks come (and something as simple as a receiver tells us this just fine).  The faster the transition, the wider the spectrum.  Our simple receiver tells us so, and it is correct.

So a string of dits is periodic and the spectrum will eventually congeal to what the Fourier transfer tells us it will be as long as we examine it for a long enough time.  It will be, by the way, about a half amplitude carrier...for the carrier to be full strength the modulating signal must have no DC component (like a sinewave).  OOK does not follow this rule.

Does this help anyone understand what is going on?  The problem seems related to the fact that Morse code has really slow modulation going on (dits and dahs) combined with something really fast (the keying edges).  The Fourier integral can sort this all out, but our receivers do a pretty good job of it already.

73,
Glenn AC7ZN
   
 

 


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KB8E on September 06, 2015, 09:04:46 PM
I've been reading this thread over the past week or so and at the risk of being flamed or drawn into a long drawn-out and potentially personal discussion, I'm going to add my two-cents worth. I think we should remember the situation being discussed, at least as I see it. An RF carrier is modulated with a keying waveform (almost square to generate key clicks) such that it becomes a CW signal. Depending upon the rise and fall times of the keying waveform, the resulting sideband structure can vary from one which has a sin x/x distribution when keyed with zero rise and fall times to one that may only have one pair of sideband in the case where the keying waveform is filtered so much that it becomes a sinewave. What is of interest is what a receiver tuned to one side of this signal would hear. If said receiver has an IF bandwidth so narrow that it only passes one of the sidebands, it would sound like a continuous tone, continuous in this case means a long time compared to the keying rate. If the receiver were tuned between the sidebands, nothing would be heard. Note that at normal keying rates this bandwidth would be quite narrow. Now consider the more common situation where we have a bandwidth wide enough that allows a group of sidebands on one side of the carrier to be received. A way to think of this is to think of the lowpass equivalent signal i.e. just a square-wave where the sidebands are at odd harmonics of the keying frequency. If we were to selectively pick out say the fifth and seventh sidebands and add them up, we would get a somewhat complex waveform, but one that would show peaks in time where the peaks of each of two sinusoids happened to line up. This would occur at the largest common denominator of the two frequencies i.e. the keying rate. As more sidebands are added up (wider receiver bandwidth), the peak would get more distinct and larger, but always at the keying rate. I postulate that this is what is happening when one has their receiver tuned to the side of a CW signal with fast rise and fall times i.e. one that appears to be modulated by a square-wave. We are picking up and summing a group of sidebands that are no longer harmonically related, but whose resulting signal has a pulse-like characteristic with a pulse repetition rate equal to the keying rate. These “key clicks” do appear to exist only during the keying transitions since only a portion of the transmitted sideband structure is being intercepted.

Sam


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 07, 2015, 03:17:43 AM
Sam & Glenn,

Thank you for your helpful summaries.

I took a few more photos this morning to illustrate some of the points you both and I have been making. In each case the top 'scope trace is the On-Off keyed waveform and the bottom trace is the keyed signal passed through a narrow-band filter. By altering the carrier frequency and the keying rate it's easy to observe one or more of the spectral components through the filter.

Here's what we get just looking at the carrier component. It's continuous as predicted by the maths:
(http://www.karinya.net/g3txq/temp/sidebands/sbc.jpg)

Now let's tune off to one side of the carrier and select a first-order sideband. It too is continuous, but of lower amplitude, again as predicted by the maths:
(http://www.karinya.net/g3txq/temp/sidebands/sb1.jpg)

Now let's tune off further and select a second-order sideband. Nothing there - as predicted by the maths:
(http://www.karinya.net/g3txq/temp/sidebands/sb2.jpg)

And tuning off even further to a third-order sideband - again continuous, but of even lower amplitude.
(http://www.karinya.net/g3txq/temp/sidebands/sb3.jpg)

Now let's tune to one side of the carrier with a bandwidth which encompasses many sideband components - in this case about 50 of them:

(http://www.karinya.net/g3txq/temp/sidebands/sbkc.jpg)

The envelope of all those sidebands added together has become impulsive; they all add in phase at times coincident with the keyed signal transitions, but for much of the time they cancel. These are the "key clicks" that we are familiar with. Note that the individual sidebands still exist throughout the period - not just at the transitions or during the On period - but because we are not separately discriminating them we just see the composite envelope.

This is the waveform we are most familiar with, because at typical keying rates we would need sub-Hz filter bandwidths to resolve the individual components; but that doesn't mean they don't exist!

The apparent paradox about how energy can be present continuously in the individual spectral components, but only appear for brief periods in the impulsive "key-click" waveform is an interesting one; the solution lies in an understanding of the redistribution of energy that occurs - spatial or temporal - when waves interfere. That's a discussion best left for another day, but "Google" 'Where does energy go in destructive interference' if you want to know more.

I would summarise things like this:

1) Individual spectral components, both the carrier and the sidebands, exist continuously - not just around the transition times or just during the On period.

2) At typical keying speeds, and with practical receiver bandwidths, we never hear individual spectral components but an aggregation of many of them. The summation of those many components is a signal that is impulsive, with peaks coincident with the OOK transitions.

Finally, we should always be very wary about looking at a time-domain representation and thinking we can subjectively deduce the equivalent frequency-domain representation. For example, if we look at an AM signal modulated with a low-frequency sine-wave we will swear the carrier is varying; but we know that in the frequency domain the carrier spectral component is not varying. Or look at the time-domain representation of a DSB signal - it's "full" of carrier signal; but we know that in the frequency domain the carrier spectral component doesn't exist. Equally, looking at an OOK signal you might be fooled into thing the carrier spectral component only existed during the On period; but as we have demonstrated and as the maths tells us, that component is continuous.

Steve G3TXQ



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on September 07, 2015, 04:47:17 AM
Steve,

It seems to me that you are intimating that a carrier turned on produces a click and sidebands but left on to infinity, the sidebands are still there.....

And when turned off, the click sidebands are also still there and of such phase and amplitudes as to cancel the carrier, no matter how long the carrier is turned off for....


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 07, 2015, 06:25:03 AM
Steve,

It seems to me that you are intimating that a carrier turned on produces a click and sidebands but left on to infinity, the sidebands are still there.....

And when turned off, the click sidebands are also still there and of such phase and amplitudes as to cancel the carrier, no matter how long the carrier is turned off for....

No - you are misinterpreting me!

I've stressed in nearly every posting I've made that I'm talking about periodic waveforms - that is waveforms where f(t+nT)=f(t) for any integer n. I've stressed that because it makes the maths easier and lends itself to simpler explanations and demonstrations - if folk can't understand the periodic case they are unlikely to understand the aperiodic case where the Fourier analysis becomes more difficult.

Your example of a carrier turned on and left on to infinity is not periodic.

Do you now accept the maths and the experimental data which shows that a periodic signal, for example a single 12wpm Dot repeated every 1 second, must have a continuous carrier spectral component and continuous sideband components spaced every 1Hz? If not, it would be helpful if your response contained objective, rather than subjective, reasoning.


Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on September 07, 2015, 09:10:53 AM
Periodic or not, the modulation waveform of an OOK keyed Morse message of some length can be approximated with sufficient accuracy for calculations of the necessary and occupied bandwidths by "1:1 reversals" or a square wave having a frequency = baudrate/2.

And this is exactly what the international radio community,embodied in the ITU and CCIR has done.

Since the very beginning of "scientific spectrum management" in the 40's, spectrum requirements and channel spacings
for aural Morse radiotelegraphy circuits have been dimensioned using the classical formula in the Radio Regulations;

"Signal with Quantized or Digital Information  
Continuous wave telegraphy,
Morse code  
Bn = BK
Bn=Necessary bandwidth in Hertz
B= Modulating rate in Bauds
K = 3 for stable Circuits"
K = 5 for fading Circuits"

Note that there is no mentioning of any rise or fall times, nor any "key clicks".

The whole "key-click problem" has been taken care of by the general clause that requires emissions to occupy no more bandwidth than necessary, and with proper erfc(t) shaping and a 10% rise/fall time of the shortest keying element, the occupied and the necsssary bandwidths are close to each other.

The higher-order sidebands that can generate "clicks" in the adjacent channels rapidly fall to insignificant levels outside the 3 and 5 order sidebands required to recreate the shape of the keying elements at the receiving end when the proper shaping is applied.

"Key clicks" are seen from a regulatory point of view as emission components outside the necessary bandwidth, capable of creating harmful interference to other regulated spectrum users. The general provisions against interference prohibit operation of transmitters which create harmful interference outside their necessary bandwidths:

"15.10    § 6 The out-of-band emissions of transmitting stations should not cause harmful interference to services which operate in adjacent bands in accordance with these Regulations and which use receivers in conformity with Nos. 3.3, 3.11, 3.12, 3.13  and relevant ITU-R Recommendations."


73/
Karl-Arne
SM0AOM


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 07, 2015, 10:00:12 AM
In case anyone is wondering... OOK is On-Off-Keying.

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on September 07, 2015, 10:11:55 AM
OK Steve,

Reductio ad absurdam.

The carrier is turned on for 24 hours and turned off for 24 hours and turned on for 24 hours etc. This is a periodic waveform meeting your criteria. I can even switch the mains off the tx for the 24 hour OFF period, and it is a set of reversals meeting Karl-Arne's criteria.

I could meet the same criteria with a tx repeatedly turned on for 1 month and off for 1 month...

For the argument to hold water, it must be valid for any value of t....


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 07, 2015, 11:29:07 AM
OK Steve,

Reductio ad absurdam.

The carrier is turned on for 24 hours and turned off for 24 hours and turned on for 24 hours etc. This is a periodic waveform meeting your criteria. I can even switch the mains off the tx for the 24 hour OFF period, and it is a set of reversals meeting Karl-Arne's criteria.

I could meet the same criteria with a tx repeatedly turned on for 1 month and off for 1 month...

For the argument to hold water, it must be valid for any value of t....


Here's the time domain formula again:
fo(t) = Ac.sin(ωct) x [0.5 + 2/π [sin(ωst) + 1/3 sin(3ωst) + 1/5 sin(5ωst) + ......  etc  ........]]

The maths doesn't stop being true just because you lower ωs to some extreme value - why would it?

If you key regularly in the way you describe, there will still be a continuous carrier spectral component and continuous sidebands. They will be spaced by 0.000005787 Hz in the 24 hour case and 0.000000187 Hz in the one month case - I've assumed July/August, but the result will be slightly different for other months :)

Of course, to detect them you will need a receiver with a bandwidth of 0.000005787 Hz or 0.000000187 Hz, and your observation time will need to be 48 hours or 2 months. If you fail to meet those requirements you will simply be looking at an aggregate of all the sidebands and you'll see a simple On-Off keyed envelope.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 07, 2015, 11:39:29 AM
OK Steve,

Reductio ad absurdam.

The carrier is turned on for 24 hours and turned off for 24 hours and turned on for 24 hours etc. This is a periodic waveform meeting your criteria. I can even switch the mains off the tx for the 24 hour OFF period, and it is a set of reversals meeting Karl-Arne's criteria.

I could meet the same criteria with a tx repeatedly turned on for 1 month and off for 1 month...

For the argument to hold water, it must be valid for any value of t....

A mixer is effectively a multiplier. You can look up the trig identities for multiplying sinewaves together on a school/college website.

Your argument amounts to:

"You can't multiply sinewaves that have a period longer than the time of the universe so the trig identities for multiplication aren't valid for ANY sinewaves"

A realistic (as in real world useful) time for sending looooong dits would be maybe 1 minute per dit.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on September 07, 2015, 12:17:43 PM
So Steve, a transmitter disconnected from the mains is still providing power hours later? That's what your mathematics assumes which, to be honest, is patently absurd.

My feeling is that your mathematical model is not applicable.

So I asked the XYL. She tells me that it is a analogous case to the calculations involved in PCB simulations for digital systems: regrettably, her full notes are at work and she won't be in there until next month. She is an Hons. BSc  in Electrical and Electronic Engineering, a CEng, MIET and a Senior Training Instructor for a well known CAD company as well as being G4FNC, so we must assume a certain degree of competence. The difficulty apparently lies in assuming that the period between the start and stop of the pulse is subject to change - which it isn't.

But please explain how this transmitter which is switched off and disconnected from the mains is still producing a carrier and sidebands? Which is the logical result of your argument.

Now if the argument is that the equations are bounded by t, justify why the bounds exist for  a long t  but not for a short t.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 07, 2015, 12:34:08 PM
Quote
So Steve, a transmitter disconnected from the mains is still providing power hours later? That's what your mathematics assumes which, to be honest, is patently absurd.

It's a case of measuring AVERAGE power over the duration of the sending session for each spectral component. For very slow OOK you would have to average the power over a long time.

You might as well argue that a 0.01Hz sinewave with average power 1W is NOT a 1W signal if you only sample it for a few nanoseconds  with a single sweep scope when the waveform is near its zero crossing point. What would this prove?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: DL8OV on September 07, 2015, 01:06:29 PM
Peter, G3RZP, I agree with your latest messages and the case where the transmitter is turned off provides an interesting mathematical challenge to the argument (a challenge far beyond my skill in mathematics). However, I also recognize that on-off keying is a way of providing modulation to an RF envelope even if this modulation is at a VERY slow rate and thus sidebands must be produced. I will therefore sit back and continue to enjoy the bun fight which at the time of writing is at 22 pages and counting.

The buzzing we can hear in the background is that of my maths teacher gently spinning in his grave. Somebody has made a fundamental error and I am not sure who.

Peter DL8OV


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 07, 2015, 01:11:21 PM
The side bands are manifested as key clicks.  Beyond that, I am also mystified as to the some of claims presented in this thread.

As W8JI said, the key clicks are ONLY EVIDENT during the transitional high to low portions of the CW waveform.  This is a fact, and is obvious to anyone who has operated on that mode.  Beyond that, the key clicks only bother other stations when the transmitting waveform is in a transitional stage. Period.

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 07, 2015, 01:19:58 PM
So Steve, a transmitter disconnected from the mains is still providing power hours later? That's what your mathematics assumes which, to be honest, is patently absurd.

Of course the transmitter can't provide power when it's off! You are trying to apply a constraint in one domain which doesn't exist in the other.

My feeling is that your mathematical model is not applicable.

Unfortunately it's not my model - it's accepted modulation theory

So I asked the XYL. She tells me that it is a analogous case to the calculations involved in PCB simulations for digital systems: regrettably, her full notes are at work and she won't be in there until next month. She is an Hons. BSc  in Electrical and Electronic Engineering, a CEng, MIET and a Senior Training Instructor for a well known CAD company as well as being G4FNC, so we must assume a certain degree of competence. The difficulty apparently lies in assuming that the period between the start and stop of the pulse is subject to change - which it isn't.

It seems a bit infantile to "swap CVs", but I'll trade your wife's qualification for my Masters Degree in Information Theory and Sytems Engineering :)

But please explain how this transmitter which is switched off and disconnected from the mains is still producing a carrier and sidebands? Which is the logical result of your argument.

Now if the argument is that the equations are bounded by t, justify why the bounds exist for  a long t  but not for a short t.

No - as I keep stressing there are no constraints in the maths.

Let's return to the simplest of examples - 100% AM modulation of a carrier by a sine wave:
Q1: Do we agree that the carrier and two sidebands exist continually?
Q2: Do we agree that for a short duration around the trough of the envelope there is no (or very little) energy supplied by the transmitter?


Please try to answer those two questions because they go to the heart of understanding what is happening.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 07, 2015, 01:22:52 PM
Look, I am only saying that the key clicks are evident during the transitional high to low, or low to high rise and fall times. Other than that, no one hears offending key clicks.

Is that hard to understand? Explain why the key clicks ONLY exist as interference during the rise and fall times of the offending signal. Anything else you have to offer does not answer the question.
Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 07, 2015, 01:46:41 PM
Look, I am only saying that the key clicks are evident during the transitional high to low, or low to high rise and fall times. Other than that, no one hears offending key clicks.

Is that hard to understand? Explain why the key clicks ONLY exist as interference during the rise and fall times of the offending signal. Anything else you have to offer does not answer the question.
Pete

I think the reason there is a major disconnect between some of us is that G3TXQ and I are just looking at the spectral makeup of the transmitted signal. Just the transmitter. I believe what we say is correct because it is based on well established theory.

However, we aren't trying to prove there are no sharp/audible clicks in a receiver. I think the reason we hear sharp clicks in a receiver has been covered already by G3TXQ. But maybe this needs some extra investigation using simulation and testing using Tx + Rx systems.

I think (hope) if we do this the final conclusions will keep everyone happy in what they observe.

I'm also hoping that at the end of it all G3TXQ and I won't be seen as being as mad as we probably appear to many of you :)





Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KI6LZ on September 07, 2015, 01:48:46 PM
To tempting to not jump in and throw more in.

Take a FT of a periodic waveform which consists of two tones, tone 1 starts at t0 of .1 sec duration, tone 2 starts at t+.5 sec duration of .1 sec. total period 1 sec.

Question?

1. Does the FT show continuous tones?

2. Does the FT show the on/off nature of the tones?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K5WLR on September 07, 2015, 01:58:52 PM
To quote a very OLD Charlie Brown.....

ARRRGHHH.... I can't stand it!

Ain't this horse dead yet??? ???



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 07, 2015, 02:11:49 PM
I think the reason we hear sharp clicks in a receiver has been covered already by G3TXQ.

Only about a zillion times!

I'll say it again: at typical keying speeds, and with practical receiver bandwidths, we are never listening to a single spectral component but many. Those many spectral components are each continuous, but they drift in and out of phase. For long periods they cancel each other, but for short periods they combine in phase to produce a narrow pulse. That's the click we hear coincident with the OOK transitions. I demonstrated this in the last photo of #316!

I'm not trying to "score points" here, I'm genuinely trying to help folk understand some of the basics of modulations theory, of which On-Off keying is just one facet.

If Peter - or anyone else - cares to grapple with those two simple questions I posed earlier, it will be very helpful - they go to the heart of some of the apparent paradoxes. I'll repeat them here for convenience:

Let's return to the simplest of examples - 100% AM modulation of a carrier by a sine wave:
Q1: Do we agree that the carrier and two sidebands exist continually?
Q2: Do we agree that for a short duration around the trough of the envelope there is no (or very little) energy supplied by the transmitter?


Steve G3TXQ

Edit: Time for bed!


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: JS6TMW on September 07, 2015, 04:09:34 PM
In case anyone is wondering... OOK is On-Off-Keying.

Pete

Oh, OK.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3RZP on September 08, 2015, 01:11:20 AM
Quote
Q1: Do we agree that the carrier and two sidebands exist continually?

With the proviso that once both carrier and modulation are removed, they no longer exist.

Quote
Q2: Do we agree that for a short duration around the trough of the envelope there is no (or very little) energy supplied by the transmitter?

Again, with the above proviso.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 08, 2015, 05:06:20 AM
Quote
Q1: Do we agree that the carrier and two sidebands exist continually?

With the proviso that once both carrier and modulation are removed, they no longer exist.

Quote
Q2: Do we agree that for a short duration around the trough of the envelope there is no (or very little) energy supplied by the transmitter?

Again, with the above proviso.

Good - we are starting from some common ground! To remind other readers, this is the signal we are considering:

(http://www.karinya.net/g3txq/temp/sidebands/am1_sm.png)

We have agreed that the energy contained in the modulated wave is relative low around the trough of the envelope, even though the individual spectral components have a constant amplitude throughout the modulation cycle.

Just extending the concept further, it's clear that the total energy contained in the first half-cycle of the modulation must be much greater than the total energy in the second half cycle. In fact if you do the sums you'll find that there is over 10 times as much energy in the first half cycle of the modulated signal as in the second half cycle .... even though the energy within the individual spectral components is "evenly distributed" throughout the whole cycle.

This principle clearly applies whatever the modulating frequency. So for fun let's assume a modulating frequency of 1Hz:

Q3: For that signal, do you agree that the energy contained within the first 0.5 seconds of a modulating cycle is much greater than the energy contained within the second 0.5 seconds, even though the energy contained within the carrier and sideband spectral components totals the same during the first 0.5 seconds as it is does for the second 0.5 seconds?

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K9AXN on September 08, 2015, 10:49:10 AM
Hi Steve,

What was the Q value of the filter that you used to separate the carrier from the sideband?  Isn't it likely the high Q would prevent any or most or the quench, allowing the carrier to ring through the open key gap?  Same thing happens when you filter an SSB tone.  Starts out modulated, losing or gaining a cycle at each trough depending on the sideband and the Q ringing of the filter fills the trough.  It's a receive not transmit function.

Have a great day Kindest regards Jim  



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 08, 2015, 11:27:21 AM
What was the Q value of the filter that you used to separate the carrier from the sideband?  Isn't it likely the high Q would prevent any or most or the quench, allowing the carrier to ring through the open key gap?  Same thing happens when you filter an SSB tone.  Starts out modulated, losing or gaining a cycle at each trough depending on the sideband and the Q ringing of the filter fills the trough.  It's a receive not transmit function.

Hi Jim,

It was a 2.6kHz wide filter centred on 1.4MHz.

It's tempting to try to find implementation explanations - such as filter ringing - for the existence of the carrier and sideband spectral components during the Off period; but if you look carefully at the waveforms I showed, there is no visible decay of the signal during the Off period.

And of course - as we keep stressing - the maths tells us those spectral components are continuous through the Off period; just trust the maths and you don't need to find implementation explanations. That said, of course we have to use practical frequency selective components to allow us to view the frequency domain.

What bothers folk is how energy can exist during the OFF period when none is being supplied from the source. My dialogue with Peter G3RZP is designed to explain how - so keep reading. We currently have agreement that the spectral components are still present - at the same amplitude - even when the modulated carrier is at an extremely low level and passes through zero, so we're making progress  :)

Steve G3TXQ



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 08, 2015, 12:12:44 PM
For those interested in the signal processing aspects, I'll give you a starting point.

https://en.wikipedia.org/wiki/Uncertainty_principle#Signal_processing

We're led immediately to the Gabor limit, also known as the Heisenberg-Gabor limit because it's related to the Heisenberg uncertainty principle. However, instead of particle position and momentum, it's about signal analysis temporal resolution and frequency resolution.

You can either have high temporal resolution or high frequency resolution. You can't have both.

This is at the core of the discussion.

BTW, Dennis Gabor is a very interesting guy. He won the Noble Prize in Physics for the invention of the hologram.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 08, 2015, 01:00:05 PM
You can either have high temporal resolution or high frequency resolution. You can't have both.
Absolutely correct!

That's why, if you want to know what are the spectral components of a single Dot repeated every second - which will be 1Hz apart - you'd better analyse the signal over a period of 1 second - not just for the Dot period and not just for the Off period.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 08, 2015, 01:32:52 PM
You can either have high temporal resolution or high frequency resolution. You can't have both.
Absolutely correct!

That's why, if you want to know what are the spectral components of a single Dot repeated every second - which will be 1Hz apart - you'd better analyse the signal over a period of 1 second - not just for the Dot period and not just for the Off period.

Steve G3TXQ
And that's exactly where you're missing the point.

If you want to create a high resolution representation of the spectral components during the key clicks, you have to sample over many dits.

But if you want to know when the key clicks exist, you have to sample the dits at high temporal resolution. The off time, the on time and the rise and fall times.

If you can't wrap your head around that, then we'll never never ever agree.

And BTW, this applies to your AM modulation example, so don't try using that as a counter argument.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 08, 2015, 01:45:51 PM
If you want to create a high resolution representation of the spectral components during the key clicks, you have to sample over many dits.

According to Fourier theory not many dits if the signal is periodic - but at least one dit and the succeeding space.

But if you want to know when the key clicks exist, you have to sample the dits at high temporal resolution. The off time, the on time and the rise and fall times.

Absolutely correct - if you want to examine those clicks with good resolution in the time domain you can't at the same time have good resolution in the frequency domain. So you can't examine the clicks in close detail and at the same time resolve the separate spectral components which it comprises.

Please see #316

Steve G3TXQ



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 08, 2015, 01:56:48 PM
So then we do agree. Holy crap! But seriously Steve, you've been using some very confusing semantics in your posts.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 08, 2015, 02:10:01 PM
Just to promote SDR technology (which I've had a ton of fun with over the last 3 years) here are the components of my test bed.

1) The transmitter was an Ettus Research B210. http://www.ettus.com/product/details/UB210-KIT

It's MIMO capable, and I bought it to do DVB-T2 MISO experiments. The single TX and RX (and cheaper) version is the B200. http://www.ettus.com/product/details/UB200-KIT

2) The receiver was a Nuand bladeRF. http://nuand.com/bladeRF

3) GNU Radio. http://gnuradio.org/redmine/projects/gnuradio/wiki

It's entirely Open Source software. https://github.com/gnuradio/gnuradio

4) GQRX receiver. Also entirely Open Source software. https://github.com/csete/gqrx

5) A workstation running Linux. Specifically Ubuntu.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 08, 2015, 02:11:10 PM
I've been absolutely consistent throughout this discussion that if you use a wide bandwidth you'll see/hear keyclicks because you are seeing the aggregation of many individual spectral components; but if you want to understand what an individual spectral component is doing, you need to use a bandwidth commensurate with the frequency spacing of those components.

That was illustrated in #316 where I showed this for the situation where the bandwidth was wide:
(http://www.karinya.net/g3txq/temp/sidebands/sbkc.jpg)

and this for the situation when we narrow the bandwidth to examine just the carrier spectral component:
(http://www.karinya.net/g3txq/temp/sidebands/sbc.jpg)

By the way - did you notice in that second photo that that carrier spectral component is constant amplitude, even through the period when the keyed signal is off?

Now, it's getting towards my bedtime  :) so does anyone care to answer my question before I turn in:

Q3: For that signal (carrier AM modulated by a 1Hz sine wave), do you agree that the energy contained within the first 0.5 seconds of a modulating cycle is much greater than the energy contained within the second 0.5 seconds, even though the energy contained within the carrier and sideband spectral components totals the same during the first 0.5 seconds as it is does for the second 0.5 seconds?

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 08, 2015, 02:35:23 PM
Well, here's 0.5 Hz AM modulation on the SDR at high temporal resolution. It's 100% modulated. At this point, I don't know if it agrees with your concepts or not.

The "cloud" around the carrier is the combined close in phase noise of the two SDR's. It's a little different in each picture because the two SDR's are not in phase with each other.

Two screenshots, one during the low amplitude part and the other during the high amplitude part.

(http://www.w6rz.net/amlow.png)

(http://www.w6rz.net/amhigh.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 08, 2015, 03:20:26 PM
Well, here's 0.5 Hz AM modulation on the SDR at high temporal resolution. It's 100% modulated. At this point, I don't know if it agrees with your concepts or not.

The modulation was 1Hz not 0.5Hz, but that's incidental. I'm not sure your displays are giving us the full picture because the resolution bandwidth appears not to be resolving the 1Hz sidebands which we know are there.

However, this was the point of the question ........

We know that there must be much more energy in the first half of the modulating cycle compared to the second half-cycle. A simple integration of the modulated signal (squared) will tell us that; try it on a spreadsheet. In fact, if you look at the energy in a small time slice - about 10% of the modulation cycle - around the peak of the envelope, it contains 3500 times as much energy as the same time slice around the trough.

So, we have a massively varying energy content in the time-domain representation, and yet we know that the carrier spectral component, and those two sideband spectral components, chug doggedly on at the same amplitude throughout. Their energy temporal profile bears no relation to that of the time-domain signal!

The lesson is that there is no need for there to be equivalent energy in a time-slice of the time-domain representation of a signal as there is in the corresponding time-slice of its spectral components. Except as below:

The only energy-conservation constraint is that the total energy in the time-domain representation, integrated over all time, must be equal to the total energy in the frequency-domain representation, integrated over all frequencies (obvious, but also Parseval's Theorem); and a tighter constraint that in the case of a periodic signal, the total energy integrated across one period must be the same in both domains.

Conceptually, the transform from one domain to the other has resulted in a temporal re-distribution of the energy. If that's a bit difficult to get your head around, it might help to consider the reverse process - going from the frequency domain to the time domain - using a very simple practical example:

Take 3 separate audio oscillators, each with the same constant amplitude; one at 500Hz, one at 1500Hz and one at 2500Hz, and add them together. The resulting waveform is the red trace here:

(http://www.karinya.net/g3txq/temp/sidebands/osc_volts.png)

And this is how the power in that resulting waveform varies with time:

(http://www.karinya.net/g3txq/temp/sidebands/osc_power.png)

So there we have an example where 3 constant amplitude sinusoids combine to form a composite waveform whose energy content is very different at different parts of the period. Understanding that, the reverse process should be a little easier to accept.

Once we understand that energy is re-distributed temporally as we move from viewing the time-domain to viewing the spectral components, it becomes just a small step to accept that spectral components can exist even at times in the modulation cycle when the  time-domain signal contains no energy.

With that I'm off to bed!

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 08, 2015, 04:01:43 PM
But if you actually listen to a 0.5 Hz (or 1 Hz) AM modulated signal with the BFO on, what do you hear? What do you see happening on the watt meter connected to the transmitter?

I guess you could try this experiment with ham equipment, but you'd have to verify that the transmitter can pass a 1 Hz audio signal faithfully. With SDR, it's no problem.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: AC7ZN on September 08, 2015, 06:12:58 PM
Hi Steve,

Wish I had more time to discuss...I should never jump into these things (until I retire!).  The discussion is most interesting.
Could you please give the pulse frequency you used in the pics in #318 (that is, the repetition rate)?

Also, please comment on what happens to the spectrum when the last dit is sent and the transmitter is turned off.  How long does the spectrum persist?  What math principle do we invoke to calculate the decay?

73,

Glenn AC7ZN


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K9AXN on September 08, 2015, 06:24:15 PM
Have to ask a couple of questions.

Remember a phrase coined by a great physicist that was drilled into the grey matter in the 60’s.

“If you have to resort to complex math to explain a concept --- you simply don’t understand it”

This list is for all sorts of folks at all levels of knowledge --- not just a few.

First someone brought up the mixer and products developed.  I would like that person to explain in detail, without calculus --- just in common terms how the two sidebands are generated along with the carrier.  I am absolutely certain this will answer the question regarding W8JI’s statement.

Next, I saw a mixer schematic that was supposed to disqualify the W8JI statement.  It’s a balanced mixer and there is a unique difference between a single ended and balanced configuration.
The fellow that drew it suggests there is no carrier present, just the sidebands, which is correct.  Again, explain in commonly understandable terms without resorting to the Calculus or math in any form just why there is no carrier present.  Don’t use phasors either just wrap words around the concept.

Thanks --- Kindest regards Jim


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K6JH on September 08, 2015, 06:40:12 PM
OK Steve, I believe I understand what you are saying for the special case where the dit stream is continuous.

and this for the situation when we narrow the bandwidth to examine just the carrier spectral component:
(http://www.karinya.net/g3txq/temp/sidebands/sbc.jpg)


Do one more photo, using the same bandwidth settings as above, a wider timescale on the scope, and this time stop the TX bit stream after a dit but look at how long it takes for the carrier detected by the filter to die down. Using the same BW please - don't go into the discussion about needing to look for a month with an extremely narrow filter, to prove that the dit rate wasn't changed to a month on, month off.   ;)

In practical terms a typical HAM is listening to CW with a 500Hz, maybe even a 100Hz filter. OK, maybe even a nice 10Hz digital filter. Not nearly narrow enough to resolve the carrier signal or the individual sidebands predicted by the math. And never listening to an infinitely long dit stream - at some point the QSO is done, and the station is QRT. Even using QRSS.

The point you are trying to make is true, and even interesting, but is too esoteric to be particularly useful in a practical sense.

Regards,
Jim, K6JH, BSEE

PS: Neat parlor trick. I'll have to try that sometime on my HP E4421B sig gen sitting in my office. But I wonder if I can get my EXA or FSL analyzers that narrow in bandwidth. Hmm, some guys down the hall have a PSA...



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 08, 2015, 07:14:33 PM
Good to see a few more folks chiming in.

@AC7ZN - I'm retired.

@K9AXN - Here's my simplest explanation. Just words, and as few as possible.

We have a CW transmitter. Attached is a hand key.

When the key is up, nothing is transmitted.

When the key is down, an unmodulated (constant amplitude) carrier is sent. The bandwidth of the carrier is very small. Theoretically, it's zero. But in the real world, due to phase noise and other oscillator instabilities it has some finite bandwidth. However, it is still very small.

During the transitions from key up to key down or key down to key up, the amplitude changes from zero to full power or full power to zero. If the change is very fast, it will create a wide bandwidth signal that we hear as a key click. If the change is slowed down with some filtering, the bandwidth will be much smaller and there will be just a very minor key click (so close to the carrier, it can't be heard with a typical receiver). These wide bandwidth signals (key clicks) only exist during the transitions.

That's it. Very easy, very simple.

The other camp is misapplying long term Fourier analysis to short term events.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 09, 2015, 01:48:44 AM
My goodness - a lot happened overnight :) It's nice to see others contributing.

Let me try to answer some of the questions posed:

Quote
W6RZ: But if you actually listen to a 0.5 Hz (or 1 Hz) AM modulated signal with the BFO on, what do you hear? What do you see happening on the watt meter connected to the transmitter?

Provided you have a receiver with the selectivity to pick out just the carrier, or just one of the sidebands, you'll hear a continuous tone and the watt meter will be steady - they won't be varying at the 1Hz rate. You'll need a bandwidth </= 1Hz, otherwise you'll be looking at the aggregate of the spectral components, not them individually.

Much easier to use something like 1kHz modulation; then it's easily proved by tuning across the AM signal.

Quote
AC7ZN: Could you please give the pulse frequency you used in the pics in #318 (that is, the repetition rate)?

I need to go back to my lab book to check, but I'm pretty sure I chose 5kHz to keep the spectral component spacing significantly more than my filter bandwidth.

Of course this raises the issue touched on by K6JH - is the presence of the spectral components during the Off period merely a manifestation of the filter ringing? I would argue "no" because the maths (and it's the whole basis of discrete Fourier analysis) predicts the spectral components exist and are constant amplitude for a periodic function. But I'm well aware this particular issue can take us quickly into the philosophical question of "what do we mean by existence?", and I've seen discussions elsewhere on that topic which make our discussion look brief :)

Quote
AC7ZN: Also, please comment on what happens to the spectrum when the last dit is sent and the transmitter is turned off.  How long does the spectrum persist?  What math principle do we invoke to calculate the decay?

Things get more complicated because we are now dealing with an aperiodic function rather than a periodic one. Put simply, we have to used the more general Fourier Transform rather than the Discrete Fourier Transform, and we no longer have discrete spectral components, but energy spread continuously throughout the frequency domain.

You can conceptualise this by taking a periodic square wave - whose discrete frequency spectrum we well know by now - and imagine making the Off period go to infinity by reducing the repetition frequency to 0Hz. Clearly, the individual spectral component spacing will then be 0Hz(!), and we no longer have a discrete spectrum.

In the discussion I've deliberately stuck to periodic waveforms - such as a repeated series of Dits - to keep things simple. I felt we needed to understand and agree a simple case before moving on. The fact that in the real world the Dits will have to begin at some point, and end at another point, doesn't invalidate the analysis of the periodic signal in between.

Quote
K9AXN: "If you have to resort to complex math to explain a concept --- you simply don’t understand it"

Let me offer another quote: "If your understanding is at variance with the simple maths, then your understanding is wrong!

Quote
K6JH: OK Steve, I believe I understand what you are saying for the special case where the dit stream is continuous.

Not just a continuous Dit stream, but anything that's periodic. It could be continuous dashes, continuous "Y"s or continuous "CQ CQ CQ DE G3TXQ G3TXQ"s. Of course you would need to adjust the bandwidth to match the periodicity of the signal.

Quote
K6JH: The point you are trying to make is true, and even interesting, but is too esoteric to be particularly useful in a practical sense.

Jim - well I'm glad someone else agrees it's true ;) Applied to Morse at typical keying rates and with practical receiver bandwidths, it probably does seem esoteric; but it doesn't seem so esoteric when exactly the same analysis is applied to the output of an On-Off RF mixer  - something we exploit on a regular basis in our radios. For me this discussion was all about getting folk to think through the underlying principles.

Quote
K6JH: Do one more photo, using the same bandwidth settings as above, a wider timescale on the scope, and this time stop the TX bit stream after a dit but look at how long it takes for the carrier detected by the filter to die down. Using the same BW please - don't go into the discussion about needing to look for a month with an extremely narrow filter, to prove that the dit rate wasn't changed to a month on, month off.

Jim - I'm not sure I can do that with the kit I have, but I know exactly where you are going  ;)  However, as I hinted in the answer to AC7ZN I suspect it will quickly get us into deep philosophical discussions. For instance, if I excite a narrowband filter with a very short single pulse, we both know I'll see a relatively slowly decaying carrier at the filter's centre frequency. Now - am I seeing the "filter ringing", or have I extracted the spectral component from the pulse that's at the filter frequency? Or both?  :)

Quote
W6RZ: The other camp is misapplying long term Fourier analysis to short term events.

If you have a periodic waveform comprising a Dit followed by a period of silence:

* Analyse it for just the Dit period if you want to know the spectrum of just the Dit - it'll be a single spectral line
* Analyse it for just the Off period if you want to know the spectrum of silence - it'll be nothing!
* Analyse it for just the keyclick period if you want to know the spectrum of just the click
* But if you want to know the spectrum of the complete signal, examine over one complete period, Dit+Silence

When I was teaching this stuff to graduates, I would use this illustration to show what a mess we can get into if we don't analyse the signal over a complete period:

I'd show them a picture something like this of a time slice of a modulated carrier, and then ask them if the signal had a spectral component at the carrier frequency:

(http://www.karinya.net/g3txq/temp/sidebands/dsb2.png)

Pound to a penny they'd tell me that it must have - you can see it! Then I'd reveal the more complete signal:

(http://www.karinya.net/g3txq/temp/sidebands/dsb1.png)

Whoops! It's a DSBSC signal which we know has no spectral component at the carrier frequency.

Fourier had it right - if you want to know the spectrum of a periodic signal, you must analyse it over one complete period, not just a part of the period.

-----------------------------------------------------------------

Like many discussions of this type, the differences of opinion hinge on different interpretations of the words used. "Sidebands" is a classic case - G0HZU and I have taken it to mean the individual frequency-domain spectral components which represent a signal; others have taken it to mean the aggregation of those components.

G0HZU said something to that effect 23 posts ago, and I said it 103 posts ago - seems a lifetime, now  :)

Steve G3TXQ




Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 09, 2015, 06:43:13 AM
Do one more photo, using the same bandwidth settings as above, a wider timescale on the scope, and this time stop the TX bit stream after a dit but look at how long it takes for the carrier detected by the filter to die down.
Jim,

This was the best I was able to do this morning with the kit available. It shows the leading and trailing edges at the output of the filter with an On-Off keyed 1.4MHz carrier. The carrier was centred in the filter bandwidth, which is about +/-1.3kHz. The keying rate was 1 second On, 1 second Off - in other words 0.5Hz. The sweep rate is 0.5mSec/division.

Apologies that the images are not better quality, but it was tough holding the camera steady with long exposure times!

(http://www.karinya.net/g3txq/temp/sidebands/edge_lead.jpg)
(http://www.karinya.net/g3txq/temp/sidebands/edge_fall.jpg)

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K9AXN on September 09, 2015, 08:35:57 AM
G3TXQ,
If you take the 180 degree phase shift mixer that for years was called a balanced mixer and inject two signals, you can prove or disprove your theory.

Inject your 1.4Mc carrier in the LO leg.  In the 0 or 180 degree leg inject a keyed signal with the rise time resembling the first quarter of a 100 Cycle Sine, peaking with the key held and ending with the key up decline, shaped like the second quarter of the 100 cycle sine.

Don't worry about the second half of the 1.4Mc cycle --- we'll just use the rise and fall of the first half.  Yes I know, this defeats the purpose of the balanced effect on carrier exclusion.
 
Shaping the rise and fall times with a 100 cycle sine shape will produce two 100 cycle side bands, unlike keying with a square wave which produces an obvious noise cluster.

The side bands will be a result of the shape of the rise and fall time, not the periodic depression of the key.

Further the two side bands will only be present during the rise time followed by only carrier until the key is released where the decline shape will again produce the two side bands.  When the decline ends with the key up, there will be no carrier present.

Kindest regards Jim


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 09, 2015, 08:54:52 AM
At this point, I'm not even sure where we stand, so I'll just stick with SDR engineering.

It's interesting that G3TXQ brought up DSBSC (Double Side Band Suppressed Carrier). Here are the GNU Radio flows for AM, DSB and SSB.

The Float To Complex block is essentially a balanced modulator. All it's doing is converting from real samples to complex samples with all the imaginary values set to zero (the Constant Source block provides the imaginary zero values). This takes the audio spectrum and creates a positive version (zero frequency and an up) and a negative version (zero frequency and down).

The Low Pass Filter (or Bandpass Filter in SSB) is optional. The filters are just there to limit the bandwidth of our emitted signal. One thing is for sure, I can send some extreme ESSB with this setup. Hi fidelity 20 Hz to 20 kHz if desired.

For AM, the Add Constant block inserts the carrier. The DSB flow is exactly the same, except the Add Constant block is deleted.

For SSB, a Bandpass filter is used instead of a Low Pass filter. LSB is the cool case, because we use a bandpass filter with negative frequencies. In other words, we're using mathematical constructs that don't exist in the real world (negative frequencies) to send real world signals. To send USB, just swap the Band Pass Filter block with the one that's greyed out.

The Rational Resampler block is just used to increase the sample rate. The SDR hardware that I'm using doesn't work at 48 kHz, it's below the minimum sampling rate. The UHD USRP Sink block sends the samples to the SDR hardware.

(http://www.w6rz.net/amflowsm.png)

(http://www.w6rz.net/dsbflowsm.png)

(http://www.w6rz.net/ssbflowsm.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 09, 2015, 09:10:22 AM
If you take the 180 degree phase shift mixer that for years was called a balanced mixer and inject two signals, you can prove or disprove your theory.

Inject your 1.4Mc carrier in the LO leg.  In the 0 or 180 degree leg, a keyed signal with the rising first quarter of a 100 Cycle Sine, peaking with the key held and ending with the key up decline, shaped like the second quarter of the 100 cycle sine.

Don’t worry about the second half of the 1.4Mc cycle --- we’ll just use the rise and fall of the first half.  Yes I know, this defeats the purpose of the balanced effect on carrier exclusion.
 
Shaping the rise and fall times with a 100 cycle sine shape will produce two 100 cycle side bands, unlike keying with a square wave which produces an obvious noise cluster.

The side bands will be a result of the shape of the rise and fall time, not the periodic depression of the key.

Further the two side bands will only be present during the rise time followed by only carrier until the key is released where the decline shape will again produce the two side bands.  When the decline ends with the key up, there will be no carrier present.
Jim,

Sorry, but I got lost somewhere in that description - could you sketch what you are describing? For one thing I didn't get what the keying period is - that will obviously determine the spectral component spacing.

Thanks,
Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K9AXN on September 09, 2015, 09:34:24 AM
It's written in as simple a form as I know.  It just doesn't get any simpler.

Have a super day and carry on old chap--- got to move on.

Kindest regards Jim


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 09, 2015, 09:58:10 AM

The Float To Complex block is essentially a balanced modulator. All it's doing is converting from real samples to complex samples with all the imaginary values set to zero (the Constant Source block provides the imaginary zero values). This takes the audio spectrum and creates a positive version (zero frequency and an up) and a negative version (zero frequency and down).
 

And I can take a constant carrier, phase locked to the desired carrier frequency, and use a combiner to add the DSB sidebands.  Does the carrier vary with the modulation intensity?

What if I use two separate antennas, one for the DSB signal, and the other for the phased locked carrier transmitter?

What if....



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 09, 2015, 11:42:55 AM
At this point, I'm not even sure where we stand, so I'll just stick with SDR engineering.

OK - I'm going to summarise my position and then call it a day. Please note that everything which follows relates to a periodic signal such as repetitive Dots, and an analysis window of one or more of those periods:

1) The whole basis of Fourier analysis is that this signal can be represented by a set of sinusoids which are constant amplitude throughout the analysis period.

2) The mathematical time-domain expression for that periodic signal contains terms which represent the spectral components, and they are time invariant

3) If we key a carrier on and off quickly enough that we can examine individual spectral components with practical filters, we observe that they are constant amplitude.

4) There is no requirement for the instantaneous power in the time-domain signal to be equal to the instantaneous power in the spectral components - we saw that in the simple example of an AM sine-wave modulated carrier. So there is no anomaly in having power in the spectral components when the modulated carrier is Off. All that is required is that the total energy in the spectral components should equal the total energy in the modulated carrier when integrated across the analysis period.

5) At typical keying speeds and with practical bandwidths we don't observe individual spectral components, but rather an aggregation of many of them. That aggregation results in the energy in the spectral components being concentrated into a small time period around the keying transitions - which we observe as keyclicks.

6) If we define "sidebands" to mean an aggregation of many spectral components, then sidebands only exist at the transitions and not between the Dots; if we define "sidebands" to mean individual spectral components, then they exist during and between the Dots.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: AC7ZN on September 10, 2015, 03:04:48 AM
I appreciate Steve's summary.  It would be presumptuous for me to summarize my take on things since I have not been an active participant here, so let me start with a summary of what the Fourier integral does.  We will need to use the Fourier integral to add insight to what is happening because one of my claims is that a CW spectrum cannot be adequately addressed by periodic analysis.

Those of you who understand and have worked with the discreet Fourier transform (DFT, or FFT) will have no trouble extending those concepts to the Fourier integral.  In the DFT we know that increasing the sample rate (using more sample points in our set) also increases the frequency resolution in the frequency domain (there are more frequencies, more closely spaced).  Lengthening the dataset in time increases the period (makes the repetition rate slower) and potentially adds lower frequency energy to the spectrum.

Now imagine increasing the both the number of samples and the time length of the dataset  to infinity.  This is the Fourier integral.  Because the time of the dataset is infinite, the period is infinite and becomes nonperiodic, since the one dataset covers all time.  Both the time domain and frequency domain have infinite resolution and are continuous, not discrete.  We still multiply all datapoints by the sine and cosine of every frequency as is done in the DFT, but we now do it analytically with calculus.  As with other calculus functions, we can often successfully approximate results with discreet computer calculations by making sure we include enough of them to give accuracy, and the DFT (FFT) can be used to do this by simply increasing sample rate and period until meaningfully accurate results are obtained.

Some common results have been tabulated.  We know the Fourier transform of a Dirac delta function is 1 at all frequencies.  Its dual, a Dirac delta function in the frequency domain is 1 for all time. A Dirac delta function is a mathematically convenient construct that has infinite amplitude, zero width and integrates to one.  On plots it is noted by an arrow pointing up. It can also be called an impulse, and the impulse response of a filter assumes this function has been applied to the input.

The unit step is of particular interest in CW, because the (clicky) CW tone can be modeled as a pulse consisting of the convolution of unit steps with the carrier.  The unit step spectrum has a Dirac delta function at zero, then a spectrum of 1/jw. We will plot this and comment on it in a future post, hopefully.

73,
Glenn AC7ZN





 




Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 10, 2015, 06:58:59 AM
Glenn,

I appreciate your input.

I agree with all that you have said, although I wonder where the third sentence is taking us -  Fourier analysis of an aperiodic 3 minute CW QSO may test my abilities  ;)

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: REASTON on September 10, 2015, 07:27:38 AM
This whole thread just cracks me up.  Lots of talk about math, but I still don't see any.  How can you people "talk" about math.  You either know it and write it or you don't know it.  All this crap in this thread is everybody blowing smoke trying to convince each other they are RF Engineers when they are not.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 10, 2015, 08:42:36 AM

I would summarise things like this:

 
2) At typical keying speeds, and with practical receiver bandwidths, we never hear individual spectral components but an aggregation of many of them. The summation of those many components is a signal that is impulsive, with peaks coincident with the OOK transitions.

 
Steve G3TXQ



Isn't that basically what W8JI was trying to convey before this topic became an advanced math discussion?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 10, 2015, 09:29:11 AM
Isn't that basically what W8JI was trying to convey before this topic became an advanced math discussion?

I don't know - I wasn't part of the early discussions so I don't know what W8JI meant.

I did say way back in #250:

"If by "sidebands" you mean a summation of several individual spectral components, then I agree they are not continuous; but if we are talking about separate spectral components, then they are continuous."

But then in #285 W6RZ challenged our claim that the spectral components were continuous; maybe he meant to say "sidebands", but the term he used was "spectral tones".

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 10, 2015, 09:35:36 AM
What are "spectral components"?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 10, 2015, 09:48:30 AM
What are "spectral components"?
The individual, frequency-domain, spectral lines resulting from a DFT.

What did you mean by "spectral tones"?

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 10, 2015, 09:49:22 AM
This whole thread just cracks me up.  Lots of talk about math, but I still don't see any.  How can you people "talk" about math.  You either know it and write it or you don't know it.  All this crap in this thread is everybody blowing smoke trying to convince each other they are RF Engineers when they are not.
Perhaps you missed these: #290, #309,#322

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 10, 2015, 10:01:57 AM
What are "spectral components"?
The individual, frequency-domain, spectral lines resulting from a DFT.

What did you mean by "spectral tones"?

Steve G3TXQ

The same thing. We just disagree on exactly when they exist. And by "exist", I mean in the real world, transferring power. Not just as a mathematical construct.

BTW, I'm not an RF engineer. I'm a software engineer with an EE background. The last 18 years of my professional career were in the video compression industry working on MPEG-2 encoders and systems. Before that, military data communications networks.

That's why I'm having so much fun with SDR. What a wonderful marriage of my professional skills and my lifelong hobby, amateur radio.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 10, 2015, 10:39:38 AM
The same thing. We just disagree on exactly when they exist. And by "exist", I mean in the real world, transferring power. Not just as a mathematical construct.

Understood - but I would point you back to a couple of examples we discussed earlier, to think about whether they are just a mathematical construct: we can extract "real" energy from a spectral component at the output of a switching mixer, even when the switch is Open; we can extract "real" energy from the spectral components of a sine-wave modulated AM signal, even through the trough where the instantaneous power in that modulated signal is zero. But it seems that folk find it difficult to accept once the modulating frequency gets low.

I can understand your fascination with this stuff. My background was RF engineering with the BBC, but then I worked on developing the convolutional error correction coding systems for the UK's military ComSat programme. That really grabbed my interest, and still does; the highlight was meeting Andrew Viterbi!

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K9AXN on September 10, 2015, 11:24:42 AM
Steve,

In #306 you have a photo containing two wave forms.  You say the top is of a carrier running through a diode switch and the lower is of the keyed signal passed through a filter selecting the carrier frequency.  The carrier is present in the gaps between key down time.  What you did is place the channel 2 probe on the input to the diode switch and of course the carrier is present, it has not been switched off.

If you review #309 and the common modulation formula that you presented, you will see that when the key is closed modulation commences generation side bands until the end of rise time where the side bands cease.  Now, with the key remaining closed, only carrier is present.  When the key is opened, modulation reoccurs during the quench time, again generating side bands.  Then there is nothing in the inter-key gaps.

There is a problem with 306 and the narrative.

Kindest regards Jim


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 10, 2015, 12:17:56 PM
I can understand your fascination with this stuff. My background was RF engineering with the BBC, but then I worked on developing the convolutional error correction coding systems for the UK's military ComSat programme. That really grabbed my interest, and still does; the highlight was meeting Andrew Viterbi!

Steve G3TXQ

Then you would absolutely love the SDR based DVB-T2 transmitter I developed over the winter. With this GNU Radio flow, you can transmit a signal exactly like the HD multiplex you receive from Crystal Palace (or wherever your regional DVB-T2 transmitter is located and assuming you actually watch TV).

The BBC was instrumental in promoting DVB-T2 technology, and I used their reference streams to verify my software. Another interesting aspect is the error correction coding technology. Instead of Viterbi codes, DVB-T2 uses a code called Low-Density Parity-Check or LDPC. The interesting part is that LDPC coding was invented in 1960 by Robert G. Gallager. Because computers were not powerful enough in 1960 to implement the codes, it was forgotten for many years. Then it was rediscovered in 1996 and was specified in DVB-T2 instead of the then popular Turbo codes.

Also, I collaborate with one of your countrymen, Charles Brain G4GUO. Charles is leading the charge for digital amateur TV or DATV with his transmitter project DATV-Express.

http://www.datv-express.com/

He also has a very nice blog at:

http://g4guo.blogspot.com/

Here's the graph with the exact BBC DVB-T2 parameters that delivers 40.2 Mbps is an 8 MHz channel. Just a little more complex than the previous ones I've posted.

(http://www.w6rz.net/bbcdvb-t2.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 10, 2015, 12:32:55 PM
In #306 you have a photo containing two wave forms.  You say the top is of a carrier running through a diode switch and the lower is of the keyed signal passed through a filter selecting the carrier frequency.  The carrier is present in the gaps between key down time.  What you did is place the channel 2 probe on the input to the diode switch and of course the carrier is present, it has not been switched off.

Of course not - the upper trace is the output of the diode switch; the lower trace is the output of the filter selecting the carrier.

That's the fundamental point - the carrier spectral component is present even when the switch is off. Its power is real - not just a mathematical construct. If the carrier frequency was 5MHz and the switching frequency was 1MHz we'd have continuous energy available at 5MHz, 4MHz and 6MHz, not just when the switch was On.

I really don't know how much plainer to put it.

If you review #309 and the common modulation formula that you presented, you will see that when the key is closed modulation commences generation side bands until the end of rise time where the side bands cease.  Now, with the key remaining closed, only carrier is present.  When the key is opened, modulation reoccurs during the quench time, again generating side bands.  Then there is nothing in the inter-key gaps.

Nope - when you expand it, that formula contains a first term of 0.5xAcxsin(ωct). That's a sine wave whose amplitude is 0.5xAc. Notice that the amplitude is not time varying - it's a constant. And that's exactly what the photo in #306 shows.

As I keep trying to explain, there is a temporal redistribution of energy between the time-domain and frequency-domain representations. The total energy is the same in each case, but in the time domain it is all contained in the first half-cycle of the modulation, whereas in the spectral components it is evenly distributed across the whole modulation cycle.

If you have trouble with that concept, just go back to the basic sine-wave AM case. We know the carrier component and the sideband components are constant amplitude throughout the modulation cycle, even though the modulated wave has many times more energy in the first half-cycle compared to the second half-cycle.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 10, 2015, 12:42:46 PM
The BBC was instrumental in promoting DVB-T2 technology, and I used their reference streams to verify my software.

Interesting stuff!

I paid a visit to the BBC's Research Department during my training with them back in 1965. We were very excited when told we were going to be shown the newly-developed Line Standards Converter which converted US 525-line TV pictures to our 625-line. We wondered what amazing new digital technology were going to see.

Imagine our disappointment when we saw a UK 625-line camera pointing at a US 525-line monitor! The "high-tech" bit was that the monitor had been modified to incorporate "spot wobble". In other words the resolution was deliberately degraded so that there was no strobing between the two line structures.

Happy days!

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: AC7ZN on September 10, 2015, 12:56:01 PM
Glenn,

Fourier analysis of an aperiodic 3 minute CW QSO may test my abilities  ;)

Steve G3TXQ

Yeah, me too.  I'm reminded of a time early in my career when I would attend meetings where spectrum analyzer measurements were presented to show runout and vibration.  I couldn't make heads or tails of them but the other engineers around me would nod sagely.  I wondered if I would ever get this stuff.  I later realized it was all gibberish and it was a case of the emperor's new clothes.

Anyway, no three minute QSOs (I get little enough info from those anyway...my signal always seems to be 599). We will be going the other direction...here is my evil plan:

I am attempting to answer the unwritten (but I think implied) question of this thread:  "I get it that an abrupt keying transition causes clicking sounds in my receiver, and know that adding a low pass filter to the keying waveform  cures it,  but what makes the actual click?  Looking at the keying waveform there is sure enough a sharp edge, but why does it sound like a click?  Can a spectrum of the click itself help me understand this?"

Now, you certainly know what makes the click, but to my knowledge, no one in the ump-t-ump posts here has presented the actual spectrum of the click.  This, then, is my intent:  to zero in on the click itself with the Fourier integral and get a spectrum that gives insight as to why we hear a clicking sound, and why we can hear it many 100's of Hz or even kHz away.  This spectrum is difficult to catch with a spectrum analyzer, even the new gated ones, because it comes and goes so fast.

If nothing else, the exercise will give us (me especially) a review of the Fourier integral and what we can do with it.

73,
Glenn AC7ZN


 


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on September 10, 2015, 01:40:00 PM
Very early in this thread there was a brief discussion of what constitutes a "key click".
To my knowledge, there is no formal definition, but i tried to predict what I believe that the ITU would answer if an Interim Working Party of the Study Group 1 should be set up:

"Higher order keying sidebands, created by deviations from the optimum shaping of the keying envelope,  causing the occupied bandwidth of an A1A or A1B radiotelegraph emission to become much greater than its necessary bandwidth as established by the relevant relations between signalling rate and bandwidth set out in Appendix 1 of the Radio Regulations and in ITU-R Recommendation SM.328"

This may serve as a starting point.

73/
Karl-Arne
SM0AOM


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 10, 2015, 01:47:51 PM
Now, you certainly know what makes the click, but to my knowledge, no one in the ump-t-ump posts here has presented the actual spectrum of the click.  This, then, is my intent:  to zero in on the click itself with the Fourier integral and get a spectrum that gives insight as to why we hear a clicking sound, and why we can hear it many 100's of Hz or even kHz away.  This spectrum is difficult to catch with a spectrum analyzer, even the new gated ones, because it comes and goes so fast.

If nothing else, the exercise will give us (me especially) a review of the Fourier integral and what we can do with it.
Glenn,

I will certainly be interested in the results - particularly in how you would then seek to apply them more generally to a complete Morse character.

The more I ponder this the more it becomes clear what separates the two "camps". I'll go back to periodic functions again (because I find the DFT easier than the more general Fourier Integral) and consider repeated Dots:

I think we all agree that if we look at the DFT of a short time slice taken around the leading edge of the Dot we'll see a wideband spectrum; let's call it Spectrum A. Then if we look at a time slice of the Dot period we'll see a different spectrum - it'll be a single spectral line at the carrier frequency; call it Spectrum B. Then we look at a time slice around the trailing edge of the Dot and get a wideband Spectrum C. And finally look at the spectrum of the Off period - nothing - call it Spectrum D.

We can look at that analysis and say that the spectrum of the periodic signal is clearly a sequence of changing spectra:
Spectrum A > Spectrum B > Spectrum C > Spectrum D > Spectrum A > Spectrum B > Spectrum C > Spectrum D ....... etc

However we can also look at that signal, recognise that it is periodic, and take a time slice equal to a complete period (Dot plus gap). In that case we get a completely different Spectrum - let's call it Spectrum X.

The spectrum of the periodic signal is then: Spectrum X > Spectrum X > Spectrum X ...... etc
In other words a steady, continuous, spectrum with constant amplitude spectral line for as long as the signal exists.

Some may then ask: "Which is the real one?" To which the answer - as so often - is "it all depends"!

W6RZ usefully reminded us earlier of the duality between time resolution and bandwidth. If we observe that signal with a wide-bandwidth spectrum analyser we will get good time resolution and will see the sequencing, changing, spectra. But if we look at it with a very narrow bandwidth spectrum analyser we would see just one continuous spectrum.

So the answer to the question  "Which is the real one?" is:  "It depends on how you observe it"!

Most of our differences have been because we've been viewing the same signal from those two different perspectives.

I have a feeling another contributor said something very similar a zillion posts ago and we ignored him. My apologies!

Steve G3TXQ



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K9AXN on September 10, 2015, 01:56:08 PM
You'll have to be patient here Steve,

Reference #302

As I understand it, you have a carrier generator A, connected to a diode switch B, and the output from the switch is represented by the top trace.

The bottom trace is rendered by connecting a crystal filter to the output of the switch which is pulsed carrier and probing at the output of the filter.

Is this correct and if so, what is the time duration of the on off bursts?  Also is the filter the 1.4Mc crystal filter that you referred earlier?


Kindest regards Jim


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 10, 2015, 02:12:14 PM
As I understand it, you have a carrier generator A, connected to a diode switch B, and the output from the switch is represented by the top trace.

The bottom trace is rendered by connecting a crystal filter to the output of the switch which is pulsed carrier and probing at the output of the filter.

Is this correct and if so, what is the time duration of the on off bursts?  Also is the filter the 1.4Mc crystal filter that you referred earlier?
All correct!

The filter is centred on 1.4MHz and has a bandwidth 2.6kHz [Post #338 refers]
The keying frequency was 5kHz [Post #353 refers]
The impulse response of the filter is shown in #354

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K9AXN on September 10, 2015, 02:52:51 PM
Steve,

Don't believe I can key that fast. 

I believe you just explained the carrier being present between keying periods.  I asked you earlier if it could have been ringing and you said no.

You drove that filter with 1.4Mc 5Kc bursts of full amplitude square wave modulation.  That's 10us on and 10us off with a gap of only 10us between exposure to full carrier amplitude.  The experiment and measurement is patently invalid.

Isn't It a bit optimistic to think that a filter with a Q exceeding 10,000 would not ring through a 10us period.  In fact it did, and that is what your basing your supposition.

Need to call the troops back and regroup.

Please have a good day.  Kindest regards Jim


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 10, 2015, 02:59:21 PM
It's getting late here, so if I may, I'll try to guess where Jim's questions are leading and pre-empt them.

Edit: posted this just before I saw Jim's question - looks like I guessed correctly ;)

We know we can extract energy from the continuous signal at the output of that 1.4MHz filter, even when the switch is Off; so to that extent it is "real" - but is it an artefact of the filter "ringing" through the Off period, excited by the trailing edge of the input signal?

I say "no" and offer the following evidence:

1) The time-domain equation says we should see a time-invariant signal without needing to invoke explanations of filter ringing.

2) There is no evidence in the filter output waveform of any "decay" during the Off period that would be consistent with the measured impulse response of the filter.

3) The amplitude of the filter output signal is exactly one half of the carrier amplitude at all times. How can that be if the filter is simply "allowing the carrier through" during the On period, and then ringing during the Off period? Why would the filter output be half the carrier amplitude during the On period of the switch?

4) The amplitude of the filter output signal is exactly one half of the carrier amplitude, confirming the 0.5xAcxsin(ωct) predicted by the maths.

5) More generally, if the existence of individual spectral components is attributed to filter "ringing", how is it that we can filter out the carrier from a sine-wave modulated AM signal and find that it is constant amplitude even through the periods when the input signal is at a trough? It's very difficult to believe that the filter is ringing as a result of being excited by a sine-wave modulated input signal.

Steve G3TXQ



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 10, 2015, 03:45:39 PM
Isn't It a bit optimistic to think that a filter with a Q exceeding 10,000 would not ring through a 10us period.

Maybe, but since the Q is 538 that's irrelevant ;)

Q= fc/Δf = 1.4MHz/2.6kHz = 538

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K9AXN on September 10, 2015, 05:15:29 PM
Isn't It a bit optimistic to think that a filter with a Q exceeding 10,000 would not ring through a 10us period.

Maybe, but since the Q is 538 that's irrelevant ;)

Q= fc/Δf = 1.4MHz/2.6kHz = 538

Steve G3TXQ

Steve,

Your calculation regarding the Q of a crystal filter is patently incorrect --- no absurd.

The apparent Q of a crystal filter is based on the dimensions of the skirts not the 2.4Kc width of the 3db pass band.
If your calculation method was correct, the shape factor of your filter would be astronomical and resemble a pair LC tuned circuits which would not even be light years from distinguishing the difference between a carrier and 5Kc  sideband @1.4Mc.  

Now that you are aware of the behavior of a crystal filter, you can retrace your steps and rerun your test using the appropriate inputs and measurement techniques.  

I would suggest a 100 cycle on off, depending on the skirt selectivity of your filter, so you don't get tangled up in your calculations.  Using 100 cycle on off times will provide a good idea as to the rise and quench time of that filter.  Note there is a spool up time in the filter as well.   Remember we were addressing key clicks not audio.  if you need further help to understand crystal filter behavior G3RZP or W8JI are good tutors.

Now lets look at #336 the, the photo you copied from a book.  You seem to understand the concept.  Now replace the red modulation line with a pair of two positive Square waves separated by an off condition.  What do you think it will morph to?  Do it in your head it's simple.  You don't have to draw it just explain it.

BTW, how do you add a photo to the text?   Someone please.

G-day to you --- Kindest regards Jim      


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K9AXN on September 10, 2015, 06:01:21 PM
It's getting late here, so if I may, I'll try to guess where Jim's questions are leading and pre-empt them.

Edit: posted this just before I saw Jim's question - looks like I guessed correctly ;)

We know we can extract energy from the continuous signal at the output of that 1.4MHz filter, even when the switch is Off; so to that extent it is "real" - but is it an artefact of the filter "ringing" through the Off period, excited by the trailing edge of the input signal?

I say "no" and offer the following evidence:

1) The time-domain equation says we should see a time-invariant signal without needing to invoke explanations of filter ringing.

2) There is no evidence in the filter output waveform of any "decay" during the Off period that would be consistent with the measured impulse response of the filter.

3) The amplitude of the filter output signal is exactly one half of the carrier amplitude at all times. How can that be if the filter is simply "allowing the carrier through" during the On period, and then ringing during the Off period? Why would the filter output be half the carrier amplitude during the On period of the switch?

4) The amplitude of the filter output signal is exactly one half of the carrier amplitude, confirming the 0.5xAcxsin(ωct) predicted by the maths.

5) More generally, if the existence of individual spectral components is attributed to filter "ringing", how is it that we can filter out the carrier from a sine-wave modulated AM signal and find that it is constant amplitude even through the periods when the input signal is at a trough? It's very difficult to believe that the filter is ringing as a result of being excited by a sine-wave modulated input signal.

Steve G3TXQ



#1 You are apparently misinterpreting the Time domain equation.  I see nothing that excludes crystal filter behavior when they are used to confirm a theory.  On the contrary, simple logic would include any components that would subvert the outcome of a test.

#2  Once you get your arms around the behavior of a crystal filter you'll change your mind.

#3  A 3db loss in a crystal filter is not so bad --- most are more.

#4  Answered with #3.  You have again forgotten to acknowledge all of the relevant considerations to a simple equation.  The math works but you have to provide correct input.

#5  You need a tutor regarding crystal filters --- Again try G3RZP he may be local.

Please have a good evening and Kindest regards Jim


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: AC7CW on September 10, 2015, 07:55:00 PM
Glenn,

Fourier analysis of an aperiodic 3 minute CW QSO may test my abilities  ;)

Steve G3TXQ

Yeah, me too.  I'm reminded of a time early in my career when I would attend meetings where spectrum analyzer measurements were presented to show runout and vibration.  I couldn't make heads or tails of them but the other engineers around me would nod sagely.  I wondered if I would ever get this stuff.  I later realized it was all gibberish and it was a case of the emperor's new clothes.

Anyway, no three minute QSOs (I get little enough info from those anyway...my signal always seems to be 599). We will be going the other direction...here is my evil plan:

I am attempting to answer the unwritten (but I think implied) question of this thread:  "I get it that an abrupt keying transition causes clicking sounds in my receiver, and know that adding a low pass filter to the keying waveform  cures it,  but what makes the actual click?  Looking at the keying waveform there is sure enough a sharp edge, but why does it sound like a click?  Can a spectrum of the click itself help me understand this?"

Now, you certainly know what makes the click, but to my knowledge, no one in the ump-t-ump posts here has presented the actual spectrum of the click.  This, then, is my intent:  to zero in on the click itself with the Fourier integral and get a spectrum that gives insight as to why we hear a clicking sound, and why we can hear it many 100's of Hz or even kHz away.  This spectrum is difficult to catch with a spectrum analyzer, even the new gated ones, because it comes and goes so fast.

If nothing else, the exercise will give us (me especially) a review of the Fourier integral and what we can do with it.

73,
Glenn AC7ZN


 


If the click comes from the ramped waveform of the keying envelope then it's pretty much a straight line ramp and it will be a fundamental and all the harmonics. How you get a fundamental frequency of a ramping signal is beyond me though. What if it just ramps forever, where is the fun(damental) in that?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 11, 2015, 02:22:39 AM
Your calculation regarding the Q of a crystal filter is patently incorrect --- no absurd.

From one of my engineering text books: "A band-pass filter can be characterised by its Q factor. The Q-factor is the inverse of the fractional bandwidth." Fractional bandwidth = Δf/fc, so Q= fc/Δf

I would suggest a 100 cycle on off, depending on the skirt selectivity of your filter, so you don't get tangled up in your calculations.  Using 100 cycle on off times will provide a good idea as to the rise and quench time of that filter.  Note there is a spool up time in the filter as well.   Remember we were addressing key clicks not audio.

I already showed you the result of that measurement!

Now lets look at #336 the, the photo you copied from a book.  You seem to understand the concept.  Now replace the red modulation line with a pair of two positive Square waves separated by an off condition.  What do you think it will morph to?  Do it in your head it's simple.  You don't have to draw it just explain it.

That's easy - it will morph to the same diagram but with many more spectral components, each spaced by the modulation frequency. Each spectral component will be continuous even through the period when the modulated carrier is Off.

Steve G3TXQ



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 11, 2015, 02:41:26 AM
#1 You are apparently misinterpreting the Time domain equation.  I see nothing that excludes crystal filter behavior when they are used to confirm a theory.  On the contrary, simple logic would include any components that would subvert the outcome of a test.

You miss my point - the time domain equation predicts the existence of a constant amplitude carrier without invoking any crystal filter behaviour.

Here's the time domain formula again:
fo(t) = Ac.sin(ωct) x [0.5 + 2/π [sin(ωst) + 1/3 sin(3ωst) + 1/5 sin(5ωst) + ......  etc  ........]]

The Ac.sin(ωct) term is the carrier; the terms in brackets are a signal toggling between 0 and +1 at the switching rate. We are multiplying the two together so that we get a carrier which toggles on and off at the switching rate. Do you notice that there is nothing here representing "filter ringing" and yet the equation predicts the existence of a 0.5 x Ac x sin(ωct) term?

I really don't know how to make it any plainer!

#3  A 3db loss in a crystal filter is not so bad --- most are more.

I'm comparing the filter output amplitude with a keyed carrier, to the filter output amplitude when the carrier is not keyed. The filter loss is common to both and therefore irrelevant to the discussion.

[PS: although it's irrelevant, since when has a halving of amplitude equated to a 3dB loss ??????]

#5  You need a tutor regarding crystal filters --- Again try G3RZP he may be local.

It would be helpful if you try to answer the questions rather than make personal comments. I'll try it again:

Consider the very simple case of a carrier AM modulated with a sine-wave. If we filter that signal to exclude the two sidebands we get a constant amplitude spectral line at the carrier frequency. It's real, we can extract energy from it, it's not "just a mathematical construct"! Try it with a receiver tuned across an AM signal. Now:

Q1: How can that carrier component be constant amplitude when the input signal is a carrier that is varying in amplitude?
Q2: How can the instantaneous power in the filtered carrier component be constant, even when the instantaneous power in the input signal is zero?
Q3: Over the first modulation half-cycle, how can the energy in the filtered carrier component be so much less than the energy in the input signal.
Q4: Over the second modulation half-cycle, how can the energy in the filtered carrier component be so much higher than the energy in the input signal.


If you can think through satisfactory answers to those basic questions you'll have a much better understanding of the underlying principles.

Steve G3TXQ





Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K9AXN on September 11, 2015, 10:06:09 AM
Your calculation regarding the Q of a crystal filter is patently incorrect --- no absurd.

From one of my engineering text books: "A band-pass filter can be characterised by its Q factor. The Q-factor is the inverse of the fractional bandwidth." Fractional bandwidth = Δf/fc, so Q= fc/Δf

I would suggest a 100 cycle on off, depending on the skirt selectivity of your filter, so you don't get tangled up in your calculations.  Using 100 cycle on off times will provide a good idea as to the rise and quench time of that filter.  Note there is a spool up time in the filter as well.   Remember we were addressing key clicks not audio.

I already showed you the result of that measurement!

Now lets look at #336 the, the photo you copied from a book.  You seem to understand the concept.  Now replace the red modulation line with a pair of two positive Square waves separated by an off condition.  What do you think it will morph to?  Do it in your head it's simple.  You don't have to draw it just explain it.

That's easy - it will morph to the same diagram but with many more spectral components, each spaced by the modulation frequency. Each spectral component will be continuous even through the period when the modulated carrier is Off.

Steve G3TXQ



Point #1 The calculation of Q.  Your engineering textbook is correct for simple resonant circuits.  It in no way considers complex clustered filters i.e crystal filters.  I was not being facetious when I recommended a tutor regarding crystal filters.  If you would kindly cite the engineering text book that you found the Q formula, I'll be happy to point you to the chapter regarding Dampened oscillations --- ringing and the formulas used to calculate the durations. 

Your assertion that your 1.4Mc crystal filter has a Q of 538 is totally absurd.  A simple way to calculate the Q is to use the shape factor of your 1.4Mc filter which I would expect to be no more than 2/1.  Separate the width of the rise and fall  from 3db to 60db.  For a shape factor of 2/1 the combined skirt width would be 2.4Kc.  Now design a single 1.4Mc series resonant LC circuit that has a skirt width of 2.4Kc at -60db.  Nope, you won't be able to do it.  but if you could you would find the 3db point to be microscopically narrow and the Q in the 10,000 range.

The net of what I'm saying is that you are using a crystal filter to validate your assertions, --- the wrong component.  The formulas are correct --- your experiment is badly flawed.  Use a heterodyne scheme, it will render correct results.

You need to think before doing proof of concept.  Like it or not that filter is ringing through with a 10us gap between 10us square bursts.  Look at your example where you used a .5 second interval.  Note the carrier did in fact go to zero.  That was because the gap exceeded the ringing effect of the filter --- or does that mean there must be some flaw in the formula or maybe the test setup was incorrect.  Formulas look good to me.     

Try an analogy.  You attach a generator to a motor and wire the generator to the motor.  Then spin it up --- voila perpetual motion.  sounds good feels good its exciting but just doesn't keep going --- resistance will stop it.  What kept it going?  Stored inertial energy ---- sorta like RINGING.  Nothing is free not even carrier Steve.

If your engineering or physics book doesn't have the chapter regarding dampening oscillations and spin up time calculations, let me know and I'll send a copy of the chapter.  You actually can do the rough calculations using a theoretical series resonant LC circuit.   


 

Have a good day Steve.  Kindest regards Jim       


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 11, 2015, 11:28:09 AM
Jim,

Are you going to answer my questions? If we can agree the answers we don't need to debate Q definitions.

Look at my 0.5 second interval results - there is a delay (not ringing) common to both the leading and the trailing edges; but the "ringing" drops significantly in the 100uSec Off period of the keyed signal. Why do you keep talking about 10uSec gaps - a 5kHz square wave is 100uSec On followed by 100uSec Off?

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K9AXN on September 11, 2015, 02:31:28 PM
Steve,

I answered all of your questions and because I don't agree you assume they are unanswered. 

The answer to your questions lies in the configuration of your test setup using a crystal filter.  Compounding the problem is your gross misunderstanding of complex filter technology.

The use of and your expectations of that filter are the fundamental error in your test setup and the genesis of your bizarrely incorrect assertions.  Using a crystal filter the way you have is patently invalid.   

I suggested using the heterodyne method which will eliminate any filter ringing from the equation.  It is apparent that you won't use it out for fear it will disprove your theory; which it will. 

I also explained a rough way to calculate the Q of a crystal filter.

That being said Steve, I wish you well and have a good day.  Until you choose to use an valid test setup there is no point in continuing this conversation. 

Kindest regards Jim

     


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 11, 2015, 02:52:23 PM
Jim,

This evening I've carefully measured again the leading and trailing edge responses of the filter. I've also looked at the carrier output from the filter with a slower, 2.5kHz, keying rate - constant amplitude with no visible modulation! It's absolutely clear from the filter's edge responses that its ringing cannot completely "fill" a 100uSec "gap" of the carrier input, let alone a 200uSec gap - there would be an amplitude variation of at least 20% during the "gap". So I'm sorry, but filter ringing cannot explain the constant amplitude carrier.

You've not answered my questions directly, so I can only assume you are avoiding them for some reason; therefore I see little point in continuing the debate.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K9AXN on September 11, 2015, 04:12:19 PM
Steve,

Use the heterodyne configuration --- get the filter out of the equation.  Remember the carrier did cease after 400us.  The quench period was 400us plenty of time to ring through a 100us gap.  What happened to the carrier after the 400us quench?  The carrier was gone?

Please explain that.  Just do the experiment correctly.

A good day to you and good luck either way.

Kindest regards Jim


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 12, 2015, 01:30:50 AM
Jim,

I see that you still choose not to answer my questions, so let's see if you will answer one of your own.

Let me remind you from your #383:

(http://www.karinya.net/g3txq/temp/sidebands/am1_sm.png)

Quote
Now lets look at #336 the, the photo you copied from a book.  You seem to understand the concept.  Now replace the red modulation line with a pair of two positive Square waves separated by an off condition.  What do you think it will morph to?  Do it in your head it's simple.  You don't have to draw it just explain it.

I'm not prepared to discuss filters further until you can answer your own question; it goes to the heart of the debate, and will reveal how much you understand about modulation theory.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K9AXN on September 12, 2015, 08:45:02 AM
Good morning Steve,

Just set the test up using the Heterodyne technique.  This will remove any ringing effects and with the results you can either learn something from it or if I'm wrong you can tell me how silly I was and I'll have learned something.

Been an interesting discussion from all the participants.  A whole lot better than talking about line cords and what you had for breakfast.

I did learn some things though, that would ordinarily bring attention to my age.  When I first saw the drawing of the "0 and 180 degree switching mixer", I thought "What the hell is this thing".  Had to stop for a moment to see it was what was called a Balanced mixer for the last 60+ years.  I'm a creature o the 60's --- graduated in 1964 and until this thread never heard of a 0/180 degree switching  mixer.  Kinda wonder what a Double balanced mixer will be called --- maybe a Grand High Exalted Mysterious device used as a stripper --- of side bands that is --- just jest.

On the lighter side, Steve.  Mathematical formulas are descriptions of very disciplined thought and are intolerant of the presence of external influences. 

A good example is the formula used to extract the pure reactance of an inductor 6.28xfxl.  In the real world it's correct but from a purely theoretical point of view it is incorrect because by inference it is used to calculate the pure reactance of an inductor disregarding all other influences.

Have some fall hang gliding waiting --- best time of year for me so have to say A great day to you and I wish you well.

Kindest regards to you and your family Steve.

Regards Jim

         


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 12, 2015, 08:51:12 AM
Hi Jim,

I did more filter measurements this morning and can readily demonstrate that the ringing from my filter cannot "bridge" a 200uSec gap, nor even a 100uSec gap in the carrier.

But as I said in my last post, I'm not prepared to discuss that further until you answer your own question:

Quote
Now lets look at #336 the, the photo you copied from a book.  You seem to understand the concept.  Now replace the red modulation line with a pair of two positive Square waves separated by an off condition.  What do you think it will morph to?  Do it in your head it's simple.  You don't have to draw it just explain it.

Surely it can't be too difficult to answer your own question; to quote you: "Do it in your head it's simple".

Steve G3TXQ



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: AC7ZN on September 13, 2015, 06:23:44 AM
Keyclicks

The original question in this post was whether poor IMD performance in an amplifier could cause keyclicks.  We can gain insight on this from the spectrum of a keyclick.  This is discussed here.

Modern commercial transceivers don't usually have keyclicks so you may not have heard many CW signals that have them.  But when a signal has bad keyclicks the QRM it causes can be particularly annoying.

Much information on the nature of keyclicks can be gleaned by simply listening to the signal in a receiver:

1.  The clicks sound like clicks (obvious I know, but sometimes things are misnamed).
2.  They occur at the beginning and end of the transmission of a dit or dah (I'll call the dit or dah a code element).. 
3.  The keyclick at the beginning of the code element sounds just like the keyclick at the end (this is usually true but does not have to be--see below).
4.  Tuning off frequency, the CW can be outside the passband of your receiver but you can still hear the clicks.  For a strong signal with bad clicks they can be sometimes heard more than a kHz off frequency.
5.  The clicks occur both above and below the transmitted frequency and appear to be symmetric with frequency.
6.  They grow weaker the further off frequency you go until they drop into the noise floor.  There are no nulls where the click goes away and then comes back as you tune further.  The decay in click strength with frequency off the carrier is monotonic.

You might be able to discern more properties.

I think keyclicks are annoying to many hams because their brains automatically engage and try to decode them just like a telegraph signal.  But because the start and end click sound the same the information of when keydown starts and stops is not there, so the ham brain automatically raises the fact that something is wrong to the ham's conscious level.

It seems incongruent to talk about the spectrum of just a click: a spectrum is a frequency domain thing and is taken over a long time on a signal that is often repetitive so it seems inappropriate to say things like 'here is how the spectrum decays over time' or 'here is the spectrum of the click'.  But in the sixties scientists and engineers realized there is meaning in these statements...signals packed with information vary over time and short-term spectra that change frequencies as the signal varies provide valuable information.  A field of study called harmonic analysis studies these time-shifting spectra and interesting transforms such as the short term Fourier integral and wavelet transforms are used as tools.

We model the click at the beginning of a code element with a unit step function, which goes instantly to one at time zero and stays at one forever.  Our code element eventually goes off so we will discuss how this changes the spectrum of the simple unit step model we are using.

Here is the unit step:
(https://lh4.googleusercontent.com/7OOaqYJyHgweFaCDLx4TGyDEQtE5PXpQjArhskK9bqqJS2XJMTutqNM_tXahhi6daOXF0j8s5HtcTRE=w978-h556)

The Fourier integral transform is used to find the spectrum of the entire unit step.  Here is the Fourier integral that converts a time function in seconds to a frequency function in hertz:

(https://lh4.googleusercontent.com/wamTpRsuFcZ77ROVG-6HK3auc3tMPy-yU88k3_6yKaMh23XkSVWKi3TQbsEuqdvmVDxa0c7HQiCPeTQ=w978-h556)

We multiply the time waveform by the sine and cosine of the frequency we want, and integrate the results to find the magnitude and phase of that frequency component.  Note the infinities in the integral limits: we must integrate the time waveform over all time. But we can simplify:  times before zero are all zero so we can move the lower limit up (as long as we are not interested in really, really low frequencies) to near zero. As we move the upper limit to smaller and smaller times, we find the only thing affected is the energy at or very near zero Hz.  So the 'leftover' portion of the spectrum must be what is making the keyclick.  Here it is, plotted as amplitude magnitude versus frequency.
 
(https://lh6.googleusercontent.com/hjGBjefFTaoHgxm518KmpdJ4IxKOu3Ww1_C3XmM0q0SAYyIXM4lmnjbpjUXAJO14pXb6vq12wG8haI4=w978-h556)

I was surprised that the absolute magnitudes seemed so low, but all the energy in that waveform that gets through your receiver passband is packed into a very short time, so you will get a nice sounding click.  Note that the energy does drop the farther you go off frequency, and that the decay is continuous, as we observed.  When we modulate the carrier with this step, we get the click both above and below the carrier as we hear.  The spectrum at the end of the code segment is identical.

We can see that a low pass filter set to the right frequency will cut out most of this spectrum and that is what is used in the transmitter keyclick filter.

Anything that enhances the rising or falling edge slope will make the clicks worse.  For an amplifier, a really bad case of power supply sag, where the transmission starts out at very high power but quickly decays to a constant value due to droop in the DC supply could in theory enhance a transmission that already is close to having keyclicks.  The power supply droop caused by too-low capacitors in the DC supply would be too slow to make something as fast as a click, but poor (resistive and/or inductive) wiring from the DC supply to the finals might be able to do it.  So I think it may be possible for an amplifier alone to generate keyclicks in unusual cases. 

73,
Glenn AC7ZN
 




 


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 13, 2015, 07:17:49 AM
Glenn, your images don't post up on the forum for me. Could you try another method to repost them?

I was kind of hoping the thread would have died away if I'd stopped posting because I think it got seriously derailed by my initial comments. But it has kept going for several days now...

My comments were based on what W8JI said in this thread and from what I saw when I looked at his analyser plots he described earlier in the thread. They can be found here:

http://www.w8ji.com/occupied_bw_of_cw.htm

I really don't want to kick off the whole debate again but W8JI presents information in a strange way that is easy to misinterpret. For example, I interpreted from his comments and the above web page that he thinks the classic definition of 99% occupied bandwidth is independent of keying speed and it is only affected by the keying waveform.

Hence all the stuff I posted up about how to work out the (average) power of each spectral component in a continuous stream of dits using Fourier analysis and also the practical examples I gave using Fourier synthesis and some test gear. Occupied bandwidth is also known as spectral occupancy and I don't think W8JI is measuring this correctly in his plots of 99% OBW.

However, I do think that what/all he is actually trying to say is that the range of frequencies you hear clicks over can be limited by shaping the keying waveform. If he reworded his page to lose the reference to 99% OBW and also explained (better) what he was really trying to portray then I think it would save a lot of confusion.

Whilst his analyser plots on his page are of no use for displaying 99% OBW they do give some indication of the influence of the keying waveform. I think the whole page needs a makeover to make it less confusing.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: AC7ZN on September 13, 2015, 07:31:45 AM
Thanks, but I don't know.  I've tried my own website, photobucket, and finally an oblique procedure using Google drive before I was able to see the images myself.  I almost gave up.  Can anyone help me with an absolute surefire method of making these display?  I personally feel eham should provide online space for these images because this seems to be an ongoing problem...

Thanks,
Glenn


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 13, 2015, 08:03:25 AM
I'm using photobucket and it seems to work for me OK :)

I  click directly on 'IMG' to the right of the image on photobucket where it says SHARE LINKS and it then says 'copied' the moment I click on the IMG logo

I then paste the clipboard text directly here

See below for an image of OOK sent at 1 dit per second. This has a 99% OBW of about 42Hz even with a square keying waveform.

But obviously, this waveform will click very badly over a very wide frequency range when listened to by a typical receiver because of the square keying waveform.  I've posted up this example to show that measuring a 99% OBW of 42Hz doesn't tell us much about keyclick performance.

Hence the reason I think W8JI needs to reword his analyser plots to lose the confusing reference to 99% OBW as a means to indicate keyclick performance for different keying speeds.

His plots don't actually measure 99% OBW properly anyway because his analyser display trace is losing detail/resolution in the critical area of the carrier. So they are worthless in terms of displaying 99% OBW anyway.

(http://i1227.photobucket.com/albums/ee439/G0HZU/OOK_1Hz.gif) (http://s1227.photobucket.com/user/G0HZU/media/OOK_1Hz.gif.html)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 13, 2015, 10:46:36 AM
I don't think there was any doubt that a square wave is composed of an infinite number of harmonics... and that the energy would occupy a vast spectrum.

The point that I see that Tom made was that the key clicks were evident during the transitional phase of the waveform.  I think anyone who has observed key clicks from an offending CW signal has noted that the interference is present to other stations during that time period. All of the math and arguments don't refute that observation.

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 13, 2015, 12:04:28 PM
Here's a nice online book about digital signal processing.

http://www.dspguide.com/

Chapter 8 discusses the DFT in depth.

http://www.dspguide.com/ch8.htm


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 13, 2015, 12:07:08 PM
I've not thought about this much, but on the question of "does the keying speed affect the 90% OBW", it seems to me important to distinguish between the theoretical square-wave keying case and the situation where there is some shaping.

If we take G0HZU's spectrum - which I assume is close to square-wave keying - when we increase the keying rate the spectral lines will move further apart but retain the same relative amplitudes. Clearly, the bandwidth containing any particular power percentage will vary directly with keying rate.

But what happens if we shape the spectrum of the baseband keying signal? Imagine we could shape it so that the amplitude of the spectral components reduced linearly with increasing frequency, and were cut off totally beyond some upper limit. When we alter the keying rate of that signal the spectral lines cluster or separate, but always within the same, fixed, frequency-domain triangular envelope. Wouldn't the bandwidth containing a particular power percentage now be largely independent of keying rate?

I guess I'm saying that we need to be careful that what we deduce from the hard-keying case - or something close to it - is still valid for more practical cases.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on September 13, 2015, 12:40:20 PM
You are quite right, shaping will change the amplitude distribution considerably.

If the optimum erfc(t) keying shape is used, the sideband amplitudes fall off very rapidly, with increasing order
and the occupied, expressed as the band containing e.g. 99% of the total power, and necessary, expressed as the amount of spectrum needed to transmit a certain keying speed with a certain quality, bandwidths become close to each other.

Without any shaping, the higher-order sidebands extend very far, and objectionable interference due to key clicks is created far outside the 99% "occupied bandwidth" and very far outside the necessary bandwidth of the A1A emission which has been long established by the CCIR to be 3 or 5 times the keying speed in Bauds.

Outside this "main lobe" formed by the first order modulation sidebands and their first 3 or 5 odd harmonics, there should ideally be no sideband energy. The width of this necessary frequency band is fixed by the keying speed, and the need for recreating the shape of the keying elements.

73/
Karl-Arne
SM0AOM



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 13, 2015, 01:10:16 PM
I think the 99% OBW is a function of both the keying speed and the key shaping. But if you fix the shaping and then reduce the keying speed by say, an order of magnitude then the 99% OBW will go down with keying speed.

But I think the important point is that the 99% OBW of a transmitter is NOT a good thing to measure if you are really interested in  how far away you can detect clicks on a receiver.

If W8JI deletes the references to 99% OBW on his web page and removes the 99% OBW marker text from his plots and just says:

"Look at the general shape of each spectrum plot for 5-40dps to see that the spectrum is limited by the keying waveform in the radio"

Then I think it will be less confusing. Otherwise, he will attract criticism for his poor use of his test equipment in his attempts to measure 99% OBW properly and also for his conclusions wrt 99% OBW vs keying speed..

It should be obvious that if he reduced the sending speed to 0.1dps on a radio transmitter that the 99% OBW of the transmission should go down. But the click phenomenon in a receiver will be the same in terms of how the clicks (and their level) sound although the click rate will obviously go down.

So I think the web page would ideally need some clarification wrt this.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 13, 2015, 02:18:38 PM
A quick test shows that occupied bandwidth is not a function of keying speed. It is only a function of the rise/fall time of the CW dit.

First set of four images are with an approximately 8 millisecond rise/fall time. 48 wpm, 24 wpm and 12 wpm.

Second set of four images are with an approximately 4 millisecond rise/fall time. 48 wpm, 24 wpm, 12 wpm.

The keying shape is the optimum raised cosine.

The frequency resolution is 300 Hz per division in all spectrograms.

(http://www.w6rz.net/risetime8.png)

(http://www.w6rz.net/occ48n.png)

(http://www.w6rz.net/occ24n.png)

(http://www.w6rz.net/occ12n.png)

(http://www.w6rz.net/risetime4.png)

(http://www.w6rz.net/occ48w.png)

(http://www.w6rz.net/occ24w.png)

(http://www.w6rz.net/occ12w.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 13, 2015, 02:44:37 PM
A quick test shows that occupied bandwidth is not a function of keying speed. It is only a function of the rise/fall time of the CW dit.
 

I'd hope so...  a square wave is comprised by an in an infinite number summation of sinusoidal wave forms.  The occupied bandwidth of the resultant side bands would, at least in theory, would go on forever.  Obviously, in practice, the tank Q and antenna bandwidth would limit the product levels and their spectrum.

but, as Tom noted:

A CB amplifier without bias will square the envelope rise at the bottom That is due to the lack of bias in a Class C transistor RF amplifier.[/b]

Thus, how much harmonic energy is produced at the lower drive levels, as compared to distortion products at the peak of the RF crest?

To quote further:


The average power in the sidebands goes up with keying rate, but the signal sideband width stays the same (since it is set by the rise and fall shapes). Faster CW with clicks is more annoying because the clicks come more often, but the distance we tune off and are bothered by the clicks remains the same as long as the shape and duration of the rise and fall stay the same.

ALC induced clicks are reduced with higher speed, which is another irony. This is because the ALC stays more constant with higher speeds. At slow speeds, the ALC can drop to zero at times, and the next element can have an exaggerated rise with overshoot.


That pretty much is in line with W6RZ's observation.

Pete  k1zjh


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 13, 2015, 02:56:56 PM
Quote
A quick test shows that occupied bandwidth is not a function of keying speed. It is only a function of the rise/fall time of the CW dit.

First set of four images are with an approximately 8 millisecond rise/fall time. 48 wpm, 24 wpm and 12 wpm.

Second set of four images are with an approximately 4 millisecond rise/fall time. 48 wpm, 24 wpm, 12 wpm.

The keying shape is the optimum raised cosine.

The frequency resolution is 300 Hz per division in all spectrograms.

Oh for goodness sake....

Turn the keying rate down to 1 dit per second and measure the 99% OBW as per the way it is supposed to be measured.

Has it gone down?

To save you the time just go back and look at my plot using the Agilent signal analyser. It shows a 99% OBW of about 42Hz for 1 Hz OOK.




Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 13, 2015, 03:11:23 PM
Occupied bandwidth is the same at 1 wpm.

8 millisecond rise time.

(http://www.w6rz.net/occ2n.png)

4 millisecond rise time.

(http://www.w6rz.net/occ2w.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 13, 2015, 03:21:42 PM
Quote
Occupied bandwidth is the same at 1 wpm.

Measure the 99% OBW. W8JI was trying to measure and report the 99% OBW to prove that it is independent of keying speed.  

The definition for 99% OBW is all over the internet and it is a function of keying speed and the keying waveform.

Would you like me to post up the equations for it?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: AC7ZN on September 13, 2015, 03:22:36 PM
licks

Thanks to G0HZU for helping with the images.  This post is now stale (it was stale when I posted it) so you don't have to read it, but if the images are not coming through please let me know.

Thanks, Glenn

The original question in this post was whether poor IMD performance in an amplifier could cause keyclicks.  We can gain insight on this from the spectrum of a keyclick.  This is discussed here.

Modern commercial transceivers don't usually have keyclicks so you may not have heard many CW signals that have them.  But when a signal has bad keyclicks the QRM it causes can be particularly annoying.

Much information on the nature of keyclicks can be gleaned by simply listening to the signal in a receiver:

1.  The clicks sound like clicks (obvious I know, but sometimes things are misnamed).
2.  They occur at the beginning and end of the transmission of a dit or dah (I'll call the dit or dah a code element)..
3.  The keyclick at the beginning of the code element sounds just like the keyclick at the end (this is usually true but does not have to be--see below).
4.  Tuning off frequency, the CW can be outside the passband of your receiver but you can still hear the clicks.  For a strong signal with bad clicks they can be sometimes heard more than a kHz off frequency.
5.  The clicks occur both above and below the transmitted frequency and appear to be symmetric with frequency.
6.  They grow weaker the further off frequency you go until they drop into the noise floor.  There are no nulls where the click goes away and then comes back as you tune further.  The decay in click strength with frequency off the carrier is monotonic.

You might be able to discern more properties.

I think keyclicks are annoying to many hams because their brains automatically engage and try to decode them just like a telegraph signal.  But because the start and end click sound the same the information of when keydown starts and stops is not there, so the ham brain automatically raises the fact that something is wrong to the ham's conscious level.

It seems incongruent to talk about the spectrum of just a click: a spectrum is a frequency domain thing and is taken over a long time on a signal that is often repetitive so it seems inappropriate to say things like 'here is how the spectrum decays over time' or 'here is the spectrum of the click'.  But in the sixties scientists and engineers realized there is meaning in these statements...signals packed with information vary over time and short-term spectra that change frequencies as the signal varies provide valuable information.  A field of study called harmonic analysis studies these time-shifting spectra and interesting transforms such as the short term Fourier integral and wavelet transforms are used as tools.

We model the click at the beginning of a code element with a unit step function, which goes instantly to one at time zero and stays at one forever.  Our code element eventually goes off so we will discuss how this changes the spectrum of the simple unit step model we are using.

Here is the unit step:
(http://i1081.photobucket.com/albums/j357/glennbdixon/unitstep_zpszznwkyab.jpg)

The Fourier integral transform is used to find the spectrum of the entire unit step.  Here is the Fourier integral that converts a time function in seconds to a frequency function in hertz:

(http://i1081.photobucket.com/albums/j357/glennbdixon/inverse%20fourier%20integral_zps0mcj1icf.jpg)

We multiply the time waveform by the sine and cosine of the frequency we want, and integrate the results to find the magnitude and phase of that frequency component.  Note the infinities in the integral limits: we must integrate the time waveform over all time. But we can simplify:  times before zero are all zero so we can move the lower limit up (as long as we are not interested in really, really low frequencies) to near zero. As we move the upper limit to smaller and smaller times, we find the only thing affected is the energy at or very near zero Hz.  So the 'leftover' portion of the spectrum must be what is making the keyclick.  Here it is, plotted as amplitude magnitude versus frequency.
 
(http://i1081.photobucket.com/albums/j357/glennbdixon/click-spectrum_zpsxbcmhexn.jpg)

Plotted in dB:

(http://i1081.photobucket.com/albums/j357/glennbdixon/click-db_zpszdnrfk8m.jpg)

I was surprised that the absolute magnitudes seemed so low, but all the energy in that waveform that gets through your receiver passband is packed into a very short time, so you will get a nice sounding click.  Note that the energy does drop the farther you go off frequency, and that the decay is continuous, as we observed.  When we modulate the carrier with this step, we get the click both above and below the carrier as we hear.  The spectrum at the end of the code segment is identical.

We can see that a low pass filter set to the right frequency will cut out most of this spectrum and that is what is used in the transmitter keyclick filter.

Anything that enhances the rising or falling edge slope will make the clicks worse.  For an amplifier, a really bad case of power supply sag, where the transmission starts out at very high power but quickly decays to a constant value due to droop in the DC supply could in theory enhance a transmission that already is close to having keyclicks.  The power supply droop caused by too-low capacitors in the DC supply would be too slow to make something as fast as a click, but poor (resistive and/or inductive) wiring from the DC supply to the finals might be able to do it.  So I think it may be possible for an amplifier alone to generate keyclicks in unusual cases.

73,
Glenn AC7ZN
 




Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 13, 2015, 03:42:03 PM

Just for the record, here are a couple of images showing the definition of 99% OBW  and the equation from the ITU for A1A showing how to work it out based on both baud rate and risetime .




(http://i1227.photobucket.com/albums/ee439/G0HZU/obw_99a.gif) (http://s1227.photobucket.com/user/G0HZU/media/obw_99a.gif.html)

(http://i1227.photobucket.com/albums/ee439/G0HZU/OBW_99.gif) (http://s1227.photobucket.com/user/G0HZU/media/OBW_99.gif.html)


Does this help?






Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 13, 2015, 03:50:51 PM
That equation is based on relative rise-time. In other words, as a percentage of the dit time. As the keying rate increases, the rise-time gets shorter.

I'm showing you constant rise time. The rise-time remains the same at different keying rates.

Please consider for a moment what I'm showing you instead of immediately attacking it.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 13, 2015, 04:11:58 PM
But if you fix the rise/fall time then I can just lower the keying rate as low as I like to keep lowering the 99% OBW.

The relative build up coefficient will then fall to nearly zero and the equation simplifies to

99% OBW = 19 * Baud rate





Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 13, 2015, 04:29:08 PM
Way back in the thread I mentioned that I did a quick and dirty excel spreadsheet to predict %OBW for a square key waveform. I think/hope it is accurate.

Maybe this will be interesting or useful.

It looks at the power sharing of a 100W OOK transmitter wrt the modulation and harmonics etc.

You can see by the 19th harmonic we are at 98.99% OBW. See the red arrow to show this.

This closely agrees with the '19' in the ITU equation.

(http://i1227.photobucket.com/albums/ee439/G0HZU/OBW_square.gif) (http://s1227.photobucket.com/user/G0HZU/media/OBW_square.gif.html)





So for 1Hz OOK the baud rate is 2 so the 99% OBW for square keying is

99% OBW = 19*2 = 38Hz according to the ITU equation.

This agrees with my spreadsheet in the /Hz column on the far right (arrowed in red).
It predicts the 98.99% OBW is 38Hz.





Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 13, 2015, 06:08:39 PM
The spreadsheet also predicts the levels of each modulation sideband in terms of dBc.

If you look at the 1Hz OOK image below you can see that the dBc levels for each harmonic agree very closely with the prediction in the spreadsheet. See the Excel column with dBc data.

eg the 19th harmonic is about -30dBc in the plot and the spreadsheet predicts -29.5dBc.


(http://i1227.photobucket.com/albums/ee439/G0HZU/1hz_00k.gif) (http://s1227.photobucket.com/user/G0HZU/media/1hz_00k.gif.html)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on September 14, 2015, 03:12:31 AM
Looking back some 400 posts, it appears that the "bottom line" of the argumentation is that the international radio engineering community and the ITU are all wrong regarding the relations between
rise time, keying rate and the bandwidth expressions for Morse emissions.

It is certainly "your call", but I do find it somewhat ineffective trying to convince especially the ITU of its shortcomings using the pages of eHam.net.
It is also highly improbable that any representative of the FCC, the ITU Radiocommunications Assembly or its Secretariat is following this debate and comes up with proposals to change its views on its own.

Instead, some guidance of a potentially more efficient way to accomplish these goals is hereby respectfully submitted:

It starts with first closely studying this document http://www.itu.int/dms_pub/itu-r/opb/res/R-RES-R.1-6-2012-PDF-E.pdf,
and then contacting the Office of Engineering and Technology at the FCC presenting your arguments.

If the FCC is convinced and finds your arguments worth-while, the FCC can make up a proposal for re-writing the affected parts of the Radio Regulations
(Appendix 1) and the related ITU-R Recommendations.
The FCC then submits this proposal to the ITU Secretariat through the US Department of State.
This is because only Governments are full voting members of the ITU.

If sufficient interest about this matter can be aroused internationally, the ITU Secretariat may work out Terms of Reference for a Working Party (WP) within the
ITU-R Study Group 1, who is dealing with spectrum efficiency and spectrum management matters.
Its objective will be to rewrite the relevant parts of ITU-R Recommendation SM.328.
All this is subject to the final approval of the Director of the ITU.

The designated members of this WP may then meet at their convenience, discussing the matters about the necessary, occupied and key-click bandwidths of Morse emissions
to their "heart's content" until a consensus within the WP has been reached about how the new wording of the relevant texts in the Radio Regulations and SM.328 should be.
Their findings have to be summarized in an input document to a future meeting of the ITU-R Study Group 1, where it will become an agenda item.

After a through discussion in the sub-, sub-sub and sub-sub-sub groups formed at this meeting, the proposal ultimately reaches the Plenary where it may be approved.
When finally approved by consensus or vote, the new text is finally incorporated in the ITU-R Recommendations, and by reference in the Radio Regulations.

A reasonable time frame for the whole process is in the order of 3-7 years, but if you are convinced that your views are right, it might be "worth waiting for".

73/
Karl-Arne
SM0AOM




Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 14, 2015, 07:05:10 AM
It's interesting to take the ITU equation and consider its derivative with respect to keying rate at various values of alpha.

W6RZ made the important observation that alpha is the "relative" rise time; for a fixed rise time - the situation in most radios I would guess - alpha is proportional to the keying rate B.

When the rise time is very short or the keying rate is very slow, alpha tends to zero; the occupied bandwidth then becomes directly proportional to the keying rate.

At smaller values of alpha the occupied bandwidth is less dependent on the keying rate; however for practical values of alpha it is never completely independent.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K9AXN on September 14, 2015, 08:16:47 AM
Perfect explanation Ron!

Kindest regards Jim



A quick test shows that occupied bandwidth is not a function of keying speed. It is only a function of the rise/fall time of the CW dit.

First set of four images are with an approximately 8 millisecond rise/fall time. 48 wpm, 24 wpm and 12 wpm.

Second set of four images are with an approximately 4 millisecond rise/fall time. 48 wpm, 24 wpm, 12 wpm.

The keying shape is the optimum raised cosine.

The frequency resolution is 300 Hz per division in all spectrograms.

(http://www.w6rz.net/risetime8.png)

(http://www.w6rz.net/occ48n.png)

(http://www.w6rz.net/occ24n.png)

(http://www.w6rz.net/occ12n.png)

(http://www.w6rz.net/risetime4.png)

(http://www.w6rz.net/occ48w.png)

(http://www.w6rz.net/occ24w.png)

(http://www.w6rz.net/occ12w.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 14, 2015, 08:58:34 AM
Is there actually any inconsistency between the ITU equation and W6RZ's plots? Taking W6RZ's rise time of 8mSec, the ITU equation predicts a tiny 2% change in occupied bandwidth when you drop the keying rate by a factor 4 from 48wpm to 12wpm - a change that it would be very difficult to resolve on those plots.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 14, 2015, 11:33:10 AM
Is there actually any inconsistency between the ITU equation and W6RZ's plots? Taking W6RZ's rise time of 8mSec, the ITU equation predicts a tiny 2% change in occupied bandwidth when you drop the keying rate by a factor 4 from 48wpm to 12wpm - a change that it would be very difficult to resolve on those plots.

Steve G3TXQ

But the plots don't prove that 99% OBW won't fall if you lower the keying rate down to a much lower level.

If people think I'm being overly pedantic here then fine. But if you work out the 99% OBW of 5dps it will be a lot less than W8JI's plots show even with no key shaping.

Not only is he (obviously) measuring 99% OBW incorrectly with his analyser he is using the WRONG measurement anyway for what he is trying to demonatrate because 99% OBW will be affected by keying speed at low key rates for a fixed keying waveform.

Like I said earlier, this could all be put right if he simply removed the markers and his tabulated data of 99% OBW and just said "look at the similar shape of the spectrum for each dps setting and you can see that the keying waveform in this radio limits the range of frequencies you may hear clicks over"

If he sticks with the claim that 99% OBW is independent of keying speed then it will just cause confusion.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 14, 2015, 11:34:09 AM
This chart may be of interest - it illustrates what I was trying to say earlier. It shows on the vertical scale the relative bandwidths predicted by the ITU equation for a range of keying speeds and 4 different rise times:

(http://www.karinya.net/g3txq/temp/sidebands/itu.png)

Clearly, for some rise times, the bandwidth is almost constant over a wide range of common keying speeds. If you do measurements in that region you will correctly deduce that bandwidth is substantially independent of keying speed; do the measurements in another region of the chart and the conclusion might be different.

It would be interesting to see if plots from W6RZ taken in those other regions support the ITU equation or not.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 14, 2015, 11:48:42 AM
Maybe if you imagine the keying waveform with a raised cosine at the start (and equivalent at the end) and have them touching each other you will have very low bandwidth similar to 100% AM with a sine wave.

But then imagine you can stretch these apart like a trombone and produce a longer and longer 'square' period between the cosines then apply this to a mixer LO port then you will see the 99% OBW of the OOK transmission start to go down as you stretch the 'trombone' more and more. So this proves that 99% OBW can not be defined as being independent of keying speed.

If it didn't go down then what would be the point in using ultra slow morse speeds? The required bandwidth of the transmission goes DOWN so you can achieve a better S/N ratio at lower sending speeds and get more range by using a narrower filter because the power is confined inside a smaller bandwidth.

Can we just let this go now because it really isn't a big deal anyway?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 14, 2015, 12:01:48 PM


...and so I repeat my earlier quote....

Quote
GoHZU:  Like I said earlier, this could all be put right if he simply removed the markers and his tabulated data of 99% OBW and just said "look at the similar shape of the spectrum for each dps setting and you can see that the keying waveform in this radio limits the range of frequencies you may hear clicks over"

If he sticks with the claim that 99% OBW is independent of keying speed then it will just cause confusion.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 14, 2015, 04:23:31 PM
Here are the results at 2 milliseconds. Pretty much the same. It should be noted that the ITU specifically states that the equation is empirical and that the error is greater when alpha is less than 0.02. It doesn't seem prudent to use an empirical equation to gain any vast insight.

BTW, the document that contains that equation can be downloaded for free here:

https://www.itu.int/rec/R-REC-SM.328-11-200605-I/en

As for the spectrograms, this is the tool that I have. In my opinion, waterfalls are a very intuitive way to display spectra. I do have a Rigol DSA 815, but never bothered to purchase (or hack) the automatic measurement suite.

The frequency scale is not changing from capture to capture, it's always 300 Hz per division. The only parameter that's changing is the keying rate. I've condensed the 12, 24 and 48 wpm captures into one picture so that they can be "eyeballed" a little easier. The last capture is 2.4 wpm.

(http://www.w6rz.net/2millirisetime.png)

(http://www.w6rz.net/2milliwaterfall.png)

(http://www.w6rz.net/2millislow.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 14, 2015, 05:27:33 PM
Quote
It should be noted that the ITU specifically states that the equation is empirical and that the error is greater when alpha is less than 0.02. It doesn't seem prudent to use an empirical equation to gain any vast insight.

Oh come on... :)  you are just blowing smoke now to cover for the fact that you really don't want to admit that the 99% OBW WILL go down if you just keep reducing the keying rate...

You don't 'need' the equation to know that the significant spectral components move in closer to the carrier as you reduce the keying rate. It's basic modulation theory.

Just look on your GNU plots. Do the significant spectral components move closer to the centre of the plot as you reduce the keying rate?

That means the bandwidth where 99% of the power is contained is slowly getting less and less.
Keep reducing the keying rate and the 99% OBW will eventually start to halve with every halving of the keying rate if you stick with a fixed keying waveform.

However, if you know I'm wrong then please show me why the 99% OBW is frozen for all keying speeds no matter how slow you key the transmitter?

Please note that I'm not talking about key clicks here. I'm talking about measuring the 99% OBW of a stream of OOK dits. I don't think the 99% OBW is a wise thing to use as a figure of merit for keyclick performance.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 14, 2015, 05:42:43 PM
Just showing my results. The waterfalls are the result of DFTs, just like any other piece of modern test equipment used for spectrum analysis.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 14, 2015, 05:51:33 PM
You don't 'need' test equipment for this type of analysis.

If you whistle into an AM transmitter and slowly reduce the frequency of the whistle does the bandwidth of the transmission go down as you lower the whistle frequency?

If the answer is YES then why can't you agree the same will happen with OOK signals as you lower the keying rate? Keep lowering the OOK rate and the 99% OBW will eventually halve for every halving of the keying rate.

Why do you need test gear for this?









Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 14, 2015, 05:55:38 PM
Dude, I'm not going to argue with you. At this point, you just want to make me out to be wrong. You don't want to learn from me. What is my motivation to interact with you?


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G0HZU on September 14, 2015, 06:11:54 PM
Quote
At this point, you just want to make me out to be wrong.

No, I'm just trying to post up basic and easily proven theory.

Go back to my post #404

Quote
G0HZU: I think the 99% OBW is a function of both the keying speed and the key shaping. But if you fix the shaping and then reduce the keying speed by say, an order of magnitude then the 99% OBW will go down with keying speed.

An order of magnitude is a factor of 10. If you optimise the keying waveform for the higher speed and then go down by a factor of 10 or more in keying speed then you will begin to see the 99% OBW go down more and more noticeably. Go down another factor of 10 and it will go down a lot. Eg it will eventually halve for every halving of the keying speed.







Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K9AXN on September 14, 2015, 07:39:57 PM
In #414 you used a square wave and consistent keying to do the calculation, is that correct?

Thanks --- Kindest regards Jim


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: K6JH on September 14, 2015, 08:04:55 PM
Why don't we move the discussion on to the second part of the thread title, i.e. amplifier non linearity?

We've pretty much established that there are additional spectra present during keying and un-keying, based on the rate and the slope of the waveform. (Via 430 posts - is this an e-ham record for much to-do about nothing?)

Now what happens when you put this through an amp with "good" or "bad" IMD performance? How much worse does it get?

Seeing as how the click bandwidth with high rate & sharp edges seems to be approaching that of SSB speech, with many more than two tones present, I bet it ain't good!


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: KB8E on September 15, 2015, 03:06:39 AM
Agree with K6JH. I'm curious about the impact of a non-linear amplifier on the spectrum of a properly shaped cw waveform. As explained by me in a much earlier post and demonstrated in a following post, the key clicks occur at the keying transitions because an off-channel receiver only picks up part of the overall spectrum. The reason a shaped cw waveform can have a bandwidth that appears independent of keying speed is that while a raw (square) cw bandwidth varies directly with keying speed, shaping acts like a bandpass filter. If the shaping is significant and fixed, its bandwidth dominates above a certain keying speed and effectively determines the overall bandwidth.

Hope this clarifies rather than confuses things.

Sam


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SM0AOM on September 15, 2015, 03:13:12 AM
This was briefly discussed at the beginning of the thread.

In order to get a quantative picture of how much clicks that are caused
by a non-linear amplifier fed by a "perfect" keying envelope, we need
some way of converting the 3 and 5 order IM performance at a
given power level into a corresponding transfer function.

Depending on the shape of this transfer function, discontinuities will have different influences.
Peak compression has a comparatively benign influence, it is estimated that a 10% gain compression is
about equivalent to -25 dBc IM3 distorsion on a two-tone signal.

Sharp bends or discontinuities in the transfer function will cause more higher-order sidebands, so cross-over or bias distorsion at the beginning of
the keying envelope is probably much worse than gain compression alone.

To find out exactly how much worse a given amount of IMD makes the key-click suppression,
it will likely be necessary to make a piece-wise linear approximation of the transfer function and
analyse the Fourier coefficients of the parts separately, and then create the total using the superposition theorem.

An alternative would be to first create a polynominal approximation of the transfer function, and then use a series expansion to find the coefficients for the
higher-order keying sideband components.

I have not seen any tabulated data for typical transfer functions for power amplifiers using different devices and classes of operation, but they may be synthesized out of the
tube or transistor characteristic curves.

73/
Karl-Arne
SM0AOM









Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 15, 2015, 03:58:52 AM
If you want to get a qualitative appreciation, here's something you can try.

Capture a sample of a shaped, keyed signal in Audacity. Install the Crossover plug-in for Audacity - it "simulates the distortion that happens in class B and AB power amps when the signal crosses zero". Look at the spectrum using Audacity's spectrum or waterfall display. Add some crossover distortion, then look at the spectrum again.

It's very revealing!

Although the Crossover plug-in adjustments do not use parameters we are familiar with, you can get a rough idea of calibration by applying the same crossover settings to a two-tone signal and looking at the level of the 3rd order intermods.

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 15, 2015, 05:09:56 AM
Just to illustrate that last posting:

This is the original keying waveform:

(http://www.karinya.net/g3txq/temp/sidebands/aud_original.png)

Here's a close up of the leading edge:

(http://www.karinya.net/g3txq/temp/sidebands/aud_undist.png)

and here's its spectrum:

(http://www.karinya.net/g3txq/temp/sidebands/aud_spec_undist.png)

Now add crossover distortion; here's the leading edge - you can just about see the distortion:

(http://www.karinya.net/g3txq/temp/sidebands/aud_distorted.png)

And here's the spectrum:

(http://www.karinya.net/g3txq/temp/sidebands/aud_spec_dist.png)

Checking afterwards, the same degree of crossover distortion applied to a 2-tone signal produced 3rd order products 30dB below PEP.

Perhaps not very scientific, but it seems to indicate that crossover distortion could produce significant artefacts.

Steve G3TXQ



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W1BR on September 15, 2015, 09:00:53 AM
Steve

I suspect that key clicks are more complex than a two signal. But, I understand you point, and it shows that the cross over distortion is real.  The amount of energy in the side bands, and the occupied bandwidth, are subject to the waveform shaping.  While  a true square wave modulated signal has side band energy going out to infinity.

My question was based on a using a Class C amp (CB, transistorized, no bias) with a CW transmitter that perhaps had a reasonably shaped waveform.  I was curious as to how much the added distortion might affect the occupied BW, as far as noticeable artifacts (key clicks) to another station--in practical terms. I realize my question is rather quantitative, since it lacks specifics. It was more practical than theoretical.

Pete


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 15, 2015, 09:26:15 AM
Pete,

With the tools I have I can't easily add class C distortion; the Audacity Crossover plug in only allows a limited range of distortion levels. I'll look to see if it might be possible to achieve the required effect by slicing and reconstructing the waveform.

What I didn't say in the earlier postings is that the effect of the crossover distortion was predominantly to increase the 3rd, 5th, 7th etc harmonic levels of the carrier. These components would presumably be eliminated by the output stage filtering. Looking in much more detail at artefacts close to the carrier which would not be eliminated by harmonic filters, the crossover distortion adds very little when the rise time is slow (10mSec), and a modest amount when the rise time is fast (2mSec).

I guess we're left to speculate what class C would do ;)

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: DL8OV on September 15, 2015, 10:20:38 AM
This experiment has already done (many posts ago)

http://www.eham.net/ehamforum/smf/index.php/topic,104806.msg864042.html#msg864042

I supplied an OOK signal to an amplifier and varied the bias until the amplifier was operating at Class C, at which point the key clicks were described as "terrifying" by the guy with the receiver a few hundred metres away.

Peter DL8OV


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 15, 2015, 01:40:20 PM
Peter,

Do you know what the rise and fall times of the keying were before being applied to the output stage?

Thanks,
Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: JS6TMW on September 15, 2015, 03:38:27 PM
This experiment has already done (many posts ago)

http://www.eham.net/ehamforum/smf/index.php/topic,104806.msg864042.html#msg864042

I supplied an OOK signal to an amplifier and varied the bias until the amplifier was operating at Class C, at which point the key clicks were described as "terrifying" by the guy with the receiver a few hundred metres away.

Peter DL8OV

And thanks for reminding everyone, Peter. It was also my original query on a different thread that led to this vast outpouring of knowledge. Ironically, the transistor amp in question self-destructed before I could incorporate biasing, and even before I could generate a single dit or dah.

Steve in Okinawa JS6TMW/QRP


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: DL8OV on September 15, 2015, 10:36:47 PM
Rise time during my experiment was 4,5 milliseconds and I used a raised cosine. Output power was only at QRP levels and the receiver was a Yaesu portable rig handled by another ham a few hundred metres away in the car park. Before the start of the experiment I reduced my output power until he saw S9 so that overload was not an issue.

Based on the results of this experiment I would NOT run CW through a Class C Amplifier.

Peter DL8OV


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: G3TXQ on September 16, 2015, 12:40:44 AM
Peter - thanks!

Steve G3TXQ


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: SWL2002 on September 16, 2015, 03:38:44 PM
Rise time during my experiment was 4,5 milliseconds and I used a raised cosine. Output power was only at QRP levels and the receiver was a Yaesu portable rig handled by another ham a few hundred metres away in the car park. Before the start of the experiment I reduced my output power until he saw S9 so that overload was not an issue.

Based on the results of this experiment I would NOT run CW through a Class C Amplifier.

Peter DL8OV

This is what I said way back only a few posts in to the original thread.  Amazing all the "theories" going around when the answer can be ascertained with a simple physical real world experiment like you did.  Bravo!

Don't use a class C amp for CW.



Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 17, 2015, 12:37:33 AM
All the waterfalls I posted were from a real world SDR transmitter and receiver except the very first one in post #128. You can see the phase noise in them. Since I'm using a real receiver, I can also record the audio.

Here's 18 wpm CW with a 1000 Hz filter and fast AGC. About halfway through, I tune up 1 kHz.

The first clip has 8 millisecond raised cosine keying shape.

http://www.w6rz.net/noclicks.wav

The second clip has no shaping at all.

http://www.w6rz.net/clicks.wav

Here's what the audio looks like in Goldwave. You can easily see the keying shape at the beginning of an element.

(http://www.w6rz.net/cwaudio.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 17, 2015, 01:14:05 AM
Here are the clicks zoomed in. Looks like a single click duration is around 2 milliseconds.

(http://www.w6rz.net/clickszoom.png)

(http://www.w6rz.net/clickzoom.png)


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: N3QE on September 17, 2015, 03:53:51 AM
Here are the clicks zoomed in. Looks like a single click duration is around 2 milliseconds.

Now that is the result of using a circa 1000Hz CW filter in the chain. You will never resolve a feature finer than a millisecond or two if you are using a 1000Hz filter.

If you had been using say a 6kHz AM filter, then the feature produced by a broadband click would be 0.2 - 0.3 milliseconds.

Tim.


Title: RE: For W8JI: key clicks and amplifier non linearity
Post by: W6RZ on September 17, 2015, 04:44:48 AM
Here are the clicks zoomed in. Looks like a single click duration is around 2 milliseconds.

Now that is the result of using a circa 1000Hz CW filter in the chain. You will never resolve a feature finer than a millisecond or two if you are using a 1000Hz filter.

If you had been using say a 6kHz AM filter, then the feature produced by a broadband click would be 0.2 - 0.3 milliseconds.

Tim.
Correctamundo. Filters are also subject to the Gabor limit. That is, you can't have a filter that has high frequency and temporal resolution at the same time. You can only have a narrow filter with a long time constant or a wide filter with a short time constant.

Here are the clicks with a 6 kHz filter.

(http://www.w6rz.net/clickszoom6khz.png)

(http://www.w6rz.net/clickzoom6khz.png)

Pretty close to your prediction. I guess it depends on exactly when the beginning and end of the pulse is defined.

Here's the audio file. I tuned away in 1 kHz steps. By the end of the clip, I'm 4 kHz away.

http://www.w6rz.net/clicks6khz.wav