Call Search
     

New to Ham Radio?
My Profile

Community
Articles
Forums
News
Reviews
Friends Remembered
Strays
Survey Question

Operating
Contesting
DX Cluster Spots
Propagation

Resources
Calendar
Classifieds
Ham Exams
Ham Links
List Archives
News Articles
Product Reviews
QSL Managers

Site Info
eHam Help (FAQ)
Support the site
The eHam Team
Advertising Info
Vision Statement
About eHam.net

donate to eham
   Home   Help Search  
Pages: Prev 1 ... 24 25 26 27 28 [29] 30 Next   Go Down
  Print  
Author Topic: For W8JI: key clicks and amplifier non linearity  (Read 64249 times)
G0HZU
Member

Posts: 163




Ignore
« Reply #420 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.
« Last Edit: September 14, 2015, 11:35:14 AM by G0HZU » Logged
G3TXQ
Member

Posts: 1845




Ignore
« Reply #421 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:



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
« Last Edit: September 14, 2015, 11:42:33 AM by G3TXQ » Logged
G0HZU
Member

Posts: 163




Ignore
« Reply #422 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?
« Last Edit: September 14, 2015, 12:39:59 PM by G0HZU » Logged
G0HZU
Member

Posts: 163




Ignore
« Reply #423 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.
Logged
W6RZ
Member

Posts: 363




Ignore
« Reply #424 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.





Logged
G0HZU
Member

Posts: 163




Ignore
« Reply #425 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... Smiley  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.
« Last Edit: September 14, 2015, 05:32:09 PM by G0HZU » Logged
W6RZ
Member

Posts: 363




Ignore
« Reply #426 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.
« Last Edit: September 14, 2015, 05:47:48 PM by W6RZ » Logged
G0HZU
Member

Posts: 163




Ignore
« Reply #427 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?







« Last Edit: September 14, 2015, 05:56:18 PM by G0HZU » Logged
W6RZ
Member

Posts: 363




Ignore
« Reply #428 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?
Logged
G0HZU
Member

Posts: 163




Ignore
« Reply #429 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.





Logged
K9AXN
Member

Posts: 442


WWW

Ignore
« Reply #430 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
Logged
K6JH
Member

Posts: 515




Ignore
« Reply #431 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!
Logged

73
Jim K6JH
KB8E
Member

Posts: 50




Ignore
« Reply #432 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
Logged
SM0AOM
Member

Posts: 257




Ignore
« Reply #433 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







Logged
G3TXQ
Member

Posts: 1845




Ignore
« Reply #434 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
« Last Edit: September 15, 2015, 04:48:49 AM by G3TXQ » Logged
Pages: Prev 1 ... 24 25 26 27 28 [29] 30 Next   Go Up
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.11 | SMF © 2006-2009, Simple Machines LLC Valid XHTML 1.0! Valid CSS!