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

   Home   Help Search  
Pages: [1] 2 3 4 Next   Go Down
  Print  
Author Topic: SS Amplifier IMD Testing (New Start rolled from Command Technologies Thread)  (Read 9927 times)
N4ATS
Member

Posts: 812




Ignore
« on: January 25, 2012, 05:17:34 AM »

One last need...

When I test IMD on Telemetry Transmitters , I usually build a super narrow notch filter to eliminate the high power CW carrier (usually a stub) , then there is no need to pad down the input to the Spectrum Analyzer while maintaining a 50 ohm match. This way you see a true phase noise reading and not a padded down one. This may be tricky with high power and no circulator protection on the amplifier for reflection.

On this test , I recommend using the amplifier into a 20 dB coupler into an attenuator into the Spectum Analyzer and the other end of the coupler to a HP load to maintain 50 ohms.

I also have some samplers , I may have one big enough here to do it , I will have to go look. I recall BIRD has some line coupler samplers , there may be some in the lab which will make life very easy.

The other way is I have a few peak power analyzers programmable to any CW order but they may not do what we want.

If you have a better way , please advise.

For the above comments from G3RZP:

IMD products generally arise when you over-drive a part.  When you do that, it departs from the “predominant” behavior and you get the higher order products.  An amplifier, for example, will stop amplifying and the output gets flat or even falls off.  So the answer to your question is “to a great extent”.

N4ATS
« Last Edit: January 25, 2012, 05:26:04 AM by N4ATS » Logged
G3RZP
Member

Posts: 4619




Ignore
« Reply #1 on: January 25, 2012, 05:57:06 AM »

>IMD products generally arise when you over-drive a part. <


I can't buy that. The old '3 for 1' dB on IMD and input  wouldn't hold anywhere if that was the case. So what is overdrive? It appears that some of these devices have  -35 or -40 dB 3rd order products (at which they can hardly be said to be overdriven) but still have significant numbers of higher order products compared with a similar power capability tube amplifier at the same IMD.

Back in the 1960's, the standard 3rd order IMD level for big (30kW) ISB transmitters was -42dB rel PEP, but the 5th and higher order reduced very quickly. These would run two voice channels on the USB and 24 voice frequency telegraphy channels on the LSB. Average power for a single VFT channel came out at about 100 watts...
Logged
N4ATS
Member

Posts: 812




Ignore
« Reply #2 on: January 25, 2012, 06:25:45 AM »

The output simply stops being a linear transform of the input, because the device runs out of steam.  You might have flat-topping for example, and that waveform contains higher order components, which means that you’ll get IM products.

In a classic class A, the sine wave clips at saturation, and that gives the classic 3rd order intercept model of saturation. It assumes a linear set of curves on the curve tracer. Practically, both tubes and modern solid state devices don’t do that - you get gain expansion and compression, the compression on MMW fets is very mushy due to the large number of fet fingers and the difficulties in driving them equally and keeping them iso thermal and even identical in litho. This is acknowledged on the use of X parameters, which map in great detail the transfer function vs. drive, frequency and I believe temperature.

Practically, you can’t rely on the old formula if you need very accurate models of intermods. They are now a rule of thumb at best.
Logged
W8JI
Member

Posts: 9296


WWW

Ignore
« Reply #3 on: January 25, 2012, 06:36:52 AM »


Practically, you can’t rely on the old formula if you need very accurate models of intermods. They are now a rule of thumb at best.


That's what I observe in tests.
Logged
W8JI
Member

Posts: 9296


WWW

Ignore
« Reply #4 on: January 25, 2012, 06:55:08 AM »

Quote
"##  OK, how about  3820.00 khz for the 1st tone..and 3820.200  for the 2nd tone.  IE: use 200 hz  tone spacing.   With 200 hz tone spacing, then we can easily see IMD-3 through IMD-15.  Most of these SS amps don't dump the higher order imd, like a tube amp does.  It's the higher order imd that causes the splatter.   With 200 hz tone spacing, then the imd 3-5-7-9-11-13  will all be in band...and  imd 15+  will be out of band."

Why on earth would anyone use 200 Hz tone spacing???

Are you looking for a bias system or power supply dynamic regulation problem?

That makes no sense to me. Unless you are looking for PS or other dynamic regulation problems, the only thing a 200Hz spaced test does is make the measurement much more difficult.

If you are looking for dynamic issues, why not use a three tone with one low pitched tone?

73 Tom
Logged
N4ATS
Member

Posts: 812




Ignore
« Reply #5 on: January 25, 2012, 07:16:54 AM »

Be warned that the theory depends on a lot of assumptions and approximations.  It doesn't hold up all that well in practice, which just means that no one can predict these things all that well.  In my experience, for example, the 3 for 1 rule (in dB absolute) for 3rd order products always holds up reasonably well over some range of input levels.  However, there are also ranges of input levels, not all of them high, where it doesn't hold very well.

The basic idea is that in a non-ideal (i.e., not perfectly linear) system, an input signal, x, will give an output signal:

a1x + a2x2 + a3x3 + a4x4 ...

In an ideal system, only the a1 term is non-zero.  An important presumption here is that ai+1 << ai for all i.  Now let's assume that the input consists of 2 in-band signals, one at frequency f and one slightly higher at frequency f+d.  That is, input signal = Asin(ft)+Bsin((f+d)t).  In general, the signal levels A and B may be different.  If we substitute this into the equation above, the a1 term, of course, gives us our fundamental response.  If you go through all the expansions using trig identities, it turns out that all the even order terms result only in DC, fundamental, and harmonic terms.  The fundamental terms are assumed negligible compared to the a1 terms.  The DC and harmonic terms are assumed to be filtered out in the end and are also ignored.  This is one of the prime areas for errors, because a multi-stage MMIC chip may actually be quite responsive to harmonics, and one stage may generate them and pass them to another stage that is also running non-linear.

The third order intermodulation products come, as you would expect, from the a3 term.  After you do the expansion and drop all kinds of terms deemed to be negiligible, you get a term at frequency f-d and one at frequency f+2d.  These are the 3rd order "flankers" immediately above and below the two fundamental tones.  The lower tone has amplitude proportional to A2B, and the upper tone has amplitude proportional to AB2 .  If A and B are equal, the upper and lower tones will be equal and will vary by the familiar 3 dB for 1 if you change both A and B the same.  However, if you change just A, then the lower tone will vary by 2 dB for 1 dB change in A, and the upper tone by 1 dB for 1 dB change in A.  The converse applies if you change B.  The initial level of the flankers, and the 3rd order intercept, depend, of course, on the value of a3.

The next farther out flankers, at frequency f-2d and f+3d, come from the a5 term.  The amplitude of the f-2d tone is proportional to A3B2 and the tone at f+3d is proportional to  A2B3.  These are 5th order intermods and change at a 5 dB to 1 dB rate when A and B are changed together.  As with the 3rd order intermods, the upper and lower tones have different individual dependencies on A and B.  The seventh order lower and upper intermods are proportional to A4B3 and A3B4 respectively, and so on. 

So basically, with respect to changing both input tones at the same time, the first flankers are 3 for 1, the second are 5 for 1, the third are 7 for 1, etc.  With respect to changing one signal at a time, the bottom flanker is more sensitive to the lower main tone, and the top flanker is more sensitive to the higher main tone.
Logged
G3RZP
Member

Posts: 4619




Ignore
« Reply #6 on: January 25, 2012, 07:51:37 AM »

Even in a class A stage at low levels, you get 3 for 1. As you go up in power, below saturation, it can be anything between 2 and 5 for 1, depending on all sorts of effects, including AM-PM affecting one side and not the other. So obviously at powers approaching saturation - whatever you define that as - the linearity will go. The question really boiled down to 'Are MOSFETS worse than tubes in terms of high order products because of larger coefficients in the higher order terms of the transfer characteristic?' That has, I suppose to be the characteristic at that particular power level.

Another point about the '3 for 1' is not infrequently 'How close to 1 dB is the step?' Attenuator errors can be magnified, so measurement needs doing with care, and it is surprising  how much error can build up over a 70dB range - often a cause of problems with log amp strips at customers incoming goods inspection.
Logged
N4ATS
Member

Posts: 812




Ignore
« Reply #7 on: January 25, 2012, 08:02:11 AM »

Another point about the '3 for 1' is not infrequently 'How close to 1 dB is the step?' Attenuator errors can be magnified, so measurement needs doing with care, and it is surprising  how much error can build up over a 70dB range - often a cause of problems with log amp strips at customers incoming goods inspection.

Very much agree , usually we calibrate them prior.

Logged
VE7RF
Member

Posts: 212




Ignore
« Reply #8 on: January 25, 2012, 09:05:22 AM »

Quote
"##  OK, how about  3820.00 khz for the 1st tone..and 3820.200  for the 2nd tone.  IE: use 200 hz  tone spacing.   With 200 hz tone spacing, then we can easily see IMD-3 through IMD-15.  Most of these SS amps don't dump the higher order imd, like a tube amp does.  It's the higher order imd that causes the splatter.   With 200 hz tone spacing, then the imd 3-5-7-9-11-13  will all be in band...and  imd 15+  will be out of band."

Why on earth would anyone use 200 Hz tone spacing???

Are you looking for a bias system or power supply dynamic regulation problem?

That makes no sense to me. Unless you are looking for PS or other dynamic regulation problems, the only thing a 200Hz spaced test does is make the measurement much more difficult.

If you are looking for dynamic issues, why not use a three tone with one low pitched tone?

73 Tom

##  what's wrong with 200 hz or similar narrow tone spacing??   For now, we will assume the VL-1000 [rated for 1 kw pep out]  will have no dynamic reg problems/bias  when run at 500/600/700/800 w pep out on ssb.


##  why would 200 hz tone spacing affect the regulation anyway ?  It's all sitting there in a steady state.  OK, I think i get ur point, use real low spacing..like 50 hz..and ur sorta modulating the vdc up and down at a slow rate..slow enough to modulate the pwr supplies , is that it ?

## what's the narrowest tone spacing we can get away with then?   This makes no sense to me..since ssb is using loads of close spaced audio freqs all the time.   Heck, he could go the other way and use 100 hz and 3100 hz..and space the tones 3 khz apart.   that will work..as long as the SA in question is broadened out enough.  I'm interested in the imd 3-15.  Most of these SS amps appear to not drop off the higher order IMD...they instead just keep going and gloing and going.

##  have u tried rob sherwoods method ?  He pumps the xcvr  with white noise..then looks at it on the SA..and measures the width to the -45 db down points.  That's not new...I tried it almost 8 years ago, since the white noise gen was one of the features in the outboard audio gear  I have. I also found that white noise results in exactly a 50% duty cycle on the plate current meter on the amp.  It's exactly 1/2 of what it is with a cxr.  You can tune up a tube amp with white noise quite easily
[with a pep meter of course].

##  here'[s something else he could try..and that's tweak the tone spacing foe say aprx 1 khz..then move the tones closer/further apart..and tweak the spacing so the imd is worse case.   He could tweak the other way too..and find a tone spacing that minimizes imd.

##  BTW.. I see on ur website u used 100 + 3100 hz tone spacing in ur argument against ESSB.  try as hard as I can, no phrases or words or groups of words or phrases will result in SIMULTANEOUS low and high freqs.  I can get one..or the other, but not both at the same time.  The only thing that will produce simultaneous high and low freqs is [a] wide spaced two tone tests..or program material...like music.

later... Jim  VE7RF
Logged
N4ATS
Member

Posts: 812




Ignore
« Reply #9 on: January 25, 2012, 09:20:00 AM »

I will bow down and let the arguments fly , However I will do some testing on the 3 amps soon (IMD) and post results as soon as they are complete (soon as possible when I can get into our lab with the radio equipment)

The bottom line is  , "IMD" applies to a limited amount of folks, Most buy an ALS-600 or FL-7000 or VL-1000 and could care less about it as long as the amp works however as an interesting task , I will take it on. As a matter of fact , the majority of ham operators (99.9%)probably think IMD some new drug to enhance hair growth...

I am already setting things up (preliminary ideas)  with a lab guy here who is a also a ham and is interested in the results...
Logged
ZENKI
Member

Posts: 938




Ignore
« Reply #10 on: January 25, 2012, 05:22:40 PM »

What would be useful is to use white noise and report the IMD for every 5 khz upto 50khz out.

If you also  then reported the signal width at -50db with ALC deactivated, mid  and at full range. We could then see the impact of ALC design issues.
It would also be interesting to produce data with all knobs to the right for both the speech processor and mic, gain and report the worst possible
IMD case for the radio. Hams who use the all knobs to the right philosophy might have a clearer picture of their stupidity.

If your spectrum analyzer does not have enough dynamic range you might have difficulty doing these measurements accurately. A crystal notch filter can  attenuate the fundamental signal sufficiently enough to spot very weak IMD products. W7ZOI published a article in QEX for some crystal notch filters.

My voice can produce 10db more IMD if i use low frequency arrrrrrrggggggg, if I go nuts into the mic making all sorts of weird sounds (dummy load!) there is always a frequency or sound that I can which makes the IMD products go much worst than reported by white noise or 2 tone  tests. There are a lot of radios that produce this sort of burst of IMD as the radio modulates, whether this is IMD or ALC causing this problem is hard to know since you cant turn of  TX ALC entirely by applying a ALC voltage.

Multiple test tones also drive the radio to the highest PEP and also have a higher peak to average ration which is a more realistic way of testing a ham transmitter because we use high levels of speech processing.


Measuring a transmitters IMD profile is a complex issue that requires a methodical approach. The more that people  measure the IMD of radios and amplifiers the better will our understanding be  of how bad our transmitters are. It would be certainly better than the ARRL's simplistic approach which tells you absolutely nothing by reporting IMD products only to the 9th order level.


SM5BSZ has produced the largest  database of the IMD  performance of radios. His LINRAD  TX test most produces meaningful results. This is the approach that the ARRL should take.

http://www.sm5bsz.com/dynrange/dubus205/dubus205.htm

Where can I buy transmitter whose IMD profile is


50db(5khz), 60db(10khz), 70db(15khz), 80db(20khz)? Only the ADAT transceiver comes close and radios like the Elecraft K3 are right at the bottom of the IMD performance table with figures that are 30db worst than the above! We have a long way to go before hams will actually see a radio on the market that is  really clean. We just using glorified Cb radios in the terms of IMD performance


Your results will be interesting I am also interested in a clean solid state amp that has exceptional IMD performance. Its really idiocy spending huge dollars on a solid state amplifier then running it at low power to achieve a clean signal when I can achieve better results with a SB220 and a clean radio driving it at low power. Amplifiers need to be designed from the ground up to produce clean signals.

I read many cavalier comments by  some hams for their enthusiasm for the new indestructible solid state devices. Why are they so enthusiastic for devices which were never designed with good low IMD performance in mind  is beyond belief. We becoming like CBérs  pursuing power amplifiers designs like this solely for  power for the sake of power and forgetting about linearity.  This  approach just encourages idiots to use things like class C CB amplifiers on the ham bands.  Rather than promoting and glorifying these industrial fets we should out rightly condemn them for what they are, industrial FETS  with poor linearity. Alternatively lobby the manufactures for devices with good linearity.  If you visit these manufacturers  web pages they dont even bother publishing IMD data which is very unprofessional. At least MACOM when they produce RF devices publishes some IMD data.

Logged
W6RMK
Member

Posts: 651




Ignore
« Reply #11 on: January 25, 2012, 09:21:21 PM »

If you visit these manufacturers  web pages they dont even bother publishing IMD data which is very unprofessional. At least MACOM when they produce RF devices publishes some IMD data.



If you're selling the part into a market which doesn't care about IMD, then why would you bother characterizing and publishing the data.  Surely not to just satisfy the curiosity of a casual observer.

This is why amplifier design isn't cookbook. The data sheet only lets you identify candidate parts, then you get to buy samples, and test them for what you really care about.

If you're smart, you'll get multiple parts from multiple lots to characterize the spread.

And pray that the mfr doesn't change the process, or that your parameter of interest happens to be tied to something that they DO care about.

And, if you want the manufacturer to guarantee performance on some new parameter, you can pay for it separately (which is what makes MIL and Space parts expensive..)   radiation tolerance is a good example of something that can change radically with small changes in process.
Logged
G3RZP
Member

Posts: 4619




Ignore
« Reply #12 on: January 26, 2012, 12:01:09 AM »

>And pray that the mfr doesn't change the process, or that your parameter of interest happens to be tied to something that they DO care about.<

After 32 years in the semiconductor business, I can only say 'Amen' to that. We had a BIG customer who wanted a different glass frit on ceramic packages. We were not told about this until all of a sudden, parts no longer had the high frequency performance because the permittivity had gone up by a factor of 40, so capacitance between leads was way up. Production blamed engineering, management who didn't even understand what permittivity was (something to do with letting people do things?) blamed everybody else, and in the end, we just withdrew the ceramic package versions.

On data sheets, 'Typical' usually means 'we tested a few from one batch'.

Logged
VE7RF
Member

Posts: 212




Ignore
« Reply #13 on: January 26, 2012, 04:04:18 AM »

I will bow down and let the arguments fly , However I will do some testing on the 3 amps soon (IMD) and post results as soon as they are complete (soon as possible when I can get into our lab with the radio equipment)

The bottom line is  , "IMD" applies to a limited amount of folks, Most buy an ALS-600 or FL-7000 or VL-1000 and could care less about it as long as the amp works however as an interesting task , I will take it on. As a matter of fact , the majority of ham operators (99.9%)probably think IMD some new drug to enhance hair growth...

I am already setting things up (preliminary ideas)  with a lab guy here who is a also a ham and is interested in the results...

##  check out rob sherwoods TX imd tests on xcvr's.  He pumps the mic jacks with white noise.  The typ ICOM is  10 khz wide to the -45 db down points.  The yaesu's in class A..are 3.5 khz wide..to the same -45 db down points !   That tells me right there we could cram almost triple the number of station's  on a given band..on ssb.

##  Your typ ham xcr  is way worse than the typ tube amp.   And there is no point in using a clean xcvr..to drive a dirty SS amp either.   -29db pep  IMD3 on the K-3 is a joke.   That's on a par with sweep tube amps.

##  No rocket science to get these SS amps  cleaner..and also xcvr's..without having to resort to class A.  use more powerful devices..then derate em to 1/3..so the imd is better...like -40db pep. imd3.  IMO,, they should all be designed around a  min imd spec.   As is, they are now designed around a max power spec..with heat rise, etc, thrown in.   The bands are not crowded..they are just full of lousy imd.

##  Zenki is correct. The ITU, etc, needs to come up with a simple, bare min imd spec for ham gear..which of course, the xcvr makers  would be able to meet easily...virtually overnight too.  Icom and yaesu already do it with their marine ssb gear.   Notice on all this commercial marine hf ssb gear, none of em have mic gain controls !

Later... Jim  VE7RF
Logged
KE5JPP
Member

Posts: 0




Ignore
« Reply #14 on: January 26, 2012, 04:57:30 AM »

Rather than promoting and glorifying these industrial fets we should out rightly condemn them for what they are, industrial FETS  with poor linearity. Alternatively lobby the manufactures for devices with good linearity.  If you visit these manufacturers  web pages they dont even bother publishing IMD data which is very unprofessional. At least MACOM when they produce RF devices publishes some IMD data.

Earth to John aka "Zenki"...  To the manufacturers of these new RF devices, Ham Radio is a minuscule fringe market.  Your whining falls on deaf ears with them.  

It is SO EASY to design from an armchair, isn't it?  

Gene
« Last Edit: January 27, 2012, 06:49:13 AM by KE5JPP » Logged
Pages: [1] 2 3 4 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!