PDM Class E IMD AM Tests

[steve_qix]

If changing the voltage/current ratio of the RF amplifier changes the audio performance, it may not have anything to do with the RF stage, but rather, may have something to do with your PWM filter.

If the input inductor isn't correct, the PWM waveform will distort, particularly at high levels of negative modulation, where there is little energy, and the pulse width is small.  Anyway, just a thought.  Back in the tube days of PWM over here at QIX radio, the input inductor value was very critical to creating a low distortion system.

What is your input inductor value?

[K1JJ]

I was thinking the RF final needs to have good, matched load, thus the L/C experiments.


Anyway, my high level PDM filter was designed by Jay / W1VD and called an "infinite impedance filter."  The three coils are air-core wound with #6 wire on 4.5" PVC plastic pipe.

L1 = 28.8 uh      C1= 10.83 uf     L2= 28.8 uh    C2= 7.55uf    L3 = 14.1 uh     (actual values)

I have some more bypassing to ground at the output of the filter.

Would you suggest some changes?


I hope to try some NPL tests later today.  

T      

[W9BHI]

I was just reading some of the earlier posts on this subject.
I work for a company that makes audio frequency filters (.1 Hz to 100Khz), we make LP,HP, BP, BR.
We also have some custom DSP filters.
Take a look on the web for Frequency Devices inc.
I'm sure that any kind of filter could be designed for whatever frequency range you need.
Just trying to help.
Don W9BHI

[W1DAN]

Hi Don:

I looked at your web site and you have some good filters to offer. Bear was mentioning a long time ago about making a passive LC low level filter. Found this calculator on the web:

http://www.calculatoredge.com/electronics/ch%20pi%20low%20pass.htm

Maybe this can be put in the middle of two op-amps that control source and load Z.

Tom:

As you are finding more distortion around -90%, this could be due to a changing pulse waveform, which as Steve said could be due to filter or load issues. A few years ago I came across this paper which describes some issues to PDM performance, and an interesting "tail biter" circuit (see figure 6), of which one person on this board who designs Class E transmitters is intimately familiar with:

http://www.analfatecnicos.net/archivos/73.AM-HDRadio-DRM-Harris.pdf

Always thought of trying it out.

I have been extremely busy and am considering doing some tests on my E-rig. I have two high priority projects on the bench right now, so it may take some weeks to get to. If I test, I would use a PC with sound card and editing software. My Adobe Audition can do a spectrum display of a recorded file. I would try to break up the test into two parts, 1: the PDM generator, where I would use a simple RC filter at the output of the generator to verify this simple circuit is doing well, then 2: at the modulated DC in to the RF deck, to verify my modulator frequency response and linearity. I know I will find filter issues as my filter is not optimized.

73,
Dan

[steve_qix]

I'm sure the filter is fine.  I use a completely different filter here - it's a standard Butterworth.  The input inductor is 11uH.

Redoing the filter is a big deal, and I wouldn't bother.  By the way, I have a 1.7 ohm terminating resistance (the RF amplifier).

[K1JJ]

As you are finding more distortion around -90%, this could be due to a changing pulse waveform, which as Steve said could be due to filter or load issues. A few years ago I came across this paper which describes some issues to PDM performance, and an interesting "tail biter" circuit (see figure 6), of which one person on this board who designs Class E transmitters is intimately familiar with:

http://www.analfatecnicos.net/archivos/73.AM-HDRadio-DRM-Harris.pdf

Always thought of trying it out.

I have been extremely busy and am considering doing some tests on my E-rig.

Thanks for the info Dan. I will read it over and see if I can try any of it.

For AM, I am hoping to use the class E rig exclusively instead of the class A linear system.  Heat is the major reason. So I am determined to get it running as cleanly as possible.

Here's what I'd like to do...  It's very easy to run THD tests on the air to see what some of the other guy's e-rigs look like. Many of the guys on 3875 are running e-rigs and last night seemed very interested in the tests I've run. All we have to do is put a 1 KC audio tone into a class e rig and run the modulation up to -99% negative and watch the SDR spectrum on the receiving end.  I'll bet there are different PDM filters being used so it will help to screen out the filter issue, if it exists.   If some look better than mine, then I know I can improve things here.

Last night one of the guys did a whistle, which is a pure sine wave.  At low modulation levels, I saw just the 1 KC tone. But at high modulation levels I could see the "xmas tree on its side" of X2 X3 X4, etc., THD peaks - just like my rig here. A "perfect" transmitter would put out just one 1 KC peak. But nothing is perfect, of course, nor should we expect it with any AM rig. It's just a matter of testing and trying to get things set up the best we can and then forget about it...

I'll see if some of the guys will run a simple signal gen 1 Khz single tone on-air test tonight or tmw.


Don: Appreciate the filter info too. As hams we would probably like to find an inexpensive active 12 pole R/C analog kit for $29...

Steve:  OK on the filter.  I'l continue on as described above... thanks for the input.  BTW, how did you calculate terminating resistance for the final... is it a simple total drain current / drain voltage or is there a factor involved?


T

[W1DAN]

Tom:

I have a couple of clean audio test oscillators. If I am around this weekend and you are interested, maybe I could insert tone. My rig should be the reference for the most distorted! :-)

Dan

[K1JJ]

Sounds good Dan!  At this point I have no data from others' rigs and desperately need it to continue on.


NPL Test Update:

Nope, it's not the internal negative peak limiter doing it. I adjusted the NPL completely out and the THD peak levels  grew similarily as they approached -99%.  However, at -100% negative, the NPL does have an effect mostly on the higher harmonics, causing them to increase slightly.  

The NPL is still a good safety net in case of stray peaks.  I wouldn't lean against it all the time, however.



T

[K1JJ]

I got some on-air test results from two of the class E guys tonight on 3875.   They both injected a 1Khz tone directly into their H-modulator class E transmitters.  These results were generally confirmed by Al/ W1VTP watching on his SDR receiver too.

These tests were done at about 95% modulation without activating the negative peak limiter.

They pretty well match my own e-rig results when hitting the audio near 95% modulation. The difference is that I'm running a PDM modulator.

At 20% to 50% modulation, (not pictured here) the THD was excellent - down 40 Db.. which is about 1% THD.  
Still trying different things...

T

[steve_qix]

Chuck / WA2ONK suggested we try a test on an H modulator e-rig. That's a great idea and will do.

You just did!  N1GTU and W1LLY are both running class H modulators.

[K1JJ]

How 'bout that.  I corrected it above, tnx.  


Since both the H amd PDM modulators show the same performance, I wonder if adding more ferrite cores to the output combiner might be the next logical test?  Possible saturation at higher modulation levels?  I have eight cores for each group of six Mosfets.  They do get warm after a transmission.



T

[N2DTS]

Funny, I did tests on 3 different transmitters today, I fed in a 604 Hz signal into the rigs and watched the spectrum, and on all three, it was clean up to about 50% modulation, above that, I started getting the additional spikes.
Plate modulated rigs all, with all different tube types, and I adjusted bias, grid leak, screen voltages, tuning and loading, and nothing made a bit of difference.

Could this be an sdr thing?
 

[Steve - K4HX]

No, it's the reality of what's been put out on the air all these years. Most of us never had equipment to take such a look at our transmitters. I've done it with a traditional swept tuned spectrum analyzer and saw the same sorts of spectra. It's not unique to the SDR.

[K1JJ]

Funny, I did tests on 3 different transmitters today, I fed in a 604 Hz signal into the rigs and watched the spectrum, and on all three, it was clean up to about 50% modulation, above that, I started getting the additional spikes.
Plate modulated rigs all, with all different tube types, and I adjusted bias, grid leak, screen voltages, tuning and loading, and nothing made a bit of difference.

Could this be an sdr thing?
 

Hi Brett,

Thanks for taking the time to give us some more data.


Usually, the harmonic spikes are always there, even at 10% modulation, but way down, maybe below the noise floor. They will increase with modulation and sometimes hit a point when the growth accelerates due to some non-linearity problem being hit.  I haven't tested any tube transmitters yet. Maybe some more guys will.

Next, try a two tone test and see if you can improve the IMD using the same things you tried already for THD. Maybe the THD is "built into" the modulator system with little adjustment available - and you can have better influence improving IMD overall, dunno.  Mod iron and tubes in general can be one reason for higher THD.  Just like any tube audio amplifier used for music...


Here's what keeps me from questioning the SDR results...

1) I ran these tests with my new class A linear system (4CX-350FJ) at 100% modulation and see excellent THD results, below 1%.  Steve/ K4HX did the same on his FT-102 and saw the same good results.  So it CAN be done.

2)  Using a single tone, push any transmitter close to 100% modulation and listen on a separate receiver as we tune up the band. The worse the THD, the louder the harmonic tones will be as they parade up the band.  The cleaner the transmitter system, the weaker the tones.  I hear this on my rigs which confirms the SDR results.

Obviously, getting a transmitter to modulate 100% at high power with THD below 2% is no easy feat. But we are all learning.

Distortion is quite normal for any power modulation system. It's just a question of degree.

T

4CX-350FJ at 100% AM modulation:

[K1JJ]

Just to prove to myself that there can be the "perfect" AM signal, I tried the test using my HPSDR low level transmitter. This is the SDR going into a MiniLabs ZHL-3A 1 watt lab amplifier.

A "perfect" amplifer would do ONE tone in... ONE tone out (one per sideband)

The pic below shows a 1 Khz tone at 140%+ modulation.  If I had more signal to work with above the noise floor, we would likely see the peaks down over -65 Db.  But as-is, this shows that an ideal AM signal has very little harmonic X2, X3, etc.  activity.

Of course it's easy to get a low level signal to be this clean - but a bear to do it with a high level modulation system and good efficiency.   The next step up is a class A linear amplifier, but lots of heat.  (Or heavy NFB or predistortion software techniques)

T

[steve_qix]

Well, it seems to me that it's "good enough", as they say 

On a practical level, you have to be generating a LOT of distortion for anyone to notice the signal is "wider" - and in a good system, that means serious clipping and/or nothing to keep the high frequencies in line at all, and probably some high boost to boot.

I think if this were a problem, we would have been hearing unusually wide signals for many, many years.  There have traditionally been a lot of high power transmitters around for a long time and I don't remember anyone being wide due to a few percent (and that's probably being generous - most are much worse) harmonic distortion unless they were overmodulating or "flat topping" in some way.  And you're well below that.

[K1JJ]

Yep, as I said back on the first page of this thread after the initial IMD tests, a good working class E rig is better than average for IMD.  When run with voice modulation, the sideband crud is way down. It's a beautiful working design....  

As for the THD numbers, I honestly thought they would test out better at high modulation levels. That's why I've been trying to determine if I had a problem with mine or not.  I'm the kinda guy who could never be satisfied if I at least didn't try to improve things.

I doubt any improvements could make any noticable difference in less bandwidth at 80% modulation or below - or even improve voice quality on freq - but when hitting 150% positive using the NPL, I think we might hear some extra crud on my e-rig due to THD.  That's why I plan to keep it down around 100% positive most of the time - unless I can eventually come up with better THD numbers when hitting it hard.

That being said, I'll bet if some of the guys do THD tests on their high level class C tube rigs, we will see similar or even poorer THD numbers.   With iron that cud saturate and tubes, run a big tube rig at 150% positive and we might be surprised by the THD numbers, I dunno.  I'd like to see some tests. Then again,  ones with modulator negative feedback may really shine here.

I'd be willing to capture a screen spectrum shot for anyone who wants to put through a 1 KHz tone directly into the audio at 95% negative modulation or whatever. (with no processing or EQ's in the way)  Let's test  a 4X1 X 833A's, a stock ricebox driving a linear, a DX-100,  Flex driving a linear , etc.   I'm really curious.
 

T

** BTW, yesterday Clark/ KB1NFS, said he was able to improve his  e-rig THD by optimizing his tank L/C ratio. He ended up with more turns on L1 and used a distortion analyzer to monitor the improvements.  I also ended up with MORE turns for L for some THD improvement. Higher Q did not work in this case.

[W1DAN]

Tom:

How about later Saturday afternoon, maybe evening?

I will be doing some testing at W1ZZZ's, and can bring my oscillator over to test his D-rig.

Maybe later we can do my E-rig? I do not have a mod-monitor, so can you talk us in to various %'s of modulation? Can you record an audio *.wav file?

If good on all, please PM me your phone number.

Dan

[K1JJ]

Tom:

How about later Saturday afternoon, maybe evening?

I will be doing some testing at W1ZZZ's, and can bring my oscillator over to test his D-rig.

Maybe later we can do my E-rig? I do not have a mod-monitor, so can you talk us in to various %'s of modulation? Can you record an audio *.wav file?

If good on all, please PM me your phone number.

Dan

Sure, no problem, Dan.  We can do it Saturday.  

Maybe we can find the "sweet spot"  settings, if any.   Be prepared to try a few things.
I haven't tried a .wav file recording on the SDR yet. Should be easy enuff to do....

I'll send ya a PM.


BTW, I read your polyphase info.  It looks impressive and complex -  though I don't think I am quite THAT motivated to pursue it right now ... :-)

T

[W1DAN]

Tom:

Maybe if you can record a wav file, then we can look at the spectrum in an audio editor application and see if the harmonics of the test tone are high at all. This might eliminate the SDR display as any potential issue.

Thanks,
Dan

[K1JJ]

Sure.  I'll get set up for recording the .mpg.   In fact, I can audio record THD of my own e-rig signal and get it to you soon.  

I would guess there is little chance of the SDR display being off, but it's worth a look to verify it in your audio editor.   I'll bet it will give us real % THD numbers too?


T

[W1DAN]

Tom:

I was thinking of importing a wav audio file into Adobe Audition, then looking at spectral view, so no analysis. I do have an AP System 1, but it has not been run in years. It would take some effort to get 'er going.

Dan

[Tom WA3KLR]

K1JJ Plot from Reply # 64:
"HPSDR near perfect" USB only, 0.91 % THD.

[K1JJ]

Today Rob/ W1AEX wanted a THD test. The band was quiet and stable. Using his Flex 5000 SDR and pair of 3-500Z's, he tested at better than  1.8%  THD at 100% modulation.   He has that system really well tuned and optimized.


T

Rob / W1AEX's   100% modulation  -   1.8%  THD pic:

[N2DTS]

I thought it was odd that I got the exact same results on various rigs:
4x150a modulated by 811's, the 3x4D32 modulated by the same 811's, a pair of 813's modulated by a pair of 4cx250b (ab1) or push pull 100TH's.

All the rigs showed the same clean below 50% modulation, all got nasty above that, and went crazy past 100% neg modulation.

Maybe I will run the tests again and actually note just how far down the spikes are.

I am also building a push pull 812a rig to see how the triodes take modulation since there is no screen to screw around with the modulation.

I used to enjoy the test suite in the flex 5000, you could actually transmit a square wave and it was square.