PDM Class E IMD AM Tests

[K1JJ]

I finally got around to doing some IMD tests on my 24 pill class E PDM rig.  It uses digital drivers and also a homebrew analog VFO with digital converter.  I've been very pleasd with it but never did any serious testing.

IMD = Inter-Modulation Distortion
THD = Total Harmonic Distortion


This is the Class E construction thread back in March, 2010.
http://amfone.net/Amforum/index.php?topic=23354.0

Here's a thread on IMD testing that describes the test set up and shows more sample graphs:
http://amfone.net/Amforum/index.php?topic=33505.0

Pic #1  shows the rig running at 1KW using a 2-tone test. Remember that with AM DSB full carrier, there is a mirror image of IMD peaks. It's like a standard ssb graph but with two sides.  Pic # 1 shows roughly -35 dB 3rd order,(down from the two main audio peaks)  while the higher orders drop off very well. I would rate this IMD slightly better than average compared to most ham signals for cleanliness. Maybe with some custom work it could be improved.

What does a 2-tone test measure?  It checks the linearity of an amplifier. We want the output to be an exact replica of the input. (Or in class C/D/E, the audio to modulate and mix linearly) When we put two tones in, we want ONLY two tones out. But mixing occurs and we end up with something less than perfect.  As long as these mixed tones that march up the band are below an "acceptable" level, then the amplifer is OK.  A standard linear ssb amplifer has to have a linear output as well as a class C or class E stage.  The two tone test does not require audio filtering into the rig... it takes no prisoners.


Voice tests are a little more forgiving than the 2 tone test:

Pic #2 shows voice speech with an "SSS."   I used a 3kc audio rolloff to show what artifacts remained.  I see the first shelf about -60Db down from the carrier peak.  I would rate this good. One could feel confident running this rig at +-3 KHz audio if the band were crowded.


Pic #3 shows an "SSS" with 5 khz of audio bandwidth.  The main point is there are little artifacts and crud as the audio rolls off into the noise.  It has a good slope.

Pic # 4  in the next post, shows a standard voice Yallo.  Up the band 6-8 Khz the sidebands are down 60 to 70 dB, which is good.


In summary, this class E rig is above mainstream "ricebox" average in cleanliness. However, when run without audio bandwidth limiting, it can naturally get very wide, even though the audio may be clean for IMD. A DC-to-light rig like this requires audio bandwidth filtering. In my case I usually run it at 4 to 5KHz using the roll off of the EQ, an adjustable digital filter and a hi-freq cut.   This gives me a sharp slope  as shown in pic #3.   The combination of a sharp audio roll off AND a clean rig for IMD that does not add artifacts can produce a signal without worry of splatter.

The class E rig will never have sharper slopes than an SDR radio with digital filters driving a super clean linear amp, but the huge increase in efficiency along with reasonable IMD of the class E rig is a good trade-off.

T

[K1JJ]

This is a voice Yallo.  Unscientific, but shows the general trend. Notice 6-8 kHz away the signal is down over -70 dB.

[N2DTS]

And when you run lots of power, the stuff off to the side needs to be WAY down to not bother other people.
With class E (and other rigs that can run wide) its typical to have strong artifacts out past 20 KHz.
With a100 watt rig that rolls off the high end and just will not pass anything very high its no issue, but people that take up 20 KHz or more are being a little bit of band hogs, no?

[K1JJ]

Yep.

When we choose to run 10 dB more power than the average low power station,  then our sidebands must be 10 dB cleaner to blend in. So high power gives us a bigger responsibility to run cleaner than the average station.

It CAN be done with good audio filtering AND a clean rig. Without both of these factors under control, we may be wide.

My class E rig can easily pass +- 12Khz, cleanly.   However, as shown in picture #2, if I choose, I could run narrow ... +- 3Khz audio and be down -65 dB  5KHz away either side of center freq.   Once the rig is clean, it's all up to the operator to decide how wide he wants to be.

It's one thing to be intentionally wide with a clean signal on a day when or where the band is not active. No problem.  It's another thing to be wide due to a rig problem on a crowded band.

If by miracle, every station on the air had a -15 dB improvement in IMD on their transmit signals, overnight, QRM from adjacent channel  splatter would mostly disappear.


T

[KK4YY]

Tom,

Having been responsible, and taken the time and effort to clean your own house first, puts a lot of clout behind the common sense opinions you've expressed here. We can keep hoping for a miracle to occur and perhaps it will, in time, but you've clearly taken the lead in trying to make it actually happen. An admirable effort. And even those who may disagree with you will have to admit that.

Don

[kb3ouk]

Something else that seems to affect how clean a high level modulated transmitter of any type is (class E PWM, class C plate modulated, etc.), is the negative peak limiters that some are running. Those things are fine if you are using it just to keep from ocassionally going over 100% negative, but there are some who feel the need to crank the audio so hard that they are hitting the limiter all the time, and that causes just as much crud as overmodulating without one does.

[K1JJ]

Tom,

Having been responsible, and taken the time and effort to clean your own house first, puts a lot of clout behind the common sense opinions you've expressed here. We can keep hoping for a miracle to occur and perhaps it will, in time, but you've clearly taken the lead in trying to make it actually happen. An admirable effort. And even those who may disagree with you will have to admit that.

Don

Thank you, Don.

This IMD journey has been a learning experience for me and I simply want to document it for others to learn.  It's for the people here, the archives and the general readers on the web.  Some of these technical threads have received 20,000++ hits over the years.  They do go mainstream after awhile.   I'll bet these IMD discussions will help some hams. There is precious little easy-to-understand  IMD testing info on the web. Most is chock with formulas and ivory tower jargon.

Even six months ago I didn't have a clue on how to run meaningful IMD tests nor set up a reasonably reliable test set up.   I would just build a rig, look at the scope and hope for the best. Sometimes I'd tune up the band and hear a QSO talking about how my signal was splattering them. Have you ever had that happen? It's the worst. And I didn't know how to fix it, though I tried.  All I had to do is look at some IMD tests and it would have showed the problem - at least a reference  baseline to know if my modifications were working.

My IMD threads are not to point fingers nor make anyone feel bad. It's just to record ideas and sample results - a progression of what I've found works for me.  I'll bet there's not a ham on the air who would not like to make his rig cleaner, if they knew how.   It's like going to a party with a "little" bad breath. We can also get on the air with a "little" splatter. Either way, we are probably going to worry about it. It doesn't have to be that way...

I think in the end, we may hear more talk about IMD and how to make our rigs cleaner. The journey doesn't end when we get a new rig on the air and ask for a report...  

Well, I'm about finished with the tests for now but might add a white noise generator graph that I just built. White noise appears to be a neat way to generate highs, mids and lows all at the same time to look at bandwidth. Doing it by voice is near impossible.   A combination of 2-tone, pulse pecker, white noise and the human voice seems to cover all bases for SDR spectrum tests.

T


BTW, here's a picture of my class E rig - dead carrier.  The SDR dynamic range is only about -75 dB at this high power level.  But I see no signs of spurs or other stuff.  I'll bet it's really down over -90dB.  The analog VFO driving a digital converter board into the digital chip drivers works for me.

[kb3ouk]

First, does that transmitter have a negative peak limiter in it? I'd like to see an A/B comparision, first with it set  so that the audio does not go over 100% negative, therefore not hitting the limiter, and then with it pushed past that point so that the limiter kicks in. I'm curious to see if that makes a huge difference.

[K1JJ]

First, does that transmitter have a negative peak limiter in it? I'd like to see an A/B comparision, first with it set  so that the audio does not go over 100% negative, therefore not hitting the limiter, and then with it pushed past that point so that the limiter kicks in. I'm curious to see if that makes a huge difference.

Hi Shelby,

I wondered the same thing and actually tested it the other day.  This is a low level diode clipper circuit that comes stock in the QIX PDM generator board.   Into a dummyload, using my voice, I tested the rig at about 90% negative. The positive peaks were about 110%.  The IMD and side products looked normal.  I ran it up to the preset clipper level (about 95% neg) and really hammered the audio to maybe 175% positive.  I mean I was abusing the rig to its limits.  Amazingly I did not see a significant increase in IMD or side products.  The noise floor did come up, but not to the point of generating complaints, in my opinion.  I've had rigs that tore up the band when I did that test, but not in this case.

At 150% pos audio and the clipper working, there was barely any difference compared to  running it conservatively  at  110% positive.

So I would say this particular negative peak iimiter works extremely well.   I have run other high level clipper circuits  on the 4X1 plate modulated rig and did not like the results at all.  I am a fan of low level negative peak clipping / limiting.  Maybe a softer limiting would be better, but my results show FB with hard clipping.

In actual use, I usually limit my positive peaks to no more than 125% due to the many diode detectors out there, thus the negative limiter, though working, is barely in use. It is a great safety net and would make life difficult without.

T

[N2DTS]

Its great to see someone take the time to generate a clean signal.

There are a lot of people on the air who do not have or do not use things to check their frequency, modulation, bandwidth.

Having an SDR can be good and bad, it can show me how my signal is (on frequency, not very wide, enough audio), but you also get to see things that you do not want to see, one guy 2 KHz low, one 1 KHz high, someone 5 KHz below who is 20 KHz wide and over modulating...

The class E stuff is great, once again its the AM guys actually building stuff, and testing it, and using it.
Ham radio is not quite dead yet.

[K1JJ]

Hi Brett,

Yes, the SDR's can be two-sided.   Just like a dream or nightmare.

I've been having a lot of fun with these projects.  In the beginning it was frustrating and I almost gave up a few times.  But when the good results started, it made all the difference.

Right now I'm a little stuck on how to further improve the class E rig.  Linear tube amps and plate modulated rigs are easier for me.  Frank/ GFZ suggested I disconnect the PDM modulator and test that separately to check its own performance.

Maybe the E Godfather (QIX) will add some input as to what areas can be beefed up, tested or changed to improve IMD in a class E PDM rig.  I am happy with what I have, but there is always room for optimization and improvement.

T

[steve_qix]

Hi Thom,

I'm glad the class E transmitter has reasonable IMD products.  It should, really.  And, the reason the negative peak limiter doesn't make the rig a lot wider is because the filtering is *after* the N.P.L., which is where it should be!

Over here, the frequency response of the transmitters rolls off reasonably significantly starting just before 6kHz.  I don't think I can actually modulate a 12kHz signal more than a small amount due to the filtering action.  I have measured it, but it's been quite a while!  That's the standard filter in all of the new designs that are published.  The PWM implementations are superior in several respects to any other of the modulators used in the class E rigs.

Over here, I have an 8 pole filter that I can turn on, with a 5kHz sharp rolloff.  It's not a DSP. 

*Nothing* is better than a DSP filter because there is no phase shift, but DSPs introduce their own unique set of issues (like delay).  I'm working on a DSP project over here, and might come out with an adjustable filter for nighttime operation someday, although that is not the goal of the project at the moment 

[K1JJ]

Hi Steve,

OK on the frequency limitation.  I meant 12 Khz based on the high level modulator filters with the low level PDM generator filters set up to cut off higher.   I will check the roll off as a stock radio. I think mine is higher than 6 Khz, however.

Questions:  

1) Besides delay, what problems do you find with digital audio filters?   I have an adjustable audio filter running two cascaded digital chips. I do hear some very slight "cell phone" type ringing, but it is not a problem when running it occasionally when needed.

2) If you were looking to improve IMD on a 24 pill PDM class E rig, where would you start looking?


T

[steve_qix]

Hi Steve,

OK on the frequency limitation.  I meant 12 Khz based on the high level modulator filters with the low level PDM generator filters set up to cut off higher.   I will check the roll off as a stock radio. I think mine is higher than 6 Khz, however.

Questions:  

1) Besides delay, what problems do you find with digital audio filters?   I have an adjustable audio filter running two cascaded digital chips. I do hear some very slight "cell phone" type ringing, but it is not a problem when running it occasionally when needed.

Ringing is always a problem with all filters that have very sharp slopes.  I generally will go with a slightly less steep slope in favor of no ringing.

2) If you were looking to improve IMD on a 24 pill PDM class E rig, where would you start looking?

I really wouldn't bother    The IMD is quite reasonable as-is, and is far better than pretty much any high level modulated system out there. 

When you start designing super low distortion, perfect audio systems you get need really high switching rates (so the filters don't introduce phase shift in the audio range), then negative feedback (not trivial to impose on a pulse width modulator), remodulation (negative feedback that includes the RF amplifier - demodulated RF is fed back into the modulation), etc. etc.

For ham radio, I don't see the return given the added complexity.

When I was in broadcast engineering, I had occasion to measure the audio performance of many broadcast transmitters.  The IMD figures on your amateur transmitter are better than any broadcast rig I've ever tested.  Keep in mind, I pretty much got out of radio before the newer solid-state transmitters hit the market.  These, no doubt, are much, much better than their predecessors.

[K1JJ]

Hi Steve,

I'd like to find an audjustable, analog audio filter. A kit form or ready-made wud be FB.  There are plenty of digital ones out there, but no serious analog types I could find.

The analog require precision resistors, so it's not a junk box deal. Also to make it adjustable would require lots of pots.  So maybe a fixed 5 KHz analog filter would be the ticket.  There is probably a good market for this if someone marketed one.


Later today I plan to run some more tests on the PDM rig. I got the white noise generator working so want to see the  bandwidth and general IMD at various modulation levels, straight in, with no external audio filtering..  So far I see pretty consistent performance from low to high modulation levels, probably due to the overkill on the PDM FETS.  The only limit is the PS...

Frank suggested the modulation degradation may show up under 10% where the filter goes discontinuous.

I plan to try loading and other HV levels to see if the IMD changes. Being a fixed digital drive system with no tuning in the lower levels, there's not a lot that can be done for tweaking.  I understand what you are saying about the more complex design if better IMD is desired.

Look for some more test graphs later....

T

[W1DAN]

Tom:

First of all, wonderful tests on your linear amplification and Class-E. An education for all of us.

I think that as a PDM pulse-width gets very narrow, the linearity would tend to decrease and at some point at say -90% to -100% modulation, the pulse sometimes tends to dissapear. If interested, you may wish to create a resistive attenuator and record a low voltage audio version of the modulated DC at the input to the RF deck. Then you could use PC-based audio analysis to see THD and IM in the PDM modulator. These results may point a finger to any modulated RF distortion.

As far as analog low level audio low-pass filters, I bet you could use a multi-switch to select different capacitors for a 6db or 12db per octave LPF using op-amps. Maybe put a couple in series at 6kc. Here is a good intro on filters here:

http://www.kennethkuhn.com/students/ee431/frequency_filters.pdf

See here (and attached image) for a basic second order low pass filter:

http://www.ti.com/lit/an/snla140a/snla140a.pdf

Walt Jung is one of my favorite tech authors on op-amps. I built an audio console in the 1980s from learning from his book "Op-Amp Applications", found here:

http://waltjung.org/ADI_Books.html

73,
Dan

[W3GMS]

Walt Jung is one of my favorite tech authors on op-amps. I built an audio console in the 1980s from learning from his book "Op-Amp Applications", found here:
http://waltjung.org/ADI_Books.html
73,
Dan

I did a similar thing Dan.  Walt came to our business and I met him in the early days of my engineering career.  He gave me his book and signed it at the same time.  He is certainly a very talented individual.

Joe, GMS

[K1JJ]

Thanks for all the info, Dan.

Actually, I have already built up an analog filter using four of the circuits you showed in cascade.  It used eight pots that were preset to various cut off freqs.  It worked rather well, but it was done breadboard and I wish to get a circuit board or equivalent kit to clean it up.


BTW, you have an H-modulated Class E rig, right?  Are you set up to run some two-tone IMD tests with it?  I would be curious what others see for IMD with their E-rigs too.  So far I am impressed with the IMD results and the rig's ability to run very heavily modulated and stay relatively clean.

T

[K1JJ]

Here's some more results and conclusions testing the 24 pill, class E PDM rig.  These tests are with audio directly into the rig, with no external audio filtering.   My PDM high level filter has a rolloff at about 10 kHz.  Other rigs may vary.

Instead of posting more graphs, I did a bunch of tests and got a general feel for the IMD and audio bandwidth...

First, I swept it with a standard audio oscillator and found the high frequencies start to drop off at about 7khz.  It is pretty flat from 5 Hz to 7 KHz.   There is still considerable energy at 8 kHz and even 9 Khz, but rolls off pretty well by 10 Khz. This is the six-pole 6-7 kHz filter after the NPL having an effect.

A series of voice S's  shows the energy extends out up to 10Khz, confirming the tone test. However, up 10 kHz, the energy is down -40dB.  So if someone where S9 +40, their S's would be about S9 10Khz up the band.

I ran some more 2-tone IMD tests and thought they still looked good. The higher order peaks drop off sharply after 5 Khz, which is what a good linear amp would do using these same tone freqs.   As a confirmation, tuning a second receiver, I hear the tones disappear 5 kHz up the band too.  A good 3-500Z linear amplifer will do about the same.

In summary, the rig is clean for IMD. However, the exisiting internal six-pole 6-7Khz audio filter after the negative peak limiter needs a sharper roll off, perhaps in the form of more poles.  I feel there is still too much energy above 7 Khz, if the goal is to run 7 Khz maximum with the rig stock.

But this is easily fixed...  In my case, I use a very sharp digital audio filter before the rig. This lets this clean rig operate well within the audio bandpass we desire, whether it be 3, 4, 5, 6 or 7 Khz.  I would rather have an analog filter, but that is another subject.

I feel the rig can handle heavy modulation with ease - the negative peak limiter with the filter works superbly... no sign of splatter. It is a great platform for high fidelity, high efficiency AM - it just needs a method of sharper audio filtering, probably in the form of a 6 or 8 pole analog audio filter placed before the audio input.  It appears the internal NPL six-pole analog audio filter is not enough.

In a sense, this is probably nit picking - like saying a dragster needs a governor to keep it from going too fast. But new users need to be aware that there is a way to get full contol of their bandwidth, if they desire.

My hat's off to Steve/ QIX for doing a marvelous engineering design that does everything it should.  

T
 

[W1DAN]

Hi:

Yes I agree about Steve's designs. Really nice and one of the best things I built. The big turn-around for me is when Steve came out with the push-pull RF deck. I had built a 6-FET single ended deck, and while it worked OK, the deck got more efficient and with a better output waveform when I converted it to push-pull.

GMS: That's cool you got to meet him. I emailed him once, and did some of his circuits, but that is about it.

Tom: I run a PDM class E rig. 6 pill. I should test it, but have been lazy (and scared to see my numbers!). Actually we can test it in three places: after my switched capacitor LP filter (the one you do not like), after the PDM generator (before the switching transistors), then after the PDM LPF. Eh, I am too lazy, but really should do it. It would take me to get the test audio file and an audio recorder set up. I do have an Audio Precision System 1 test set, but the PC has not been fired up in a few years.

Not much time these days either.

So, why not do THD as well as IM?

Thanks for all the good info.

Dan

[K1JJ]

Cool on the PDM rig, Dan.  Maybe we can do some tune-oriented things to improve the IMD.  I still want to go back and try some stuff now that I'm caught up.

How would you go about testing THD in a linear and class E rig?  I'm game to see how the harmonic distortion looks.   Would THD show something that does not show up in IMD?  IE, if the THD got worse, wouldn't the IMD results get worse too?

T

[steve_qix]

Oh, I forgot to  mention - the PWM generator board has all of the necessary traces, etc. for an 8 pole filter.  I have schematics for this.  The same board is just configured differently, and poof - it's got an 8 pole filter instead of a 6 pole.

I use the 8 pole configuration in one of my PWM rigs, however it is switchable.  So, I switch in the ADDITIONAL 8 poles of filtering ahead of the 6 pole filter in the main PWM generator board (I use a 2nd, old-rev PWM generator board I had lying around for the 8 pole filter - with the filter only installed and no other components).  Gosh, that's 14 poles of filtering!!

[K1JJ]

Steve,

Yep, I can tell when you have that second 8 pole filter inline. The highs are attenuated very well and your bandwidth drops to a hard 5-6 Khz, brickwall.  I heard you using it the other night when the band was crowded.  Good show.


Do you have any more of those old-rev PDM boards?   I'd like to build up an 8 pole filter.  Do you have the parts values available for 4.5  to 5 Khz or is it all standard for the 5.7 KHz?    As I mentioned, my rig starts to roll off at about 7 Khz, so maybe I have an older version of the parts from three years ago.  IE, maybe you tightened up the bandwidth to 5.7 Khz since I bought it.


T  

[W1DAN]

Tom:

THD testing is really a modulator issue. Say you put in a 500kz tone. Look for the presence of 1k, 2k, etc.

I'd like to see the response graph of a 6 or 8 pole filter. Wonder if any peaking or early roll-off occurs.

Dan

[K1JJ]

Dan,

Here's something maybe you can explain....

I put in a single audio tone of 1 KHz and then another test at 7 KHz.  These are SINGLE tones.  On ssb I get just a single peak.  But on AM I get a spectrum as shown in the two attached pics.

What are we actually seeing here?  Is this a 1-tone test with the tone beating against the carrier or what?


Pic #1 is AM using a 1KhZ tone at 100% modulation.

Pic #2 is using a 7KHz tone at 75% modulation.

T