AM Demodulation Quality Flex-6000 VS Others

Bob-N4HY

You have described almost perfectly what happens behind the code veil on AM and Synchronous AM in the code.Thank you for explaining our work and agreeing with what we have done since PowerSDR.

N7AIG

ooohh... interesting web page. I'd love to read your dissertation!

Ken - NM9P

Check the end of this thread. https://community.flexradio.com/flexradio/topics/6300-echo?topic-reply-list[settings][filter_by]=all&topic-reply-list[settings][reply_id]=15379398#reply_15379398

Bob-N4HY

That page hasn't been updated in over 5 years! Hahaha. Let's see if I can even dig up the pdf, which is available from Brown University but I think they charge.

N7AIG

okay... this result is quite disconcerting to me... not yet finished.

When you look further, that square root is a real problem when applied to autocorrelation spectra. It is a nonlinear function with a branch cut.

In simpler terms, going back to the math expansion of I^2 + Q^2 of a double-sided spectrum, and using half-angle identities, e.g., 1 + cos(2*x) = 2*cos^2(x), you actually do find that the end result is a squared quantity. For a single tone we end up with (1 + a*cos(w*t+phi))^2. And taking the positive square root of that produces just 1 + a * cos. High-pass filter away the DC and you have exactly what we always thought just worked. It really does, after all that nonsense.

So, now with differential propagation between the two sidebands and carrier of an AM transmission, you end up with some phase shifting, possible phase cancellation among certain components, and possibly some time delay / advance of some spectral components. But I still don't see anything that might be called phase distortion.

N7AIG

Perhaps, what you call phase distortion, arises from selective attenuation of the carrier when the modulation index is close to 100%. At the receiver there isn't enough carrier present, and it sounds like overmodulation. In that case, you will experience phase flips at random times, and that could sound like phase distortion?

Steve N4LQ

You're way over my head. I just think there are interactions between  dissimilar sidebands. The sidebands are dissimilar as a result of selective fading or multipaths. I called it phase distortion because audio distortion is produced due to the mixing of out of phase sidebands. 
From what I observed, the carrier remained fairly constant in amplitude however I was easy able to observe significant visible difference between the sidebands and when I did, I heard distortion often to the point of total cancellation. Eliminating one sideband seems to reduce the effect. 
I don't have the lab equipment here to simulate this. What's the best way to simulate a multipath AM signal?

N7AIG

My understanding of multipath is that it is sometimes simulated by way of a Rician distribution of noise in phase and amplitude.

Thanks for clarifying your meaning of phase distortion. Agreed, you would see phase cancellation from time to time. You could also hear overmodulation if the sidebands rise and the carrier remains constant -- still produces overmodulation effects on demodulation, since it appears that there is more sideband power than should be permitted for the power in the carrier.

Product detectors, of old, tried to overcome these effects, and you still find the most immunity from phase cancellation and differential attenuation by listening only to one sideband with SSB demodulation. So that much still remains true.

I think what I just found out about envelope detection also offers an argument to give yourself some fading margin by never going fully to 100% modulation. Perhaps the engineers ahead of me were clever enough to realize this, and that's why so many international SW stations appear to be using low modulation indexes?

N7AIG

I just did an experiment here with two signal generators locked to a Rb reference. Tune both to 5.015 MHz, and AM modulate one of them with a 500 Hz tone. I feed both generators to separate short wire antennas and pick up their signals on my receiver in AM detection mode. By setting both to the same received amplitude, and varying the phase of one of them, I can go from doubling up the carrier to completely knocking it out.

Indeed, as I begin knocking down the sum carrier at the receiver, you begin to lose the 500 Hz tone, and a 1 kHz tone arises. 2nd harmonic distortion - exactly as you'd expect to see with overmodulation.

By lowering the modulation percentage from 100% down to 80% and lower, I do indeed buy some fading margin, as it shows I can get closer and closer to knocking out the carrier before the received tone disappears and harmonic products begin to appear.

As a double check, I **** the knock-out carrier, and listen to only the AM modulated tone. I can raise the modulation to a max of 120% on this generator, and it shows the formation of 2nd harmonic distortion.

N7AIG

I can repeat the experiment, but this time use the knock-out generator tuned to the lower sideband at 5.015 MHz - 500 Hz. I adjust the level of the knock out carrier to the same amplitude of the sideband signal.

As I begin knocking down the lower sideband to simulate fading, the 500 Hz tone remains pretty much intact until I get to extreme levels of knock down, at which point I do develop some 2nd harmonic distortion.

But the fading margin for sidebands seems much greater than for knocking down the carrier. I suppose that makes sense.

N7AIG

Actually, I misspoke. I can knock out the lower sideband almost entirely and no distortion appears. The 2nd harmonic distortion appears when I enhance the lower sideband to levels that would correspond to greater than 100% modulation.

Even when the lower sideband is completely gone, I still see a nice clean and strong 500 Hz tone coming from the carrier and upper sideband alone.

N7AIG

So, what these experiments suggest to me, is that for best operation in AM, use somewhat less than 100% modulation to buy yourself some fading margin. Fading is worst between the carrier and the sideband levels. Even losing one sideband to fading has almost no effect on the received signal.

Steve N4LQ

What kind of receiver are you using?

N7AIG

I was using my IC-7700.

N7AIG

I was also monitoring the signal, visually, with a Funcube Dongle Pro+, fed off a splitter just ahead of the IC-7700, and the signal was received from a Pixel Tech mag loop located about 30 meters away from the signal generators.

N7AIG

and for completeness sake, the FCDP+ was watched with HDSDR running on a Mac Mini in Bootcamp Win7/64 mode.

Steve N4LQ

I assume you don't have a Flex-6000x to compare with?

Bob-N4HY

N7AIG, Steve,  NOW do the same lower the carrier with SAM please and report to us the results.  THANKS.

Bob

Steve N4LQ

David. Are you saying whatever these test show applies to the Flex 6000? How's that work?

N7AIG

I'm saying that anyone using the demodulation technique described by Flex should get these same results. Two different SDR demodulators in my shack got the same results. So if the Flex-6000 has a problem then it must be due to something else.

Steve N4LQ

So you were injecting single tones. What about wide-band noise as found on 80 meters instead? Or voice 

N7AIG

Well, the process described is a linear process, insofar as not mixing frequency components in the final result. And the math verifies this to be the case, despite the production of 2nd harmonic products during the demodulation when we form the I^2+Q^2 quantity, those components are also accompanied by DC components so that their sum becomes the square of a DC plus original audio content. That was the point of my comment about half-angle identities. The final square root and HPF gets back the expected audio.

So while I did indeed inject a single tone, there is no reason to expect it to behave differently with a multiplicity of tones.

Unfortunately, I don't have any way of injecting a cancelling sideband of random noise. But I can AM modulate with random noise and **** off the carrier, which as I found, is the most damaging to AM reception. Or I could modulate with a pulse wave and generate a multiplicity of audio sideband components.

Note that my experiment showed only that AM distortion can indeed occur in *any* receiver using the HPF(Sqrt(I^2 + Q^2)) envelope detector that Flex also claims they use. And that distortion arises when the carrier gets selectively faded in amplitude relative to sidebands that then appear to have too much relative amplitude at the receiver.

What actually appears to happen is that the AM demodulation takes as much of the carrier as it can, and the carrier deficiency leaves remaining upper and lower sideband power to beat against each other to produce the 2nd harmonic products. No receiver can be immune to this effect as long as AM envelope demodulation is used. You can defeat this distortion by using SSB demodulation of one or the other sideband.

The experiment showed only that distortion can occur, and under what conditions. We found that using less than 100% modulation gives some margin against carrier fading, and separately that destroying one sideband still allows faithful reproduction of audio.

I don't know the FPGA internal code of the Flex-6000, but I do have tons of experience with fixed-point DSP's, which might behave similarly. Perhaps Flex has instances of amplitude saturation or wrap in their intermediate products and that might be causing the audio distortion that you hear. 

N7AIG

... or perhaps their square root routine is faulty?

N7AIG

... or perhaps there is some nasty interaction with their AGC to cause amplitude modulation of the audio?

KY6LA_Howard

@N7AIG although I am not an AM guy, I am really enjoying reading your analysis. Keep up the good work

Steve N4LQ

Ok...Sounded bad twice.

Steve N4LQ

Do you want to try 20m tomorrow?