Even Better PureSignal Performance on your Apache Labs radio

[K1LSB]

Wow!!

After some experimenting with two-tone frequencies I was able to improve my IMD3 by more than 6 dB on 40 meters (as measured in Thetis)!

Per my Thursday post in this thread I had measured an IMD3 of -63.26 dBc on 40 meters at 81 watts into a dummy load, using 1143 and 1969 as my two-tone frequencies.

Today I ran the exact same test again, but this time I was using 943 and 1969 as the two-tone freqs, and now Thetis says my IMD3 is -69.61 dBc!

Capture-1.jpg (81.71 KiB) Viewed 128 times

That snapshot was taken after letting the test stabilize for 5 seconds.

The only change I made between yesterday and today was to modify one of the two-tone frequencies.

FWIW, here's my AmpView graph at the time of the test.

Capture-4.jpg (45.51 KiB) Viewed 128 times

Mark

[w-u-2-o]

You are not "improving" your results by doing this unless that you can show that performing a PS single cal operation with those specific tones causes the IMD products with a real-world signal, like human speech in SSB mode, are better than performing a single cal with the standard 700 & 1900 Hz tone pair. And, if during normal operations, you are using PS automatic mode instead of single cal, the performance obtained with a given tone pair (standard or otherwise) is only an approximation of the real world performance.

To compare results both parties must use the same tone pair. The selection of tone frequencies is somewhat arbitrary and may vary depending on the specific measurements being made, but in all cases the tones should be harmonically unrelated, and in most cases (not all) the same amplitude.

Just to muddy the waters further consider the following list of commonly used tone pairs:

- 700 Hz and 1900 Hz: Common for audio equipment (and the default for Thetis).
- 60 Hz and 7 kHz: Standard SMPTE IMD test.
- 19 kHz and 20 kHz: Standard for measuring high-frequency IMD in audio.
-Fc ± 5-10 MHz: Common for high-frequency RF systems.

[K1LSB]

Okay, here's a screenshot of a full 100 watts out to a dummy load on 80 meters just a moment ago, using Single Cal (notice the "feedback Level" in the Linearity window is not lit, proof that Single Cal is in effect).

The -65dBc IMD3 reading is much better than any of the charts in this thread show for 80 meter measurements.

Capture-3.jpg (145.35 KiB) Viewed 123 times

And you haven't explained why using different two-tone frequencies can produce different results, you simply implied that the improved results I'm seeing aren't meaningful.

Mark

[w-u-2-o]

Mark--unfortunately that screen shot proves nothing, other than the PureSignal algorithm can obtain better results with that tone pair vs. other tone pairs. It's irrelevant if spoken audio is used in automatic mode, and probably irrelevant in any case.

To prove your selected tone pair has any real value for IMD improvement then you have to perform the experiment that I described. This is not that hard. You can use the peak hold feature of Thetis to capture the spectral response obtained with human speech on SSB while using a given single cal result. Ideally it would be best to use the same recording for each measurement, but I suspect it won't matter that much.

[K1LSB]

Scott,

I'm not sure what experiment it was that you described...if you're referring to comparing test results using my new two-tone frequencies to the old test results using different frequencies, I've already done that on multiple different bands and posted the results in multiple threads in this forum.

I have no idea why you (or anyone else) are such an advocate of Single Cal. That feature is nothing more than a one-time correction curve calculator, and thus renders the very notion of "adaptive predistortion" effectively meaningless.

As I've said repeatedly, I always let PureSignal correct in real-time for any changes in the measured nonlinearity in my voice transmissions. That's what it's made for!

My strong contention has always been that Single Cal will not produce the cleanest signal during voice TX at any given instant, precisely because Single Cal is not real-time "adaptive" correction!

With that said, I do of course realize that any particular choice of two-tone frequencies has absolutely no bearing on the real-time (not Single Cal) operation of PureSignal.

Mark

[K1LSB]

I have to go ride my bicycle while this awesome weather is here, I'll be back later.

Mark

[K1LSB]

.
.
You can use the peak hold feature of Thetis to capture the spectral response obtained with human speech on SSB while using a given single cal result. Ideally it would be best to use the same recording for each measurement, but I suspect it won't matter that much.

Here's what I think you're asking for: You want a snapshot of the cumulative TX Peak Hold envelope across a swath of spectrum scope, to include the passband and extending for a few KHz on both sides of the passband, representing a given duration of voice transmission. Is that correct?

If so, I can't find any Peak Hold option for TX, or am I just missing it?

Is there maybe another way to accomplish the same thing?

Mark

[w-u-2-o]

Here's what I think you're asking for: You want a snapshot of the cumulative TX Peak Hold envelope across a swath of spectrum scope, to include the passband and extending for a few KHz on both sides of the passband, representing a given duration of voice transmission. Is that correct?

Correct.

But it's only worth doing the experiment if you feel that adjusting the tone pair frequencies has some sort of value with respect to PureSignal performance.

As you've pointed out, using the automatic, real-time correction mode uses the human speech audio stream to guarantee the amplifier is driven over it's entire dynamic range, a requirement for PureSignal to make the necessary measurements. This will, of course, result in results that are slightly different than with two-tone. This is true regardless of the selected two-tone frequencies, default, custom, or otherwise.

Thus the custom two-tone frequencies you've selected are only valuable if you believe the performance achieved with them carries over to normal operations. And the only way I can think of to do that is to use single cal to capture those measurement results and then apply that cal to normal voice SSB operations. There would need to be two such trials, one with the results of the default two tone pair, the other with your custom tone pair.

If so, I can't find any Peak Hold option for TX, or am I just missing it?

The "Peak" button is adjacent to the CTUN button. The screen capture has to be done before switching back to receive.

[K1LSB]

Thanks Scott, I'd hidden that Peak button long ago because I never used it and I've since forgotten all about it, lol.

You are of course 100% correct that no combination of tone pairs will have any effect on PureSignal voice correction.

One of the reasons I was playing with the tone pairs was to see if that had any effect on the results of a two-tone test, and it certainly does, as I've observed on all the bands I've tested. And I don't believe that is a consequence of the particular test instrument used (external SA versus Thetis' internal test utility), but rather a consequence of the particular amplifier being tested -- it's been seen repeatedly that some amplifiers just don't like certain two-tone combinations. And IMO to use any particular arbitrary tone pair as "the de facto standard" is doing a disservice to any particular amp that doesn't like that tone pair, as simply using a different tone pair will show improved results with that amp in an IMD test.

I know that using a non-standard tone pair will negate any comparison of results between different machines, but that's actually only true in the case that either single party is forced to use the "standard" tone pair -- if both parties are free to use whatever tone pair their amp is happy with, then a comparison of test results between parties is again valid.

The other reason I was playing with tone pairs was to try to find out what is the best IMD numbers that my radio is actually capable of producing in a two-tone test. And in that light I'm very pleased with the improved numbers I'm seeing.

Regardless, I must concede that your suggestion of using the Peak Hold button to gather a full history of the corrected transmitted signal is probably the most meaningful way to capture real-world results (though that approach would also necessarily capture any instantaneous energy "flyers" that would unavoidably cloud the resulting picture).

With that said, here's a snapshot of 40 seconds of voice transmission at 100 watts on 75 meters into my cantenna dummy load. I'm awarding PureSignal a solid "60" for its efforts, given that the slope of the IMD skirts intersect the passband at roughly -80 on the scale, and the passband peaks are at roughly -20 (IMO any very slight bulging at the skirt slope intersect points is due to artifacts from the Thetis display Windowing function and not IMD, but I could be wrong).

NOTE: I have my waterfall adjusted to flow very slowly ... my monitor is capable of displaying over 2 minutes of waterfall history so what you're seeing in the screenshot is my full 40 seconds of voice transmission.

Capture-3.jpg (58.3 KiB) Viewed 42 times

Mark

[Edit] That transmission was not made using Single Cal, it was made using PureSignal in normal real-time adaptive correction mode.