My bet with WA2CYT and modulation patterns

[steve_qix]

Modulation patterns - or are you in a betting mood 

Dirk, WA2CYT (a VERY old call !) likes to watch people's modulation patterns on his receiver connected oscilloscope.  He is very keen on pointing out when one's modulation is "out of phase", and there have been NUMEROUS discussions to this effect on 75 meters during the day. We have had many fun and lively discussions about this.

For SEVERAL years now, Dirk has labeled the modulation patterns of myself (WA1QIX), Bob K1KBW and several others as
"the class E modulation pattern".   This pattern can best be described as a Christmas tree pointing to the left (the big end is to the right, in other words).  This pattern shows up in audio systems that have little or no phase shift, flat low frequency response, and generally a small to moderate amount of bass boost.

Now, I have had this pattern literally for 20 or more years (at least since 1982 or so) when I started using the audio processing equipment I'm still using today.

So, today old Dirky was saying that someone had the "class E modulation pattern" and it must have to do with the class E rigs because all the class E rigs show that pattern, etc. etc. etc.  So *I* broke in and made a bet (witnessed by sevaral other operators).  Dirk, if I set up a vacuum tube transmitter, and it has the same pattern, will you promise never to call that pattern the "Class E pattern" again ?? 

Well, good old Dirk took the bet, and I said I'd be back in an hour.  I went to work modifying an old DX-60 (that I had never fired up before) for good audio.  An interesting task, although quite straight forward.  After rewiring the modulator section and performing some simple tests into a dummy load, I fired up the rig after about an hour and ten minutes - and - lo and behold !!!  The "Class E Modulation Pattern" on a tube rig !!!  Gasp !!

Needless to say, this was not a big surprise.     The transmitter in question (the now modified DX-60)was flat down to at least 10 Hz, and the positive peak capability was WELL over 150%.  The distortion was low, and the sound, as heard in the modulation monitor output, was reasonably transparent.

However, it does raise an interesting question about modulation patterns, and "phase".  I always run, regardless of the shape or orientation of the pattern, my audio, such that there are more positive peaks than negative peaks.  That's how I do it.  The shape is secondary.   And, it does not appear to be affected by voice.  On my system, everyone's voice is "in phase".  Some have more asymmetry than others, but NONE have even required a phase reversal.

Thoughts?  Ideas?

Steve C.....

[K1JJ]

Steve,

Interesting.

I know this subject has been debated for a long time. I'd like to hear more opinions. I'll start...

First of all, the look of the waveform phase has nothing to do with class E or tube, that's for sure. It's all about the particular voice characteristics - assuming the rig is clean and flat.

When I switch 180 degrees phase here on any transmitter, I can produce the same "shark fin" audio pattern OR the more "symmetrical" pattern. (for lack of better labels) It doesn't matter if it was my previous class E rig or a flat tube rig.

However, for me, the shark fin, so called, "class E pattern" is out of phase with my voice. I must use the other (180 degree out) symmetrical pattern to get the highest peaks. But my particular voice is only about 110% asymetrical according to the tests we ran at your place with the group that time. So it can be very difficult for some people to decide which phase is correct for me by listening on the air. I can easily see it on the scope.

Both phases each have a distinctively different sound too. Once the headphones are swiched to compensate, I find the shark fin pattern to have a more resonant sounding low end, for better or worse.

In fact, the phase difference with my voice is so subtle that there have been months I've run it one way with most guys saying it's in-phase... while the next month I'd run it in the other phase and they would say it still sounds correct.

Whereas, when Dirk looks at my voice on the scope he will always say the shark fins are out of phase, while the class E guys say the shark fins are IN phase for me. I notice almost every class E guy runs his audio with the shark fin phase look. I know some guys say this is out of phase...

Go figure. I prefer to run mine with the symmetrical looking phase regardless of the rig, class E or tube... No shark fins-   "symmetrical" gives me the best sound and highest peaks here..

73,
T

[WA1GFZ]

All I know is Steve and I both sound better on his audio system. I think maybe PDM or dc coupled series modulation has the advantage of no phase shifts due to transformer responses. I would take a bet that it has nothing to do with the RF stage. ..

[steve_qix]

Last night I was too tired to write this, but something else I've observed over the years.

Assuming a flat, phase linear modulator - if I apply bass boost, or use a "basey" microphone, my audio is in phase with the so-called class E pattern (the shark tooth - Christmas tree pointing to the left).

If, on the other hand, I *cut* the low end, my audio wil need to be reversed - the opposite pattern is "in phase".

In phase being defined as that phase which yields more positive modulation, on average, than negative.

I've speculated for many years that different parts of the human voice's frequency spectrum are in different phases.  I suppose it makes sense, being that it is a very complex waveform full of multiple frequencies.

On the dx-60:

The DX-60 I modified for this original test puts *OUT* about 10 watts of carrier for about 36 watts in.  These operating parameters being necessary to achieve the high positive peaks I want.  I can get more power output (and somewhat higher efficiency) at a sacrifice in positive modulation capabilities.  Not being overly familiar with the DX-60, are these power levels more or less customary?  The thing runs at about 60mA at 600 volts input, and I measure about 10 watts into a calibrated dummy load.

The dx-60 design has a number of flaws, not the least of which is the resistor/capacitor combination in series between the modulator output and the screen.  This is necessary, given the modulator design, to allow the transmitter to be modulated 100% negative - where the scrren of the 6146 output tube must be driven approximately 20 volts negative, with respect to the cathode.  Since the screen characteristic curve is non-linear, the average screen voltage will change with modulation  - this due to the non-linear screen voltage/current curve, causing the average voltage drop across bypassed dropping resistor to change, with average modulation, resulting in a certain degree of carrier shift.

I fixed this by using a 36 volt zener diode bypassed with a capacitor in place of the resistor/capacitor combination.  I also added a negative peak limiter, clamping the negative peaks at around 90%.  With these, and a redesign of the voltage amplifier preceeding the modulator tube, the transmitter generates a very respectable, albiet low power, signal.

Regards,

Steve

[steve_qix]

All I know is Steve and I both sound better on his audio system. I think maybe PDM or dc coupled series modulation has the advantage of no phase shifts due to transformer responses. I would take a bet that it has nothing to do with the RF stage. ..

Oh yes !  Nothing whatsoever to do with the RF stage - that was the point of the "bet" 

[KL7OF]

Very interesting.......Any chance to see some "shark fin" scope shots?

[WBear2GCR]

Hear! Hear!

Please post some shots of the scope display(s)!
Identifying the X & Y input signal source would be helpful as well...

     

          _-_-WBear2GCR

[K1JJ]

Steve/QIX said:

"I've speculated for many years that different parts of the human voice's frequency spectrum are in different phases.  I suppose it makes sense, being that it is a very complex waveform full of multiple frequencies."


I think that's it exactly!!!

If you listen to a voice like W1VZR, you will hear a very powerful peak around 125hz??-    Like some others with his strong resonant, bassy voice, Pete has a very dominant peak there that is likely very asymetrical. His is an audio that needs to be phased correctly.

But just as probability allows, every voice is different, thus there will be voices that need one phase for the lows and another phase for higher freqs, etc. The wisdom of running a real-time phase sensing and switching device (like the Optimod) can make sense.

There are certain voices like third BA Bob/KBW that do great justice to the shark fin pattern. And then there's voices like mine where it makes little difference.

This would explain why there is so much debate about this subject. If everyone had the same voice characteristics, there would be no debate and we all would require the same phase. ie, On AM, tailor your phase settings to your own voice.  (BTW, it makes no difference on ssb)

73,
T

[Steve - WB3HUZ]

Yup. I'm having trouble visualizing the scope display based on the descriptions given.

Ref phasing of different frequencies of the voice: Bacon designed a processor many years ago that had bass phase advance. It would change the phase of the bass frequencies so you could achieve more positive peaks. Hopefully, he will jump in here and explain it. Also, many commercial processors use a series of all-pass filters to remove voice asymmetry before doing any dynamics processing, Might be interesting to see which sounds better or louder, an asymmetrical processed voices or a symmetrical processed voice.

It would be good to know how Dirk was monitoring AM received signals, i.e. what was the IF bandwidth, shape factor and phase delay, was the receiver AGC engaged, if so what were it's response time characteristics, etc? I've found many receivers mangle the transmitted waveform to the point that giving anyone a report based on a scope display from the IF is about useless.

[steve_qix]

Here are the two modulation patterns - the one of the right is the so-called "class E pattern" also the "Sharktooth" pattern.

The one on the left is more of what you will see with most transmitters.

Bottom is the modulation monitor display of the right-hand (sharktooth) pattern. Notice the very high asymmetry!  Both patterns show a similar asymmetry.  Both patterns are of my voice, with different equalizations.  The one on the left is my voice with the bass cut, and in phase.  The pattern on the right is with normal equalization (bass bosted somewhat), and in phase.

[Steve - WB3HUZ]

Thanks for the pix. Make sense, more bass, a wider horizontal waveform. Wouldn't expect anything different.

[steve_qix]

Thanks for the pix. Make sense, more bass, a wider horizontal waveform. Wouldn't expect anything different.

The other key feature is the DIRECTION and phase of the waveforms. The two waveforms shown are out of phase with each other !! But, they are "in phase" with respect to modulation.  I should have made that more clear!  I had to flip the phase when I cut the bass.  I do have the waveform out of phase (with respect to modulation) also.

Regards,

Steve

[Bacon, WA3WDR]

I think it's the extended low end that causes the waveform seen there.

Steve, you use an omnidirectional electret condenser mike, if I recall.  I find that a little bass rolloff enhances asymmetry, like a 1 meg load on a D-104, etc.  The rolloff results in some low-end phase advance that affects the relative timing of the voice fundamental and its first overtone (second harmonic), and this usually adds to voice asymmetry.  But the deeply extended low end that comes from your omnidirectional electret condenser mike does not advance the low frequency phase, and this produces a disitinctly different waveform.  Of course different voices produce different waveforms as well.

If anyone wants to play with low frequency phase shifting, it is possible to adjust low frequency phase without having to use low-end rolloff.   Here is a circuit that does this, and it is not hard to build.

AM Press/Exchange, July, 1988
"IMPROVING POLAR ASYMMETRY, The Phase Shift Technique"
http://amfone.net/AMPX/62.html

The relative timing of the voice fundamental and its second harmonic results in the slant of the composite asymmetrical waveform.

[Steve - WB3HUZ]

Good stuff. Thanks Steve and Bacon!

[K1JJ]

If anyone wants to play with low frequency phase shifting, it is possible to adjust low frequency phase without having to use low-end rolloff.   Here is a circuit that does this, and it is not hard to build.
AM Press/Exchange, July, 1988
"IMPROVING POLAR ASYMMETRY, The Phase Shift Technique"
http://amfone.net/AMPX/62.html

Interesting.

This is a problem I've seen for years when boosting the low end. I will build up this circuit after I get the main rigs going here. It looks simple enuff. Being able to slide the bass phase around to compensate might be worthwhile.

TNX.

T

[Bacon, WA3WDR]

I have been using that circuit since around 1988.  It works.  The weak point in it is the mid-voltage reference; it is brutal to put a capacitor on an op-amp output like that.  It might be good to add a few hundred ohms in series with the op-amp output to the rest of the circuit, equivalent to lifting the output pin out of the socket and sticking a 330 ohm resistor in series with it.  But that exact circuit works fine for me!  I think I have a TL-082 in it.

[WA1GFZ]

There is an op amp application around to drive a high C load putting the series resistor in the loop. I think it was in the old national books. It is done to keep the amp from high frequency oscillatois.

[W2VW]

Does this all mean that Dirk has to wear a dress to the next Hosstraders?

[steve_qix]

Does this all mean that Dirk has to wear a dress to the next Hosstraders?

Gee, I never thought of that !!!

I should have thrown in that he has to remove his hat 

[Bacon, WA3WDR]

The combination of the low end response of omnidirectional electret condensor mics, and AM transmitters with bottom ends down below 20 Hz, is unprecedented in amateur radio.  The resulting speech waveforms are relatively new, and pretty much only seen on PDM stations.  It is the reason for this that was really in question.

[K1JJ]

The combination of the low end response of omnidirectional electret condensor mics, and AM transmitters with bottom ends down below 20 Hz, is unprecedented in amateur radio.  The resulting speech waveforms are relatively new, and pretty much only seen on PDM stations.  It is the reason for this that was really in question.

Very well stated.

I have four FT-102's here (with full NE-602 DUQ mods) that use their balanced modulators to go down to a few cycles - cleanly. The shark fin pattern is quite evident when switching phase and passing extreme voice lows..

T

[Bacon, WA3WDR]

Hmm, shark-fin.  (Imagines du-dat du-dat du-dat from 'Jaws')

True, that's where else it would be seen - low-level modulators with that sort of microphone.  Those mikes have been around for a while now.  So maybe it's not as new as I was thinking.

It is possible to break that waveform down into sine waves of various amplitudes and phases.  Then by advancing the phases of the lowest frequency components and recombining them all, other possible waveforms can be produced, that have the same spectral characteristics.  That's basically what I did on my Commodore-64 back around 1988, using graphics, and adding various combinations of sin and cosine of f, 2f, 3f, 4f, etc, and displaying the results.

I have been thinking of speech and audio processing that dynamically adjusts the relative phases of various frequency bands for desired polar characteristics.  It might be possible to optimize for lower peak-to average, possibly allowing for a certain degree of asymmetry, etc.

One thing that I noticed - my ear definitely likes audio properly polarized on the air.  Since the mike and the speaker may well have reversed polarity, and since the frequency response is definitely not flat, I have to assume that this effect has to do with detector linearity, or maybe peak distortion that results from weak delayed components attacking the sharp negative peak if the waverform is polarized wrong.  And maybe it's all in my head - because I knew if the peaks pointed up or down, and of course I prefer the upward direction.  I don't know.  Upward-pointed peaks sounded less distorted, that's all I know.

To some extent it's like the worry about downward modulation. We get all upset if a modulated signal drops a bit with modulation - but we are perfectly happy if it bumps up a bit with modulation.  Meanwhile, it can be better if the level drops a lot with modulation - because then the receiver AGC will introduce significant gain reduction between words and syllables, and the signal will sound quieter to the ear.

[Ian VK3KRI]

Last night I was too tired to write this, but something else I've observed over the years.

Assuming a flat, phase linear modulator - if I apply bass boost, or use a "basey" microphone, my audio is in phase with the so-called class E pattern (the shark tooth - Christmas tree pointing to the left).

If, on the other hand, I *cut* the low end, my audio wil need to be reversed - the opposite pattern is "in phase".

In phase being defined as that phase which yields more positive modulation, on average, than negative.

I've speculated for many years that different parts of the human voice's frequency spectrum are in different phases.  I suppose it makes sense, being that it is a very complex waveform full of multiple frequencies.

Thats interesting.  Ive noticed that running recorded audio into the TX , that the assymetry driection seemed to change with eq.  I made a mental note to sit down and think about exactly what was happening , but I never did.

I wonder if fiddling the eq/phase shift at various freqs to increase assymetry  might result in an increase in percieved loudness or S/N ?                                                      Ian VK3KRI

[Bacon, WA3WDR]

Some people talk about higher modulation density with phase randomization.  ('Phase randomization' is done with a number of all-pass phase shifters in a row, set to a few different frequencies.)

Probably phase optimization for some moderate asymmetry would produce even better results.  It may be that some frequency ranges have different asymmetry, with some voices and some microphones.  Lining that up would be a good idea as well.

I had 1.8:1 to 2:1 asymmetry, and there were problems.  Of course there were big positive peaks to support, and also many modern receivers with peak-oriented AGC compressed it, and made it sound like crap. One guy really complained about the compressed quality, and I told him to turn off his AGC and use manual gain, and that fixed it.  The problem was the SSB-oriented, peak-oriented solid-state receiver AGC, compressing me all wrong, even while receiving in AM mode!  In comparison, AM-oriented receivers tend to have average-oriented AGC that doesn't bounce with modulation.

W2ZM Bob was saying that yes, you could have those skinny stringbeans sticking way up, but was that really louder than phase randomization with higher modulation density?

I have been thinking that it is really better to optimize with a more moderate asymmetry like 1.4:1.  With PEP-oriented power limits and the realities of peak power handling in a station, it is probably better to have reasonable asymmetry with good modulation density.

[K1JJ]

Speaking of experimenting....The Huzman once designed me an "artificial asymetrical audio enhancer". (For lack of a better name)

It consisted of two op amps with audio fed in parallel and summed at their outputs, with one running more gain than the other. You could artificially increase the audio peaks of one side vs: the other.

I built it up and got it working FB. The result was the ability to take my voice which was naturally 110% positive and make it look like 180%++... :-)

It looked quite good on the scope, but in the real world it produced a crackling effect due to whatever... overloaded detectors, intentional distortion, etc.

It was an interesting concept and idea, anyway... but now sits in the cellar crap pile with some of my other bright ideas.

T