The AM Forum

I’ve got to ask a question that is somewhat embarrassing, but I always remember what my Mom told me – the only dumb questions are those you won’t ask! Just hooked up a scope to monitor the transmitter output. On AM, I’m wondering if the balanced audio cables are wired correctly, or is the end result audio out of phase. Exactly what am I looking for that will tell me if I’m in phase?

One way is to note on the scope whether your negative peaks are cutting off the carrier before the positive peaks reach 100%. If they are just swap the wires at the connector and try again.

Very Good explanation.....Get a feel for the way the display looks on the 'scope. Doesn't take much audio to see that you're hitting excessive NEGATIVE peaks. Negative modulation is hitting the center line of the 'scope OR the BASELINE.
Positive peaks are extending to the top and bottom of the CRT display.

Hitting that baseline on a high powered, or even desk top TX, is shortening the life of the mod transformer and casing splatter on the Ham bands

The two screenshots are from a video I made of Dirk - WA2CYT as he flipped the phase of his Viking II back and forth. The scope is plugged into the 455kc IF of my FRG-7700 and the audio in the recording is speaker-to-microphone. In the video you can readily see the positive peaks rise when the phase is set optimally and how they are restrained when the phase is not set optimally for Dirk's voice. Another observation that has always held true when observing the envelope display on a scope is that when the phase is optimally set the pattern looks like a Christmas tree that is on its side with the top pointing to the right. When you look at the two screenshots it should be easy to spot which one displays the phase setting that is optimal. The link to the video is below.

Rob W1AEX

https://www.youtube.com/watch?v=pvowzE_XYl0 (https://www.youtube.com/watch?v=pvowzE_XYl0)

(Note that Dirk's transmitter uses the WA1QIX 3-diode "keep alive circuit" so you will not see it reaching 100% negative in either phase.)

Thanks so much, that really makes it easy to identify

There's a wild card in this voice audio phase business...

Everyone's voice is different AND our voices often change throughout the day and over the years. My phase could very well be the opposite of yours, even on the same rig.  This may mean that at times our phase selection is optimum and then it seems the other way is correct. (at various frequencies). This can cause confusion and frustration.  For example, when I wake up in the morning and test my phase, the low end is heavier and will often look out of phase compared to the higher freqs.  This is when a phase sensing circuit will actually select the proper phase for our voice on the fly.

Then there's some lucky guys who never have to touch anything for years. They usually sound like K1KBW and there is no doubt where the testosterone is banging the vocal cords and phase is a cinch to determine.... :-)

In addition, if the low end of the transmitter is poor and has distortion, a certain phase choice for our voice may be best to cover up problems. But put it through a class E rig with response down to 1 Hz and you may see the "shark fins" effect where the phase looks totally different.

In my case, I know I have it correct when I see the shark fins and the positive peaks can kiss 170% easily.

Some guys have voices that are easy to determine the correct phase, while others have hardly any asymmetry in their voices.

We once did a voice test with about eight hams all standing in the shack (was it at QIX's QTH?) using the same mike. We measured NATURAL voice asymmetrical effects. Voices ranged from a low of 110% (me) to as high as 140% (Big Al JCL) asymmetrical.  

Bottom line is to experiment after getting your rig to pass audio as cleanly and flatly as possible. The cleanliness of the rig has as much effect as our voice characteristics.  Once you get the best combination of clean equipment that will pass 1 Hz to 12 KHz, equalizing suited to your voice, microphone, etc., you will know it via heavy positive peak capability and great unsolicited audio reports on the air.

T
 

I added a phase reversal switch on the mixer so it's easy to see what's going on. Thanks!

Here is a related article that is hot off the press by Jim Tonne w4eme:

http://tonnesoftware.com/appnotes/allpass/Allpass.pdf

Jim
Wd5JKO

Jim is a genius.

Rod

Vocal asymmetry is frequently opposite at lower vs higher pitches as a result of the changing physical shape of the vocal system and the difference is often apparent when uttering deep "aaaahs" vs higher "eeeeeeees".

One way to get a more definitive phase display is to use a High Pass Filter or Low Cut Filter to eliminate the conflicting (opposite) phase that works against the mid energy frequencies.

Also be sure to set the mic phase with all processing either bypassed or below any compression, limiting, or clipping levels for an accurate reading. Some processors sample 1/2 the waveform for gain reduction causing a difference in output based on phase. A negative peak clipper will completely mask an out of phase condition by simply cutting off the greater negative peaks (with lots of distortion).

I actually have one of those all-pass units that makes the audio symmetrical. They are available from China for about $50 or so.

Symmetrical audio gives the ability to run a denser waveform, which makes sense. So I gave it a try.  After playing around with it a few days, I pulled it off line and went back to my standard negative peak limiter (CRL PMC-300A) and large positive peaks.

It is probably just a matter of my liking to watch the peak reading wattmeter go higher and seeing the big positive peaks on the scope.  Ham Fantasies.  (HF)

Clark - very true about our voice low frequencies often having the opposite optimum phase than the highs. This can cause frustration for sure. I see it on my voice.   Without a limiter, the lows will do a natural 140% but the highs do 110% before negative peak clipping.  The CRL helps a lot here and works harder to keep the highs under control.

T

You have any info (links)? Aside from the visualized symmetry, is there any audible difference, and if so can you elaborate?

I bought an All-Pass filter for less than $20 with free shipping...

http://www.ebay.com/itm/Audio-All-Pass-Filter-Phase-Rotator-Module-Board-/130681036896

...although I don't see one posted today. The board is mounted in a 1RU rack chassis with an active balanced I/O circuit. It needs to precede any compression.

Every b'cast processor made in the past 30+ years has one as an aid to competitive loudness without adding compression artifacts. In other words, it gives your compressor less to do, so less gain reduction and recovery are needed.

I have never heard any difference in the audio. The wrong phase might cause a little distortion on the bass frequencies,,,,,,,,,,,,dunno

Positive peaks are what we're shooting for.
Thanks "QIX" for the youtube link. It's very visibly obvious what happens.

If the received signal-to-noise ratio is high (e.g. lot's of radiated output power from the transmitter, and good propagation conditions)... as it was for people listening to my signal a few years ago, on 75M, when I was running legal limit power from by Bernardsville location:

Sounding good was (for me) a matter of:

a)  having a transmitter with a reasonably linear relationship between the audio input amplitude and the modulated RF envelope output amplitude

b) having a flat frequency characteristic from 50Hz to 5kHz (i.e. the sine wave audio input amplitude at frequency f vs. the corresponding modulation of the amplitude of the RF envelope at frequency f), and

c) Using a properly adjusted (to match my voice and my microphone) audio chain.

Back then, I adjusted the "phase" and amplitude of my audio input to obtain ~125% positive peaks and ~100% negative peaks... without the requirement for negative peak limiting to prevent over modulation in the downward direction.

The loudness (audio density) of my signal, at the receive end, was not very critical (because of the high received SNR)... and I could adjust the modulation for 125% positive peaks just because I liked to see 125% positive peaks on my oscilloscope and on my modulation monitor.

That was then...

Now, with propagation conditions being what they are (and using a lower power transmitter and a not-so-efficient antenna at my new location):

I wander if "sounding good" necessitates a different tradeoff between having larger positive modulation peaks vs. symmetrical modulation (e.g. using the phase rotator that is built into my audio chain) with a higher density of modulation of the carrier.

Stu

Due to their short time constant, peaks contribute nearly nothing to perceived volume and can destroy mod iron in tube transmitters or RF components in any transmitter. Most receivers clip excessive positive peaks anyway, making a station sound distorted.

Positive peaks look great on mod monitors but that's a hell of a tradeoff.

Actually, correct phase typically causes distorted lows because the negative clipper gets driven significantly harder. All-pass filters reduce that effect.

Are the issues surrounding "proper phase", and having lots of asymmetry remnants of a time long past (Auld Lang Syne) when converting an asymmetrical waveform to a symmetrical waveform was not technically feasible (as per Jim Tonne's article)

Stu

I'm interpreting this thread as a discussion of "If asymmetry then how much?"

Rather than take my word for it, here's the lesson learned 50+ years ago in the professional broadcast world:

http://www.w3am.com/SymmetraPeakBrochure.pdf

For those who desire a total understanding, here's the patent:

http://www.w3am.com/SymmetraPeakPatent.pdf

100% of AM broadcast processors include an all-pass filter. I like experimenting but not reinventing the wheel...

One of the interesting things that occurs with the AM mode when running the pre-distortion linearization protocol (Pure Signal) with OpenHPSDR and the ANAN hardware is that the algorithm perceives asymmetry as distortion. While running AM without Pure Signal engaged you can create a signal with very impressive positive peaks that looks very pretty on a scope. With Pure Signal engaged, the correction algorithm will assert itself so that the same signal will hold the line at 100% positive peaks with a typical 3rd order IMD improvement of around -20 dB and an additional bonus with the transmitted THD that drops from ~ 3% to 0.2%. I can't really hear any difference in the received signal when a station toggles the algorithm on and off but it is very evident when you watch the signal on a scope.

Rob W1AEX

 "like experimenting but not reinventing the wheel..."

From the DSP based Optimod 9400 manual:

"OPTIMOD-AM precisely controls peak levels to prevent overmodulation.
Asymmetry in the analog processing channel is adjustable from 100% to 150%
positive peak modulation."

The "wheel" has remnants of Asymmetrical capability for a reason, OM!

One thing is for sure: The cost of "loudness" by stripping asymmetry to achieve a tightly compressed signal very close to 100% constantly (with less gain reduction), is the loss of natural dynamics! After all, we aren't selling advertising spots to listeners using 2" speakers commuting to work!

The one huge positive (no pun intended) for the all pass symmetrical approach (at 100% is the improvement of selective fading with a conventional detector. The higher the sideband amplitude (at higher positive peaks), the more the selective fade effects.

Makes me wonder if all the folks listening to Orson Welles would have been even MORE frightened if ONLY he had the symmetry right..

That I agree with and I'm not advocating -100/+100. My point is that all-pass filters reduce asymmetry going in to the processor which lets the final clipping stages provide more consistent positive modulation. If nothing else, the dramatic opposite polarity negative peaks from lower vocal tones won't reduce gain or drive the negative clipper into noticeable distortion and resulting splatter.

Yes, I use -100 to +125...

Quote from Jeff... W2NBC

"OPTIMOD-AM precisely controls peak levels to prevent overmodulation.
Asymmetry in the analog processing channel is adjustable from 100% to 150%
positive peak modulation."

The "wheel" has remnants of Asymmetrical capability for a reason, OM!"
..........................

Could it be that the folks at Orban include, in their product, those capabilities that their target customers want (expect to find)... independent of whether they (the folks at Orban) believe those capabilities are useful or helpful.

As an example: my Optimod 9200 (as of software update version 3) includes the capability to set a positive peak limit for "night" operation that is independent of the positive peak limit that can be set for "day" operation.

Stu

I own a 9200 as well Stu. The Orban method and description of "loud" audio hasn't changed much through the years.. Based on their own analysis, I think you're right on the money! They may have the 150% as a positive limit for the warm and fuzzy feeling of familiarity ..   ;D


 

When using processors that allows you to set a percentage of assymetry it sure is nice having an all pass in front of it. 

The all pass I use has from 4 to 8 poles and accomplishes what I would equate to almost a couple dB of multiband compression.  Maybe 2 channels.

It sure makes my single band compressor work better as to spectral density when I preceed it with the APF.

It does add a bit of Optimus Prime sound to the audio,  though. 

--Shane
KD6VXI

Group delay (e.g. due to the transmission of a sound wave through air) is a specific type of all-pass filter... that rotates the phases of the different frequency components of a multi-frequency waveform by different amounts. But, group delay will not change the shape (e.g. the asymmetry) of a multi-frequency waveform... it will only produce a time delay. I.e. the amount of phase shift experienced by each frequency component is linearly proportional to the corresponding frequency.
 
One could reduce the asymmetry of a sound pressure wave... that is produced when someone speaks... by passing the sound pressure wave through a multipath acoustic channel (which is not an all-pass filter, because components of the sound pressure wave at different frequencies will experience differing amounts of multipath fading) ... such as a typical room with acoustic reflections from walls and other objects.

Alternatively, as has been discussed,   one could employ an electrical all-pass filter... following a microphone... that rotates the phase of each component of a multi-frequency waveform by an amount that is not linearly proportional to the frequency of each of the multi-frequency waveform's components.

Stu

The best way to be louder is to run more power and/or have a better antenna. The rest is mostly a matter of taste.

The distortion compensation of the recent SDR transmit chains is not to improve the on channel audio quality. It's to reduce/remove the off-channel RF that could result in interference. Given most receive audio chains have several percent (or more) THD, even if there were on-channel improvement, you couldn't hear it.

Steve

I respectfully disagree, in part, with your post (above)

I agree that pre-distortion is primarily employed to reduce out of channel odd order distortion products that are introduced in a non-linear RF amplifier chain.

However, if the RF amplifier has a significant amount of non-linearity (i.e. output envelope vs input envelope)... as in the case of my Elecraft KXPA 100 when operated at 100W peak output on 40m...
then there will also be significant amounts of in-channel odd order distortion products that correspond to distortion of the demodulated audio.

It is easy to see (on a scope)/hear (with my off air monitor) this distortion when operating 40m AM. The KXPA100 produces about 3dB of compression at full 100 watt output (3rd order IMD only about 20dB down in a two-tone test).

See the attached curves of amplifier output envelope v. input envelope (blue) and Pure Signal (adaptive predistortion) linearity correction (red).


Without pre-distortion applied, the off-air monitor output sounds very distorted.
With pre-distortion applied (around the KXPA100 amplifier), the off air monitor output sounds excellent.

Stu

If the pure signal digital / mathematical algorithm has little problem removing distortion from RF then it has even less problem removing distortion from the concomitant modulation envelope controlling the RF at much lower frequencies.

By superposition theorem, the Audio and RF waveforms are now one waveform, choppable in minute discrete intervals for manipulation.

My technology tale of woe:    ;)

As a side note about "Pure Signal" and the new RF digital technology...  I find it curious that I spent a ham lifetime experimenting with many, many conventional homebrew RF amplifier configurations looking for the cleanest, perfect high power signal without splatter, etc.  The best I could do was about -30db 3rd order IMD at best. When the amplifier was pushed hard, it got nasty down below -25db 3rd.

After almost giving up over 30 years, I eventually settled on a Class A low level system that drove cascaded class AB systems, giving over -50dB 3rd IMD at high power. In 2012, I had finally arrived! For once my high power linear signal was truly "linear."

Then the signal processing software became mainstream. Now a crap CB splatter-maker amplifier can be made to run as clean or cleaner than my low efficiency, pristine nightmare. It's not fair!  :D

I thought I had finally achieved something that was unattainable to the average ham and I could have pride in running it on the bands - the elite few!  But time and technology march on. Nobody cares or knows anymore how the signal was generated.

This pattern has repeated itself throughout history and will continue. Just think of the ham rigs we thought were hot-sh@t in the 50's, 60's, 70's, etc.

T   (dragged thru time kicking and fighting)

It is interesting to think about (but perhaps not for too long) the difference between

case 1: y(t) = [a m(t) + b m(t)**2 + c  m(t)**3 ...] x [cos(2pift)]  

and

case 2: y(t) = a [m(t)cos(2pift)] + b [m(t)cos(2pift)]**2 + c [m(t)cos(2pift)]**3 ...


where m(t) is: 1+ an audio waveform whose peak magnitude is less than 1

Case 1 corresponds to the low-pass filtered output of a high-level AM modulated (e.g. a Class C plate modulated transmitter or a Class E drain modulated transmitter) transmitter... where the modulation process is not perfectly linear.

Case 2 corresponds to the output of a (not perfectly linear) RF linear amplifier driven by an ideal AM modulated RF input signal (before using a low pass RF filter to attenuate RF harmonics)

It is even more interesting to think about (but definitely not for too long) how adaptive predistortion would be implemented in Case 1 or Case 2.

Stu

It's a perfect solution but the units are rare now and the good quality inductors are as rare in the values that work. As an exercise I drew up an electronic version using tubes in a Philbrick GAP/R K2-W op-amp circuit and otherwise based on that 8 op-amp circuit board shown. It would not be exactly compact with sixteen 12AX7s and a regulated +/-300V 50mA supply. I have not built it so anyone who uses the schematic would have to debug it themselves. It's mainly presented an example of what could be done with tubes if someone wished it.

All radio receivers contain "automatic volume control" functionality. In most radios (naturally including traditional AM radios), the AVC circuitry does not respond to (capture and hold) the actual peaks of the i.f. output signal

Depending upon the details of how the actual "peak" detector is designed, it responds to a short term average of the envelope of the received signal.. which lies somewhere between the true peak and the long term average (carrier) level. [Note that, by analogy, most power meters, including high-end power meters, do not respond to and/or display the true the peak envelope power of an AM signal... even though they are specified to be peak reading... because their internal peak detector is not fast enough to capture and hold the actual peaks].

If you reduce the carrier level (the average of the received envelope) of an AM signal... the AVC will compensate by increasing (to some extent) the gain of the i.f. and/or audio amplifiers.... even if the peak RF envelope stays the same.

When someone tunes across the "stations" with an AM radio... without touching the volume control... the stations with a higher ratio of peak envelope power to carrier envelope power will sound louder... due to the action of the AVC circuitry.


That says nothing about the effect of having high positive peaks (v. negative peaks) on whether the received audio sounds good or not.


This effect will hold for SSB as well as AM if the AVC/ALC/AGC circuitry of the receiver is not fast-acting enough to capture and hold the actual peaks of the modulated received signal.

Stu

The above (my last post) indicates that if you want to fight in the loudness wars... with a given PEP limitation... you want adjust your voice processor to produce the largest ratio of positive peaks-to- whatever characteristic of the modulated signal produces the smallest response from a typical AM radio AGC "peak" detector.

Taking advantage of the AGC behavior, in this way, is conceptually equivalent to reaching across to my receiver, and turning up the audio gain control, when it is your turn to transmit.

Stu

Analog AVC in those am radios are r/c integrators, that is they use a real interval or passage of time as the needed dimension to operate.
Come to think of it, we all do.  ;D

Information is therefore averaged .  Informatiom integration also allows lower signal levels to be utilized.

Further support for thoughtfully clipping the positive peaks.

On another web site: Yahoo Groups... Apache Labs Group,

Paul (W9AD) posted the following excerpt from the Orban 9300 manual (note the double negative at the beginning of the excerpt)



"Have a look at the Orban Optimod-AM 9300 operating manual, referring to page 1-15:

 “There is really nothing lost by not modulating asymmetrically: Listening tests easily demonstrate that modulating symmetrically, if time dispersion has been applied to the audio, produces a considerably louder and cleaner sound than does asymmetrical modulation that retains the natural asymmetry of its program material.”

 http://www.orban.com/orban/products/manuals/9300_1.0.5_Operating_Manual_rev_02.pdf "

This is an example of a customer driven approach to selling a product:

My interpretation: 

Our processor can be adjusted for 125% positive peaks... as you have requested, when talking to our sales and marketing folks (but its really not good for you).