kahn Symmmetra-Peak

[W2PHL]

I just installed one of those ebay all pass filters in my CRL PMC-300a peak limiter. Wow it works!

My voice must be very asymmetrical. In the past, even with the asymmetry control on the 300a turned completely down, I would have to modulate about 140% positive to see negative peaks in the high 80's. After installation I am seeing negative peaks in the high '90s with the positive peaks in the 110-120 range. From what I understand, this is a good thing, increased loudness with less positive peaks required.

The 300a has a +15/-15 PS so installation was a snap. The board interfaced directly to the balanced input of the 300a. Best ten bucks I ever spent!

Question for the group: Where is the optimum place in the audio chain to install this board? It's obviously working in the peak limiter but would it be better at the output of the mic preamp, before the EQ and the Aphex Compellor?

Phil

[ke7trp]

Please take sound recordings phil!

C

[W4NEQ]

I just installed one of those ebay all pass filters in my CRL PMC-300a peak limiter. Wow it works!

My voice must be very asymmetrical. In the past, even with the asymmetry control on the 300a turned completely down, I would have to modulate about 140% positive to see negative peaks in the high 80's. After installation I am seeing negative peaks in the high '90s with the positive peaks in the 110-120 range. From what I understand, this is a good thing, increased loudness with less positive peaks required.

The 300a has a +15/-15 PS so installation was a snap. The board interfaced directly to the balanced input of the 300a. Best ten bucks I ever spent!

Question for the group: Where is the optimum place in the audio chain to install this board? It's obviously working in the peak limiter but would it be better at the output of the mic preamp, before the EQ and the Aphex Compellor?

Phil

All of Bob Orban's hardware processor designs begin with a subsonic low pass filter then proceed directly to the allpass, phase rotator, phase scrambler, or whatever the trendy name is.  This is prior to EQ, compression, limiting, or clipping.

With my own very lopsided voice, I have found that one section helps and two sections completely fix the asymmetry.   I can't imagine needing 8 sections.
 

[k4kyv]

I have looked at recommendations in some of the Kahn literature that I could find on-line and comments on a couple of broadcast message boards.  The consensus seems to be to put the all-pass after equalisation and high/low pass filtering, but before any clipping or peak limiting.  Anything that affects frequency response may affect the phase relationships and thus the symmetry of the signal, but you want the waveform to be as symmetrical as possible before any limiting or clipping action.

That makes perfect sense to me.  I have a tube type Altec multi-channel mixer that I used as a mic pre-amp while I was transitioning my station from the upstairs bedroom in the house to the separate shack, and all my audio stuff was still in the old location.  I recall having to keep the tone controls set to neutral; otherwise the symmetry would be affected.  Turning up the bass would cause low frequencies to extend asymmetrically in one direction, while higher frequencies would extend asymmetrically in the opposite direction.

Something else I have observed is that even if you have the head-room, and phase the positive peaks in the proper direction, when you turn the audio level up high enough to average close to 100% negative, you still get frequent short-duration negative peaks that far exceed 100% negative. The  lop-sidedness of the natural audio waveform of the human voice is not perfect.  Probably not enough to cause a severe splatter problem, but the distortion is still there, along with the danger it poses to the modulation transformer.  A negative peak clipper would be useful, since hard clipping would occur relatively infrequently.

[W2PHL]

All of Bob Orban's hardware processor designs begin with a subsonic low pass filter then proceed directly to the allpass, phase rotator, phase scrambler, or whatever the trendy name is.  This is prior to EQ, compression, limiting, or clipping.

Good to know!

The consensus seems to be to put the all-pass after equalisation and high/low pass filtering, but before any clipping or peak limiting.  Anything that affects frequency response may affect the phase relationships and thus the symmetry of the signal, but you want the waveform to be as symmetrical as possible before any limiting or clipping action.

I figured there might be a difference of opinion on a topic like this. Before peak limiting makes perfect sense to me. Before or after EQ is not as clear. I have another all pass board that I'll test in the mic preamp. Thanks for the input!

Phil

[W2VW]

The eq will change phase relationship within the audio passband. Probably best to use the all pass after it.

[Opcom]

The op-amp and passive solution do the same thing so it might be wise to adhere to the advice from Kahn's brochure and put it at the output of the studio mixer or master control, ahead of the first AGC amplifier or program limiter.

He didn't mention EQ but it seems the rotator should go in before the mike signal is perverted by equalizers or the other instruments.

Is the concern that the EQ will undo the work of the rotator? With these things that each have a definite fixed curve of degrees vs frequency, a result of the presence of the EQ as well as any adjustments made with it are not predictable and the setup must be tried by experiment.

Perhaps a way to test it would be finding a voice sound (a continuous vowel?) that gives a worst case of the 'problem', where low and high frequency peaks coincide repeatably in time, sampling it, and then playing it back in a continuous loop through the audio chain will show via a scope how it works as a whole for each individual case.

[W4NEQ]

Anytime I develop or test audio processing for my own voice, I used a digital recording of my voice made with the same mic I will be using.  Then have a good scope / amp / monitor speaker set up on a high impedance probe to view and listen to various points in the chain.  Changes in symmetry aren't hard to see.

It is quite educational to have the number of allpass poles easily selectable during this process.  The cheap commercial opamp unit has 8 poles, but I'm betting few people will actually need more than 3 to fix symmetry.

Chris