Audibility of Absolute Phase

[RHosch]

I know this has been discussed numerous times. I know that tests have established that absolute phase is audible with certain signal types, such as (IIRC) triangle waves.

Have there been any good explanations offered for how this is possible with steady state AC signals of any form? It is intuitive that transients might have absolute phase audibility... but steady state signals? Even asymmetry in the driver's forward/rearward excursion behavior shouldn't create audible differences if the signals are symmetric (which all are, if decomposed).

[SY]

Polarity, not phase!

Any asymmetry between push and pull of a driver will manifest itself as a differing distortion spectrum and quantity with changes in absolute polarity, assuming an asymmetric signal. The data showing audibility on specially created test signals is solid; the data on audibility in real rooms with music and reasonably low distortion speakers is much more ambiguous.

Dick Greiner, who wrote one of the definitive papers on the subject, concluded that on real-world music reproduction, it's probably a minor issue, maybe even a non-issue, but so easy to get right that it's worth doing.

[Bill Fitzpatrick]

Originally posted by SY
so easy to get right that it's worth doing.

How so? I have a scope so I can check the amp and preamp for inverting or not but what about the signal leaving the CD player?

[SY]

Greiner was speaking about the entire reproduction chain, from mike to speaker.

[RHosch]

Polarity, not phase. Check.

How can you assure anything at all about the recording chain? I'm not just talking about making sure the polarity of all equipment is correct, but more fundamentally how do you know that the microphone is picking up the "correct polarity," whatever that is. I mean, with wavelengths being a few inches or so, on average, typical differences in mic placement could easily have the mic pick up a vocalist on a particular note "180 degrees out of phase." If she moves her head, does that not change the "polarity?"

But that really wasn't my question. Why is it audible? I assumed that the explanation involved asymmetric signals and non-linear driver (and room, obviously) response in the two directions... but how are the signals asymmetric? Decomposed any signal is just sine waves. Within the bandwidth of any speaker, any signal is just the sum of symmetric signals. Of course transients can be different, but I'm talking about steady state repeating signals. Hasn't audibility been established for those as well?

[analog_sa]

An interesting topic. I agree that absolute polarity is audible but not more so than the difference between passive components or wire. It seems to be 'officially' recognised as audible only because it's easy to understand and detect.

[MBK]

How can you assure anything at all about the recording chain? I'm not just talking about making sure the polarity of all equipment is correct, but more fundamentally how do you know that the microphone is picking up the "correct polarity," whatever that is. I mean, with wavelengths being a few inches or so, on average, typical differences in mic placement could easily have the mic pick up a vocalist on a particular note "180 degrees out of phase." If she moves her head, does that not change the "polarity?"

No. A varying distance to the microphone will produce group delay ( delay of all frequecies by the same time). You could also have phase shift, often varying with frequency. But polarity refers to whether the microphone picked up a pressure or a vaccum. No matter what distance you are at, the pressure part of the sound wave will remain the pressure part, and the vaccum part will remain the vacuum part.

Some instruments, say, horns, seem to have severely asymmetric wave forms. For instance the wave could be quick on the up stroke and slow on the down stroke. If you change polarity, you exchange pressure with vacuum, and regardless of driver distortion you now have a completely reversed soundfield *if* your wave was asymmetrical to begin with. And recording studios, good ones, go to great lengths to preserve absolute polarity - if only to avoid cancellations due to time delay when mixing.

As to

Decomposed any signal is just sine waves.

... *in the long run*. You can decompose any signal into a combination of signals *steady state*. The Fourier analysis serves as a mathematical tool, not as a sensory explanation of what actually happens.

For instance you excluded transients, but for the duration of the "upstroke" of a hypothetical asymmetric triangular wave, even a steady state one, you do have a transient upstroke which has nothing to do with a sine wave, and which doesn't necessarily look like like the downstroke. A hypothetical square wave of 0.1 Hz fundamental, while "composed of nothing but sine waves", actually consists of 10 second alternations of DC. Not that I claim audibility of that. Just as a thought experiment.

Why is it audible?

Because the mammal ear seems to work as a highly nonlinear, actively amplified zero-crossing detector, see for instance

Hearing mechanisms

which explains in one stroke potential tolerance of fairly high THD with simultaneous low tolerance for crossover distortion, and the audibility of absolute polarity, and the subjective preference some people have for single ended amplifiers.

[SY]

Originally posted by analog_sa
An interesting topic. I agree that absolute polarity is audible but not more so than the difference between passive components or wire. It seems to be 'officially' recognised as audible only because it's easy to understand and detect.

It is "officially" recognized to exist because it is detectable under controlled (i.e., blind) conditions, at least with certain signals and certain systems.

How can you assure anything at all about the recording chain?

You can't. Otherwise I wouldn't have so many records and CDs that sounded so terrible.

but how are the signals asymmetric? Decomposed any signal is just sine waves.

Yes, and each of those sine waves has a coefficient! If you've got a mike, an oscilloscope, and a trombone, you can convince yourself that asymmetrical sounds exist.

[planet10]

Originally posted by MBK
... *in the long run*. You can decompose any signal into a combination of signals *steady state*. The Fourier analysis serves as a mathematical tool, not as a sensory explanation of what actually happens.

A crazy Viking chats with Mr. Fourier

dave

[SY]

I'd hate to correct Fourier (and even more so, a math whiz like you), but you can indeed`deconstruct the symphony into a Fourier series. You merely assume that the wave's period is the length of the symphony. That's how a single impulse standing all alone among 4096 points can have an FFT...