Use of phase rotation

[Jeraldo]

Now that a number of software packages offer various forms of phase rotation, I'm interested in knowing what the potential uses are for this feature, apart from fixing problems associated with closely spaced microphones.

Are there uses related to preparing program material, especially classical, and do radio stations (again oriented toward classical music) employ such rotation in transmission?

And then there is that Neve "stereo field" MS box.

Thanks.

[damoore]

It does not fix phase problems associated with differences in mic positions.

It changes relationships between harmonics in the input.

Simple example just in the frequency domain:

Suppose you have a signal in which all the even harmonics are in phase and there are no odd ones. There is a peak twice per fundamental oscillation at which all the harmonics are adding up.

If you shift the phase of the harmonics you can avoid that peak. Which means you can go louder.

In practice (i.e. given that real sounds don't have constant Fourier transforms over time) there are common signals, like voice, that have strong (positive) correlations across frequencies. That translates into high peaks relative to the overall level. Phase rotation breaks up those correlations by shifting the time at which certain frequencies are heard.

Conversely if the correlations are already weak for a given sound source, its not going to do anything useful.

[dmills]

It is one of the more benign things you can do in a broadcast airchain, so most broadcast processors are configured to do it (It makes quite a big difference in speech broadcasting apart from anything else).

Regards, Dan.

[Hugh Robjohns]

I'm interested in knowing what the potential uses are for this feature, apart from fixing problems associated with closely spaced microphones.

As others have said, it doesn't fix problems with spaced mics -- that's a pure delay issue! Phase rotation is about altering the coherence of signal elements at different frequencies to alter the waveform shape (but not the perceived sound). Typically it is used to help make asymmetrical waveforms more symmetrical, which generally allows more gain before clipping. it is a routine and essentially benign form of processing employed by virtually all FM radio broadcasters.

H

[Mixedup]

I'm interested in knowing what the potential uses are for this feature

I tend to use these things quite a bit when dealing with multi-miked electric guitar recordings. It gives you control over the phase relationship between the signals from different mics/DI inputs... which in turn can alter the sound in a manner not dissimilar to EQ, but which often seems a little less intrusive to me.

[Jeraldo]

........

Phase rotation is about altering the coherence of signal elements at different frequencies to alter the waveform shape (but not the perceived sound). Typically it is used to help make asymmetrical waveforms more symmetrical, which generally allows more gain before clipping. it is a routine and essentially benign form of processing employed by virtually all FM radio broadcasters.

H

So if this were applied to stereo material, each channel would by nature be independently processed, and this doesn't do anything perceptually to the stereo image? (Probably the answer is no because of its use by broadcast media?)

And if the results are benign, why isn't this routinely done for any program material for any kind of distribution (CD's, downloads, etc.) where compressing dynamic range by peak limiting might be useful?

(And have you or anyone figured out the Neve box which purports to use phase rotation on the mid component of a stereo source to perceptually move the center forward or backward? Evidently it does do something perceptually in this case, even only if relative to the side element.)

There are some issues I may be confusing:
Phase rotation employed by radio transmission I would assume to be a dynamic, constantly changing scenario.

With rotation with a software plugin, is this generally a fixed amount for all the material-a non dynamic process?

Is "dynamic rotation," as employed by radio transmission, commonly available in a plugin?

I don't quite understand the "changed waveform but lack of perceptual effect." I may be confusing individual sources with mixed sources. Some signal processing boxes (the Mico) uses rotation to produce perceptual effects (I think).

Sorry my limited understanding of the issues is leading to questions based on limited or incorrect assumptions.

Perhaps there is an SOS article waiting to be written that would explain everything to know about phase rotation.

Thanks to you and others for the information and clarification.

[Jeraldo]

One more:
Phase rotation offered in Rx2, for example, offers a few choices.

Each channel individually may be rotated by a chosen degree. When might one want to rotate the phase of one or more channels by an arbitrary amount?

There is also a "Vari Phase" mode to maximize levels in a stereo file. This is what I meant by a dynamic process. Why isn't this procedure used as a matter of course-or is it? I've never used it, so I can't comment on any perceptual effects. Same question, though-is each channel being rotated independently or is this a parallel operation based on the greatest peak level of either channel?

BTW, some time ago I experimented very briefly and not at all systematically with rotating the phase of the mid component of a stereo program, with little apparent effect. So the Neve box remains a bit of a mystery. But I also don't understand why rotating the phase of the mid component would lead to varying depth effects.

[dmills]

I think we are seeing several meanings here:

As used in broadcast a 'phase rotator' is classically nothing more then a allpass filter set up to have a steep phase shift at about 1Khz or so.
The effect is simply to prevent the even harmonics from stacking up in the time domain allowing you get a little more loudness for a given average level of limiting. Not so much benign as benign compared to hitting the limiter that much harder.
The early ones were entirely passive RLC networks placed in a 600 ohm audio line feeding something like a CBS Volumax or such.

In the plugin sense, the term usually refers to using a Hilbert transform to convert the signal to an analytic pair, then rotating the resulting vector by multiplying with a unit vector or appropriate angle before outputting the real part. This does not have the group delay change that is an essential part of the broadcast phase rotator, and so is quite a different effect.

An interesting application can be in reducing the level differences caused when panning a mono source and then reproducing in mono, as introducing a 90 degree phase shift between the two outputs drops the level change to 3dB from 6dB.

Note that if playing with this stuff in M/S space, you need to carefully compensate the delays if the result is to work halfway right.

Given how precious folks get about audio, would you want to introduce something that takes actual grown up math to understand even the basics of into a place where the audiophile crowd will notice and have all sort of conspiracy theories about, you think the great 'records Vs cd' debate has more legs then is ever reasonable......

Regards, Dan.