Does Phase Distortion/Shift Matter in Audio? (no*)

[j_j]

Also, it is interesting that people will argue about audible differences between cables, where there is pretty much no differences to measure, and at the same time have no issue with nonlinear phase, where the temporal distortion is easy to illustrate. Audibility is of course up for discussion, but it would be easier to argue for audibility of phase than of cables.

Well, my policy on cables remains, if a cable makes an audible difference compared to another cable, at least of them is bad.

As to temporal distortion audibility is an interesting problem, because, in fact, most (but not all) physical processes are minimum phase, which has phase shift, but which tends to create slowly changing temporal distortion across frequency. You don't get pre-echo, you don't get weirdness from allpass filters, and the like. (Note, USUALLY.) Digital processes can do many things a native acoustic process can (literally) not do (maximum phase filtering, for instance, a sound that often weirds people right out, all still remaining in the domain of strictly linear processes.

But again, until you look at how things like onset and envelope change inside ERB's and across adjacent ERB's, the question is very, very difficult to even start to examine, especially in the lack of substantial results on measuring co-articulation.

 

[Holmz]

I am on board with this. Also, it is interesting that people will argue about audible differences between cables, where there is pretty much no differences to measure, and at the same time have no issue with nonlinear phase, where the temporal distortion is easy to illustrate. Audibility is of course up for discussion, but it would be easier to argue for audibility of phase than of cables.

And its sister argument of…
- The audio signal not needing to match the electrical signal in phase/polarity
- Cables that measure exactly the same, can have additional magic which makes them sound different.

 

[René - Acculution.com]

Well, my policy on cables remains, if a cable makes an audible difference compared to another cable, at least of them is bad.

As to temporal distortion audibility is an interesting problem, because, in fact, most (but not all) physical processes are minimum phase, which has phase shift, but which tends to create slowly changing temporal distortion across frequency. You don't get pre-echo, you don't get weirdness from allpass filters, and the like. (Note, USUALLY.) Digital processes can do many things a native acoustic process can (literally) not do (maximum phase filtering, for instance, a sound that often weirds people right out, all still remaining in the domain of strictly linear processes.

But again, until you look at how things like onset and envelope change inside ERB's and across adjacent ERB's, the question is very, very difficult to even start to examine, especially in the lack of substantial results on measuring co-articulation.

Well, to be clear, the pressure output of a loudspeaker is not minimum phase in general, although it is a physical system. The crossover alone ensures that, as it sums several transfer functions, and although they may be minimum phase on their own, their sum is not in general min ph (as the crossover typically aims to be an allpass filter when summed). Something like a step response will look very distorted and clearly shows the temporal issue, but we are lucky enough that we are not too sensitive to that. I will show a lot of this in the upcoming presentation. I am just afraid that people will not like a 2 hour lecture on this, as I have seen complaints about much shorter videos. But there is just no way around getting in to complex numbers and transfer functions to properly explain the many aspects of phase.

 

[j_j]

Well, to be clear, the pressure output of a loudspeaker is not minimum phase in general, although it is a physical system. The crossover alone ensures that, as it sums several transfer functions, and although they may be minimum phase on their own, their sum is not in general min ph (as the crossover typically aims to be an allpass filter when summed).

Not all crossovers, no. Sorry.

There are many ways to build filters, analog or digital, and it is not necessary to have that pi/2 phase shift at the crossover frequency. (or 3/4 pi, or some other options) It just isn't. There are other considerations, as well, involving radiation pattern, etc, to deal with. There is much beyond 2nd (minimum acceptable steepness) or 3rd order butterworth to be had, as well as considerations of how much peaking, etc, is permissible.

Phase is a very simple thing, I'm not sure why you argue otherwise. On the other hand, unwrapping phase in an observation can be very interesting in some systems. Ever done radio equalizers?

 

[René - Acculution.com]

Not all crossovers, no. Sorry.

There are many ways to build filters, analog or digital, and it is not necessary to have that pi/2 phase shift at the crossover frequency. (or 3/4 pi, or some other options) It just isn't. There are other considerations, as well, involving radiation pattern, etc, to deal with. There is much beyond 2nd (minimum acceptable steepness) or 3rd order butterworth to be had, as well as considerations of how much peaking, etc, is permissible.

Phase is a very simple thing, I'm not sure why you argue otherwise. On the other hand, unwrapping phase in an observation can be very interesting in some systems. Ever done radio equalizers?

It is not about the phase shift at the crossover frequency. The phase as a function of frequency for most loudspeakers cannot be calculated from their corresponding magnitude response, so it is not a minimum phase system. The sum on axis will approximate the sum of the crossover plus driver sections. If you for example take a linkwitz riley two way filter, and ignore for simplicity the drivers, the magnitude sum is flat, but the phase is nonlinear, whereas it should have been flat (zero) were it a minimum phase system. I think you have misunderstood something here? This will all be cleared up in the presentation. I think @KSTR has already made posts on this topic already, so I don’t plan to clear it up here.

 

[amirm]

I am on board with this. Also, it is interesting that people will argue about audible differences between cables, where there is pretty much no differences to measure, and at the same time have no issue with nonlinear phase, where the temporal distortion is easy to illustrate. Audibility is of course up for discussion, but it would be easier to argue for audibility of phase than of cables.

Many crimes are committed in the name of correcting phase/timing in audio such as low order crossovers, stepped drivers, etc. That is the problem and focus of my video.

 

[j_j]

It is not about the phase shift at the crossover frequency. The phase as a function of frequency for most loudspeakers cannot be calculated from their corresponding magnitude response, so it is not a minimum phase system. The sum on axis will approximate the sum of the crossover plus driver sections. If you for example take a linkwitz riley two way filter, and ignore for simplicity the drivers, the magnitude sum is flat, but the phase is nonlinear, whereas it should have been flat (zero) were it a minimum phase system. I think you have misunderstood something here? This will all be cleared up in the presentation. I think @KSTR has already made posts on this topic already, so I don’t plan to clear it up here.

Funny, you brought up crossovers. Then you don't want to talk about crossovers. Since I've measured a few hundred drivers, loudspeakers, systems, etc, yeah, I'm well aware that there are some very interesting storage elements in loudspeakers, as well as delay shifts between drivers. That leaves out port resonances, etc, which are at best "interesting" but that are quite easily observed in many loudspeakers.

I am rather well aware of the performance of quite a few physical systems, from loudspeakers to the human ear, and I have the credentials to prove it, so yes, I do understand what "phase" is. You continue to direct me to your basic tutorial, but I suggest that you might want to move to another level to hold this discussion in the future.

 

[j_j]

Many crimes are committed in the name of correcting phase/timing in audio such as low order crossovers, stepped drivers, etc. That is the problem and focus of my video.

Well, and then one must ask "and, if you do this correction, over how much space does this correction work". Further, do we measure pressure, velocity, or all 4 variables?

(I'm well aware you know about this, but it is truly discouraging how many people don't consider more than just pressure.)

 

[gnarly]

I'm guessing there is some simple miscommunication going on.....
I've been following you guys for a while, and know dang well that you all completely understand that drivers are predominately minimum phase devices, and IIR xovers with their all-pass phase rotations aren't. Kindergarten for you guys !
Only wish I could follow all of you folks deeper stuff....

Anyway, I do play with phase a lot...FIR and linear phase in particular.
Here's an example of the sections of a DIY 4-way main tuned to a 0-30 deg horiz average (for use with a sub crossed at 100Hz).
I'm showing it to try to gain credibility for the attachment after it, which i think you'll find funky and hopefully interesting.

Ok, this one is a 3-way's acoustic output, where each section has a fader on its output level.
The xovers are linear phase 96 dB/oct LRs.
I wanted to see how phase would hold together with rather drastic magnitude boosts and cuts.
Green trace is of course level outputs.
Isn't purple magnitude really whacked out? As phase stayed together.

I realize there are probably some averaging/math issues that muck phase up more than is shown...no doubt more to learn.
Hope that was interesting...

 

[KSTR]

Many crimes are committed in the name of correcting phase/timing in audio such as low order crossovers, stepped drivers, etc. That is the problem and focus of my video.

Be that as it may, would you consider to change the title to "Does Phase Distortion/Shift Matter in Audio? (it depends)"?

 

[thorvat]

You do that test the other way round. Take a system with proper "full-blown DRC" and resulting smooth FR and flat phase response from each speaker. Then introduce a analytically calculated phase distortion which mimics for example the phase rotations from a typical 3-way passive speaker. This can then be A/B tested (or ABX) and then most people will find that phase distortion is audible and can be detected (once you know what to listen for and have the right music tracks that highlight the effects) but the difference is very subtle, the proverbial icing on the cake.

I tried exactly that. It didn't even come close to a proper A/B test as I was switching the filters myself, but I would say I wouldn't be able to recognize the difference. On the other hand, my room is quite "reflection happy" and my speakers have wide dispersion. For example, I need to apply FDW of 3 cycles if I want a phase recorded from my LP app 4 meters from speakers to be presented in REW without rolling.

I would say that the difference would be more easilly spotted listening from shorter distances, with narrow directivity speakers and with a less "lively" room. It would still be very subtle, but my guess is it would be noticeable.

That is probably the reason why sound engineers aim for such setup in their recording studios. While I personally find that setup of course optimal for that purpose I know more than a few sound engineers who don't actually use that setup in their homes for casual listening, where they prefere speakers with wider directivity and untreated room (often their living room) which is of course more "reflection happy" than studio rooms. IMHO in such setups some of the perfect imaging is sacrificed for wider and "livelier" soundstage, so one should choose according to own personal preferences, as no universal solution will ever be perfect for everybody.

 

[Thomas_A]

FYI I made a few ABX of polarity here

Absolute Polarity - Myth or "Important"?

I'm really confident it's a matter of training and preconditioning. The polarity ABX test is my go-to warm-up test when doing any kind of analytical listening and testing. On some days it just doesn't work... Number #1 rule of ABX'ing: The moment you feel any, even subtle kind of stress or...

www.audiosciencereview.com

 

[gnarly]

FYI I made a few ABX of polarity here

Don't mean to be cutting you off, but i think polarity is its own topic, and doesn't really fit into discussions about phase audibility.
But maybe that's just me.....

 

[Thomas_A]

Don't mean to be cutting you off, but i think polarity is its own topic, and doesn't really fit into discussions about phase audibility.
But maybe that's just me.....

Well, it has nothing to do with the claims made by Paul in the video that Amir debunks regarding requiring extreme bandwidth of amps etc. But have you read this whole thread?

 

[thorvat]

Well, it has nothing to do with the claims made by Paul in the video that Amir debunks regarding requiring extreme bandwidth of amps etc. But have you read this whole thread?

I have read the whole thread, and absolute polarity has nothing to do with it.

In fact, to anyone even remotely familiar with the mixing process it is clear that "absolute polarity" has nothing to do with anything.

 

[Thomas_A]

I have read the whole thread, and absolute polarity has nothing to do with it.

In fact, to anyone even remotely familiar with the mixing process it is clear that "absolute polarity" has nothing to do with anything.

So an allpass filter is not about phase?

 

[thorvat]

So an allpass filter is not about phase?

"Absolute polarity" is not about phase. It is meaningless term.

 

[Thomas_A]

"Absolute polarity" is not about phase. It is meaningless term.

But introducing a filter that shifts the phase is, right?

 

[gnarly]

But introducing a filter that shifts the phase is, right?

Hi, first, pls know i meant no offence with saying polarity is not really about phase.

And yes, a filter that shifts phase is about phase...... when by 'shift' we mean on a relative frequency basis...one frequency's phase shift vs another.
A polarity switch does not shift the relative phase between phases...it simply inverts everything across the spectrum.

Another yep, I read the whole thread....great thread

If the definition of phase distortion is any rotation or warping of phase between frequencies / across the spectrum
Then for me, posts addressing phase ability like J.J.'s in #429 when he talks about phase shift within an ERB, are the kind i think gets to the heart of the matter...

 

[Thomas_A]

Hi, first, pls know i meant no offence with saying polarity is not really about phase.

And yes, a filter that shifts phase is about phase...... when by 'shift' we mean on a relative frequency basis...one frequency's phase shift vs another.
A polarity switch does not shift the relative phase between phases...it simply inverts everything across the spectrum.

Another yep, I read the whole thread....great thread

If the definition of phase distortion is any rotation or warping of phase between frequencies / across the spectrum
Then for me, posts addressing phase ability like J.J.'s in #429 when he talks about phase shift within an ERB, are the kind i think gets to the heart of the matter...

My point was that of audibility of asymmetric signals when the waveform is changed in critical bands as demonstrated in thread. Flipping it will also be audible. For some at least.