• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

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

It was largely accurate, it just reeked of a sales pitch or AI.

It just reeked of 'not what we want at ASR' to me.

We have plenty of real intelligence here.
 
The technical term for “frequency dependant phase” is “Group Delay”.
Slight correction: frequency-dependent phase is normal and is not directly equivalent to group delay; group delay is the (negative of) the change in phase over frequency. If the change in phase over frequency is linear, then the group delay will be a constant (single number), and all frequencies will be delayed equally.

HTH - Don
 
Slight correction: frequency-dependent phase is normal and is not directly equivalent to group delay; group delay is the (negative of) the change in phase over frequency. If the change in phase over frequency is linear, then the group delay will be a constant (single number), and all frequencies will be delayed equally.

HTH - Don

Exactly, which is why I say endlessly that you unwrap phase, then do a first-order fit, and subtract that out. The remains is the frequency-dependent part that is not pure delay.
 
you unwrap phase
Only if it were that easy. I could not yet find a single tool that unwraps phase correctly for all possible impulse responses and Matlab's unwrap is one of the worst.
 
I could not yet find a single tool that unwraps phase correctly for all possible impulse responses and Matlab's unwrap is one of the worst.
Agreed! It gets especially bad where there are multiple wraps between measured points.
 
Agreed! It gets especially bad where there are multiple wraps between measured points.
This is true. What I've found is that you're usually fine as long as you exclude any part of the phase response that has close to no energy in the spectrum.
 
Reading this and trying to understand all the different terminology and theory and possible issues is challenging and makes me want to try some speakers with first order crossovers :). I know that higher order crossovers are easier and the mechanical requirements of the drivers much lower so much cheaper drivers can be used, but if modern materials and engineering were applied to the drivers and the enclosures and the wave guides would it really be that much harder and more expensive? Seems like first order filters eliminate a lot of potential issues and problems. Unless of course the OP is correct and phase does not matter.
 
Reading this and trying to understand all the different terminology and theory and possible issues is challenging and makes me want to try some speakers with first order crossovers :). I know that higher order crossovers are easier and the mechanical requirements of the drivers much lower so much cheaper drivers can be used, but if modern materials and engineering were applied to the drivers and the enclosures and the wave guides would it really be that much harder and more expensive? Seems like first order filters eliminate a lot of potential issues and problems. Unless of course the OP is correct and phase does not matter.

There's more to it than that. You also have much, much wider overlap between drivers, which means much more potential strangeness in radiation pattern, more likelihood in the lf driver (or the lower frequency driver in a mid/tweet combo) to be operating above the point at which the driver is acting as a piston, also creating uncertainty in radiation patterns, protection for higher levels in the HF driver.

It's not just the question of phase response.
 
First-order crossover filter pairs are also not in-phase at all frequencies.
Interesting, how does that work if a first order filter doesn't change the phase of either signal of a pair? It make sense to me that mechanical differences in the drivers could cause relative phase changes but electronically it seems like they should stay in phase? Filter theory reminds me of thermodynamics where intuition is not helpful.
 
Interesting, how does that work if a first order filter doesn't change the phase of either signal of a pair? It make sense to me that mechanical differences in the drivers could cause relative phase changes but electronically it seems like they should stay in phase? Filter theory reminds me of thermodynamics where intuition is not helpful.
Here is a simple first-order crossover with low-pass (LP) and high-pass (HP) filters at 1 kHz. The HPF starts at 90 degrees falling to 0 degrees at high frequency, while the LPF starts at 0 degrees and falls to -90 degrees. The phase is not the same though the slopes are the same.
1729301373416.png


1729301391177.png
 
Last edited:
Interesting, how does that work if a first order filter doesn't change the phase of either signal of a pair? It make sense to me that mechanical differences in the drivers could cause relative phase changes but electronically it seems like they should stay in phase? Filter theory reminds me of thermodynamics where intuition is not helpful.
Their sum (at least with one of the two choices of polarity a) s+w0/s+w0 and not b) s-w0/s+w0) is exactly one with zero phase. So in total, the phase is unchanged. It is often said that you can only do this with first-order, but that is not true. You can do it to any order, but as m-way, meaning no longer two-way.

Filters that have all parts be "in-phase" will have a total of that phase also, so the sum might be one in magnitude but not zero phase, such as LR filters. So the overall question in this thread is whether or not it matters if the total is one exactly with zero phase, or one in magnitude and any old (time-constant) phase. I am pretty sure that a time-varying phase would be audible...
 
@j_j "Is it audible? I believe so, actually, on very "pitchy" sounds, perhaps a clarinet, maybe some very bright brass, or some individual's voices. Is it very audible, probably not, and we are all trained to hear right through it."
You are right, the phase shift at the speakers crossover is fully audible.
I have only one mitigation of this problem: I need to design a new passive crossover that is 3rd order (I.e.: C L C) for each driver.
It is taking time as I have some 3 way speakers but the end result is a real improvement of the sound quality that you described in your post.
 
Exactly, which is why I say endlessly that you unwrap phase, then do a first-order fit, and subtract that out. The remains is the frequency-dependent part that is not pure delay.
How about fitting a 3D Heyser spiral curve based on the magnitude and phase of the target transfer function?
1729329353614.png
 
Phase shift does nothing for me, sounds the same. Some reviewer claimed to recognize if phase were switched in a blind test. This made me wonder if this test could be beaten?
With known musical material, trained analytical hearing, maybe it can. I dont have the language or understanding to analyze sound at a higher level, but to spot difference in music material could be easier even for a layman. So maybe you could spot a single passage there phase shift is pronounced, even in a very, very subtle manner.
I got this Idea trying to arrest a sampleconverter from altering sounds, playing reference music over and over at different clock rates. Inconclusive.

This would only and eventually work with music more than well known to the auditor.
 
So maybe you could spot a single passage there phase shift is pronounced, even in a very, very subtle manner.
In my opinion, absolutely yes to this point, as this was my own experience, being highly critical and "knowing" that this should not be audible. When several instruments had to 'hit' (whatever the musical term is) at the same time, you could hear them "in-sync" in one setting (Kii3s, phase linear), and not in the other. Very subtle, yet something that made all the difference in the blind test, and perhaps something you could grow sensitive to(?).

Regarding phase, all of the information regarding delay is hiding in it. If there is any 'strict' delay in a system, there will be a corresponding linear phase. This goes into the so-called Excess part of the phase, that might contain some other non-minimum aspects. They add up, as I illustrate below, where the excess phase from a LR4 filter (allpass system) can add to the phase from a delay and make up the total phase.
Unavngivet.png


Phase is tricky to understand. Perhaps this helps: https://audioxpress.com/article/simulation-techniques-misconceptions-in-the-audio-industry
 

Rene, first off let me say that I am appreciative of your efforts to educate us. But I have to say something about your article: I am sure there is something valuable in there, but all that maths makes it impenetrable for those of us who do not have much math literacy. Would you consider another article but written in a more accessible way?
 
In my opinion, absolutely yes to this point, as this was my own experience, being highly critical and "knowing" that this should not be audible. When several instruments had to 'hit' (whatever the musical term is) at the same time, you could hear them "in-sync" in one setting (Kii3s, phase linear), and not in the other. Very subtle, yet something that made all the difference in the blind test, and perhaps something you could grow sensitive to(?).

Regarding phase, all of the information regarding delay is hiding in it. If there is any 'strict' delay in a system, there will be a corresponding linear phase. This goes into the so-called Excess part of the phase, that might contain some other non-minimum aspects. They add up, as I illustrate below, where the excess phase from a LR4 filter (allpass system) can add to the phase from a delay and make up the total phase.
View attachment 399974

Phase is tricky to understand. Perhaps this helps: https://audioxpress.com/article/simulation-techniques-misconceptions-in-the-audio-industry
It is tricky, byt your example points to a miniscule difference in presentation. So it just might be it
 
Back
Top Bottom