Is phase shift audible?

[Conrad Hoffman]

I recently worked on a highly regarded amp, the brand of which will remain unsaid, and even though it had flat response and very low THD, I thought it was a tad bright or fatiguing. Just for fun I tried to do an input to output null test and found it had enough phase shift to make that difficult. The only place where it would null was about 150 Hz. I need to make a phase plot, as the amount isn't (I don't think) all that much, but the question is, is phase shift, IOW an all-pass filter, audible?

 

[jcx]

very odd for a amplifier - if it has flat frequency response then flat phase response is pretty much "free"

all-pass filters have parallel non flat in frequency and phase paths summed to give flat frequency response with phase shift


there is a "phase shift deniers" club - unfortunately seems to include many mixing engineers, some audio writers

but it turns out that neurophysiology has caught up with some careful psychoacoustic researchers who have long been able to demonstrate human phase discrimination in audio listening tests

google "auditory nerve phase locking" - its now textbook knowledge - below ~4 kHz nerve impulses are aligned with the positive wave of the pressure in the cochlea

audibility of relative phase of harmonics, "absolute polarity" of tones with high even order harmonics, polarity of transients like kick drum can be easily demonstrated with test tones and headphones, harder with speakers in rooms - but with training even LR4 XO phase shift can be identified with music

and seems to be of very little importance in music listening - allowing sound engineers to largely ignore it in production with very little consequence. likewise multiway speakers with considerable phase shift over each XO frequency are still sold, praised by reviewers

 

[sreten]

Hi,

If you can't null it THD+N by definition will be awful.

rgds, sreten.

I suspect the question has nothing to do with the answer.

 

[DF96]

I am not sure I understand the question. A first-order roll-off will have significant changing phase for about a decade in frequency either side of the corner. That means that an amp which goes from, say, 10Hz to 30kHz will have noticeable phase shift below about 100Hz and above about 3kHz - this is, of course, quite harmless to audio. To do an amplifier null your reference path has to have the same frequency response. Otherwise, you will find that there is just one frequency where amplitude adjustment alone will give a good null; this will usually be in the vicinity of the geometric mean of the two corner frequencies.

Failure to get a null simply means that one or more of the following does not match
- amplitude
- phase
- distortion
The first two require accurate frequency response matching.

 

[Conrad Hoffman]

I am not sure I understand the question. A first-order roll-off will have significant changing phase for about a decade in frequency either side of the corner. That means that an amp which goes from, say, 10Hz to 30kHz will have noticeable phase shift below about 100Hz and above about 3kHz - this is, of course, quite harmless to audio. To do an amplifier null your reference path has to have the same frequency response. Otherwise, you will find that there is just one frequency where amplitude adjustment alone will give a good null; this will usually be in the vicinity of the geometric mean of the two corner frequencies.

Failure to get a null simply means that one or more of the following does not match
- amplitude
- phase
- distortion
The first two require accurate frequency response matching.

You're absolutely right and looking again, that's where I see the phase shift. The amps I've looked at in the past, mostly my DIY stuff, had wider bandwidth and would null more easily. THD is low and not an issue, amplitude is adjustable, so phase remains the culprit. The damping factor is a bit low, so the speaker impedance is also getting involved more than I'd like.