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

[j_j]

Are you using symmetric or asymmetric FIRs?

You said "constant delay", right? Haven't you answered your own question then?

 

[gberchin]

(Re: Pre-ring) Definitely not, using proper filter design. Been there, done that, works fantastically.

@j_j , I'm aware of temoral masking, which leads to the question: In the real world of linear-phase crossovers, how much pre-ring is tolerable? Or to put it another way: How much pre-ring is too much? (I understand that the answer has a lot of "It depends" caveats.) Has this actually been studied, or are we limited to anecdotal evidence?

 

[ferrellms]

^That^ looks like a cut-n-paste sales pitch.
FIR would be dandy, but the pre-ringing is something that seems like a bit of an Achilles heal.

Cut and paste of accurate info beats BS from humans. I trust GPT4o more than almost any human, did you not find it helpful even a bit? I learned a lot.

At any rate, I feel fine as I have Kii 3 which have the best bass I have ever heard, so after reading the below cut, from GPT4o, I will skip the audiolense.

The DSP (Digital Signal Processing) in Kii Three speakers does introduce group delay, but it is managed and minimized carefully to ensure it does not negatively affect the listening experience. Group delay refers to the time it takes for different frequency components of a signal to pass through a system. In typical speakers, group delay can distort the timing of different frequencies, especially in the low end, making bass sound "slow" or "boomy."
Kii Three speakers use advanced FIR filters for their DSP to control the phase response and minimize group delay, especially in the lower frequencies, which are more susceptible to this effect. The Kii Three system is designed to provide linear phase response, meaning the timing of all frequencies is aligned to ensure accurate, coherent sound reproduction across the frequency range.
The speakers utilize an intricate algorithm that corrects for the inherent group delay caused by the filters, especially in the bass region, resulting in a very tight and accurate low-end performance. While no system is completely free from group delay, Kii Three's DSP is optimized to keep it at a level that is essentially inaudible to most listeners.
So, in summary:

• Yes, the DSP in Kii Three speakers technically has group delay.
• However, it is very well managed to the point where it should not be perceptible in normal listening conditions.

4o
But I use the PEQ built into the Kii 3...
Yes, the Parametric EQ (PEQ) in the Kii Three speakers can affect aspects of the system's performance, including phase response and group delay, depending on how it's applied. However, the extent of the effect depends on the specific EQ settings and how aggressively the filters are used.
Here’s how the PEQ might impact these factors:

1. Frequency Response and Group Delay​

• When you apply PEQ to adjust the frequency response of the Kii Three speakers (e.g., boosting or cutting certain frequencies), you are introducing additional filtering. These adjustments can alter the system's phase response, which can, in turn, affect the group delay.
• The more extreme the EQ settings, particularly in the low-frequency range, the more likely you are to introduce group delay. For instance, strong low-frequency boosts can increase the delay for those frequencies, as it takes time for the DSP to process the boost without introducing significant distortion.

2. Phase Coherence​

• The Kii Three’s DSP is designed to maintain phase coherence across all frequencies, which is one of the reasons for its precise imaging and time alignment. The built-in DSP compensates for phase shifts introduced by the room correction and internal filters. However, when you apply PEQ, depending on the settings, it can introduce additional phase shifts.
• If the EQ is used conservatively (small boosts or cuts), the phase and group delay changes are likely to be minimal and may not be perceptible to most listeners.

3. Interaction with DSP Room Correction​

• Kii's DSP is already performing room correction and time alignment, particularly for the bass frequencies. When PEQ is applied, it interacts with these filters, and excessive adjustments in the same frequency range might disrupt the careful balance the DSP has achieved.

Summary of PEQ Effects:​

• Minimal PEQ adjustments: Likely to have negligible impact on group delay and phase coherence.
• Large PEQ adjustments (especially in the low end): Could introduce noticeable group delay and affect phase coherence, especially for transient sounds or in critical listening situations.

Kii Three’s DSP is quite advanced, so it’s designed to handle most adjustments smoothly, but like any system, extreme EQ settings can push the limits of phase and group delay correction.
,
So, don't use the Kii PEQ but use the EQ in Audiolense?
,
I don't hear any problems, so I will sit tight

4o

 

[antcollinet]

Cut and paste of accurate info beats BS from humans. I trust GPT4o more than almost any human, did you not find it helpful even a bit?

How do you know it is accurate?

I dont trust AI generated text at all. A lot of the time it is accurate. And a lot of other times it regurgitates bullshit it's been trained on, but does it in a very confincing and confident manner. Like your uncle who knows nothing but thinks he is gods gift to enlightened understanding.

And the real problem - there is no way of telling when it is accurate and when it is bullshitting you. it can generate garbage at a rate that makes it impossible to fact check. Like your wall of text up ^ there.

 

[sejarzo]

I suspect what you are asking is whether they curated the tracks to show off the speakers? I have no idea. I did not get to play any of my music. They were playing their own tracks. It was mostly standard audio show fare - jazz, female vocals.

I was wondering how much phase distortion was already baked into the recordings. Were they all close miked?

 

[milosz]

I think that frequency-dependent phase shift has an impact on the perception of stereo image and depth. I have conducted experiments comparing two amplifiers - a high-quality solid state amp ( Forte 4a ) against a good quality hybrid amp with a tube output stage ( AMC CVT2030 MOSFET driver / EL34 output stage ) on several pairs of good quality 2-way speakers. ( ADS & Selah speakers ) and consistently noted by multiple listeners was the immediately audible greater "depth" of the image when using the tube amp.

In-room frequency response measurements of the systems should no differences - less than 1 dB differences between 100 and 15,000 Hz between the solid state and tube amps.

So why the increase in depth effect? ( and it is an effect ) This isn't about transparency or accuracy. The amps have some difference that the human hearing system ( ear / brain system ) is decoding as depth. At the moderate playback levels I was using, I could not measure much THD or IM in the amplifiers outputs. And the tube amp had a low enough output source impedance that I don't think I was hearing an ohms-law variation in response to the speakers impedance curves - which would have shown up in frequency response plots had it existed in any event. I did measure two differences between the amps when driving the speakers:

1. The tube amp had slightly higher THD, especially the 2nd harmonic. 60 dB down as opposed to the solid state amp which was over 80 dB down.
2. The tube amp shifted phase of signals as a function of frequency- more phase shift as the frequency rose, something like 60° between 1000 and 10,000 Hz. I suspect this was from the output transformers.

Nelson Pass has stated that he has found that adding a small amount of 2nd harmonic distortion to signals gives a greater sense of depth, but his findings were related to 2nd harmonic levels higher than -60 dB; more like -40 dB. But the additional 2nd harmonic content might be a contributing factor.

I think it may well have been the phase shift in the tube amp that created this effect- a shift in relative phase as frequency rose. THis just an opinion informed by a guess. So, not even a theory.

Euphonic tube amp effect here? I guess so. I LIKED the way the tube amp sounded in terms of "depth," other things being equal. Is the "enhanced depth" sound more ACCURATE? I have to say NO, because it results from something the amplifier does to the signal. Accurate, no. Fun - YES.

Depth of stereo image is one of those things one can easily hear but as yet there is no way to measure.

DISCLAIMER: my listening sessions were not scientific tests, not double-blind. I knew which amplifier was playing during a given listening session, but the guest listeners did not. As far as statistics for my little experiment go, guest listeners identified "greater depth" at a rate of 100% when the tube amp was driving the speakers. Like I said, the effect was NOT subtle.

 

[Sokel]

I think that frequency-dependent phase shift has an impact on the perception of stereo image and depth. I have conducted experiments comparing two amplifiers - a high-quality solid state amp ( Forte 4a ) against a good quality hybrid amp with a tube output stage ( AMC CVT2030 MOSFET driver / EL34 output stage ) on several pairs of good quality 2-way speakers. ( ADS & Selah speakers ) and consistently noted by multiple listeners was the immediately audible greater "depth" of the image when using the tube amp.

In-room frequency response measurements of the systems should no differences - less than 1 dB differences between 100 and 15,000 Hz between the solid state and tube amps.

So why the increase in depth effect? ( and it is an effect ) This isn't about transparency or accuracy. The amps have some difference that the human hearing system ( ear / brain system ) is decoding as depth. At the moderate playback levels I was using, I could not measure much THD or IM in the amplifiers outputs. And the tube amp had a low enough output source impedance that I don't think I was hearing an ohms-law variation in response to the speakers impedance curves - which would have shown up in frequency response plots had it existed in any event. I did measure two differences between the amps when driving the speakers:

1. The tube amp had slightly higher THD, especially the 2nd harmonic. 60 dB down as opposed to the solid state amp which was over 80 dB down.
2. The tube amp shifted phase of signals as a function of frequency- more phase shift as the frequency rose, something like 60° between 1000 and 10,000 Hz. I suspect this was from the output transformers.

Nelson Pass has stated that he has found that adding a small amount of 2nd harmonic distortion to signals gives a greater sense of depth, but his findings were related to 2nd harmonic levels higher than -60 dB; more like -40 dB. But the additional 2nd harmonic content might be a contributing factor.

I think it may well have been the phase shift in the tube amp that created this effect- a shift in relative phase as frequency rose. THis just an opinion informed by a guess. So, not even a theory.

Euphonic tube amp effect here? I guess so. I LIKED the way the tube amp sounded in terms of "depth," other things being equal. Is the "enhanced depth" sound more ACCURATE? I have to say NO, because it results from something the amplifier does to the signal. Accurate, no. Fun - YES.

Depth of stereo image is one of those things one can easily hear but as yet there is no way to measure.

DISCLAIMER: my listening sessions were not scientific tests, not double-blind. I knew which amplifier was playing during a given listening session, but the guest listeners did not. As far as statistics for my little experiment go, guest listeners identified "greater depth" at a rate of 100% when the tube amp was driving the speakers. Like I said, the effect was NOT subtle.

If that's true,the same happens with old,50's and 60's Lew Layton's recordings for example,where the depth is far more evident than newer ones (except when it's intentional) and it's system independent (at least at systems where can do depth,not all can) .

 

[ferrellms]

How do you know it is accurate?

I dont trust AI generated text at all. A lot of the time it is accurate. And a lot of other times it regurgitates bullshit it's been trained on, but does it in a very confincing and confident manner. Like your uncle who knows nothing but thinks he is gods gift to enlightened understanding.

And the real problem - there is no way of telling when it is accurate and when it is bullshitting you. it can generate garbage at a rate that makes it impossible to fact check. Like your wall of text up ^ there.

Well, if I am suspicious there is Wikipedia. (But how do you know that that is true? How do you that what that ARS post says is true???) And every edition of OpenAI bots gets better in my experience d

How do you know it is accurate?

I dont trust AI generated text at all. A lot of the time it is accurate. And a lot of other times it regurgitates bullshit it's been trained on, but does it in a very confincing and confident manner. Like your uncle who knows nothing but thinks he is gods gift to enlightened understanding.

And the real problem - there is no way of telling when it is accurate and when it is bullshitting you. it can generate garbage at a rate that makes it impossible to fact check. Like your wall of text up ^ there.

My personal experience is that GPT4o is more accurate than humans, but some skepticism is warranted, but not as much as with humans.

 

[Blumlein 88]

I think that frequency-dependent phase shift has an impact on the perception of stereo image and depth. I have conducted experiments comparing two amplifiers - a high-quality solid state amp ( Forte 4a ) against a good quality hybrid amp with a tube output stage ( AMC CVT2030 MOSFET driver / EL34 output stage ) on several pairs of good quality 2-way speakers. ( ADS & Selah speakers ) and consistently noted by multiple listeners was the immediately audible greater "depth" of the image when using the tube amp.

In-room frequency response measurements of the systems should no differences - less than 1 dB differences between 100 and 15,000 Hz between the solid state and tube amps.

So why the increase in depth effect? ( and it is an effect ) This isn't about transparency or accuracy. The amps have some difference that the human hearing system ( ear / brain system ) is decoding as depth. At the moderate playback levels I was using, I could not measure much THD or IM in the amplifiers outputs. And the tube amp had a low enough output source impedance that I don't think I was hearing an ohms-law variation in response to the speakers impedance curves - which would have shown up in frequency response plots had it existed in any event. I did measure two differences between the amps when driving the speakers:

1. The tube amp had slightly higher THD, especially the 2nd harmonic. 60 dB down as opposed to the solid state amp which was over 80 dB down.
2. The tube amp shifted phase of signals as a function of frequency- more phase shift as the frequency rose, something like 60° between 1000 and 10,000 Hz. I suspect this was from the output transformers.

Nelson Pass has stated that he has found that adding a small amount of 2nd harmonic distortion to signals gives a greater sense of depth, but his findings were related to 2nd harmonic levels higher than -60 dB; more like -40 dB. But the additional 2nd harmonic content might be a contributing factor.

I think it may well have been the phase shift in the tube amp that created this effect- a shift in relative phase as frequency rose. THis just an opinion informed by a guess. So, not even a theory.

Euphonic tube amp effect here? I guess so. I LIKED the way the tube amp sounded in terms of "depth," other things being equal. Is the "enhanced depth" sound more ACCURATE? I have to say NO, because it results from something the amplifier does to the signal. Accurate, no. Fun - YES.

Depth of stereo image is one of those things one can easily hear but as yet there is no way to measure.

DISCLAIMER: my listening sessions were not scientific tests, not double-blind. I knew which amplifier was playing during a given listening session, but the guest listeners did not. As far as statistics for my little experiment go, guest listeners identified "greater depth" at a rate of 100% when the tube amp was driving the speakers. Like I said, the effect was NOT subtle.

You can get Pkane's Distort and add 2nd harmonic under your complete control to see if achieves the result claimed or not. As well as finding exactly which level of 2nd harmonic is optimum for you.

Now most cone speakers are primarily 2nd and 3rd harmonic distortion with not much else. 2nd is highest. It is also level dependent. Not very many speakers manage less than -60 db levels of 2nd harmonic.

 

[Sir Sanders Zingmore]

The second is, of course, the phase shift resulting from an IIR crossover. There are designs that are not power complimentary that can avoid this, but at the cost of creating some pretty "interesting" radiation pattern

Ok I’m going to show my ignorance
What does “not power complimentary” mean and which designs show this?

 

[j_j]

@j_j , I'm aware of temoral masking, which leads to the question: In the real world of linear-phase crossovers, how much pre-ring is tolerable? Or to put it another way: How much pre-ring is too much? (I understand that the answer has a lot of "It depends" caveats.) Has this actually been studied, or are we limited to anecdotal evidence?

At what frequency. Figure out the length of the main lobe of a given ERB filter, you can use bandwidth as a reasonable approximation, and you have your answer.

How much is too much? What frequency?

 

[j_j]

Ok I’m going to show my ignorance
What does “not power complimentary” mean and which designs show this?

Power complimentary means that the sum of the POWER out of all the filters equals the power in. A subtractive crossover, say one using 3rd order butterworth HP, is not power complimentary. An LR crossover is, but has phase shift. With IIR that's your choices.

 

[gberchin]

How much is too much? What frequency?

Exactly the kind of "It depends" caveat that I expected. It has long been my suspicion that pre-ring at frequencies above perhaps the mid-bass is likely masked. My concern is that a linear-phase crossover between, for example, a subwoofer and a woofer, in the 50Hz - 100Hz region, might produce sufficient pre-ring to be audible. However, I've never tested it myself, and never seen any reports of someone else who did.

 

[gberchin]

A subtractive crossover, say one using 3rd order butterworth HP, is not power complimentary. An LR crossover is, but has phase shift.

A subtractive crossover using a Gaussian (FIR) or Bessel (IIR) LPF has lowpass and highpass responses in-phase with each other at all frequencies (Gaussian) or from DC to a half-octave or so above crossover frequency (Bessel). Both are -6 dB at the crossover frequency. As such, they are amplitude-complimentary but not power-complimentary.

I believe that the Linkwitz-Riley is the same -- in-phase at all frequencies, -6 dB at the crossover frequency, amplitude-complimentary but not power-complimentary.

 

[j_j]

Exactly the kind of "It depends" caveat that I expected. It has long been my suspicion that pre-ring at frequencies above perhaps the mid-bass is likely masked. My concern is that a linear-phase crossover between, for example, a subwoofer and a woofer, in the 50Hz - 100Hz region, might produce sufficient pre-ring to be audible. However, I've never tested it myself, and never seen any reports of someone else who did.

You have it backwards, actually. Narrow ERB's, lower sensitivity to pre-echo. Wider ERB's, more sensitivity. Fortunately, one only needs steep (in frequency, not log-frequency terms) filters at low frequencies.

Come listen to our stuff sometime if you're around.

 

[Chazz6]

Nelson Pass on phase, from the manual for his FirstWatt F8 amplifier:

The Q3 'Mu follower” biases not only the output power Jfet Q2, but also the input Jfet Q1, and this bias feed is also the feedback signal of the CFA circuit. Both devices operate with the same phase with respect to their characteristics. This congruence helps give rise to the nice “negative phase” second-harmonic character that you can see here:

You can see that it is a nearly pure second harmonic and that its phase alignment is “negative” at both peaks of the fundamental waveform. No spikes or weird looking stuff.

When I began working with SIT circuits, I discovered that there was a consistent preference for this sort of characteristic in amplifiers over a range of distortion levels from about .01% to 1%. Somewhere below that, it was not usually heard - and above that it could be excessive. Not only was second harmonic preferred over other harmonic content, but its phase seems to matter. I used blind evaluations for the SIT-1, the First Watt amplifier, and on the front panel there was a knob that allowed adjustment of this characteristic in distortion amplitude and phase. I found that listeners tended to cluster on the same setting, which was made the center of the adjustment meter. The F8 has this same characteristic, but down at the lower edge of audibility.

 

[gberchin]

You have it backwards, actually. Narrow ERB's, lower sensitivity to pre-echo. Wider ERB's, more sensitivity.

I'm gonna have to ponder that for a while.

EDIT: When did ERB replace Bark? I got accustomed to the Bark scale with our MPEG 1 audio work back in the 1990s, but haven't revisited it since then.

Come listen to our stuff sometime if you're around.

I really wish I could.

 

[j_j]

A subtractive crossover using a Gaussian (FIR) or Bessel (IIR) LPF has lowpass and highpass responses in-phase with each other at all frequencies (Gaussian) or from DC to a half-octave or so above crossover frequency (Bessel). Both are -6 dB at the crossover frequency. As such, they are amplitude-complimentary but not power-complimentary.

I believe that the Linkwitz-Riley is the same -- in-phase at all frequencies, -6 dB at the crossover frequency, amplitude-complimentary but not power-complimentary.

Hmmm. I'd have to look. Is it in phase at all frequencies? It's been about 30 years since I used an analog crossover. They are just too )(*(&( fussy to deal with.

Certainly an (s^2 -1) / (s^2 +2s + 1) crossover has constant amplitude, but ...

So, you may need to refresh my memory on the various alignments since I got very tired of the whole analog crossover thing a long, long time ago.

 

[j_j]

I'm gonna have to ponder that for a while.

Consider. Narrow bandwidth ::= long impulse response. Detection depends on some level at the start of the impulse response. Longer impulse response, less sensitivity.

 

[Keith_W]

Thank you for your response, @j_j

I just re-read what Toole said about the audibility of phase shift (Ch 4.8). He made the following points:

1. in his own study, there was correlation between listener preference and a linear amplitude response, but no correlation between preference and phase shift. Part 1 of the study can be found here (free PDF download), Part 2 is behind a paywall.
2. recordings are unlikely to have preserved phase integrity
3. it is difficult to design a speaker that can deliver phase integrity over a larger listening area,
4. normally reflective rooms destroy phase integrity.

I have my own criticism of the points that Toole made. If he reads this (gulp) he might be wondering who this idiot is. This idiot is a scientist, albeit a scientist in a different field.

For point 1, this is what is known in medical studies as a secondary endpoint. Unless a study is specifically controlled to deliver a meaningful result for a secondary endpoint, interpretation should be treated with caution. In this case, the study was for listener preference for amplitude response. I did read the study, and I saw that no measures were put in place to specifically control for phase integrity. So it is misleading to say "our study found no correlation between listener preference and phase integrity" when the study was not specifically designed to study the effect of phase integrity on listener preference.

For point 2, loudspeakers are there to reproduce the recording. Recordings do not necessarily need to have phase integrity, that is what produces the spatiality of the recording. Perhaps some recordings need phase integrity, e.g. solo singers. If a soloist is not mic'ed properly, phase interference from multiple mics can produce undesirable effects like comb filtering, phantom centre image drift, and so on. However, I do acknowledge that minimum phase systems in the recording chain (mixing consoles, tape recorders, etc) may introduce additional phase shift that was not captured by the microphone.

Points 3 and 4 - agree, but it does not answer the question nor does it provide thresholds.

Toole then goes on to provide some references to studies that have studied the audibility of phase shift.

Hansen and Madsen 1974
Lipshitz, Pocock, Vanderkooy 1982
Van Keulen 1991

All the above studies were behind a paywall and I was not able to read them.

Greenfield and Hawksford 1990 (free PDF download!). A linear phase FIR was used for phase and amplitude equalisation and listening performed in "Celestion's listening room". No details of listener distance, room treatment, etc. was given. Although measurements of the phase equalised speakers were provided, no measurements were shown of the actual phase response at the listening position to show if phase integrity was preserved. Despite this, this very small study (4 listeners) found subtle but audible differences in the test tracks under blind listening conditions. The phase equalised speakers were thought to have better spatial characteristics.

Conclusion. After all this reading, it seems to me that not only is phase audible, lack of phase distortion seems associated with better spatial characteristics. As a medical scientist, I would say that this does not even reach Level C evidence (i.e. the published evidence is very weak, no consensus between experts).