Question about amplifier phase measurements

[Melba59]

Hello Amir,

Can an interpretation of an amplifier measurement suggest a potential absolute phase variation on the bandwidth?

Or what measurement procedure would allow it?

Indeed I suspect a phase variation on the passing BP on the class D with analog input, which is unbearable for my ears. What I do not refocus on the class D with a direct digital source, nor on the empirical or today's classes A, AB, B

PS: I hope I posted on the right thread
I will move my contribution if necessary

 

[Jimbob54]

I moved this here from the index thread .

 

[Jimbob54]

Also you need to tag @amirm if you want him to have a chance if seeing your query . But I'm sure others will pipe up .

 

[MaxwellsEq]

Indeed I suspect a phase variation on the passing BP on the class D with analog input, which is unbearable for my ears

Please can you explain what you mean?

A well designed amplifier used within its correct operating domain, will never be unbearable to anybody's ears, including yours. So there must be something else affecting the sound for it to be unbearable. I can understand "unbearable" speakers or headphones, but not amplifiers.

 

[DVDdoug]

I'm not sure what you're asking but you can't hear a phase shift without a "reference". If there is a phase-difference between the left & right channels you might hear the results of the phase shift when the soundwaves combine in the air.

A few examples -

The crossover in a speaker often introduces phase shifts. The woofer tends to have a phase-lag and the tweeter a phase-lead (at the crossover frequency). In a 2-way system the woofer might have a 90 degree lag and the tweeter a 90 degree lead at the crossover frequency. That puts the soundwaves out-of-phase and they will cancel, creating a dip in frequency response. The solution is to flip the polarity of the tweeter which makes the drivers in-phase at the crossover frequency and the tweeter 180 degrees from where it would otherwise be at higher frequencies. In a 3-way speaker the midrange connections would be reversed.

If you reverse the connections to one stereo speaker (making a 180 degree phase inversion) you'll get a weird "spacey" sound and the bass soundwaves will mostly cancel, giving you weak bass. If you invert both connections it will sound normal.

If you flip the polarity of one headphone you (probably) won't hear the effects because the soundwaves don't combine.

A lot of equipment flips/inverts the phase. There is no "rule" that an amplifier can't invert the phase and sometimes it's more convenient to build an inverting amplifier. Equipment in the studio might invert the phase so some recordings (or some instruments or voices on recordings) are inverted.

The process of cutting and playing a vinyl record (including the RIAA record/playback EQ) introduces different phase shifts at different frequencies.

Delay creates phase shifts. The distance from the speaker and the speed of sound means that high frequencies are "shifted" by thousands of degrees at your ears. The bass is probably "shifted" only a few degrees (with the short distances in your living room).

Because of the distance between your ears, the phase is different in each ear. And if you play a 10kHz test tone (about 1.3 inch wavelength) from both speakers you'll hear it get stronger or weaker as you move your head slightly and the waves from both speakers, and the reflected waves, go in-and-out of phase with each other. The same thing happens with regular music but we don't notice it because music contains many frequencies, and we are used to it.

Here are a couple of posts by Amirm -

Does Phase Distortion/Shift Matter In Audio (no)

Understanding Phase

 

[NTK]

@pma's measurement of his implementation of Hypex's NC252MP.

PMA-NC252MP power amplifier

P.S.: the small spectrum peak at 8kHz that is present in all measurements does not come from the amplifier, but it is a bug from my measuring system. Haven't fixed it, yet :(. 8 kHz is the frequency of USB noise as far as I know. Try a USB filter maybe it can fix.

audiosciencereview.com

 

[Melba59]

Please can you explain what you mean?

A well designed amplifier used within its correct operating domain, will never be unbearable to anybody's ears, including yours. So there must be something else affecting the sound for it to be unbearable. I can understand "unbearable" speakers or headphones, but not amplifiers.

It's not unbearable, but it does result in fatigue over long listening sessions, regardless of the speakers (Focal Custom OD8, Chora, Triangle Genesis, Eltax, DIY, and others of all generations).

This doesn't happen with Class A or AB amplifiers of all generations, nor with Class D TDA ICs with an integrated audio processor/DAC that handles D/A conversion: Example: Airpulse 80 system, discrete Marantz AVR (no indication of the class used; it could be Class G, and the AVR has a room corrector).

To describe my dissatisfaction with Class D amplifiers with analog inputs (without A/D conversion at the input, Purifi, Nuprimes):

- Either the output filtering is average, and the highs are shrill (e.g., Nuprime STA9, Chinese BT60s).
- Or the highs are well defined, with a smooth SPL, but the highs seem... Withdrawals, with a lack of dynamism, the attacks seem softened (Audiophonics HTA400ET, Tangent)

This results in a less deep soundstage in the background.
And this is perfectly correctable via BruteFir convolution, for example.

As I experience this with very good Class D speakers, it's not a question of THD at the end of the band, but rather of the "audible" end-of-band processing. All "bandpass/high/low" filtering causes a progressive rotation of the phase.

I hypothesize that the high-pass or any other end-of-band processing of Class D speakers further accentuates the phase rotations of speakers with more than one period. But that may be an entirely different reason.

 

[Melba59]

I'm not sure what you're asking but you can't hear a phase shift without a "reference". If there is a phase-difference between the left & right channels you might hear the results of the phase shift when the soundwaves combine in the air.

A few examples -

The crossover in a speaker often introduces phase shifts. The woofer tends to have a phase-lag and the tweeter a phase-lead (at the crossover frequency). In a 2-way system the woofer might have a 90 degree lag and the tweeter a 90 degree lead at the crossover frequency. That puts the soundwaves out-of-phase and they will cancel, creating a dip in frequency response. The solution is to flip the polarity of the tweeter which makes the drivers in-phase at the crossover frequency and the tweeter 180 degrees from where it would otherwise be at higher frequencies. In a 3-way speaker the midrange connections would be reversed.

If you reverse the connections to one stereo speaker (making a 180 degree phase inversion) you'll get a weird "spacey" sound and the bass soundwaves will mostly cancel, giving you weak bass. If you invert both connections it will sound normal.

If you flip the polarity of one headphone you (probably) won't hear the effects because the soundwaves don't combine.

A lot of equipment flips/inverts the phase. There is no "rule" that an amplifier can't invert the phase and sometimes it's more convenient to build an inverting amplifier. Equipment in the studio might invert the phase so some recordings (or some instruments or voices on recordings) are inverted.

The process of cutting and playing a vinyl record (including the RIAA record/playback EQ) introduces different phase shifts at different frequencies.

Delay creates phase shifts. The distance from the speaker and the speed of sound means that high frequencies are "shifted" by thousands of degrees at your ears. The bass is probably "shifted" only a few degrees (with the short distances in your living room).

Because of the distance between your ears, the phase is different in each ear. And if you play a 10kHz test tone (about 1.3 inch wavelength) from both speakers you'll hear it get stronger or weaker as you move your head slightly and the waves from both speakers, and the reflected waves, go in-and-out of phase with each other. The same thing happens with regular music but we don't notice it because music contains many frequencies, and we are used to it.

Here are a couple of posts by Amirm -

Does Phase Distortion/Shift Matter In Audio (no)

Understanding Phase

Thank you for taking the time to answer me.

I know.

 

[solderdude]

the attacks seem softened

'attack' in music seems to be in the 4-8kHz range, which appears to be in a region where amps do not have to much issues with.
Generally we see a slight rise in distortion in class D for higher frequencies but IM products appear to not reach very audible levels as it is generally masked by music content.

It would be interesting to 'record' the output of an afflicted amp and a non-afflicted amp (using some attenuators at the output of the amp under actual load) if you want this to be examined.

Would be easiest to use a mono signal. Left into amp1 connected to left speaker, right into amp B connected to the right speaker and record the output of both amps.
Then record the source signal directly.

This way you can end up with a difference signal between the amps and can listen to the null and analyze that null.
You can also null the source to the amp signal.
That could tell you which amp works best with your speakers and how both amps change the sound.

 

[voodooless]

Would be easiest to use a mono signal. Left into amp1 connected to left speaker, right into amp B connected to the right speaker and record the output of both amps.
Then record the source signal directly.

This way you can end up with a difference signal between the amps and can listen to the null and analyze that null.

What you'll end up with is the difference is impedance between the two speakers. This will certainly trump any deviations the amps will have.

It's not unbearable

Well, you said it

is unbearable for my ears

- Either the output filtering is average, and the highs are shrill (e.g., Nuprime STA9, Chinese BT60s).

I don't know about these; they might be pre-filter feedback systems, so they may have an impedance dependency at the high-frequency end. That may be audible, but your 49, so you'll hear about 15 kHz if you're lucky. Audibility would be a very small chance.

- Or the highs are well defined, with a smooth SPL, but the highs seem... Withdrawals, with a lack of dynamism, the attacks seem softened (Audiophonics HTA400ET, Tangent)

The Audiophonics amp is flat and transparent. It should be indistinguishable from any other properly performing amp.

Did you ever perform a blind, level-matched (within 0.1 dB) comparison of the amps?

What I do not refocus on the class D with a direct digital source

Digital in or not, it is totally irrelevant. If you have one box or two to do the same jobs, it doesn't matter.

And this is perfectly correctable via BruteFir convolution, for example.

What did you do to fix this?

very good Class D speakers

What is that?

Amp classes do not have a sound. Any competency-designed amp class can give you transparent output and will sound identical.

 

[solderdude]

What you'll end up with is the difference is impedance between the two speakers. This will certainly trump any deviations the amps will have.

unlikely as both speakers will provide a very similar load with only small differences due to component tolerances.
It is unlikely that if 'attack' and 'dynamics' are affected by an amp (which is what must be proven before being admissible as a fact) then the null must reflect that.
The tolerance differences between both speakers are unlikely to sound like 'dynamic effects' and 'differences in attack'.

If people make such claims (being able to hear differences between classes of amps using music on their speakers one must test that condition.
An audible difference means there must be a (quite) measurable electrical difference.
Such differences could occur under specific load differences with certain amps and must be ruled out.

The most likely scenario is that it is bias and proper blind testing 2 amps is difficult to do.

 

[voodooless]

unlikely as both speakers will provide a very similar load with only small differences due to component tolerances.

Yes, exactly

It is unlikely that if 'attack' and 'dynamics' are affected by an amp (which is what must be proven before being admissible as a fact) then the null must reflect that.

Sure, I agree. But we both know that no such thing will be found.

The tolerance differences between both speakers are unlikely to sound like 'dynamic effects' and 'differences in attack'.

I'm not claiming anything of the sort, am I? I'm saying that these tiny differences will trump any of the amp's differences.

 

[solderdude]

Sure, I agree. But we both know that no such thing will be found.

You are removing all of the educational parts out of my test proposal...
The idea is to let the claimant provide some 'hard evidence' which only the claimant can do with his gear.
I am quite certain there will be low level null differences between 2 amps and their loads but the question is how will that 'sound' and how high are those differences.

When you state 'we both know that no such thing will be found' you remove the will to even bother to test.
The only way to convince people is to experience things, not to tell them up front what the expected result will be.

If the claimant wants to 'prove' he heard certain effects then the only way is to 'record' the differences.

 

[voodooless]

I am quite certain there will be low level null differences between 2 amps and their loads but the question is how will that 'sound' and how high are those differences.

Guess what will be the first reaction when seeing those? "Look, there are differences, I can hear those!". We've seen it multiple times over and over, like this extreme recent example:

Does upgrading my center speaker wire make a difference?

Ok. Well, I don’t clip the amp but wanna help the sound and maybe get more sound volume without cranking up the dB on the center. I just ordered one and will see how it does and measure. It won't (Can't) make any difference. Unless it has cost trivial money I'd cancel that order if I were...

www.audiosciencereview.com

When you state 'we both know that no such thing will be found' you remove the will to even bother to test.
The only way to convince people is to experience things, not to tell them up front what the expected result will be.

No, not at all. I would like to minimize any deviations such a test might provide. And that is by testing a single speaker.

 

[solderdude]

But then you have the problem of the timing of 2 clocks being different.

Nulling isn't perfect either as phase differences will be converted to level differences and one would need the test signal as well (so would need 3-channels recorded at the same time) to determine what part of the null is phase and what is amplitude related.

So ... replace the speakers with resistors and you eliminate the differences between the speakers. That would be preferable over adding timing differences between files to be compared.

The OP expects phase differences to be the culprit and to prove that nulling may not be right tool unless the source signal (not just the digital file) is taken into consideration.

I hypothesize that the high-pass or any other end-of-band processing of Class D speakers further accentuates the phase rotations of speakers with more than one period. But that may be an entirely different reason.

Having a suspicion is great... having actual data that proves it is better. Requires the speaker + amp in question.

Of course one can always measure both amps under the same speaker load but demonstrating the audible effects is another thing.

 

[voodooless]

But then you have the problem of the timing of 2 clocks being different.

Deltawave doesn't care

So ... replace the speakers with resistors and you eliminate the differences between the speakers. That would be preferable over adding timing differences between files to be compared.

Not sure. If the amp does have impedance related differences due to output filter stuff, it won't show up.

The OP expects phase differences to be the culprit and to prove that nulling may not be right tool unless the source signal (not just the digital file) is taken into consideration.

Luckily Detawave does more than just nulling.

Phase is easy to rule out: a minimum phase system, like an amp, will have phase differences when the frequency response differs. That is why most class D amps will have a slight phase shift at the end of the spectrum due to their limited bandwidth. This, however, is inaudible, for one; because the OP is too old, and secondly; it's only a small, gradual shift, and thirdly; your room will mess up any phase shift at those frequencies anyway. It is far more likely that an output-filter-related high-frequency bump is audible, but not with the Audiophonics amps, because it doesn't have this.

 

[wwenze]

and the AVR has a room corrector

Room correction = lots of phase changes (for IIR filters, anyway). You didn't hear that?

 

[amirm]

Deltawave doesn't care

It does actually. Depending on what you are doing, getting deep nulls is very difficult. I have literally spent hours and hours trying to null the same input twice in a row. To be sure, it tries to reduce clock drift but it is no panacea if you want to prove 100% inaudibility down to resolution of the system.

 

[solderdude]

Deltawave doesn't care

Not sure. If the amp does have impedance related differences due to output filter stuff, it won't show up.

Luckily Detawave does more than just nulling.

Phase is easy to rule out: a minimum phase system, like an amp, will have phase differences when the frequency response differs. That is why most class D amps will have a slight phase shift at the end of the spectrum due to their limited bandwidth. This, however, is inaudible, for one; because the OP is too old, and secondly; it's only a small, gradual shift, and thirdly; your room will mess up any phase shift at those frequencies anyway. It is far more likely that an output-filter-related high-frequency bump is audible, but not with the Audiophonics amps, because it doesn't have this.

The idea is to get the OP to prove to himself what the reasons might be for what he is perceiving.

My preference would be to properly blind test this by switching amps but that isn't easy to do because of tells and possible L-R differences reaching audible thresholds.
With recording actual output signals this is easier to do.
Deltawave is a great tool for this but have seen many threads where there were operational issues by less well versed folks using the program which leads to incorrect conclusions.
Nulling isn't easy. One of the main issues is still clock drift which is 'solved' with my experiment. Of course there will be small differences due to component tolerances in amps and loads which will influence the nulls and you can only show the existence of differences and would still have to find out where the differences come from.

Listening to the nulls can be revealing. If for instance one can clearly hear all 'attacks' in the music then there is merit to the claims if that null has enough amplitude.
If it does not sound like that chances are that what is perceived is caused by other effects (which OP did not rule out).
One could even 'mix' the null + afflicted amp output and see if it sounds 'similar' to the the 'not afflicted amp'.

'Attack' is in a much lower frequency band as the entire audible bandwidth so is likely not a 'problem' for amps anyway. It isn't in the >10kHz range where amps may well perform differently under complex loads. Technically the claims OP makes do not make sense. The only way would be to have the OP do experiments and my suggestion was just one of them where results can be shared.

 

[voodooless]

It does actually. Depending on what you are doing, getting deep nulls is very difficult. I have literally spent hours and hours trying to null the same input twice in a row. To be sure, it tries to reduce clock drift but it is no panacea if you want to prove 100% inaudibility down to resolution of the system.

Sure, but unless you're suggesting people can distinguish run to run variation of the same content on the same system with the same settings, this is largely irrelevant.