Question about amplifier phase measurements

[Melba59]

'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.

It's more of an impact on the treble register, beyond 5 or even 8 kHz, and probably much higher, but it would definitely have repercussions on the upper midrange: a band from 5 kHz to 8 kHz that is part of the most sensitive range for instruments (acoustic guitar, snare drum, saxophone, human voice), the slightest difference is perceptible.
This variation affects the amplified audio signal, well before the waveform is emitted by the speakers, and the phase is disrupted by the speakers and even more so by the room.

Why do sound engineers attach great importance to phase management in mixing, "phasing," if it can't be heard?

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.

Yes, that's what I meant by "good Class D amplifiers": No more slight increase in Class D distortion for high frequencies to explain this annoyance observed for several years during long listening sessions

Perhaps this fatigue during prolonged listening of these various Class D amplifiers is due to clipping, much like the low sweep of a screen can give you a headache. However, why does using a DRC convolution filter to correct only the speaker phase, when properly implemented, significantly reduce this listening fatigue?

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.

Interesting, i'll look into it and run some tests.

Unfortunately, my level of electronics measurement is insufficient to prove scientific truths. Nor can I be certain that the measurements I perform on the amplifiers I own can be reliable and usable for others.

My question to Armir wasn't intended to help me choose an amplifier/speaker pair.

My question was to find out if we could detect potential phase non linearities in the tests already performed in amp reviews.
Or if it can't be analyzed, is there a standardized test of the AP software that can be performed?
It would still need to be relevant information to add to future amplifier reviews.

 

[KL....]

For consideration.... could this be due to reactive impedance affecting the LC Modulation causeing spurious Hysteresis/Timing/Phase issues/errors with the device it is feeding back too and perhaps even with the device that is being feed?.... just a pondering :=)

 

[Melba59]

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

No, that's not the point. The Marantz digitizes the analog inputs. Its ADCs are high-quality, and I can't certify that it's a Class D amplifier, just a discrete amplifier.

As mentioned, I don't experience this annoyance when the input source is digital and converted by an integrated chip/DAC. I persist (after 10 years of wondering) in noticing this fatigue on Class D amplifiers with analog inputs. For those with distorted, shrill treble, the explanation is widely documented, but with my Purifi, I don't understand.

I don't use the correction; my living room is acoustically treated for a volume of 100 cubic meters with a subwoofer equipped with a DSP, but it doesn't go down to the first octave.

My reference system is on my terrace under an awning. And the no-compromise one has been in development for 20 years. Just selling my Audiophonics HPA400ET 4-way and two Nuprimes STA-9s to finance four Topping LA90Ds or B200s.

It's been over 10 years since I've really used a EQ.
Twenty years ago, I used Sbragion's DRC: https://drc-fir.sourceforge.net/
With Angelo Farina's Aurora plugin (http://pcfarina.eng.unipr.it/Aurora_XP/index.htm)

I had developed a very effective adjustment from the low mids to the extreme treble to eliminate primary reflections that are detrimental to intelligibility and to straighten the phase as close as possible to the sound wave's emission.
No gain to compensate for the dips, only smoothing the bumps.
However, the results at the low end were not always satisfactory; the bass was often "too tight" and the tonal balance a little too high. I gained some experience in what to listen for, like my teenage hobby of DIY PA speakers, initially, then hi-fi speakers (it's been going on for 35 years...).

Today, my audio systems satisfy me greatly without correction. The balance of benefits/disadvantages is no longer favorable, at best nonexistent, but with the concern of being at the listening point, whereas without it, I can move around.
I know that things have changed significantly with REW and that some people work miracles with their subwoofers, but listening to the first octaves is of little interest to me; a -6dB FC at 40Hz suits me very well.

 

[pkane]

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.

DeltaWave will remove clock drift, but will not (by default) remove differences in phase that are due to filters or random phase variations. When looking at the true transparency to signal (and not just the audibility of the transfer function), these phase differences can't be ignored.

For example, a minimum phase filter applied to the output signal compared to a linear phase of the test source will produce a worse null. To get a better sense of what's audible and what's not, I use either the dBA null for A-weighted RMS of the difference, or use PK-Metric to apply a number of other audibility weights, including temporal and frequency masking, along with equal loudness curves.

 

[Miley_B]

Why do sound engineers attach great importance to phase management in mixing

Typically so that when combining multiple channels, such as individual microphones around a drum kit that are being combined with the ambient overhead microphones, it isn’t destructive.

 

[Melba59]

Typically so that when combining multiple channels, such as individual microphones around a drum kit that are being combined with the ambient overhead microphones, it isn’t destructive.

Thank you for your contribution.

The latter is subject to sound recording, such as drums, which require several well-positioned microphones and phase inversion (or non-phase inversion) to avoid any phase cancellation.
This correlates with the influence of the room on the transmitted waveform, not with the phasing processing during mastering, which can be correlated with distortion of the recorded source signal during reproduction/playback.

 

[KL....]

For consideration.... could this be due to reactive impedance affecting the LC Modulation causeing spurious Hysteresis/Timing/Phase issues/errors with the device it is feeding back too and perhaps even with the device that is being feed?.... just a pondering :=)

To clarify, Inductors/L are magnetic/EMF components (Capacitors, less/much less so, although they do vibrate) and can cause chaotic results, can't they?, and depends on how they are implemented/positioned. which is why I mentioned it.

 

[MaxwellsEq]

not with the phasing processing during mastering, which can be correlated with distortion of the recorded source signal during reproduction/playback.

Correlated? What do you mean?

 

[solderdude]

My question was to find out if we could detect potential phase non linearities in the tests already performed in amp reviews.

Yep, this is possible and shown in post #6.
Of course this is with a resistive load and could be a little different with a complex load.

The thing for slow and small phase shifts is that it is not audible. Substantial phase shifts and sharp ones (over a small bandwidth) could be audible.
Such effects can occur with speakers and rooms but not in amplifiers.

So the theories you have, I suspect, are not related to your perceived effects with amplifiers.
When you are able to prove it (repeatably supported with measured evidence) in well performed blind tests it would be interesting.

The thing is ... you can 'perhaps', 'maybe' and 'think' something you perceived might be causing something but evidence is what needs to be there. That evidence, currently, is extremely thin (or non existent).

 

[Melba59]

Correlated? What do you mean?

Analog isn't a better term.
What I mean is that the phase processing of each track assembled on the master (source recording) must not be altered by an electronic device for certain registers, and the upper midrange (2KHz-8Khz) is essential.

Once emitted by the speakers (which themselves vary the phase "slowly and smoothly" (gradually) when the passive filters are correctly designed, i.e., when the time alignment of the speakers at the crossover frequencies is in phase)
But this part, and even the phase variation of the speakers, is no longer important. Indeed, mastering is done with monitors and headphones. (The signal is processed "universally" to limit masking effects, for example, so that compression doesn't deteriorate the phase, etc.)

The case of drums is well known, but it's during sound recording. Many microphones in different locations and the risk of phase cancellation of the recorded signals: Similar to room effects that color or cause cancellations/bumps (standing waves, reflections, diffuse waves interfering with the direct wave, etc.)

Best regard

 

[KL....]

@Melba59
Your posts are a little confusing but to clarify.... are you suggesting that there are phase issues in the recorded music and when played back classAB handles these in a better less irritating way, in the audio treble/high frequencys, than classD?

 

[Miley_B]

@Melba59
Your posts are a little confusing but to clarify.... are you suggesting that there are phase issues in the recorded music and when played back classAB handles these in a better less irritating way, in the audio treble/high frequencys, than classD?

I'm not clear if the OP feels this is source dependant however he is clear that the issue is with class D amps when used with an external DAC, integrated DAC & class D amps (such as a typical AVR) are fine if I understand the OP correctly.

@Melba59 Attached are two amplifier phase traces from Archimago.blogspot.com
An AB from Emotiva and a Fosi class D. (the early Fosi V3 Mono had a polarity inversion as clearly shown here). Is there anything in these, or the class D Hypex trace in post #6, that could lend evidence to your hypothesis?

 

[Melba59]

The thing for slow and small phase shifts is that it is not audible. Substantial phase shifts and sharp ones (over a small bandwidth) could be audible.
Such effects can occur with speakers and rooms but not in amplifiers.

Not directly audible, indirectly audible. This doesn't sound like crossover distortion or a rise in THD+N at higher frequencies.
The renderings when comparing the two amps all sound equally natural, both of them. Therefore, it's not distortion, just an auditory effect.

@Melba59
Your posts are a little confusing but to clarify.... are you suggesting that there are phase issues in the recorded music and when played back classAB handles these in a better less irritating way, in the audio treble/high frequencys, than classD?

Not. Just to demonstrate that very small changes can have a significant impact on the sound we perceive.

Indeed, going back to the mastering example, if the sound engineer wants to amplify the depth of the soundstage by placing an instrument further away from the listening point, he can play around with attenuating the treble very slightly (it only takes a very small amount for it to be audible). Since the attenuation of his mixing console is a low-pass or band-pass filter, this attenuation has a direct influence on the phase. The summing with the other instrument tracks that would play the same frequencies but at a position further forward from the listening point would not take the same sinusoidal shape on the master recording, and a partial masking of the most distant instrument occurs.

I'm not clear if the OP feels this is source dependant however he is clear that the issue is with class D amps when used with an external DAC, integrated DAC & class D amps (such as a typical AVR) are fine if I understand the OP correctly.

@Melba59 Attached are two amplifier phase traces from Archimago.blogspot.com
An AB from Emotiva and a Fosi class D. (the early Fosi V3 Mono had a polarity inversion as clearly shown here). Is there anything in these, or the class D Hypex trace in post #6, that could lend evidence to your hypothesis?

Post #6 and the two tests make the possibility of a low-pass filter effect unlikely. Indeed, the Emotiva has a more pronounced attenuation of its SPL curve at 9 kHz than the two Class Ds. Following my reasoning, the effect should be more significant on the Emotiva than on the Class Ds.

The suggestion of the increase in THD+N and IMD at higher frequencies was lost when listening to the Audiophonics HPA-400ET, and this fatigue persisted over extended listening sessions.

Back to the first suggestion about the pulse width modulation frequency of the Class Ds? Perhaps I dismissed it too quickly when listening to my Nuprimes STA-9s (higher modulation frequency).
But it's strange because the Edifier Airpulse 80s with a USB input don't suffer from it; with the digitized analog input, the same observation was made.
And the Texas TAS5754s used perform PWM conversion regardless:

https://www.ti.com/lit/ds/symlink/tas5754m.pdf

With the Airpulse 80, the analog input is probably digitized; a purifier has no ADC; the analog signal is directly converted to PWM. Would this acquisition suffer from the same quantization errors as ADCs?

 

[solderdude]

Not directly audible, indirectly audible. This doesn't sound like crossover distortion or a rise in THD+N at higher frequencies.
The renderings when comparing the two amps all sound equally natural, both of them. Therefore, it's not distortion, just an auditory effect.

What would cause this auditory effect and would that be provable in a properly done blind test as well ?

I am getting the idea all of the observations were done when knowing what was being used.

 

[MaxwellsEq]

Not directly audible, indirectly audible

I'm struggling with this. As I understand it, things are audible, or they're not: there is no directly or indirectly. Perhaps you mean your hearing can't detect it but other senses can?

The suggestion of the increase in THD+N and IMD at higher frequencies was lost when listening to the Audiophonics HPA-400ET,

Used within its current range, an 1ET400A has THD and IMD that are inaudible to humans at all frequencies in the audio domain. More confusingly, your sentence has a double negative - "suggestion of the increase... ...was lost" - so are you are confirming there is no increase?

So far, you have offered no measurable and repeatable facts. You seem to believe you are on to something, but you have failed to convince me that you understand the words you are writing. Please try harder with facts and offer fewer, unsupported opinions. If you do that, we will all benefit!

 

[KL....]

Not directly audible, indirectly audible. This doesn't sound like crossover distortion or a rise in THD+N at higher frequencies.
The renderings when comparing the two amps all sound equally natural, both of them. Therefore, it's not distortion, just an auditory effect.

Not. Just to demonstrate that very small changes can have a significant impact on the sound we perceive.

Indeed, going back to the mastering example, if the sound engineer wants to amplify the depth of the soundstage by placing an instrument further away from the listening point, he can play around with attenuating the treble very slightly (it only takes a very small amount for it to be audible). Since the attenuation of his mixing console is a low-pass or band-pass filter, this attenuation has a direct influence on the phase. The summing with the other instrument tracks that would play the same frequencies but at a position further forward from the listening point would not take the same sinusoidal shape on the master recording, and a partial masking of the most distant instrument occurs.

Post #6 and the two tests make the possibility of a low-pass filter effect unlikely. Indeed, the Emotiva has a more pronounced attenuation of its SPL curve at 9 kHz than the two Class Ds. Following my reasoning, the effect should be more significant on the Emotiva than on the Class Ds.

The suggestion of the increase in THD+N and IMD at higher frequencies was lost when listening to the Audiophonics HPA-400ET, and this fatigue persisted over extended listening sessions.

Back to the first suggestion about the pulse width modulation frequency of the Class Ds? Perhaps I dismissed it too quickly when listening to my Nuprimes STA-9s (higher modulation frequency).
But it's strange because the Edifier Airpulse 80s with a USB input don't suffer from it; with the digitized analog input, the same observation was made.
And the Texas TAS5754s used perform PWM conversion regardless:

View attachment 436356

https://www.ti.com/lit/ds/symlink/tas5754m.pdf

With the Airpulse 80, the analog input is probably digitized; a purifier has no ADC; the analog signal is directly converted to PWM. Would this acquisition suffer from the same quantization errors as ADCs?

@Melba59
The graphs are not (really) showing an issue (are they?) but a thought that came mind is.... there are 1k steps from 1-10khz then 5/10k steps. Perhaps the input signal tested needs to be limited, 900/1Khz- 6/8khz (mentioned range of (experienced) irritation), with 20hz steps shown on the graph to more visually show any deviations/interactions?

edit: If nothing out of the ordinary is showing then at least that aspect, by measurement, is resolved and you would need to look elsewhere, wouldn't you?
Fyi... just recently (yesterday) I was listening to a fellow who wirelessly used a (modifyed) Dayton classD amp in his Sub and the main point he (extremely) briefly mentioned that I (intently) remember, was how he stressed how critical the choice of Inductor was in the LC modulation but he never indicated why....

 

[Miley_B]

Do setups that you find hard to listen to involve an external DAC with a USB, SPDIF etc. input or is it analog all the way?

In the time since you started this thread have you made any attempt to verify that this effect is real? Under more rigorous scrutiny I often (but not always) find that a previously clearly audible difference is indistinguishable.

 

[Melba59]

Do setups that you find hard to listen to involve an external DAC with a USB, SPDIF etc. input or is it analog all the way?

In the time since you started this thread have you made any attempt to verify that this effect is real? Under more rigorous scrutiny I often (but not always) find that a previously clearly audible difference is indistinguishable.

Analog only

Attempts to describe subjective and personal feelings would only be consequences that we could then deduce if we "document" a constraint related to Class D.
It's also a way to find avenues to explore in the hope of finding an objective measurement that would shed light on this phenomenon. I agree, in this difficult-to-determine case, they lead to a waste of time.

I therefore stick to the only observation shared by others, including my 16-year-old son, namely systematic auditory fatigue after one, two, or three hours of listening with a Class D amplifier and an entirely analog signal.

A little history seems useful to me because this is an unexplained observation that I have made over several years, many hours of listening, and which is repeated every time I insert a Class D amplifier into one of my hi-fi systems.
However, it's long to write; I'll come back to it later, if it's really useful.

This unexplained, repeated, and systematic observation can no longer be explained by auditory illusion, psychoacoustics, or any third-party factor (speakers, driver type, quality of power supply filtering, including poor impedance matching; this avenue has been explored at length). Comparative listening (I have source and amp inverters) offers only subjective assessments and subtle differences. It's useless to demonstrate anything.

This fatigue is not observed with a USB connection on:
- Airpulse 80 (which I listen to every day),
- a Tangent amplifier, also Class D CI
- all other Class A/AB amplifiers (Kora Explorer 90SII CSC, Topping LA90, Topping LA90D, an old Sheerwood AVR, an 8W Class A amplifier from the 70s (DIY), and many others over nearly 20 years.

In any case, the avenues explored are dwindling. This is not:
- An increase in THD+N, IMD of the audible upper frequencies, the cause of elimination: This increase is inaudible on Nuprimes STA-9 and Purifi)
- Phase variation, low-pass filter effect. Cause of elimination: The measurement of the Emotiva compared to the two Class D amplifiers tested.

From there, back to square one: the PWM signal slicing. Problem:
1/ 16-bit CDs are perfect for me...
2/ The analog input of Class D ICs is generally digitized, and this fatigue is also felt on these devices.

This suggests that the problem lies in the acquisition of an analog signal:
Either by digitization (documented quantization error)
Or by conversion to a PWM signal (Same quantization error, is this possible?)

 

[voodooless]

2/ The analog input of Class D ICs is generally digitized, and this fatigue is also felt on these devices.

No, it’s generally not.

 

[SIY]

Hypothesizing about the mechanism of a sensory phenomenon not yet demonstrated to exist is useless. First things first- demonstrate that the phenomenon is objectively and physically real and not a psychosensory effect.

Until then, people should stop waving hands, IMO.