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

[Digital_Thor]

Ok.. this peaks my curiosity
Back in the mid 90's where I bought my first PC with a dedicated sound card - motherboard had almost nothing onboard back then - I had an analog volume knob on the rear of the sound card, which had no issue with sound quality, when I turned it to max. With a jack I send the sound to my old Rank Arena integrated amplifier and 2 small speakers.

Ever since, with any other digital sound card - I've noticed that to avoid heavy distortion, you should always keep volume just around 5% down from max - to be sure.
Since these sound cards are essentially a DAC with volume, then is this effect similar to this case?

 

[Sokel]

For a while, they made a push into consumer market resulting in some amount of sales to high-end consumers. That has faded though from what I can tell.

The only one I remember is NADAC which is now discontinued.
And if I remember well it had very decent performance back then, specially noise.

They probably got bored by the royal treatment they had to give, as audiophile market is get used to it.
Pro market don't do that.

 

[Holmz]

No I haven't. To wit, I give a pass to DACs that have their maximum attenuation at 24 kHz instead of 22.05 kHz at 44.1 kHz sampling.

An impulse signal as I have explain, is wildly illegal. Using it then showing that the response of the system is not the same as the said impulse, will cause alarm in viewership who doesn't understand the nature of this signal. What is the correct output of the system, will seem totally wrong to them. Take a look at this:

Person looks at the 48 KHz response and immediately thinks that is completely broken. Yet as you well know, that is the correct, mathematical response to a system that is bandlimited to 48 kHz sampling. Going to higher and higher sampling rate allows more basis functions to be integrated and hence, approximation to impulse increases. But that is simply because the bandwidth is increased. That concept, in frequency domain, is easily understood that we don't need that extra bandwidth. Presenting it in time domain though, causes rampant confusion making folks think need Megahertz sampling rates (not realizing there is a cost to that elsewhere).

Above graph is from the marketing material of an ADC/DAC company by the way (who is completely in DSD camp): https://www.merging.com/highlights/high-resolution

That is the direction I saw @pma going and so I objected to make sure he and others understand what they are doing. As otherwise, they sell you down the path of needing every component in the system to have megahertz bandwidth.

But it is totally reasonable to present it to a DAC.
If the DAC is letting through power above the sampling rate then that says that it’s treatment of the single sample is not correct.

In the case of FFT on the file, we only see energy between DC and Nyqvist.

It’s only the DACs treatment of it that is being done with excess bandwidth in analog.
And that excess energy outside of the band, is not there mathematically, and should be largely rolled off well before Nyqvist…
If it is not, then the impulse shows it is a problem,

But white noise that is band limited from DC to just under Nyqvist should also not be Nyqvist. So the Impulse is not overly special… It is just common in control theory.

 

[Kal Rubinson]

The only one I remember is NADAC which is now discontinued.
And if I remember well it had very decent performance back then, specially noise.

They probably got bored by the royal treatment they had to give, as audiophile market is get used to it.
Pro market don't do that.

They discontinued the NADAC line at the same time as they were purchased by Sennheiser and integrated into Neumann in 2022. https://www.merging.com/news/press-releases/neumann-and-merging-technologies-join-forces.

Subsequently, NADAC has reappeared here: https://www.master-fidelity.com/nadac-d/

 

[pma]

But white noise that is band limited from DC to just under Nyqvist should also not be Nyqvist. So the Impulse is not overly special… It is just common in control theory.

Exactly. And I do not understand that strange differention to valid and seemingly invalid. It is nothing but a personal subjective feeling of the poster.

 

[bmc0]

And that excess energy outside of the band, is not there mathematically

After ideal interpolation, to be more precise.

As a (silly) aside, the "analog" impulse response shown in that marketing image has pre-ringing. What's up with that?

 

[amirm]

As a (silly) aside, the "analog" impulse response shown in that marketing image has pre-ringing. What's up with that?

Simple. Assuming it is not a fake chart, they used a digital synthesizer that uses a DAC so naturally creates pre and post ringing. Even though it is likely much higher bandwidth than an audio generator, it still has to obey the laws of universe and create that post and pre ringing. You see why I say this is a dangerous path to go down on?

 

[amirm]

But it is totally reasonable to present it to a DAC.

So is a picture of yours encoded as numbers in the wave file. Doesn't make it proper.

If the DAC is letting through power above the sampling rate then that says that it’s treatment of the single sample is not correct.

The DAC has no obligation to produce correct output when the input is not bandlimited as it should have been per sampling theorem. A DAC is a dumb device so runs the samples through regardless of what they are. That ability should not be the reason to run illegal samples through it and with it, cause concern as the marketing graph tried to do. And @pma was doing as well.

In the case of FFT on the file, we only see energy between DC and Nyqvist.

That's right yet folks were running off saying the signal was band limited by its nature and was perfectly valid as a result. You best argue with them than quoting me.

 

[amirm]

If an antialiasing filter has its cutoff (half magnitude) at Fs/2, then it is theoretically possible to get an "illegal" impulse from it.

You can't answer an impossible condition with another impossible condition. The fact that you can't have an impulse in real life means you can't have that sharp filter in real life either. They are two sides of the same coin. Nature abhors discontinuity like that.

 

[levimax]

A DAC is a dumb device so runs the samples through regardless of what they are.

Hmm.... maybe it is time for a "Smart DAC" that corrects for "intersample overs" and "illegal samples" and "increases headroom" (The current cure for intersample overs is reduce digital volume by 3-4 dB which is actually quite a bit of lost headroom). Probably not needed but it kind of markets itself and lots of scary pictures and diagrams could be part of the marketing.

 

[j_j]

Hmm.... maybe it is time for a "Smart DAC" that corrects for "intersample overs" and "illegal samples" and "increases headroom" (The current cure for intersample overs is reduce digital volume by 3-4 dB which is actually quite a bit of lost headroom). Probably not needed but it kind of markets itself and lots of scary pictures and diagrams could be part of the marketing.

There are DAC's that shift all data down 1 bit. This costs them a bit of noise floor, which is not necessarily a big deal today, in exchange for no weirdness in intersample overs. As to illegal samples, I have no problem with testing a DAC with a unit impulse.

I would, however, prefer to test it with an allpass sequence. That way the SNR of the measurement is much better.

 

[Sokel]

Out of curiosity I did a little test to see how IR looks like at a real-world output between PCM 96k and DSD64.
(it was a pain to do it but I did got results)

Settings and labels at the chart:

not much of a difference

(there's a good chance to mess things up big time testing like this, I hope I didn't)

 

[gberchin]

not much of a difference

Can you plot the impulse response magnitudes on a logarithmic scale, such as dB? That will really show the differences.

 

[Sokel]

Can you plot the impulse response magnitudes on a logarithmic scale, such as dB? That will really show the differences.

You mean like that?

 

[gberchin]

You mean like that?

No, it should look like a series of bumps that pass through -infinity at every zero-crossing, with a single 0dB peak at the center. Use the magnitude (absolute value) of the impulse response so that negative portions of the impulse response don't show up as "undefined".

EDIT: It will look something like the attached graphic.

 

[Sokel]

No, it should look like a series of bumps that pass through -infinity at every zero-crossing, with a single 0dB peak at the center. Use the magnitude (absolute value) of the impulse response so that negative portions of the impulse response don't show up as "undefined".

EDIT: It will look something like the attached graphic.

I don't think MTA has such an option for Y axis.

 

[gberchin]

I don't think MTA has such an option for Y axis.

A pity; viewing on a log scale does a great job of demonstrating the duration of the impulse response.

 

[j_j]

Out of curiosity I did a little test to see how IR looks like at a real-world output between PCM 96k and DSD64.
(it was a pain to do it but I did got results)

Settings and labels at the chart:

View attachment 470966

not much of a difference

(there's a good chance to mess things up big time testing like this, I hope I didn't)

The red impulse has a slightly lower cutoff frequency and a more gradual cutoff above that. Other than that?

Why does it say "multitone" I'd think allpass sequence would be the thing.

 

[bmc0]

Simple. Assuming it is not a fake chart, they used a digital synthesizer that uses a DAC so naturally creates pre and post ringing.

That was just a joke mocking the silly marketing material—I fully agree that it shows nothing useful for the layman and only serves to mislead. The reason I pointed to that though is that analog systems tend to not be linear phase, so chances are that their "analog" impulse isn't that of a fully analog system (as you said). They could have gotten a clean pulse with no notable overshoot/ringing from a high-bandwidth analog system, but they didn't. Maybe the goal was to make DSD "look" better?

You can't answer an impossible condition with another impossible condition.

In that case, I was not. It works just the same with a finite-length antialiasing filter as long as its cutoff is at Fs/2 (like the one in the AKM datasheet I linked). The filter's ringing has a period of Fs/2, so a sufficiently well-aligned pulse with sufficient bandwidth (infinite is not required) will result in a set of output samples where all are approximately zero except for one.

 

[amirm]

It works just the same with a finite-length antialiasing filter as long as its cutoff is at Fs/2 (like the one in the AKM datasheet I linked).

Now your argument is circular. You are assuming that the impulse is band limited therefore a band limited signal can create it.

he filter's ringing has a period of Fs/2, so a sufficiently well-aligned pulse with sufficient bandwidth (infinite is not required) will result in a set of output samples where all are approximately zero except for one.

"Approximately?" So you can't create that impulse unless you have the impossible filter. Which is what I said.