Skip to main content

Topic: Is this digital bashing true or false? (Read 1126 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.
  • board
  • [*]
Is this digital bashing true or false?
Hi guys.
I hope I put this in the right forum.


First, here's the video, and then afterwards I'll elaborate. The point in question is by David Robinson at 11:46, but this is a continuation of an answer to a question that was asked at 5:23 and then elaborated on by Robinson at 8:47

https://www.youtube.com/watch?v=FEr5UfBjDSc

So, my question is:
Is what David Robinson claiming true or false, and if what he is claiming is true, is it also audible?

hopefully you can tell me something along the lines of:
"yes, what the person says in the video is completely correct, and that phase shift is always audible on all CDs", or "no, his claim is completely false and it's never audible under any legitimate circumstances, unless a CD is intentionally made to be like that, but nobody would do that when releasing music", or "his claim of audibility only applies when something has gone wrong when making a CD, so under legitimate circumstances where nothing has gone wrong, there is neither any 'drift' in phase above 1 kHz, and nor can there be any audible effect, since there is no issue", or "in most cases, phase on CDs doesn't deviate to an extent that is worth worrying about" or another more fitting description.

It's also worth noting that David Robinson actually says "in the beginning", implying that it might be different now, but he still seems to be a CD hater.

I am aware that this topic may very easily go off-topic, so hopefully we can leave aside all the other claims made in the video, about 192 kHz sampling rates being audibly superior to 44.1 kHz ones, as I believe that's just another case of them not understanding the Nyquist theorem, and they haven't done any blind tests, and hopefully we can also leave out the claim that PCM and DSD sounds notably different, and that a recording of a vinyl disc will sound different than the vinyl disc itself, as the converter will leave it's sound imprint on the recording. While strictly speaking the latter might be true, then I like to think that most properly constructed converters nowadays are extremely close to 100 % transparancy and certainly transparent within audibility.
  • Last Edit: 17 February, 2017, 01:20:25 PM by kode54

Re: Is this digital bashing true or false?
Reply #1
That panel would likely not accept the results of a double-blind/ABX test anyway.  The fact they claim audible differences that couldn't be proven in an ABX test is also expected.  The explanation of phase shift caused by PCM sampling is complete nonsense.  They presume vinyl/analog sound "better" because it doesn't suffer from that nasty digital phase shift problem.  Except...they clearly have never bothered to look at the phase response, or square wave response of an analog tape recorder, or that tape cut to vinyl. Spoiler: Yes, it's different than PCM, No, it's not better than PCM.

  • KozmoNaut
  • [*][*][*][*][*]
Re: Is this digital bashing true or false?
Reply #2
The explanation of phase shift caused by PCM sampling is complete nonsense.

I think this type of misinformation happens because a lot of people think PCM audio works in discrete 'ticks', and that if the ratio between sampling rate and frequency becomes too low, you don't have enough "temporal resolution" to fully capture the sound wave.

It's all based on a gross misunderstanding of the Nyquist-Shannon sampling theorem.

I would love to play back a format that actually worked like that to these people, just to see their absolutely horrified expressions.

Re: Is this digital bashing true or false?
Reply #3
I would love to play back a format that actually worked like that to these people, just to see their absolutely horrified expressions.

Be the most entertaining thing in the world for me to witness as well.

  • ajinfla
  • [*][*][*][*][*]
Re: Is this digital bashing true or false?
Reply #4
they haven't done any blind tests
There goes your fishing expedition. Next.
Loudspeaker manufacturer

Re: Is this digital bashing true or false?
Reply #5

But I don't really understand this, so hopefully you can tell me something along the lines of:
"yes, what the person says in the video is completely correct, and that phase shift is always audible on all CDs", or "no, his claim is completely false and it's never audible under any legitimate circumstances, unless a CD is intentionally made to be like that, but nobody would do that when releasing music", or "his claim of audibility only applies when something has gone wrong when making a CD, so under legitimate circumstances where nothing has gone wrong, there is neither any 'drift' in phase above 1 kHz, and nor can there be any audible effect, since there is no issue", or "in most cases, phase on CDs doesn't deviate to an extent that is worth worrying about" or another more fitting description.


And the right answer is:

...there is neither any 'drift' in phase above 1 kHz, and nor can there be any audible effect, since there is no issue...

For example, consider a modern converter that is designed to be Linear Phase, which is to say it has a phase versus frequency response that is practically the same as that of an ideal delay over the entire audible range. 

These guys were as aware of the phase response of analog tape or vinyl, or microphones, or speakers or the general run of high quality studio gear that is in general use, they'd "Have a cow".  Then there is the slight matter of the fact that the human ear does not perceive phase changes applied to both channels equally above about 1 KHz.  There's no process in place to do so - the ear functions like a spectrum analyzer without any way to determine or pass on phase response.  Finally, their claim that DSD is somehow free of these problems is yet another fantasy.

Re: Is this digital bashing true or false?
Reply #6
I would love to play back a format that actually worked like that to these people, just to see their absolutely horrified expressions.

Be the most entertaining thing in the world for me to witness as well.

Actually,  analog media works much that way.

  • KozmoNaut
  • [*][*][*][*][*]
Re: Is this digital bashing true or false?
Reply #7
I would love to play back a format that actually worked like that to these people, just to see their absolutely horrified expressions.

Be the most entertaining thing in the world for me to witness as well.

Actually,  analog media works much that way.

In discrete 'ticks'? I'm... not reeeaaaally 100% onboard with that.

Sure, once you get down to the level of the individual PVC molecules or magnetic particles, I guess you could use that description. But it definitely doesn't sit quite right with me.

  • board
  • [*]
Re: Is this digital bashing true or false?
Reply #8

But I don't really understand this, so hopefully you can tell me something along the lines of:
"yes, what the person says in the video is completely correct, and that phase shift is always audible on all CDs", or "no, his claim is completely false and it's never audible under any legitimate circumstances, unless a CD is intentionally made to be like that, but nobody would do that when releasing music", or "his claim of audibility only applies when something has gone wrong when making a CD, so under legitimate circumstances where nothing has gone wrong, there is neither any 'drift' in phase above 1 kHz, and nor can there be any audible effect, since there is no issue", or "in most cases, phase on CDs doesn't deviate to an extent that is worth worrying about" or another more fitting description.


And the right answer is:

...there is neither any 'drift' in phase above 1 kHz, and nor can there be any audible effect, since there is no issue...

For example, consider a modern converter that is designed to be Linear Phase, which is to say it has a phase versus frequency response that is practically the same as that of an ideal delay over the entire audible range. 

These guys were as aware of the phase response of analog tape or vinyl, or microphones, or speakers or the general run of high quality studio gear that is in general use, they'd "Have a cow".  Then there is the slight matter of the fact that the human ear does not perceive phase changes applied to both channels equally above about 1 KHz.  There's no process in place to do so - the ear functions like a spectrum analyzer without any way to determine or pass on phase response.  Finally, their claim that DSD is somehow free of these problems is yet another fantasy.
Thanks for your help :-).
So just to make sure I understand you correctly, although you were pretty clear about it:
What David Robinson is saying is basically just made up? There's no phase drift/skewing?

I've seen people make up these things before, but as I didn't know much about this issue it sounded "scientifically enough" to make me doubt whether it was true or false. I've also looked at Ethan Winer's "Audio Myths" video, and in there he plays clips with phase shift at certain frequencies with no audible effect.
Robinson did also say "in the beginning", so I figured maybe this was before linear phase filters became common, which I think was in the early 90s (but I could be mistaken).
  • Last Edit: 17 February, 2017, 12:00:53 PM by board

  • Wombat
  • [*][*][*][*][*]
Re: Is this digital bashing true or false?
Reply #9
There goes your fishing expedition. Next.
Rich B couldn't do better.
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!

  • eric.w
  • [*]
Re: Is this digital bashing true or false?
Reply #10
The statement that, with 44.1kHz sampling, a 20 Hz sine is well sampled, but above 10k the phase is "totally random", sounds to me like a "grid snapping" mental model, where you imagine that sine wave peaks must be snapped to the nearest sample point. e.g. at the Nyquist frequency of 22.050kHz, you would imagine that there is only one possible sine wave phase.

This is incorrect, though. I suggest watching Monty's "Digital Show and Tell" video, but in particular the part at 20:55 addresses this idea: https://xiph.org/video/vid2.shtml

  • board
  • [*]
Re: Is this digital bashing true or false?
Reply #11
I would be happy to hear more comments about the phase issue in particular (instead of sample rates etc.) :-).

JJ pointed me to some of his papers available publicly, and I think I might finally understand those points in his papers filtutv1 and src1_9_16:

Quote
“Linear Phase” (constant delay)
- If a filter has a constant delay, the phase shift of the filter will be t*w, where t is the time delay, and w the natural frequency (2 pi f).
* This means that a delay can exhibit enormous phase shift.
* This phase shift, however, is ONLY delay.


Quote
* In most cases, a symmetric FIR (convolutional) filter is used. This kind of filter has a fixed (constant) delay over all frequencies, which means
* It has a phase shift, relative to the input, of 2*pi*f*t, where f is the frequency of interest, and t is the time delay. This “linear phase” means that the signal is purely delayed, all frequencies arrive at the same instant.
* It has a substantial amount of energy before the middle (main lobe) of the filter, being symmetric.
* If it’s poorly designed, or is too short, you can get pre-echo.
      *Oddly, that doesn’t happen if it’s not too short.
      *That’s another story
* This kind of filter design accounts for most filters in use


So, if I understand this correctly (and I might not), then the phase skewing that David Robinson talks about can only happen in other types of filters than linear phase filters.

  • kode54
  • [*][*][*][*][*]
  • Administrator
Re: Is this digital bashing true or false?
Reply #12
This sort of phase issue "can" happen with any sampling frequency not perfectly factoring against the sample rate of components of the signal. The thing is, due to how signal reconstruction works, this does not matter.

Re: Is this digital bashing true or false?
Reply #13
This sort of phase issue "can" happen with any sampling frequency not perfectly factoring against the sample rate of components of the signal. The thing is, due to how signal reconstruction works, this does not matter.

I don't think so.  I've never seen this happen, and I've looked for it. Please provide an independent reliable reference or a test sample.

There are some odd effects such as those where quantization distortion can be zero with some precondidtions including sample rate being a strict multiple of the signal frequency, but that is amplitude distortion, not phase distoriton.   Also, if your signal frequency is within a few Hz of the Nyquist frequency, things can get strange, but again that is a very special case.

Please educate me! ;-)

Re: Is this digital bashing true or false?
Reply #14
I would be happy to hear more comments about the phase issue in particular (instead of sample rates etc.) :-).

JJ pointed me to some of his papers available publicly, and I think I might finally understand those points in his papers filtutv1 and src1_9_16:

Quote
“Linear Phase” (constant delay)
- If a filter has a constant delay, the phase shift of the filter will be t*w, where t is the time delay, and w the natural frequency (2 pi f).
* This means that a delay can exhibit enormous phase shift.
* This phase shift, however, is ONLY delay.


Quote
* In most cases, a symmetric FIR (convolutional) filter is used. This kind of filter has a fixed (constant) delay over all frequencies, which means
* It has a phase shift, relative to the input, of 2*pi*f*t, where f is the frequency of interest, and t is the time delay. This “linear phase” means that the signal is purely delayed, all frequencies arrive at the same instant.
* It has a substantial amount of energy before the middle (main lobe) of the filter, being symmetric.
* If it’s poorly designed, or is too short, you can get pre-echo.
      *Oddly, that doesn’t happen if it’s not too short.
      *That’s another story
* This kind of filter design accounts for most filters in use


So, if I understand this correctly (and I might not), then the phase skewing that David Robinson talks about can only happen in other types of filters than linear phase filters.

True.

If the phase distortion due to time delay were an audible problem duing music playback, then all forms of media, not just digital would be greatly affected because we listen to recordings minutes to decades after the live performance, and the phase shift due to delay ranges from astronomical to even a whole lot bigger.

Furthermore I got the impression that the speaker was referring to a phase shift that wanders around, while the phase shift due to the usual run of digital filters is fixed in place, rock solid. This contrasts with analog media, where the phase shift (and amplitude response) wanders around, often fairly audibly.

May he's objecting to digital's inherent constant phase shift as opposed to his preferred analog media phase shift that wanders around.  And wander around, analog media does. It is not unusual for there to be 1 dB and greater variations within the audible range with both vinyl and analog tape playback.

  • kode54
  • [*][*][*][*][*]
  • Administrator
Re: Is this digital bashing true or false?
Reply #15
Probably quantization effects, then. Not lining up with the frequency of the waveform, it can result in sample points appearing to be some sort of gibberish, until you do something like intelligently plot a curve between the points, like a reconstruction filter would.

Don't mind me, I only have a rudimentary understanding of this stuff, and have zero college education to boot. I pick up bits and bobs as I crawl the web and read people's code, and maybe even learn something!

I did manage to sort of guess an algorithm that turned out to be really close to what Paulstretch does, though. I guess I'm not totally out of touch.