Time resolution of digital sampling

[Miska]

9 hours ago, pkane2001 said:

... and Redbook format can record delays of much less than 5 microseconds.

 

Down to about 346 ps...

 

[pkane2001]

1 hour ago, sandyk said:

 Ask Miska or Paul R, as their Naval service let them clearly hear differences below the noise level (Submarines) 

Miska provided training in that area , IIRC.

 

This is like saying you can train people to bend the bullet trajectory by flipping your wrist quickly while firing a gun. Sounds plausible and looks great when Angelina is doing it, but it's simply impossible in real life.

 

PS: To make it explicit, I’m referring to being able to train someone to hear a -200dB signal below a -120dB noise floor.

 

[Don Hills]

9 hours ago, pkane2001 said:

...

Hearing down to -200dB means (to me, at least) being able to detect, with ears, a signal (or its effects/modulation of another signal) at -200dB. No? Maybe it's that language barrier thing that Peter was talking about.

...

 

I interpreted his meaning as being what you wrote in parentheses. Of course, for the effect to be audible it has to result in an audible change of the "main" signal, and therefore have caused a much higher level of distortion than -200dB. 

 

[Jud]

On 10/4/2020 at 6:31 AM, pkane2001 said:

Delay isn’t the same thing as a transient.

 

Yes, I agreed with this earlier in the thread.

 

I of course agree with all the proven mathematical theorems that have been brought up in the thread. I have no interest in rehashing the math.

 

My interest lies much more along the lines of whether there are differences in the ways the ear-brain perceives these various acoustic phenomena. For example, 1/20,000th of a second is 50 microseconds, yet even if one is duly skeptical of Kunchur's 5 microsecond figure (thanks to whoever pointed out I was incorrectly stating nanoseconds), the consensus of prior experiments was that humans were capable of sensing delays on the order of 10 microseconds. I am virtually certain this doesn't imply we can hear 100kHz sounds, so the ear-brain must be doing something different with delays than with frequency perception. Fourier analysis is undoubtedly mathematically correct, yet at least some aspects of our acoustic perception don't depend on it (https://phys.org/news/2013-02-human-fourier-uncertainty-principle.html), i.e., our ear-brain system isn't doing Fourier analysis in processing at least some aspects of acoustic signals.

 

And so I wonder whether the ear-brain's perception of transients similarly is based on something distinct from frequency perception occurring in the brain (while noting that this, like the perception of delays, does not imply and cannot require that we hear sounds above 20kHz).

 

If it is the case, I don't know that this ought to surprise anyone. Certainly our visual frequency range (the range of the visible spectrum) has nothing to do with how quickly our visual system can react to light, or what the briefest flash of light we can perceive might be. So long as it doesn't require us to perceive signals above 20kHz (or in my case 16kHz as of a few years ago, very likely lower now), there's no necessity for evolution to have made our brains use Fourier analysis to perceive the sharpness of a transient.

 

I'm interested in transients (as part of the inharmonic attack phase of a musical sound) because it's been known for decades that they are critical to aspects of music listening as fundamental as knowing what instrument is playing.

[pkane2001]

1 hour ago, Jud said:

prior experiments was that humans were capable of sensing delays on the order of 10 microseconds. I am virtually certain this doesn't imply we can hear 100kHz sounds, so the ear-brain must be doing something different with delays than with frequency perception.

 

Yes, the lower threshold of ITD (interaural time difference) has been measured to be about 10 microseconds. This has to do with spatial perception and the evolutionary need of our ancestors to be able to tell where the sound is coming from. Probably to detect a possible dinner, or to not become one.

 

Quote

i.e., our ear-brain system isn't doing Fourier analysis in processing at least some aspects of acoustic signals.

There's no reason to think that the ear or the brain are using Fourier analysis here -- that'd be silly. But, it's just as silly to assume that because the ear is using another type of phase difference detection, that Fourier analysis somehow doesn't apply to the original signal. 

 

1 hour ago, Jud said:

And so I wonder whether the ear-brain's perception of transients similarly is based on something distinct from frequency perception occurring in the brain (while noting that this, like the perception of delays, does not imply and cannot require that we hear sounds above 20kHz).

Sure. A leading edge detection or a first zero crossing detection can be used without involving frequency analysis. There are other ways. It is very impressive, though, that the human brain can detect such tiny differences, considering the relatively much slower speeds of electrochemical processing in neurons.

 

1 hour ago, Jud said:

I'm interested in transients (as part of the inharmonic attack phase of a musical sound) because it's been known for decades that they are critical to aspects of music listening as fundamental as knowing what instrument is playing

 

The main point of the discussion up until now was that transients can be analyzed with Fourier, sampled up to any desired frequency, and are just as subject to Nyquist-Shannon as any other signal. Microsecond, nanosecond, or even picoseconds delay does not require higher sampling rates to be captured in a PCM recording. Redbook is sufficient for that. 

[Jud]

11 hours ago, pkane2001 said:

13 hours ago, Jud said:

 

Sure. A leading edge detection or a first zero crossing detection can be used without involving frequency analysis. There are other ways. It is very impressive, though, that the human brain can detect such tiny differences, considering the relatively much slower speeds of electrochemical processing in neurons.

 

13 hours ago, Jud said:

I'm interested in transients (as part of the inharmonic attack phase of a musical sound) because it's been known for decades that they are critical to aspects of music listening as fundamental as knowing what instrument is playing

Expand  

 

The main point of the discussion up until now was that transients can be analyzed with Fourier, sampled up to any desired frequency, and are just as subject to Nyquist-Shannon as any other signal. Microsecond, nanosecond, or even picoseconds delay does not require higher sampling rates to be captured in a PCM recording. Redbook is sufficient for that. 

 

All agreed. Combining these first and second points, though, leads me to wonder if the system (not forgetting mics and speakers as well) should be set up to reproduce frequencies in excess of 20kHz in order to give us the proper perception of some of the sharper/faster initial attack transients. I have no idea of the answer (haven't found any relevant academic work), just wondering.

[jabbr]

6 hours ago, Jud said:

 

All agreed. Combining these first and second points, though, leads me to wonder if the system (not forgetting mics and speakers as well) should be set up to reproduce frequencies in excess of 20kHz in order to give us the proper perception of some of the sharper/faster initial attack transients. I have no idea of the answer (haven't found any relevant academic work), just wondering.

You really want Kunchur’s work replicated, so you could use the corner frequency of the LPF at which the minimally detectable difference occurs. This would tell you what sampling rate you need.

[sandyk]

10 hours ago, Jud said:

Combining these first and second points, though, leads me to wonder if the system (not forgetting mics and speakers as well) should be set up to reproduce frequencies in excess of 20kHz in order to give us the proper perception of some of the sharper/faster initial attack transients.

Jud

 IIRC, you have wider bandwidth speakers than many members (To 40K?)

I bet that if you replaced them temporarily with typical speakers that start to die off rapidly at 20K you would have at least part of your answer

 Kind Regards

Alex

 

 P.S. 

 Have you tried listening to some of the comparisons from Barry Diament's Soundkeeper format 

comparison page ? 

[Josh Mound]

On 2/20/2020 at 5:14 PM, Don Hills said:

Another way of looking at it is to consider an "intersample over" as being no different than a lower level signal. Provided that the signal amplitude at the moment of sampling is no greater than can be represented by the digital value (is accurately captured), the values of the signal in between the samples will be accurately reproduced after reconstruction. (Provided, of course, the hardware is designed to handle intersample overs.) The Benchmark article makes this point with the Steely Dan track - no clipping, all of the samples accurately captured the signal amplitude, but many DACs clip the intersample overs.

Note that filterless NOS DACs don't accurately reproduce values in between the samples. Far from being a barrier to accurate reproduction, the reconstruction filter is mathematically essential for accurate reproduction.

Don, your posts here are wonderful.
 

How much do you think we should worry about intersample overs? Are they audible? Should we apply digital volume attenuation in Audirvana, HQ Player, etc. to avoid overs? Or does the extra processing involved in using digital volume attenuation outweigh the benefits of avoiding overs?

 

I’ve been using a true peak plugin with Audirvana to see which albums I have clip, then applying digital volume attenuation. However, I still debate whether I’m doing the right thing, as then I’m no longer sending “bit perfect” data to the DAC. 

[Jud]

11 hours ago, JoshM said:

 

How much do you think we should worry about intersample overs? Are they audible?

 

Not Don, and this is completely anecdotal, but:

 

Depends what you mean by "audible." As an audibly evident defect, I haven't experienced it. I assume a difference of 1 or more dB in response would be audible if, for example, carefully comparing two files, which I have not done.

 

This was with HQPlayer, fooling around with volume levels. I don't know whether HQPlayer has any sort of "graceful failure" mode for intersample overs that may have affected my experience.

[Don Hills]

18 hours ago, JoshM said:

Don, your posts here are wonderful.
 

How much do you think we should worry about intersample overs? Are they audible? Should we apply digital volume attenuation in Audirvana, HQ Player, etc. to avoid overs? Or does the extra processing involved in using digital volume attenuation outweigh the benefits of avoiding overs?

 

I’ve been using a true peak plugin with Audirvana to see which albums I have clip, then applying digital volume attenuation. However, I still debate whether I’m doing the right thing, as then I’m no longer sending “bit perfect” data to the DAC. 

 

Thank you.

 

I don't worry too much about intersample overs. I feel that music with aspirations to high fidelity won't have been pushed to the limit in mastering, and conversely music that has been pushed hard will likely already have more audible damage.

 

As to audibility, it of course depends on the specific case. I could probably generate a test signal where it was quite audible, but I don't think it's a big problem in everyday music. It will also depend on your DAC's behaviour when processing intersample overs - some will handle them better than others.

 

The simplest prevention method is to attenuate the digital signal before the DAC by exactly 1 bit (6.02dB). This requires minimum processing and should handle all except pathological cases. Personally, I feel that maintaining "bit perfectness" is overhyped. A competently implemented digital volume control will be effectively as transparent as an equivalent analogue control.

 

For those albums where you find overs, have you tried comparing the before and after attenuation signals? Note that you need to apply attenuation post-DAC to make both signals the same volume for comparison. My feeling is that you'll be hard put to hear a difference.

 

[Jud]

33 minutes ago, Don Hills said:

I feel that music with aspirations to high fidelity won't have been pushed to the limit in mastering, and conversely music that has been pushed hard will likely already have more audible damage.

 

From your lips to producers' ears. 😉 I found they happened occasionally in my listening without attenuation (admittedly not all of which has aspirations to high fidelity).

 

33 minutes ago, Don Hills said:

 

The simplest prevention method is to attenuate the digital signal before the DAC by exactly 1 bit (6.02dB).

 

With HQPlayer, -3dB seemed to be enough for virtually everything I tried.

[sandyk]

1 hour ago, Jud said:

 

From your lips to producers' ears. 😉 I found they happened occasionally in my listening without attenuation (admittedly not all of which has aspirations to high fidelity). - Do Hills

 

 

With HQPlayer, -3dB seemed to be enough for virtually everything I tried.

I feel sure that Barry Diament won't mind me quoting him here.

 

Quote

It wasn’t long after the earliest days of digital that engineers that were paying attention started to stay clear of 0 dBFS.  Nowadays, I’d never let a digital transfer go over -0.3.  That is digital to digital.  For analog to digital (in other words, the first time a signal is digitized) I always stay below -6 dBFS.  Why?  Because every monolithic A-D converter of which I’m aware will exhibit considerably lower distortion at -6 (or -10) than it will at higher levels.  Especially since we’re using 24-bits today, I’d rather hit digital at -20 maximum level than -2.  Final levels can be adjusted later, in mastering.

 

[Miska]

On 10/8/2020 at 6:05 AM, JoshM said:

How much do you think we should worry about intersample overs? Are they audible? Should we apply digital volume attenuation in Audirvana, HQ Player, etc. to avoid overs? Or does the extra processing involved in using digital volume attenuation outweigh the benefits of avoiding overs?

 

Yes you should. And you have no way of avoiding "extra processing involved in using digital volume" with HQPlayer.

 

On 10/8/2020 at 5:35 PM, Jud said:

I don't know whether HQPlayer has any sort of "graceful failure" mode for intersample overs that may have affected my experience.

 

It has, every time "Limited" counter is incremented you have triggered the soft-knee limiter that has been designed to be as inaudible as possible.

 

On 10/9/2020 at 12:59 AM, Don Hills said:

I don't worry too much about intersample overs. I feel that music with aspirations to high fidelity won't have been pushed to the limit in mastering, and conversely music that has been pushed hard will likely already have more audible damage.

 

To what I've seen, over 90% of RedBook material has overs. And for over 90% of it, -3 dBFS is enough to avoid it. And there are clear mathematical reasons why this is the case.

 

There are some albums though where -3 dBFS is not enough.

 

[Don Hills]

19 hours ago, Miska said:

...

To what I've seen, over 90% of RedBook material has overs. And for over 90% of it, -3 dBFS is enough to avoid it. And there are clear mathematical reasons why this is the case.

...

 

 

Personally, I don't spend much time worrying about them. By definition, an intersample over occurs between 2 samples. Therefore, its frequency content is above 22 KHz. Provided that the DAC clips cleanly and the ultrasonic content doesn't upset the following equipment, it should be inaudible. (Of course, in real life things are rarely that ideal.) But I'd still rather see them avoided at source than having to allow for them in playback.

[fas42]

1 hour ago, Don Hills said:

 

Personally, I don't spend much time worrying about them. By definition, an intersample over occurs between 2 samples. Therefore, its frequency content is above 22 KHz. Provided that the DAC clips cleanly and the ultrasonic content doesn't upset the following equipment, it should be inaudible. (Of course, in real life things are rarely that ideal.) But I'd still rather see them avoided at source than having to allow for them in playback.

 

What??!!  An intersample over is above 22kHz? ... A trivial, manufactured example of what can happen, at say 11kHz,

 

[Miska]

Intersample overs happen at least because someone decides to run "normalize" function for music at 44.1k sampling rate. Since actual sampling points rarely coincide with the actual waveform peaks, this results in values higher than 0 dBFS when the actual waveform is reconstructed. The reconstructed values are more likely to reach the highest point, higher the digital filter oversampling factor is.

 

Another common reason is RedBook content driven to clipping, which also seems to be about 90+% of modern content.

 

[Don Hills]

1 hour ago, fas42 said:

 

What??!!  An intersample over is above 22kHz? ...

 

Yes. In your example with an 11 KHz sine wave, clipping of the intersample peak will result in harmonic distortion. And those harmonics begin at 22 KHz...

 

Edit: I think I can come up with special cases where there may be distortion products below 22 KHz. I'll see what I can do with Audacity.

[danadam]

2 hours ago, Don Hills said:

Edit: I think I can come up with special cases where there may be distortion products below 22 KHz. I'll see what I can do with Audacity.

https://forum.cockos.com/showthread.php?t=201868

[Don Hills]

31 minutes ago, danadam said:

https://forum.cockos.com/showthread.php?t=201868

 

Thanks. I'll try it.

[opus101]

4 hours ago, Don Hills said:

Edit: I think I can come up with special cases where there may be distortion products below 22 KHz. I'll see what I can do with Audacity.

 

From a quick play with Audacity, I can show them. Project rate here is 44.1kHz.

 

1. Generate white noise at 99% full scale.

2. Low pass filter with 48dB slope, cut off 8kHz

 

You'll see plenty of red in the waveform, these are the overs. Plot the spectrum and you'll see there's grunge above the cut-off of the filter but below 20kHz.

[Don Hills]

If I recall correctly, the red in Audacity indicates where at least 3 consecutive samples are at the digital limit ("0dB"). This is usually real clipping, where the level of the signal being sampled at the sample moment is greater than that represented by the sample value. It is possible to have inter-sample overs where all of the sample values are valid (the sample value accurately represents the value of the input voltage at that moment). 

 

Try generating and filtering the white noise at a lower amplitude than 99%. Now normalise it. There should be no red, but there may be intersample overs. I'll try it myself hortly.

[opus101]

9 minutes ago, Don Hills said:

Try generating and filtering the white noise at a lower amplitude than 99%. Now normalise it. There should be no red, but there may be intersample overs. I'll try it myself hortly.

 

I've done this and its possible to get no red, yes. But then how to see when there are intersample overs?

 

There's another way to 'see red' with this - don't do filtering just upsample the white noise. When there's no clipping (red) at 44.1k but the red appears at higher rates would you say that's an instance of intersample overs?

[fas42]

The first example I posted above I just grabbed from a quick search of images ... but note that a comprehensive overview was done by @Archimago , http://archimago.blogspot.com/2018/09/musings-measurements-look-at-dacs.html. And specifically he notes what can occur with, as examples, fs/4 and fs/6. Here the simple, potential HD is well and truly in the audio range - and a real world DAC does make a mess of it.

 

Also note that he points to a discussion of worst case scenarios ... https://hydrogenaud.io/index.php/topic,98753.0.html.

[Don Hills]

18 hours ago, opus101 said:

 

I've done this and its possible to get no red, yes. But then how to see when there are intersample overs?

 

There's another way to 'see red' with this - don't do filtering just upsample the white noise. When there's no clipping (red) at 44.1k but the red appears at higher rates would you say that's an instance of intersample overs?

 

Yes, that makes sense. "True peak" meters do this to indicate intersample overs.