Sampling Rates and Music Quality

[Lee]

But I'm talking more specifically about music to be released (and some of it also recorded) in the future, and the huge existing library has little or nothing to do with that. If a recording is mastered at 24/192, why should it be any easier to release at 16/44.1? Slightly cheaper, perhaps, but right now CD's are often cheaper than MP3's, so file size alone is only part of the retail cost (and I don't want to get into discussing the reasons for that); releasing the 24/192 will bypass several processing steps.

I doubt that file size will be a huge issue in the future. We are likely to see audiophile downloads become a key source of better quality music as physical formats get more scarce over time. Also, the march of progress to more capacity/lower cost storage continues.

I think the bigger issue is likely to be how comfortable the record labels (which imho continue to act like dinosaurs) are with releasing near master tape quality in digital format that can be easily copied.

 

[DonH50]

@Frantz -- Thanks very much for your kind words! IME these chapters never seem to close...

@Lee -- The "phase, ripple and transition regions" are the in-band filter artifacts I was referring to, and a higher sampling rate does get those out of the audio passband. Integer increases, and power-of-two even more so, makes the digital processing much, much easier when converting among, or just having to handle, different sample rates.

 

[Lee]

@Frantz -- Thanks very much for your kind words! IME these chapters never seem to close...

@Lee -- The "phase, ripple and transition regions" are the in-band filter artifacts I was referring to, and a higher sampling rate does get those out of the audio passband. Integer increases, and power-of-two even more so, makes the digital processing much, much easier when converting among, or just having to handle, different sample rates.

Don, is it not better to keep the chapter open when there are divided opinions so that we can all gain understanding?

Also, I don't understand your response. Isn't Stuart in the quote suggesting that higher sampling rates do in fact begin to address these artifacts?

He says, "Increase the sample rate by a margin sufficient to move the phase, ripple and transition regions
further away from the human audibility cut-off. "

 

[kevinh]

Question are there algorithms that allow say a 44.1 16 bit file to be up sampled to a 88.2 24 bit for example?

Would the quality of the resulting D/A conversion be improved?

This would allow streaming audio from sources like Pandora or other internet raido sources with their huge libraries to be more usable.

 

[DonH50]

I just need more time to read all the chapters, Lee. I am still working (have some stuff to finish up) and need to practice yet tonight, been up since 5 AM and I am tired!

I agree with the quote; increasing the sample rate moves the filter cut-offs higher, thus reducing filter artifacts ("distortion") in the audio band. I think that is what I said, or implied, in my earlier post. I do think that is the primary benefit of increased sampling rate, and one of the advantages of delta-sigma modulators since they move the sampling rate (and thus Nyquist point and filter cut-offs) well above the audio band. Higher bandwidth comes at a cost, including higher noise and (typically) higher distortion.

 

[DonH50]

Question are there algorithms that allow say a 44.1 16 bit file to be up sampled to a 88.2 24 bit for example?

Would the quality of the resulting D/A conversion be improved?

This would allow streaming audio from sources like Pandora or other internet raido sources with their huge libraries to be more usable.

There are many. The simplest is interpolation, and that progresses from linear to sinusoidal to high-order predictive systems. They would not improve the fundamental DAC, but do allow filters to be moved above the audio band, with benefits described above. The catch is the algorithm is trying to "guess" what data to add, and does not always guess rightly...

 

[Lee]

I just need more time to read all the chapters, Lee. I am still working (have some stuff to finish up) and need to practice yet tonight, been up since 5 AM and I am tired!

I agree with the quote; increasing the sample rate moves the filter cut-offs higher, thus reducing filter artifacts ("distortion") in the audio band. I think that is what I said, or implied, in my earlier post. I do think that is the primary benefit of increased sampling rate, and one of the advantages of delta-sigma modulators since they move the sampling rate (and thus Nyquist point and filter cut-offs) well above the audio band. Higher bandwidth comes at a cost, including higher noise and (typically) higher distortion.

No worries Don. Get some sleep, since 5am is a long day! Appreciate your comments.

 

[Lee]

Question are there algorithms that allow say a 44.1 16 bit file to be up sampled to a 88.2 24 bit for example?

Would the quality of the resulting D/A conversion be improved?

This would allow streaming audio from sources like Pandora or other internet raido sources with their huge libraries to be more usable.

Yes, our local audio club is experimenting with Izotope. It's not as good as the original recording in hirez but it helps!

 

[kevinh]

DonH50 & Lee thanks for the info.


Lee which of the Izotope products are you guys playing with?

 

[Bruce B]

Bruce, this is different from my experience. How specifically did you come to the conclusion that bit depth trumps sampling rate from a sonic standpoint?

When I sit down to listen to a client's files, I can immediately within a few seconds tell if it's a 16 or 24 bit file. I can't do that with a high samplerate file. I'm just saying that it takes longer for me to hear the difference between a 44.1 file as opposed to an 88.2 file. Don't get me wrong, samplerate DOES make a difference. I'm just saying it's not as big of impact going from 48 to 96/192 as it is from 16 to 24bit.

 

[Lee]

DonH50 & Lee thanks for the info.


Lee which of the Izotope products are you guys playing with?

The latest version that comes with SoundForge.

 

[Lee]

When I sit down to listen to a client's files, I can immediately within a few seconds tell if it's a 16 or 24 bit file. I can't do that with a high samplerate file. I'm just saying that it takes longer for me to hear the difference between a 44.1 file as opposed to an 88.2 file. Don't get me wrong, samplerate DOES make a difference. I'm just saying it's not as big of impact going from 48 to 96/192 as it is from 16 to 24bit.

Thanks Bruce, that makes sense. I guess when I hear a 24/48 DVD-Audio disc, I immediately feel cheated a bit as the sound quality is not on par with my 24/96 or 24/88 discs. I guess the comparison is difficult in that it is tough to find 16/88 or 16/96 source material to compare to 16/44 to sort of isolate the sampling rate impact. Perhaps we both agree that 24/96, all else being equal, is a lot better than 16/44.

 

[rbbert]

IIRC, Soundforge has Izoptope's mbit+ for decimating/dithering, but uses a different SRC (than Izotope's). In any case, have you tried the downsampling and dithering comparison I suggested earlier to see which makes more of a difference in your own mind?

 

[Lee]

IIRC, Soundforge has Izoptope's mbit+ for decimating/dithering, but uses a different SRC (than Izotope's). In any case, have you tried the downsampling and dithering comparison I suggested earlier to see which makes more of a difference in your own mind?

I have not tried that but one minor correction. We used both the SoundForge version and the full version which another member in the audio club had. They created redbook and upsampled redbook files for comparison.

 

[rbbert]

I have not tried that but one minor correction. We used both the SoundForge version and the full version which another member in the audio club had. They created redbook and upsampled redbook files for comparison.

But to really address your original hypothesis you want to start with a hires master (and make sure it really is) and then either downsample or truncate/dither and see which has more effect on the sound quality.

 

[Lee]

But to really address your original hypothesis you want to start with a hires master (and make sure it really is) and then either downsample or truncate/dither and see which has more effect on the sound quality.

That's one way to do it I guess.