another good article by Justin Coletti
http://trustmeimascientist.com/2013/02/04/...d-when-it-isnt/ (http://trustmeimascientist.com/2013/02/04/the-science-of-sample-rates-when-higher-is-better-and-when-it-isnt/)
Too long; didn't get through it all.
I essentially stopped reading when he failed to explain that content beyond 17kH would likely be masked for those teenagers who can hear test tones at those frequencies, instead suggesting that their mere attention would grant them the ability.
IOW, he still left the door too far ajar.
From another post in a binned discussion before the off-topic split:
FWIW, I did appreciate everything else I had read up to the point where I stopped reading. While masking seems to be missing, the author deserves props for addressing oversampling (which is more germane to the topic). I may pick it back up at a later time, though I honestly don't imagine I'm going to read anything I haven't already read either here or elsewhere, though I'll try to keep an open mind. wink.gif
I did later finish it up and there isn't anything new. Still, it's a good article, though I'd like to a larger set of citations for all the claims that are made, especially concerning audible differences between DACs.
Thanks for the kind words Greynol.
I appreciate the critique as well. Still, I might add that there's a potentially positive side effect to 'leaving the door open' just a crack: More folks may be likely to enter through it, and get interested in evidence-based thinking. And I think that's good for everybody.
To be fair, I'd also argue that I and others have in some cases differentiated between converters and converter settings in blind tests. Although I'd also be the first to admit that this could very well be due to design differences, intermodulation distortion, or even design choices intentionally made to color the sound of the converter -- often at the analog stage.
(There's at least one popular "high-end" converter than even adds a transformer input in an effort to audibly color the sound of their unit. I'd be happy to do some listening tests with you on this kind of thing if you're interested!)
Placebo effect and confirmation bias are huge issues, but if we do leave the door cracked ever so-slightly -- especially for some genuine or reasonable caveats -- I figure we may be glad that we did in the end.
To me, the problem is not that too many people are interested in or educated about the science side of audio and perception. If anything, I'd say it might be the opposite. And I think we need to invite more of them to the party. That's my take anyway!
Thanks again for taking the time, and for weighing in.
Best wishes,
Justin Colletti
Audio Engineer, Educator, Journalist
http://justincolletti.com (http://justincolletti.com)
http://trustmeimascientist.com (http://trustmeimascientist.com)
http://sonicscoop.com (http://sonicscoop.com)
Some of the things you cite there are also said in a similar way by some man i also think fully deserves the title Audio Engineer.
The Daniel Weiss interview is still there http://ukiro.com/2011/05/12/interview-daniel-weiss/ (http://ukiro.com/2011/05/12/interview-daniel-weiss/)
Of course several statements will collide with our TOS but one thing should become clear.
There is also created a market by charlatans and clowns that always push every number higher only to make profit, nothing else.
Another phenomena that makes me shiver is the dsd crowd that clearly hears the much better sound again, taking really off starting at 5.6MHz.
another good article by Justin Coletti
http://trustmeimascientist.com/2013/02/04/...d-when-it-isnt/ (http://trustmeimascientist.com/2013/02/04/the-science-of-sample-rates-when-higher-is-better-and-when-it-isnt/)
As far as I can see, the Colletti article is just a compendium of old articles. The Lavry 2012 article that is based on his 2004 article. The Xiph article. No new experimental work. It is terribly old news that nonlinear processing can benefit from high sample rates by avoiding aliasing. Nothing that relates to ordinary playback of music on home systems. Can be of interest to electronic musicians.
Footnotes:
Lavry 2004:
http://lavryengineering.com/pdfs/lavry-sampling-theory.pdf (http://lavryengineering.com/pdfs/lavry-sampling-theory.pdf)
Lavry 2012:
http://www.lavryengineering.com/pdfs/lavry...ality_audio.pdf (http://www.lavryengineering.com/pdfs/lavry-white-paper-the_optimal_sample_rate_for_quality_audio.pdf)
Xiph 2012:
http://xiph.org/~xiphmont/demo/neil-young.html (http://xiph.org/~xiphmont/demo/neil-young.html)
He's endorsing sensible things others have said, like peer-review. It's a good thing.
Quite a good article.
The only thing I don't get is the part about sampling "too fast". There is no such thing as a "loss of accuracy" when sampling at 192 kHz rather than 96 kHz, not generally at least. While it is certainly imaginable that a practical ADC would be performing worse for all kinds of reasons, the very fact that the actual converter is running at rates well into the MHz range nowadays (with the final sample rate obtained by decimation and filtering) should be telling you something. If filtering is not up to snuff or jitter varies significantly, that's a converter / implementation problem.
Plus there usually is precious little going on at 48+ kHz to begin with, with shaped noise arguably contributing the most, and even that is at very low levels. If that bit of noise should be causing trouble on the playback side, this may be an argument for lower distribution sample rates - but only if reconstruction filters are up to par.
It's nice to see that the old trick of oversampling during recording seems to be acknowledged and useful in practice. Looks like there are a fair few (usually inexpensive consumer-level) ADCs with substantially less-than-ideal anti-alias filtering out there.
I do consider 192 kHz to be genuinely useful - for measurement and other specialty applications though. It's handy for high-quality FM stereo demodulation and generation, for example. The current Accuphase FM tuners employ 192 kHz audio ADCs for this purpose, and I bet the combination of Airomate or Stereo Tool with an ESI Juli@ enjoys decent popularity among broadcasters of one kind or another, as well as those poor souls faced with the daunting task of realigning a good FM tuner to spec (finding a top-quality FM RF generator is hard enough!). Inexpensive hardware stereo generators generally seem to be, err, unexciting.
The only thing I don't get is the part about sampling "too fast". There is no such thing as a "loss of accuracy" when sampling at 192 kHz rather than 96 kHz, not generally at least. While it is certainly imaginable that a practical ADC would be performing worse for all kinds of reasons, the very fact that the actual converter is running at rates well into the MHz range nowadays (with the final sample rate obtained by decimation and filtering) should be telling you something. If filtering is not up to snuff or jitter varies significantly, that's a converter / implementation problem.
Exactly what I was thinking every time I read that statement. This is in fact how it works and the article and Dan Lavry are wrong on that point, correct? How could Lavry get such an idea?
... Exactly what I was thinking every time I read that statement. This is in fact how it works and the article and Dan Lavry are wrong on that point, correct? How could Lavry get such an idea?
He was talking about the performance of specific chip implementations, not D/A in general. I'll look for the discussion, I think it's somewhere in the ProAudio archives.
Good. He quoted Monty. In fact his entire article could have been truncated to a link to xiph.org's video. Once I saw that 1khz tone digitized, and perfectly reproduced in amplitude and frequency using ancient analog/consumer grade equipment, I was sold. That video is easy to understand for non tech people and makes me wish my $0.99 sub to stereophile was reduced to $0.50 a month. I do like looking at shiny things once in a while.