Focus on 24/192 Misguided?.....


As I've upgraded by digital front end over the last few years, like most people I've been focused on 24/192 and related 'hi rez' digital playback and music to get the most from my system. However, I read this pretty thought provoking article on why this may be a very bad idea:
http://people.xiph.org/~xiphmont/demo/neil-young.html

Maybe it's best to just focus on as good a redbook solution as you can, although there seem to be some merits to SACD, if for nothing else the attention to recording quality.
128x128outlier
Short high frequency bursts like cymbals will suffer the most of distortion.
Kijanki, this one is for you: here is a wonderful thread showing frequency spectrum of various brand of cymbals: http://www.drummerworld.com/forums/showthread.php?t=66957

The top quarter of the thread shows some really very good spectra of various cymbals. You can see that by 40KHz the spectrum has died down to 30-40dB SPL. The major part of a cymbal crash freq content is in the 20-20K range & the content falling off rapidly thereafter. I agree there is content beyong 20KHz but atleast 30dB by the time you hit 30KHz.
So, one could make the case for a 96KHz sampling rate wherein all the freq content upto 48KHz would be included. This sounds reasonable. At 48KHz the analog filter spec becomes reasonable too. Looks like it's a win-win situation....
Lots of technical info (on paper), but what really matters is how it sounds to each person. If you cannot hear that 24/192 WAV or FLAC sounds better than redbook CD, you may want to start selling off that expensive gear and get a boombox to listen to. My LINN Akurate DS playing 24/192 WAV or FLAC will sound better (everything else in the system being the same), than almost any CD player you put next to it.
i disagree with the article and the premiss behind it. nonsense sounds about right. i have nothing to offer in dispute of it except three years of listening to hi res (burned dvd's and streaming). the improved sound quality of hi res has been obvious to me in my set-up (given well recorded music from the start).

if you're not sold on hi-res.....fine. just don't tell me what i hear isn't for real and i'm "wrong".
04-20-12: Kijanki
............
Second nonsense is that digital filter can suppress 96dB within 4kHz without any problem. There is no filter in the world, digital or analog, that can do that (no matter how many poles) with even group delays (or linear phase if you prefer).
.........
not true! Here is a link to paper written by Dan Lavry of Lavry Engineering who wrote this paper in 1997 that shows a 500-tap FIR filter that has a passband of 15KHz & a transition band of 1KHz & stopband starting at 16KHz. The attenution achieved in the 1KHz transition is a whopping 100dB!! See page 3 of 7:
http://www.lavryengineering.com/white_papers/fir.pdf
yeah, it came at a price: 88.2 million operations per second using a dedicated DSP. Very high # of MOPS but do-able.
If one opens the transition band to 4KHz like the paper referenced by the OP then I'm sure that the # of taps will come down.
The paper also goes not to say that the group delay of the FIR filter is flat all the way out to 15KHz.

here is another digital filter paper (from the AES) that shows brickwall digital FIR filters:
http://www.nanophon.com/audio/antialia.pdf.
it's possible to have these brickwall FIR filters with reasonable DSP capacity.
"nonsense! The Nyquist criteria applies to any signal that needs to be quantized. The Nyquist criteria only gives the minimum requirement; it does not say that one is forced to have only 2 samples per highest frequency."

Yes, you can have more samples (for instance 192kHz) but he claims that 44.1kHz (two samples) is all you need.

Again, Nyquist applies to continuous waves ONLY.

from Wikipedia:
"The theorem assumes an idealization of any real-world situation, as it only applies to signals that are sampled for infinite time; any time-limited x(t) cannot be perfectly bandlimited."

Perfect reconstruction of continuous signals close to Nyquist frequency (for instance 15-20kHz) is possible but when signals become very short, reconstruction is much less than perfect.

As for filters - look at typical response of 2and 8 pole 20kHz Bessel filter in dB:

2pole 8pole
20kHz -3 -3
22kHz -3.63 -3.67
40kHz -9.82 -13.68
80kHz -20.32 -51.81

As you can see there is very little attenuation difference at 44.1kHz/2=22kHz with 4x higher number of poles. You would perhaps need hundreds of poles and still not get -96dB. Dramatic difference shows at higher frequencies beyond the "knee" of the filter (160dB vs 40dB per decade). Whole purpose of converting analog to digital at higher rate is to represent bandwidth of 20kHz more accurately and not to extend bandwidth. Downsampling 24/192 master tapes to 16/44 removes some information, (audible or not) but to claim that 24/192 is inferior to 16/44 is complete nonsense.

As for dynamic range, again the point is resolution of the signals above noise floor. According to this article if I listen at 85dB peak and have 35dB ambient noise at home I should not be able to tell the difference between 16 and 8 bit recording (corresponding to about 50dB range). That's nonsense as well.

What about 192kHz being harmful? It doesn't get more silly than that.