"SACD and DVD-A proven no better the CD in a year of listening tests"

If you put all those amplifiers in a sealed room, the pressure will rise... pV=nRT.

Put green-marker-loving audiophiles in the same room with Steve or a few others around here, and the pressure will rise too!

Maybe we could record that jam session in SACD format?

You could use one of these for that!

http://www.korg.com/gear/info.asp?A_PROD_NO=MR1&category_id=3

And anyone wanting to know how users have fared with the Korg machines could grind through the posts at http://taperssection.com/index.php/topic,80529.0.html and http://taperssection.com/index.php/topic,93710.0.html

Put green-marker-loving audiophiles in the same room with Steve or a few others around here, and the pressure will rise too!

And don't forget to give them enough big beans to eat !

I’m not sure exactly what I want to ask. It is something along the lines of “what is the meaning of this?” or “what is the significance of this?” This needs a little explanation. It started with a another forum’s thread on this thread’s topic. Someone posted saying he could consistently differentiate the 96kHz version vs the resampled 44.1kHz version of the triangle sample  (triangle-2_2496) from the PCXAB site
http://64.41.69.21/technical/sample_rates/index.htm

He resampled with Audition 3 using a quality setting of 999. Remembering the admonition about overly high quality settings, I ran a few tests. I only have CE2K, so I don’t know how results might be different with Audition 3.

Comparing resampling with quality 999 vs quality 250 by means of Mix Paste Inverted, I found a very large residual signal. A little more fiddling led me to make the same comparison between quality 250 and quality 251. Again there is a large data difference, much more than between a CD track and an mp3 encoded version of same, for instance.

I know that misalignment by even one sample makes for a major difference, but that does not seem to be the problem here. The samples line up, one to one, and the wave form appears to make all the same changes on corresponding samples. The sample values themselves are the difference.

Comparing the 250 and 251 versions (via Synchronize Cursor across Windows), the individual peaks as shown by the waveform drawing appear to be just about the same, even though the samples nearest those peaks have different values. Between the 250 and 999 versions, however, the individual peaks could differ by at least 2dB. Analyze/Statistics on the entire file (all 4.4 seconds of it, most being the fade out) gave almost identical peak and RMS values, in spite of the difference in individual peaks.

We know that making this kind of data comparison with mp3s and other perceptual encoded audio is meaningless. It isn’t the numerical dissimilarity of the data that is significant. That apparently is also true here, but perhaps based on something other than psycho acoustics? I don’t think the program is concerned with that kind of algorithm for resampling, is it?

Hmm... you have to bear in mind that the quality settings on SRCs don't relate to the conversion itself, but to the residual calculations in the pre/post filtering. And I have to say that if you are going to present a sample that has significant content above your resampled Nyquist frequency, then altering the filter characteristics could potentially make more detectable difference to those frequencies - simply because there is rather more content than there otherwise would be.

As for somebody saying that they can consistently tell the difference between two of those samples - well, that says rather more about their reproduction chain than anything else. Since it's well known that sound cards don't behave identically by any means at different sample rates, I'd say that this 'result' doesn't mean anything at all. You cannot possibly achieve a valid result with these tests unless you use a known-adequate replay mechanism. I'm sure that there are several options here, but I'm pretty sure that none of them involve 'domestic' soundcards. Even with some pretty good cards, you can see significant differences in the performance at different sample rates, simply by using RightMark.

That poster finally revealed he is using an Audigy 4 soundcard. That alone makes his claims suspicious, but I haven't been following the ins and outs of Creative soundcards. Do they still all resample everything to 48kHz? Creative seems to have kept high quality in the Emu line. Perhaps some of that quality has migrated to their consumer cards?

He claims that an Audition 3 resampling quality settings of 700 does not sound "as good as" one of 800 which does not sound as good as one of 999, and so on as the setting is lowered to 150. The absolute differences in sound are supposedly supported by ABX tests.

Anyway, this is not about the hardware, only about the program. Independent of the playback chain, the data is different. To change the presentation slightly, I used a lowpass FFT filter (Blackman Window Function, FFT Size 24,000) cutting off at 21kHz to hopefully eliminate any considerations about higher frequencies effecting lower frequencies. Saving that as a new file, I first resampled to 44.1kHz with a quality setting of 250, then with setting 251. There is still the same major data difference. This makes sense to me as I would not expect the program to know if there is anything at any particular frequency. It should do the same thing every time when resampling, regardless.

Comparing the setting 250 results from the two sources (original and low passed), I find no difference below somewhere in the vicinity of 21kHz, except for a faint haze of color (in Spectral View) at all frequencies during the first 50 samples. This seems to mean that (probably) every change of quality setting, no matter how small, makes for numerically large data differences. Perhaps someone with Audition 3 can report as to whether of not it also functions this way.

Since the resampled sample values are different with different quality settings, is there any way to say that one result is superior to another? If people can truly hear a difference, then each can judge which is "best," but is there a mathematical evaluation (or any other data oriented consideration) that makes any sense?

That poster finally revealed he is using an Audigy 4 soundcard. That alone makes his claims suspicious, but I haven't been following the ins and outs of Creative soundcards. Do they still all resample everything to 48kHz? Creative seems to have kept high quality in the Emu line. Perhaps some of that quality has migrated to their consumer cards?

No such luck - they all resample. But some have a 96k pass-through. And this passes all of the processing through an entirely different route, so his claims would stand up to nothing as far as SRCs are concerned - but hearing a difference in results wouldn't be at all surprising. Different reason though...

Well, the soundcard information is interesting, but does anyone have any thoughts about the topic of my post: CoolEdit’s (and possibly Audition’s) largish resampling data differences from even small changes in the quality setting?