5121cdd@houxm.UUCP (C.DORY) (01/24/85)
The following is an open letter to the "Technologists" and "Armchair Audio Engineers" of the net in response to the resurgence of CD vs. analog "discussions" (arguments) of late: On the issue of "linearity". No argument from me that according to the definitions commonly used that the anti-aliasing filters currently employed are linear (superposition and homogeneity hold). This is not the whole story. By looking at the output spectrums (BOTH magnitude AND phase) before and after the filter transform, several interesting things can be noted if we allow ourselves a look. The magnitude spectrum looks like the input spectrum up to the corner frequency of the filter as we would expect. However, by looking only this far we are ignoring delay distortion. The phase spectrum illustrates nonlinearity resulting in group delay (if I remember what a derivative is). On output, then, we have waveform distortion -- the output DOES NOT look like the input in the pass band. You can't argue this point. What I am saying is, then, quite simple: the sampling theory argument used most often to beat the "Golden Ears" into submission is not the whole story. What the main bone of contention seems to be is the human perceptibility of these waveform distortions. To my knowledge, there has been no definitive scientific study in this area (remember, we're talking full-fidelity music, not bandwidth-limited telecommunications). The limited testing that I have done (by no means definitive) indicates that given certain types of source material and extremely good playback equipment this waveform distortion is audible. Let's look at a string section for a minute: the violins are divided into two sections each having independent musical parts. looking at one of the sections, we have several violinists (as many as 15 - 20 in large symphonies) playing in unison. Well, what that means for a section as large as this is that you have several musicians playing slightly out of tune (fractions of a semitone) all with a different relative phase. In concert, our ear combines this into a lush string tone. My mics do a very good job of capturing this sound, however PCM digital mucks it up somewhat. This, and a few other combinations, actually do sound better on an analog deck, the Nagra T-Audio (and careful, if your not familiar with this deck don't be hasty in dumping on it!). As well, the dbx 700 (CPDM Digital) sounds much better on massed strings, brass, massed voices, etc. And, by the way, what I mean by "better" in this context is "more like the input". Since most of you, I believe, are not actively involved in recording music, you don't have the opportunity to compare the input with the output on your CD player. If you have read my articles and subsequent discussions, I think you'd have realized that I'm not anti-digital -- remember I use the stuff! The point I'm trying to make is that it is not as good as it can be and definitely not good enough yet. I am incensed by the egotistical, patronizing, "holier than thou -- I've been in digital signal processing for umpty-ump years" attitude presented by a few of the netland contributors. Full-fidelity audio recording is a relatively new application for digital signal processing -- maybe what is tried and true is no longer the most optimum in this application. A secondary point is that those of us who fail to look at audio engineering following good scientific methodology make progress truly difficult. The statement "good enough for audio" is used entirely too much as well as being based on archaic concepts of what "audio" perception really is. "Digital" audio is providing us with good reasons to examine this. What we have traditionally measured to determine the "performance" of an audio system is not necessary all inclusive. As well, the philosophy of actually interpreting the results needs to undergo some evolution. Craig Dory AT&T Bell Laboratories
ed@mtxinu.UUCP (Ed Gould) (02/02/85)
> I am incensed by the egotistical, patronizing, > "holier than thou -- I've been in digital signal processing for > umpty-ump years" > > ... > > A secondary point is that those of us who fail to look at > audio engineering following good scientific methodology make > progress truly difficult. The statement "good enough for > audio" is used entirely too much as well as being based on > archaic concepts of what "audio" perception really is. > "Digital" audio is providing us with good reasons to examine this. > What we have traditionally measured to determine the "performance" > of an audio system is not necessary all inclusive. As well, > the philosophy of actually interpreting the results needs > to undergo some evolution. > > > Craig Dory > AT&T Bell Laboratories Hear, hear! I'm glad someone was able to say this better than I did! -- Ed Gould mt Xinu, 739 Allston Way, Berkeley, CA 94710 USA {ucbvax,decvax}!mtxinu!ed +1 415 644 0146