jaw@ames-lm.UUCP (James A. Woods) (02/21/84)
# Somebody's had to much too think. -- Captain Beefheart, "Ashtray Heart" from "Doc at the Radar Station" During a previous incarnation (circa 1975), I drafted an M. S. thesis with the fanciful title "Digital Data Compression of Music with emphasis on Adaptive Transform Coding." It was a curious mixture of experiment and theory. The experimental component involved programming (in C, on version 5 (!) UNIX), FFT's for a non-floating-point PDP 11/40, twiddling phase information, testing quantization and windowing methods, etc., for semi-original psychoacoustics work. The machine (at U. C. Berkeley, now replaced by Ernie CoVax) did real-time sampling to disk and was and was connected to the ARPA net (thank you Tovar, wherever you are). My tests met an untimely demise when fiberglass from the acoustic wall baffling drifted into the cartridge drive, and everyone learned the meaning of "head crash" in the absence of a maintenance contract. The project quickly shifted into a theoretical investigation, addressing topics such as Hackenbush quantizing, Huffman coded run lengths, adaptive thresholding, and Chaitin/Kolmogorov complexity theory. The upshot of all this is that I discovered that there is a LOT of redundancy in PCM encoded signals. It was fun to be able to easily change a few parameters and tell that phase really doesn't count for virtually all signals of musical interest, that distortion under 0.5% doesn't matter, blah, blah. I've been out of the specialty since (too weird for a C. S. dept., let alone the real world), and since I didn't have the foresight to learn Japanese or laser optics, became a UNIX hack. At any rate, I'm wondering what has happened in the compression field recently. Is the CD the ultimate format? I know that Bell Labs is finally picking up on transform coding for speech, but where are the music coders? There have been great strides in compression for video (witness the Widcom codec with a 1440 to 1 ratio: Electronics, Jan. 26, 1984, p. 113). Many of the techniques employed there are directly applicable to audio. Apparently the audio time domain people have a few tricks up their sleeve too, (see Electronics, Feb. 9, 1984 for discussion of a 24 : 1 speech coder). With compression, we may envision systems with no moving parts. If you agree with the above, CD's seem unnecessarily large: at least they track from the inside out so that smaller CD "45's" are possible. Oh, one thing about not being a "golden ears"--it's cheaper. I enjoy being able to use inexpensive TDK D instead of something like Maxell UDXL2 (can you A/B the difference, with one of the standard NR chips?) It's nice that digital sound directs attention to the convenience features of a system, and not the snake oil hype of those tuned to the vagaries of the analog dinosaur. -- James A. Woods {hao | menlo70}!ames-lm!jaw