Signals & Noise (Letters to Editor) (Aug. 1992)

Masur Looks Good in Profile

Dear Editor:

Robert Angus' article about Kurt Masur (December 1991) was a terrific look at the man and his musical priorities. His projected series of over 30 recordings with the New York Philharmonic on Teldec has already begun, so I'm sure your readers found the pro file timely as well as informative.

I thought it was curious that John Eargle's sidebar on Masur's recording of the Tchaikovsky Fourth referred to its Eterna release. That CD, issued by the former East German state-owned record label, received virtually no distribution in the United States, but it is identical to the widely distributed Teldec release (it was a joint venture).

That selection (Teldec 43339) is still very much available and is part of a complete Masur Tchaikovsky Symphony cycle on Teldec.

Keep up the good work! Robert S. Goldfarb Director of U.S. Operations Teldec New York, N.Y.

New Accustoms

Dear Editor:

I read my friend John Eargle's "Currents" in the March issue ("Data to the Masses") with interest. I have been following the progress of compression algorithms closely for several years on behalf of a client and, perhaps, am privy to some information otherwise not widely available.

Although I agree with many of Eargle's comments, I do not agree that systems with reduction factors greater than four-to-one "are generally relegated to speech applications." I have experienced perceptual encoders-I prefer that term to "data reduction" operating at data rates as low as 64 kilobits/S (12-to-1 "compression" compared with 16-bit PCM with a 48-kHz sampling rate, approximately 11 to 1 referenced to CD sampling) that are capable of very fine music reproduction and certainly better than the typical commercial broadcast. At slightly higher data rates-say, 88 kilobits/S-certain algorithms approach CD quality now and should achieve it in the relatively near future. These statements are based on double-blind testing with a wide variety of stimuli, using technical people and musicians.

Unlike linear PCM encoding, on which sampling rate and quantization error impose unalterable theoretical limits on quality, perceptual encoders can be improved after a standard is established. As psychoacoustic knowledge increases, improved encoding algorithms can be created that are fully compatible with the decoders already existing in the field. Eargle's concern about "How much work can be done later to 'clean up' a marginal system after it has been accepted?" may be unfounded in reality.

I wish to point out that perceptual encoding requires a new way of thinking. We must free ourselves of the comfortable assumption that preservation of "waveform" or other easily measured technical parameters is the route to the Holy Grail. Like the perception of the sound of a tree falling in the forest, the perception of music exists only when it is heard-and the ultimate criterion of sound quality has always been the human ear. In the case of perceptual encoding, the ear is, at pre sent, the only criterion. We engineering types are just going to have to get accustomed to that.

-Edward J. Foster; Diversified Science Laboratories; West Redding, Conn.

Author's Reply: I appreciate Ed Foster's added insights into the subject of audio data reduction. "Perceptual en coding" is an appropriate term in that it positions this new technology in a special light. Instead of merely reducing data requirements for audio transmission, the aim is to optimize that transmission through application of what psychological acoustics tells us about the perception of sound. I certainly agree that traditional numerical notions of what constitutes "channel capacity" must be retired in favor of phenomenological ones.

Regarding bit-rate reduction well in excess of four to one, the examples I have thus far heard fall short of reason able expectations for music. But we are in a rapid development phase of this new art and science, and I would expect many of the remarkable strides that Ed states are already here.

-John Eargle

(adapted from Audio magazine, Aug 1992)

= = = =