max@trinity.uucp (Max Hauser) (08/07/88)
I think this might be of interest to those of you interested in hard
facts on digital audio and its technology. It's posted separately
here from rec.audio to avoid spillage of the inevitable followups there.
Lipshitz and Vanderkooy, whose names will familiar to those who are
engineers interested in digital audio, presented a paper at the March
AES convention in Paris titled "Are D/A Converters Getting Worse?" I
obtained a preprint of this paper several weeks ago. It is a massive
report, 103 pages, the largest AES preprint by far that I have ever seen.
In it they compare exhaustively the technologies, electrical performance,
and particularly the low-level quantization effects in 17 CD players ("of
various vintages") and suggest that the audio performance in these, on
the whole, is getting worse. They include some experiments on distortion
audibility also.
What's most impressive is the enormous volume of supporting data in the
report.
The authors studied samples of the following players, though their data
are ordered chronologically by model release date:
Sony CDP-35
Sony CDP-101
Sony CDP-200
Sony CDP-302
Sony CDP-310
Sony CDP-350
Magnavox FD1041
Philips CD 650
Technics SL-P7
Yamaha CD-X2
Yamaha CDX-500U
Yamaha CDX-900U
Yamaha CDX-1100U
JVC XL-V400
Denon DCD-600
Denon DCD-800
Nakamichi OMS-2
(I note with an element of professional delight that Lipshitz and
Vanderkooy emphasize a number of technical points, not heard nearly
enough, about digital audio that I have made repeatedly in this august
electronic forum; indeed these authors use enough of my favorite
terminology employed in the last year of audio usenet postings to make
me wonder just a bit ... Even better, they avoid the annoying and
confusing habit that many engineers have of referring to a "Hanning"
window for spectrum estimation (the correct and unconfusing name is
Hann), though I admit that this point will be of interest to very few.)
The conclusions? Beyond the main one that I opened with, far too complex
to summarize here with any justice. But the preprint is available for $4
from The Audio Engineering Society, 60 East 42nd Street, NY, NY 10165.
S. P. Lipshitz and J. Vanderkooy
"Are D/A Converters Getting Worse?"
AES 84th Convention, Paris, 1-4 March 1988,
preprint number 2586.
I would assume that they will have submitted it as a regular AES-journal
paper too, but the publication lag there is many months.
The authors do note that full-spec 16-bit digital audio performance,
especially with properly randomized quantization effects, is extremely
good audio; all of the defects that they identify in their paper arise
from D/A converters that, in various ways, FAIL to meet the full spec.
Max Hauser / max@eros.berkeley.edu / ...{!decvax}!ucbvax!eros!max
"... and received an order from the Archbishop for a set of six duos."
(Dianne Nicolini, KKHI, 31 July 88, recounting a story of Mozart)jshelton@zodiac.ads.com (John L. Shelton) (08/07/88)
I believe there was a summary of the Lipshitz/Vanderkooy paper in the July issue of Stereophile. The summary indicated that a major problem is that audio manufacturers are using D/A converters without making the necessary trim adjustments; as a result, the D/As are not behaving linearly. The addition of $1 in parts, plus a little labor to make the correct trimming would solve the problem. =John=