max@uwm.UUCP (Max Hauser) (04/24/91)
I wish that I also had time to write to popular publications upon becoming
aware of situations where it is evidently needed, but I don't and will
therefore be content (for the moment) to use this newsgroup to respond to
comments, some of them insightful, submitted by John Shriver in article
<11263@uwm.edu> and attributed to Stan Curtis in High-Fi News & Record
Review. Characteristically, those comments accurately address some myths,
but by straying out of the author's evident expertise, or broadcasting the
author's pet notions, they also create or propagate other misinformation.
I am a research engineer and chip designer by training, and have worked in
the last several years to popularize oversampling data conversion among US
engineering communities, as well as to spread information about its
limitations and pitfalls. You must understand that I also agree with much
of what the posted article said -- of COURSE there is a lot of gushing
uncritical acclaim of "digital" technology in audio, some of it even more
exasperating to those who develop the technology than to those who listen
to it! But let us try to focus on the genuine problems.
In the following, quotations are indented.
I'm touched by the simple trust that [some have] in digital technology ...
Perhaps the biggest fallacy is that in digital logic a `1' is a `1'
and is `0' is a `0' ...
Like many critics, this one refers over and again to noise-shaping CD
systems and the like as "digital" and their circuitry as "digital
technology." Which is untrue, of course; many of the very problems that
he alludes to -- problems simplistic digital-audio-advocacy explanations
omit to mention -- are analog in nature. It is precisely because these
products entail intimate interdependent analog and digital circuitry that
they exhibit not only all the separate problems of analog and digital
electrical systems but even some new problems peculiar to the mixture.
[Richard Cabot] was amazed to find that if he `improved' the quality of
the square waves between the digital outputs of a digital filter and the
digital inputs of a converter then the measured linearity improved by
6dB at low levels -- a dramatic improvement. Yet these are the same `1s'
and `0s' that allegedly have no influence.
Whoever supposes even that digital logic's `1s' and `0s' are incorruptible
is making the armchair-expert assumption that these bits are Boolean rather
than electrical -- is neglecting an entire constellation of issues like
logic levels and timing and electrical noise and BER -- and I applaud the
author for attacking such nonsense. (I think that far too many US engineers
might make that assumption today -- a reflection on the state of engineering
education, when "computer engineers" routinely balk at learning Maxwell's
equations and even, sometimes, ordinary differential equations, both of
which pervade electronics and digital-audio electronics especially.)
But there's much more to the story here, and the author misses a golden
opportunity. First, a data converter as cited above is of course an
emphatically analog circuit, not a digital one, and is thus susceptible to
a gamut of undesired analog-digital coupling mechanisms, some of them
dependent on the detailed waveforms of "digital" inputs (a classic example
is ringing edges transiently forward-biasing substrate diodes and hence
injecting noise charge). The author mentions none of this but instead
uses the Cabot data-converter case to support the corruptibility of logic
ones and zeroes (it is not at all clear from the compressed vignette above
that this is what's occurring). Similarly he lampoons "simple trust ... in
digital technology," and the belief that digital data waveforms "have no
influence" on audio performance, but in a context where much of what he's
complaining about is really people applying "digital" intuition to non-
digital electronics. He throws in, for full flavor, the yet separate
problem area of digital-filter arithmetic. By muddying together these very
distinct technical problem areas he ironically furthers some of the
"mythology" that he is claiming to refute.
There seems little doubt that noise-shaping `bitstream' technology is
where the future lies. It was first conceived in the early 1970s but
only the availability of low cost 20-50MHz chip technology has allowed
it to become a commercial reality.
I don't know where on earth the author got those details. The seminal
patent (US 2,927,962 -- write for a copy, it only costs you $1.50; see
below) was filed by Cassius Cutler in 1954; the popular research of Inose
and Yasuda appeared in 1962 and 1963. And it is certainly at least as true
that evolving IC technology mandates oversampling as permits it (as I have
explained in far more concrete detail elsewhere). Small points, but
annoying in writing purported to debunk "myths."
Yet for the armchair theoreticians even a study of the mathematics will
show you that you can't get everything for nothing ...
I think this approaches the heart of the matter. First, what I call
"modern" data-conversion techniques (dithering, oversampling, and noise
shaping -- completely separate principles, even though often combined --
and "modern" only in a formal sense, all of them are decades old) are
not only unfamiliar to many people but also, often, extremely complex and
not always theoretically understood (in some cases even assertively
misunderstood, by "experts" -- but this isn't the forum to get into that).
The subject is inherently complex and doesn't lend itself to glib
explanations.
Second, when a technical development is oversold and misleadingly
marketed, legitimate complaint owes to those who oversell and mislead,
not, of course, to the technology itself.
Finally we come to a test which will show that any two well-engineered
noise-shaping CD players will sound the same in that their synchronised
outputs will be identical to -90dB. I'm sorry, but what a load of twaddle!
I agree in general with this sentiment, even if the author then goes on
to include some red herrings in his rationale -- suggesting the sort of
"armchair" understanding that he decries above -- though undoubtedly
impressive to some smug contrarian audiophiles, who will quote him,
uncritically, as an authority. Just as others uncritically trumpeted the
advent of perfect "digital" sound and led to Curtis's letter. La Ronde!
Here's a final aside. One of my colleagues is US marketing manager for the
European Philips digital audio chips. The poor guy gets phone calls all the
time from audiophiles who want, not long-term contracts for many thousands
of (which is this guy's job), but rather one or two premium data converters
(and free, of course) for their own CD players. A small but pertinent
example of the kind of forces that encourage those people in this business
who really know what they are talking about to keep their mouths shut.
For US patent copies write: US Patent and Trademark Office, Washington, DC
20231. Mark the envelope Attention: Order Patent Copies. (I do this often
enough that I had preprinted labels made up). Enclose a check for the total
fee; as of my last order it was $1.50 per copy, for North American orders.
Max W. Hauser
{mips,philabs,pyramid}!prls!max
Copyright (c) 1991 by Max W. Hauser. All rights reserved.