DELTA-SIGMA (not sigma-delta) It's official
Charles Thompson
the IEEE has just released a compendium of papers on the subject
edited by well known Temes and Candy. It's a nice book and
costs about $50. It seems that Mr. Candy now agrees the
terminology is Delta first not Sigma first! Max was right :-).
-Charlie Thompson
Jim Roth
In article <93...@cactus.org>, thom...@cactus.org (Charles Thompson) writes...
>For those interested in Delta-Sigma conversion techniques
>the IEEE has just released a compendium of papers on the subject
>edited by well known Temes and Candy
Will wonders never cease? A useful piece of info!
I've been meaning to learn more about this subject but haven't really had
time to ferret out and copy all the papers. My experience with other IEEE
collections is that they are often a great way to quickly penetrate
the literature.
I just want to thank you for passing this on...
- Jim
Subhash Chandra Roy
It was my understanding that Delta-Sigma basically stood for the D/A conversion
and Sigma-Delta for the A/D conversion. That was the spiel given at the
Analog Devices Mixed Signal Seminar.
Subhash
r...@mcnc.org
Max Hauser
Contents: 1. Recent revelations.
2. Subhash Chandra Roy's comment, and certain machinations at
Analog Devices.
3. What exactly does "delta-sigma" denote.
1. Recent revelations about terminology.
In <93...@cactus.org>, Charlie Thompson writes:
|
| ... IEEE has just released a compendium of papers on the subject
| edited by well known Temes and Candy. ... It seems that Mr. Candy now
| agrees the terminology is Delta first not Sigma first! Max was right :-).
Thank you Charlie.
It is good to know that the terminology "delta-sigma" is now "official."
Of course for 29 years we have had the inventors' own publications naming
it that (actively cited, even today, by many people, including me, though
apparently the inventors' own writing is not "official"). The origin and
history and offshoots of this terminology are all evident in the literature
record, obvious to anyone working in the field. In the unlikely event that
this were not enough, we have occasional emphasizings of it, such as a memo
I sent in 1990 to 27 principal specialists in the field (including Charlie,
his copy sent 13 April 1990) reviewing the history of titles on the subject,
encouraging checking of this information, and all but begging anyone to give
a concrete reason for turning around the inventors' coinage to the much
newer "sigma-delta."
(Of the 27 recipients of that April 1990 memo, by the way, 20, including
Charlie, remained mute; five responded corroborating what I wrote and two
responded still preferring the reversed form "sigma-delta." Neither of
these two offered new information; one cited Candy's titles as his source
-- Candy, as above, has repudiated the usage -- while the other offered an
imaginative _ad-hoc_ argument for "sigma-delta," one of many such, which
however are incompatible with each other [another appears in Subhash Chandra
Roy's comment, below], and which presume to second-guess the originators of
the terminology, usually at least originally out of ignorance of them.)
Of course none of this is important; what is important is that "well known"
authorities have publicly endorsed the form "delta-sigma." [I'll leave as
an exercise for the lively imagination the information of who prodded them
to do so.] It must be accurate because well-known authorities endorse it,
not because it is verifiable in its own right. Similarly we may reasonably
conclude that if well-known authorities failed to endorse it, it could not
possibly be true and sensible folks would continue to write "sigma-delta."
(Let us hope that well-known authorities don't declare that pi=3.)
Of course, Candy's acknowledgement of the "delta-sigma" terminology is not
news either; I mentioned it in a JAES paper that appeared publicly 1/91 and
that Charlie has seen (or should have seen). But that's not "official." :-)
If Charlie and others conclude that "Max is right" I do wish it were based
on the detailed hard evidence sent them 17 months ago rather than because I
say I am right or because senior authorities recently concurred with my
position. After all as I wrote 17 months ago, precisely such complacent
deference to "authorities" led to the whole "controversy" in the first place,
and to a quite astonishing resilient willingness to deny credit to the
original authors. However, the episode does provide a fertile case study
for students of human nature or of modern engineering research.
2. In <31...@speedy.mcnc.org> Subhash Chandra Roy writes:
|
| It was my understanding that Delta-Sigma basically stood for the D/A
| conversion and Sigma-Delta for the A/D conversion. That was the spiel
| given at the Analog Devices Mixed Signal Seminar.
This is one of the more colorful of the many after-the-fact suppositions
that I have heard on the origin of this terminology, and I have been
collecting them since before Temes or Analog Devices or Crystal or
Motorola was prominently involved with these products. I can assure
Mr. Roy and anyone else that the same term is normally used both for A/D
and for D/A conversion (it is a generic technique useful in the
*implementation* of both converters) and that the "confusion" or
"controversy" concerns *which* single name to use at both ends.
It is hard for me to imagine an Analog Devices rep advocating opposite
terms for A/D and D/A conversion since Analog Devices has already gone to
extraordinary and distinctive lengths to rationalize the term "sigma-delta"
for both A/D and D/A conversion. The marketing house organ, "Analog
Dialogue," edited by Dan Sheingold, went so far as to quote me (vol. 24
no. 2, 1990) on the terminology issue, even though the evidence I had and
was circulating for comment overwhelmingly pointed to "delta-sigma."
Originally Sheingold had written me, saying that AD had "standardized on"
the form "sigma-delta" already, but implying that he was unfamiliar with and
interested in the origin of the earlier term "delta-sigma." In contrast the
resulting article in "Analog Dialogue" made it appear that AD had chosen
"sigma-delta" with knowledge and careful consideration of the history of
the terms. This in turn led to some colorful and most entertaining
correspondence between Crystal Semiconductor and AD, in which I was kept
on as a spectator through cc's by both sides. Moreover, a professional
acquaintance formerly with AD has told me with glee that AD, in order to
decide which term to use, had simply polled internally and found that
"sigma-delta" was the more popular. However by the time Sheingold wrote
to me, AD was already producing products under the label "sigma-delta"
(while their competitor Crystal was using "delta-sigma"), and I gather
that it would have been something of an embarassment for AD to switch
around especially if this entailed an admission of ignorance. In any
event Sheingold's piece quoting me implies a considerably more scholarly
basis for AD's choice than the process related by the former AD employee.
So the last thing I would expect today is an Analog Devices speaker
asserting one form for A/D and the other for D/A. If that is what you
heard then other aspects of the seminar may possibly have been equally
off-the-wall. (Come to one of my seminars instead -- I don't even have
much marketing interest in the subject. The next one publicly is in the
Analog and Mixed-Signal Design Conference, Santa Clara, CA, 30 Oct 1991.
For information call the registration office in Oregon, 503-624-2118.
Or contact me with a postal address for a popular overview reprint.
An audio tape of a past AES overview seminar of mine is also available
and I can dig up the information if anyone is interested.)
3. What does it mean
"Delta-Sigma" is the generic term for a one-bit, oversampled, noise-shaped
digital encoding of an analog signal, such that its average value tracks
the analog signal. This terminology is due to Inose and Yasuda in 1962 (I
have posted references to them incessantly on the Usenet). The term is
derived from the related but philosophically very different "delta
modulation." The added "sigma" is more or less arbitrary terminology and
certainly does not have any of the (diverse) concrete meanings so eagerly
ascribed to it by engineers who have not read the literature. The term
"delta-sigma" applies equally to A/D and D/A conversion; it represents a
technique used as part of the converter implementation on either end.
The term was reversed more or less inadvertently to "sigma-delta" on an
afternoon in 1963 at the former Bell Telephone Laboratories (as one of the
principals present related it to me) and it propagated through the AT&T
publications in that reversed form, becoming prominent circa 1974.
There you have the whole history of the terminology in a nutshell. The
thing I find ironic and amazing about all of this is that the origin of the
terminology is completely open and obvious; Inose and Yasuda continue to be
actively cited on the subject today. It is only possible to get confused
about the terminology or to consider it "controversial" if you have not
looked at the major published literature on the subject. This is
understandable for casual readers or newcomers but not among people setting
themselves up to be authorities. (Fortunately now it's all "official.")
Given the years and years of astounding history of this terminology I hope
that Charlie will forgive me for a lengthy followup.
Max W. Hauser prls!m...@mips.com {mips,philabs,pyramid}!prls!max
Copyright (c) 1991 by Max W. Hauser. All rights reserved.
Arthur Kreitman
stuff deleted
>
>
> Max W. Hauser prls!m...@mips.com {mips,philabs,pyramid}!prls!max
>
> Copyright (c) 1991 by Max W. Hauser. All rights reserved.
Where do you get the gall to put a copyright notice on netnews. The
copyright laws are intended to insure reasonable compensation for the
creation of a copyrighted work. You are well aware that a netnews posting
is repeatedly copied from system to system. Every poster understands that
there is no compensation involved. I'd appreciate it if you would stop
this clear misuse of a copyright.
Charlie Mingo
> > Copyright (c) 1991 by Max W. Hauser. All rights reserved.
>
> Where do you get the gall to put a copyright notice on netnews. The
> copyright laws are intended to insure reasonable compensation for the
> creation of a copyrighted work.
No, copyright laws permit one to own the expression of one's own ideas.
Nothing in there about compensation, "reasonable" or otherwise.
> You are well aware that a netnews posting
> is repeatedly copied from system to system. Every poster understands that
> there is no compensation involved. I'd appreciate it if you would stop
> this clear misuse of a copyright.
Every post to rec.audio (indeed, everything you write) is subject to
common law copyright, whether you "(c)" it or not. Like Moliere's Bourgeois
Gentilhomme who had been speaking prose all his life without knowing it, you
have been publishing copyright material for some time now.
The only practical significance of copyrighting a rec.audio posting is to
prevent someone from quoting your words in another publication without your
permission. (This actually happened in comp.risks, where selections from the
RISKS-Forum Digest were published in book form.)
How are you affected by his use of "(c)"? Then why are you complaining?
* Origin: A point in my favor. (1:109/716.4218)
Andrew Koenig
> Where do you get the gall to put a copyright notice on netnews.
I got news for you...
it's copyrighted whether there's a notice on it or not.
--
--Andrew Koenig
a...@europa.att.com
jj, like it or not
> The only practical significance of copyrighting a rec.audio posting is to
>prevent someone from quoting your words in another publication without your
>permission. (This actually happened in comp.risks, where selections from the
>RISKS-Forum Digest were published in book form.)
It's also happened (long ago) in rec.audio, where I saw some
explainatory messages I posted to net.audio (note newsgroup
name for the time frame here) attributed to me (by real name)
in the media, taken in full.
Hence, and since, my copyright notice. As the *targate
project also was started about that time, and was presented
in a way I didn't like, I also made sure that I included it
as a clear exclusion in my license.
--
An alarm it *Copyright alice!jj 1991, all rights reserved, except transmission
did sound in the *by USENET and like free facilities granted. Said permission is
still of the night,an*granted only for complete copies that include this notice.
alarm that for years... *Use on pay-for-read services specifically disallowed.
Kent L Shephard
It is clear that it will be copied from system to system. Copyright
notices show that you have taken care to make it know that the material
is original and belongs to you. If someone tries to copy and make a profit
from his work they are in violation of a copyright. The writers of code
for use in the public domain use copyright notices. As long as you
distribute the stuff freely there is no problem, as soon as you try to
mske a profit then there is a problem.
Anyway, don't get your shorts in a wad over nothing.
I think it is juat a .signature.
--
/* -The opinions expressed are my own, not my employers. */
/* For I can only express my own opinions. */
/* */
/* Kent L. Shephard : email - kl...@DUTS.ccc.amdahl.com */