20 or 24 bit digital recording?

t880...@phillip.edu.au

unread,

Jul 20, 1992, 4:53:56 PM7/20/92

to

I heard not long ago that some recording engineers were using digital
recorders that record in greater than 16 bit, ie 20 bit etc...

I have often wondered why it is that this has not been done before,
since a lot of the anti CD brigade are of the opinion that CD is
not able to reproduce the transparency of vinyl. Would switching
to a greater word length (ie 20 or 24 bit sampling) solve the
problem?

This might sould like a dumb question, but since the socalled 1 bit
players get the data off a disk as a constant stream of data, would
they be able to benifit immediately from this change in recording?,
or would it be necessary to change the player to full 20 or 24 bit

thanks.

# T880...@phillip.edu.au # in the sweetest bud. #
#########################################################################

W.G.Day

unread,

Jul 22, 1992, 5:32:38 AM7/22/92

to

>Would increasing the resolution of CD increase its transparency?

As it is unlikely that complete sonic transparency is just related to
resolution, I suspect the answer to your questio is no. Increasing
resolution will give more detail though.

Current vinyl is apparently about twice the resolution of CD (assuming
you've got a clean record, but we've talked about that). If you did
increase the CD's resolution you need to invent a new completely format
basically. Even though you are only changing the 16 bits to 24 bits you
need to increase the size of your CD or have only a 55 minute limit. Also
you need to completely change your error correction software. Increasing
your resolution just gives you more detail. However every so often people
come up with new plastics which can form more complex shapes and these can
be used to press records without changing your turntable.

>This might sound like a dumb question, but since the so-called 1 bit

>players get the data off a disk as a constant stream of data, would
>they be able to benifit immediately from this change in recording?,
>or would it be necessary to change the player to full 20 or 24 bit

The 1 bit players get the data off in the 32 byte packets just like
multi-bit players do. Data off a CD is data off a CD. The 1-bit
players however use the 1 bit (delta sigma) system to reconstruct
the sound without having large errors in the resistors that multi-bits
can have.

16384 ohms +/- 5% is about 800 ohms error. In a multi-bit player there
are also resistors of 1 ohm, 2 ohm, ... , 256 ohm , 512 ohm , 1024 ohm.
So the errors really corrupt the sound produced.

>thanks.

># T880...@phillip.edu.au # in the sweetest bud. #

Warren

Richard d Pierce

unread,

Jul 22, 1992, 8:02:04 AM7/22/92

to

In article <13...@eagle.ukc.ac.uk> w...@ukc.ac.uk (Warren Day) writes:
>>Would increasing the resolution of CD increase its transparency?
>
>As it is unlikely that complete sonic transparency is just related to
>resolution, I suspect the answer to your questio is no. Increasing
>resolution will give more detail though.

As it has never been demonstrated the detail is just related to resolution,
I strongly suspect that this is not the case either.

>Current vinyl is apparently about twice the resolution of CD (assuming
>you've got a clean record, but we've talked about that). If you did
>increase the CD's resolution you need to invent a new completely format
>basically. Even though you are only changing the 16 bits to 24 bits you
>need to increase the size of your CD or have only a 55 minute limit. Also
>you need to completely change your error correction software. Increasing
>your resolution just gives you more detail. However every so often people
>come up with new plastics which can form more complex shapes and these can
>be used to press records without changing your turntable.

The discussion about "LP vs CD" resolution that has gone flying back and forth
has to be one of the more amusing and ridiculous discussions I have recently
been saddened to see on this newsgroup (I guess that's saying a lot!). I
have no idea where the basis for your assertion about comparitive resolution
comes from, but I would certainly be interested in seeing both some measurments
and references to the relevent literature. But enough of that for the moment.

>The 1-bit
>players however use the 1 bit (delta sigma) system to reconstruct
>the sound without having large errors in the resistors that multi-bits
>can have.
>
>16384 ohms +/- 5% is about 800 ohms error. In a multi-bit player there
>are also resistors of 1 ohm, 2 ohm, ... , 256 ohm , 512 ohm , 1024 ohm.
>So the errors really corrupt the sound produced.

Well, you've made an assertion here which is easily and unequivocably testable.
That is that CD using parallel DACS ought to have whopping transfer errors.
This should lead to easily measurable converter errors, even non-monotonic
performance. The sorts of errors you describe here should be obvious and gross
yet not a single CD, DAT or, for that matter DAC of any kind produced in the
last 30 years exhibits the kind of errors you contend must exist.

Most D/A converters do not use the binary weighted resistors you discuss
here, they use what's refered to as a "R/2R" scheme. This prevents just
exactly the sort of problem you describe here (some do use it, but they
can be laser-trimmed to an accuracy far better than the 5% you assert).
So instead of a ladder that looks like:

+V ----+--32768--
|
+--16384--
|
+--8192---
. . .
+---4-----
|
+---2-----
|
+---1-----

We end up with:

+V----+--20K--
10K
+--20K--
10K
+--20K--
. . .
+--20K--
10K
+--20K--
10K
+--20K--

(I may have the pair swapped, If so, fix it yourself, I ain't gonna do it
in emacs at 2400 baud! :-))

The type of gross non-linearity is a problem that was solved in DACS long
before the CD player ever came along.

There are other real advantages to single bit DACS (for example, no glitch
at the zero crossing, etc.), but the issue you metioned is certainly not
one of them.
--
| Dick Pierce |
| Loudspeaker and Software Consulting |
| 17 Sartelle Street Pepperell, MA 01463 |
| (508) 433-9183 (Voice and FAX) |

Stuart Woolford

unread,

Jul 22, 1992, 8:04:20 PM7/22/92

to

w...@ukc.ac.uk (W.G.Day) writes:

>>Would increasing the resolution of CD increase its transparency?

>As it is unlikely that complete sonic transparency is just related to
>resolution, I suspect the answer to your questio is no. Increasing
>resolution will give more detail though.

>Current vinyl is apparently about twice the resolution of CD (assuming
>you've got a clean record, but we've talked about that). If you did
>increase the CD's resolution you need to invent a new completely format
>basically. Even though you are only changing the 16 bits to 24 bits you
>need to increase the size of your CD or have only a 55 minute limit. Also
>you need to completely change your error correction software. Increasing
>your resolution just gives you more detail. However every so often people
>come up with new plastics which can form more complex shapes and these can
>be used to press records without changing your turntable.

twice the resolution of CD??? no..not even close....

>>This might sound like a dumb question, but since the so-called 1 bit
>>players get the data off a disk as a constant stream of data, would
>>they be able to benifit immediately from this change in recording?,
>>or would it be necessary to change the player to full 20 or 24 bit

>The 1 bit players get the data off in the 32 byte packets just like
>multi-bit players do. Data off a CD is data off a CD. The 1-bit

yes.

>players however use the 1 bit (delta sigma) system to reconstruct
>the sound without having large errors in the resistors that multi-bits
>can have.

no.

>16384 ohms +/- 5% is about 800 ohms error. In a multi-bit player there
>are also resistors of 1 ohm, 2 ohm, ... , 256 ohm , 512 ohm , 1024 ohm.
>So the errors really corrupt the sound produced.

so, you actually think that they use 5% resistors in the DAC ???
no...more like 0.001% ( which I can even purchase in 5Watt discrete components
let alone on-chip film resistors! )
and you actually know the resistor values in a DAC???, interesting...

>>thanks.

>># T880...@phillip.edu.au # in the sweetest bud. #

>Warren

-----------------------------------------------------------------------------
stu...@ccu1.aukuni.ac.nz

>>>>In VI Where Available<<<<
-----------------------------------------------------------------------------

John Adams

unread,

Jul 22, 1992, 11:05:33 AM7/22/92

to

In rec.audio, t880...@phillip.edu.au writes:

>I heard not long ago that some recording engineers were using digital
>recorders that record in greater than 16 bit, ie 20 bit etc...

>I have often wondered why it is that this has not been done before,
>since a lot of the anti CD brigade are of the opinion that CD is
>not able to reproduce the transparency of vinyl. Would switching
>to a greater word length (ie 20 or 24 bit sampling) solve the
>problem?

>This might sould like a dumb question, but since the socalled 1 bit
>players get the data off a disk as a constant stream of data, would
>they be able to benifit immediately from this change in recording?,
>or would it be necessary to change the player to full 20 or 24 bit

20 bit recording is only begining to appear in recording studios
producing the master copies. The 20bit master MUST transfered to a 16
bit format as in the case of a CD. The current technology (DCC omitted)
CANNOT handle the 20 bit format. However the accuracy of the 16 bit
information on the CD may be truer to form when originating from 20 bit
master than from a 16 bit.

CD and DAT will continue to be 16 bit regarless of the resolution of the
master.

John Adams

jj, curmudgeons for trvth in audio

unread,

Jul 23, 1992, 12:08:25 PM7/23/92

to

In article <13...@eagle.ukc.ac.uk> w...@ukc.ac.uk (Warren Day) writes:

>Current vinyl is apparently about twice the resolution of CD (assuming
>you've got a clean record, but we've talked about that).

No, Warren, we haven't talked about that. The best vinyl in the world
has a noise level well aove the quantizing noise level of a CD,
and in no way can in considered to have 'twice the resolution
of vinyl'.

Since you state otherwise in such an authoritive fashion, I now request your
references, controls, experimental data (you have to have measured noise
levels somewhere to do make the claim you make) and so on.

I find (having done the appropriate measurements many times over
many different systems) the statement that "lp has twice the
resolution of CD" utterly, hysterically laughable, in fact,
absurd and bordering on misrepresentation.

>16384 ohms +/- 5% is about 800 ohms error. In a multi-bit player there
>are also resistors of 1 ohm, 2 ohm, ... , 256 ohm , 512 ohm , 1024 ohm.
>So the errors really corrupt the sound produced.

Warren, Warren, Warren, what do you think they use for resistors
in the DA convertor ladder in a CD player? The whole point of
laser trimming and such is that (some) manufacturers reach about 15 bits
of accuracy with their 18 bit convertors (I'm referring to parallel,
not delta-sigma convertors), and they're getting better all the time.

If they used YOUR example, they couldn't get past 4.3 bits.
--
Bush *Copyright alice!jj 1992, all rights reserved, except transmission
vs *by USENET and like free facilities granted. Said permission is
Clinton -- *granted only for complete copies that include this notice.
Just Say !NO! *Use on pay-for-read services specifically disallowed.

Kim Letkeman

unread,

Jul 23, 1992, 12:59:39 PM7/23/92

to

In article <13...@eagle.ukc.ac.uk> w...@ukc.ac.uk (W.G.Day) writes:

| Current vinyl is apparently about twice the resolution of CD (assuming
| you've got a clean record, but we've talked about that).

I believe that you mean "current vinyl opinion is apparently about
twice the circumlocution of CD".
--
Kim Letkeman k...@Software.Mitel.COM

Stuart Woolford

unread,

Jul 23, 1992, 5:20:18 PM7/23/92

to

j_a...@hpfcbig.SDE.HP.COM (John Adams) writes:

>In rec.audio, t880...@phillip.edu.au writes:

until they start using non-lossy compression for DAT B-)
( ie:on audio data, you can EASILY get 1:8 compression NON-LOSSY ..
anyone for an 8hour CD format??? )
-------------------------------------------------------------------------------
stu...@ccu1.aukuni.ac.nz

unread,

Jul 24, 1992, 8:51:25 AM7/24/92

to

In article <23...@alice.att.com> j...@alice.UUCP writes:
>I now request your references

Vinyl resolution reference: apparently someone has performed measurements
of the complex shape forming properties of polyvinyl and it comes out at
around 16.5, I was also told that CD is 15.5 as you always have at worst
a .5 bit error, which makes sense. Shane...?

>The whole point of
>laser trimming and such is that (some) manufacturers reach about 15 bits
>of accuracy with their 18 bit convertors (I'm referring to parallel,
>not delta-sigma convertors), and they're getting better all the time.

>If they used YOUR example, they couldn't get past 4.3 bits.

Yes it was an example, I used 5% as I know the resistors you can pick up in
your hand can do this and I don't know what other technologies can do, although
I expected it to be better as it is. 5% also made my maths easier.

As was also said combinations of 10K and 20K (or something similar) are used.

> Bush
> vs
> Clinton
> Just Say !NO!
And bring back Perot

Warren

Curt Welch

unread,

Jul 24, 1992, 10:50:39 AM7/24/92

to

In rec.audio, MAT...@LURE.LATROBE.EDU.AU (BYRNES,Graham) writes:
>I'm still surprised that if it is technically feasible to do as you say,
>Philips et al are playing about with PASC etc, which IS lossy,
>although perhaps not audibly. In fact, Pasc relies on the fact that only
>a small fraction of the bandwidth is audible at a given time to ditch the
>rest.

My guess is that you can probably do something like 8:1 loss free
compression on CD's but designing and building the hardware to do this
in real time is currently too expensive.

Also, using a lossy algorithm makes the format far more appealling to
the record companies - so it increases it's chance for success in the
market place. The manufactures don't want the next format to go the
same way that DAT has gone. They are trying to find a format that
everyone, including the record companies, will like.

But here's another angle I'm waiting for the marketing types to come up
with. As I understand it, both the DCC and MD compression algorithms
eliminate some sounds based on the human ears ability to here them
because of masking effects. That is, if you have two tones, close in
frequency, but one much louder than the other, it will mask out the
second tone. Humans can't hear it even if it's there, so why record
it.

So my thought - if the compression algorthim is removing sounds that you
can't hear - instead of saying the algorithm is throwing away part of
the orignal recording (i.e. distoring it) - they should say that it is
cleaning the signal of unwanted noise. Removing low level noise that
causes the non-linear speaker systems to distort the true high level
signals that you can hear. The cleaner the sound, the better the
speakers will be able to reproduce it. So, I'm just wondering when
people will start claiming that compressed audio sounds better than
non-compressed becauses it has been purified?

Curt Welch

Peter Monta

unread,

Jul 24, 1992, 1:02:26 PM7/24/92

to

In article <13...@eagle.ukc.ac.uk> w...@ukc.ac.uk (W.G.Day) writes:

> In article <23...@alice.att.com> j...@alice.UUCP writes:
> > I now request your references

> Vinyl resolution reference: apparently someone has performed measurements
> of the complex shape forming properties of polyvinyl and it comes out at
> around 16.5, I was also told that CD is 15.5 as you always have at worst
> a .5 bit error, which makes sense. Shane...?

(This is a reference?) First, the 0.5 bit error is likely a distorted
echo of the 0.5 LSB max error of a uniform quantizer; it does not
apply to SNR specs. CD is 16-bit linear PCM, end of story.

Second, let's take a look at the 16.5 bit number. (Disclaimer: I
don't know the LP system very well, so take this with a grain of
salt.) 16.5 bits means about 100,000 levels. If we assume a
modulated groove in the vinyl with, say, 1mm maximum extent, then
we need to measure this to the tune of 0.01 um. This would be
rather difficult even at DC, with interferometric measurement, and
over large areas (square mm). To do this over small areas, at
multi-kHz rates, and with a piece of diamond scrabbling over the
surface, is hopeless; you would destroy the surface even if you could
press it to that accuracy in the first place.

Now if we rate the LP at 70 dB SNR at best (is this generous?), i.e.
11 bits, then the physical resolution is 0.5 um, which is at the
edge of reasonableness. There remains piezo noise and distortion,
thermal noise in the preamp, etc. which I don't really know any
numbers for.

Peter Monta mo...@image.mit.edu
MIT Advanced Television Research Program

Peter Monta

unread,

Jul 24, 1992, 1:19:20 PM7/24/92

to

In article <22...@oasys.dt.navy.mil> cu...@oasys.dt.navy.mil (Curt Welch) writes:

> My guess is that you can probably do something like 8:1 loss free
> compression on CD's but designing and building the hardware to do this
> in real time is currently too expensive.

If you really mean "loss free", i.e., the original 16-bit data is
reproduced, then this is impossible. Any audio recording on CD has
an entropy larger than 2 bits/sample.

> ... As I understand it, both the DCC and MD compression algorithms

> eliminate some sounds based on the human ears ability to here them
> because of masking effects.

> ...

> So, I'm just wondering when people will start claiming that
> compressed audio sounds better than non-compressed becauses
> it has been purified?

Enhancement isn't the goal of these algorithms; they attempt to
minimize the loudness of quantization noise, so to the extent that
the compressed audio is indistinguishable from the non-compressed,
these algorithms are good. Enhancement/restoration is an interesting
topic, but it's separate from source coding.

Stuart Woolford

unread,

Jul 26, 1992, 8:39:39 PM7/26/92

to

mo...@image.mit.edu (Peter Monta) writes:

>In article <22...@oasys.dt.navy.mil> cu...@oasys.dt.navy.mil (Curt Welch) writes:

>> My guess is that you can probably do something like 8:1 loss free
>> compression on CD's but designing and building the hardware to do this
>> in real time is currently too expensive.

>If you really mean "loss free", i.e., the original 16-bit data is
>reproduced, then this is impossible. Any audio recording on CD has
>an entropy larger than 2 bits/sample.

why?? ( no, really, I would like to hear you're reason... )

jj, curmudgeon

unread,

Jul 26, 1992, 10:04:55 PM7/26/92

to

In article <13...@eagle.ukc.ac.uk> w...@ukc.ac.uk (Warren Day) writes:

>In article <23...@alice.att.com> j...@alice.UUCP writes:
>>I now request your references
>
>Vinyl resolution reference: apparently someone has performed measurements
>of the complex shape forming properties of polyvinyl and it comes out at
>around 16.5, I was also told that CD is 15.5 as you always have at worst
>a .5 bit error, which makes sense. Shane...?

Again, I ask, REFERENCES, please. I read most of the major scientific
and industry journals, and I haven't seen what you report for vinyl
anywhere.

As to CD, you're wrong. The .5 bit error is, as far as I can determine
from your example, really the .5 delta quantization error that
RESULTS from 16 bits. There need be no 15.5 bits of real
information on a CD.

As far as real vinyl, I haven't seen any vinyl yet that has
demonstrated anything over about 55dB SNR, even at low frequencies,
and that limits vinyl (note, SNR, not dynamic range, is what
is important here for measuring "bits"), even ignoring
frequency response problems, separation problems,
degratation problems, and the like, to less than 10 bits.

jj, curmudgeon for trvth in audio

unread,

Jul 27, 1992, 10:20:35 AM7/27/92

to

In article <1992Jul27....@ccu1.aukuni.ac.nz> stu...@ccu1.aukuni.ac.nz (Stuart Woolford) writes:
>mo...@image.mit.edu (Peter Monta) writes:
>
>>In article <22...@oasys.dt.navy.mil> cu...@oasys.dt.navy.mil (Curt Welch) writes:
>>> My guess is that you can probably do something like 8:1 loss free
>>> compression on CD's but designing and building the hardware to do this
>>> in real time is currently too expensive.

If by "loss-free" you mean "perceptually loss-free", some algorithms
have been controversially verified by the ISO-MPEG-Audio committee
to be "transparent" at 2.66 bits/sample, when compression 16 bit samples.
Nobody I'm aware of (and I do study the field a wee bit) has reported
transparency <including myself, to the present> below that
rate, although some algorithms (PAC, ISO-Layer III, for instance,
there may be others) may provide such performance.

>>If you really mean "loss free", i.e., the original 16-bit data is
>>reproduced, then this is impossible. Any audio recording on CD has
>>an entropy larger than 2 bits/sample.

>why?? ( no, really, I would like to hear you're reason... )

Well, because many people have measured it. If you'd like some
confirmation, look in Chapter 4 of the book "Advances in Speech
Signal Processing" by Furui and Sondhi. Chapter 4, written
by James Johnston of AT&T Bell Labs, and Karlheinz Brandenburg
of the University of Erlangen, FRG, is titled "Wideband Coding-
Perceptual Considerations for Speech and Music".

On Page 110 and 111, they list some LPC gain statistics and Spectral
Flatness Measure (SFM) statistics that show a largest SFM
gain of 45.6 dB at a 512 sample Hann Window. This, if you
consider its entropy considerations, means that about half of the
bits in the signal are actually information-carrying, i.e. you can
encode this signal in about 8 bits, give or take, and wind up
with an exact reconstruction.

Other work suggests that for most audio, 5-7 bits/sample is the known
limit for exact reconstruction, using the most modern techniques.

If one measures the single-sample entropy in a common audio signal,
perhaps the first 20 seconds of "Tom's Diner" from Solitude
Standing (Susan Vega), one might find that the single-sample entropy
is about 8.3 bits/sample. This, of course, doesn't remove
intersample redundancy.

Good luck using LZW.

>
>>> ... As I understand it, both the DCC and MD compression algorithms
>>> eliminate some sounds based on the human ears ability to here them
>>> because of masking effects.
>>> ...
>>> So, I'm just wondering when people will start claiming that
>>> compressed audio sounds better than non-compressed becauses
>>> it has been purified?
>
>>Enhancement isn't the goal of these algorithms; they attempt to
>>minimize the loudness of quantization noise, so to the extent that
>>the compressed audio is indistinguishable from the non-compressed,
>>these algorithms are good. Enhancement/restoration is an interesting
>>topic, but it's separate from source coding.
>
>>Peter Monta mo...@image.mit.edu
>>MIT Advanced Television Research Program

Stuart Woolford

unread,

Jul 27, 1992, 5:53:14 PM7/27/92

to

j...@alice.att.com (jj, curmudgeon for trvth in audio) writes:

>In article <1992Jul27....@ccu1.aukuni.ac.nz> stu...@ccu1.aukuni.ac.nz (Stuart Woolford) writes:
>>mo...@image.mit.edu (Peter Monta) writes:
>>
>>>In article <22...@oasys.dt.navy.mil> cu...@oasys.dt.navy.mil (Curt Welch) writes:
>>>> My guess is that you can probably do something like 8:1 loss free
>>>> compression on CD's but designing and building the hardware to do this
>>>> in real time is currently too expensive.
>If by "loss-free" you mean "perceptually loss-free", some algorithms
>have been controversially verified by the ISO-MPEG-Audio committee
>to be "transparent" at 2.66 bits/sample, when compression 16 bit samples.
>Nobody I'm aware of (and I do study the field a wee bit) has reported
>transparency <including myself, to the present> below that
>rate, although some algorithms (PAC, ISO-Layer III, for instance,
>there may be others) may provide such performance.

I did acutally mean perceptually-loss free..and should have made that clear..
I have not experienced the ISO-MPEG, so will not comment on it's veracity,
but the theory looks interesting ( but not the way I would go.. ).

>>>If you really mean "loss free", i.e., the original 16-bit data is
>>>reproduced, then this is impossible. Any audio recording on CD has
>>>an entropy larger than 2 bits/sample.

>>why?? ( no, really, I would like to hear you're reason... )
>Well, because many people have measured it. If you'd like some
>confirmation, look in Chapter 4 of the book "Advances in Speech
>Signal Processing" by Furui and Sondhi. Chapter 4, written
>by James Johnston of AT&T Bell Labs, and Karlheinz Brandenburg
>of the University of Erlangen, FRG, is titled "Wideband Coding-
>Perceptual Considerations for Speech and Music".

>On Page 110 and 111, they list some LPC gain statistics and Spectral
>Flatness Measure (SFM) statistics that show a largest SFM
>gain of 45.6 dB at a 512 sample Hann Window. This, if you
>consider its entropy considerations, means that about half of the
>bits in the signal are actually information-carrying, i.e. you can
>encode this signal in about 8 bits, give or take, and wind up
>with an exact reconstruction.

It seems to me that you are suggesting that at a MAXIMUM (well, you know
what I mean.. ) entropy, remember, I do not look for the constant bit
rates that MPEG ( and MPEG-Audio ?? ) require, as buffers are cheap, and
constant bit rate is not required for locally stored data ( ie:on a CD )
and constant bit rate can KILL compression..

>Other work suggests that for most audio, 5-7 bits/sample is the known
>limit for exact reconstruction, using the most modern techniques.

again, I look for exact AUDIBLE reconstruction.. ie: undetectable by
ANY human ( ie: add something like 10% to top listeners and use that as
limit.. )

>If one measures the single-sample entropy in a common audio signal,
>perhaps the first 20 seconds of "Tom's Diner" from Solitude
>Standing (Susan Vega), one might find that the single-sample entropy
>is about 8.3 bits/sample. This, of course, doesn't remove
>intersample redundancy.

then measure the intro to Pink Floyds `sheep` and find that it is about
0.01 bits/sample....

>Good luck using LZW.

personally I prefer Wavelet/VQ followed by a 1MB Markov Chain driving Arith..

>>
>>>> ... As I understand it, both the DCC and MD compression algorithms
>>>> eliminate some sounds based on the human ears ability to here them
>>>> because of masking effects.
>>>> ...
>>>> So, I'm just wondering when people will start claiming that
>>>> compressed audio sounds better than non-compressed becauses
>>>> it has been purified?
>>
>>>Enhancement isn't the goal of these algorithms; they attempt to
>>>minimize the loudness of quantization noise, so to the extent that
>>>the compressed audio is indistinguishable from the non-compressed,
>>>these algorithms are good. Enhancement/restoration is an interesting
>>>topic, but it's separate from source coding.
>>
>>>Peter Monta mo...@image.mit.edu
>>>MIT Advanced Television Research Program

-------------------------------------------------------------------------------
stu...@ccu1.aukuni.ac.nz

>>>>In VI Where Available<<<<
-------------------------------------------------------------------------------

Chris Christensen

unread,

Jul 28, 1992, 12:39:43 PM7/28/92

to

In article <1992Jul27.2...@ccu1.aukuni.ac.nz> stu...@ccu1.aukuni.ac.nz (Stuart Woolford) writes:
>j...@alice.att.com (jj, curmudgeon for trvth in audio) writes:
>>In article <1992Jul27....@ccu1.aukuni.ac.nz> stu...@ccu1.aukuni.ac.nz (Stuart Woolford) writes:
>>>mo...@image.mit.edu (Peter Monta) writes:
>>>>In article <22...@oasys.dt.navy.mil> cu...@oasys.dt.navy.mil
(Curt Welch) writes:
>>>>> My guess is that you can probably do something like 8:1 loss free
>>>>> compression on CD's but designing and building the hardware to do this
>>>>> in real time is currently too expensive.

>>>>> So, I'm just wondering when people will start claiming that
>>>>> compressed audio sounds better than non-compressed becauses
>>>>> it has been purified?

This thread runs deep!

Personally I *like* 3-6dB of compression at a 2:1 ratio on almost
anything and it only cost me $460 for my dbx 166..........:-)

I now return you to your regularly sceduled thred.

(Oh and BTW, and just in case you wanted to know, jj *does* speak from
a solid technical perspective, seasoned with a dose of reality I
expect :-)
--
Chris Christensen The opinions I express are my own,
chr...@gold.gvg.tek.com and sometimes they are wrong!
916-478-3419 FAX 916-478-3887 After all, I *AM* only human.

Seth Breidbart

unread,

Jul 28, 1992, 7:02:17 PM7/28/92

to

In article <23...@alice.att.com> j...@alice.UUCP (jj, curmudgeon) writes:

>As far as real vinyl, I haven't seen any vinyl yet that has
>demonstrated anything over about 55dB SNR, even at low frequencies,
>and that limits vinyl (note, SNR, not dynamic range, is what
>is important here for measuring "bits"), even ignoring
>frequency response problems, separation problems,
>degratation problems, and the like, to less than 10 bits.

That brings to mind an interesting question. DAT uses 16 bits, so it
has 96 dB snr for a full level signal. For a signal at -50 db, it has
46 dB snr. For a signal at -90 dB, it has 6 dB snr.

Therefore, it is correct to say that DAT has, simultaneously, X dB
dynamic range and Y dB snr, whenever X + Y = 96.

Now consider a floating-point system with 16 bits, 4 bits of exponent
and 13 bits (using the standard IEEE cheat) of mantissa. That system
has 90 dB of dynamic range (using the exponent only), and 78 dB snr.
Similarly to 16-bit linear, we can increase the dynamic range at the
expense of snr (but not vice versa).

Is that analysis correct?

Seth se...@fid.morgan.com

Lon Stowell

unread,

Jul 28, 1992, 10:25:06 PM7/28/92

to

In article <13...@eagle.ukc.ac.uk> w...@ukc.ac.uk (Warren Day) writes:
>
>Vinyl resolution reference: apparently someone has performed measurements
>of the complex shape forming properties of polyvinyl and it comes out at
>around 16.5, I was also told that CD is 15.5 as you always have at worst
>a .5 bit error, which makes sense. Shane...?
>

<sigh> I doubt if you'll get away with "someone" as being a
reference. As for your flowery phrase "complex
shape forming properties of polyvinyl..." care to explain what
the flock it means in ordinary speech...and what bizarre
relationship you think it has to the "resolution" of a vinyl
record?

>
>Yes it was an example, I used 5% as I know the resistors you can pick up in
>your hand can do this and I don't know what other technologies can do, although
>I expected it to be better as it is. 5% also made my maths easier.
>
>As was also said combinations of 10K and 20K (or something similar) are used.

And a doggoned poor one. If you've ever been within a kilometer
of an electronics shop, you would be aware that 5% resistors are
rarely used in anything other than non-critical cirtuits....and
that 1% resistors are readily available, and that 0.1% resistors
are available from several suppliers.... and obviously you
missed JJ's mention of precision laser trimming entirely...or are
blissfully unaware of the high degree of RELATIVE value accuracy
available with thin film resistors....

5% made your math easier, but totally removed your posting or
supposed "facts" from any connection to reality....but of course
it MIGHT just look good to someone who has never opened an EDN in
their life...or walked into a Radio Shack.

jj, audio curmudgeon

unread,

Jul 29, 1992, 10:22:47 AM7/29/92

to

In article <1992Jul28....@fid.morgan.com> se...@fid.morgan.com (Seth Breidbart) writes:

>That brings to mind an interesting question. DAT uses 16 bits, so it
>has 96 dB snr for a full level signal. For a signal at -50 db, it has
>46 dB snr. For a signal at -90 dB, it has 6 dB snr.

More or less, there is some complication in the assumptions one makes on
quantizer occupancy and peak level vs. rms level.

>Therefore, it is correct to say that DAT has, simultaneously, X dB
>dynamic range and Y dB snr, whenever X + Y = 96.

No. Dynamic range doesn't add to SNR. Part of the problem
is that many people have used different (so different in some
cases as to be unrecognizable) meanings for dynamic range, and
you are probably trying to sort the world out. Sorry, but
"dynamic range" isn't particularly well defined, even where
it IS well defined, so to speak. The Peak/Noise Floor dynamic
range (that's what I call dynamic range) of both DAT and CD is
96 dB, considering the adjustments above.

The SNR is indeed level sensitive, just like you'd expect with
a medium with ONLY a noise-floor, and no signal-correlated distortions.
This confuses some people who think that means that vinyl is better,
but some people desparately want to be confused, too.

>Now consider a floating-point system with 16 bits, 4 bits of exponent
>and 13 bits (using the standard IEEE cheat) of mantissa. That system
>has 90 dB of dynamic range (using the exponent only), and 78 dB snr.
>Similarly to 16-bit linear, we can increase the dynamic range at the
>expense of snr (but not vice versa).

Well, the numbers are approximately ok.

>Is that analysis correct?
Except that the 78 dB SNR across the band will not always be enough to
prevent unmasked "pumping" of the quantizing noise.

You need more mantissa bits, at least 14, arguably 16. 16 also
covers the necessary reproduction dynamic range (although not the
recording dynamic range).

Bill Vermillion

unread,

Jul 31, 1992, 10:52:17 AM7/31/92

to

In article <MONTA.92J...@image.mit.edu> mo...@image.mit.edu (Peter Monta) writes:

>Now if we rate the LP at 70 dB SNR at best (is this generous?), i.e.
>11 bits, then the physical resolution is 0.5 um, which is at the
>edge of reasonableness. There remains piezo noise and distortion,
>thermal noise in the preamp, etc. which I don't really know any
>numbers for.

One of the engineers at Micro-Accoustics (years ago) told me
their cartridge would respond to 1/2 micron.

Now have I gotten totally screwed up (that is highly possible)
but 1/2 micron is 1/2 of a millionth of an inch, which is a bit
smaller than 1/2 of a millionth of a meter.

70db is a reasonable figure for a carefully crafted state of
the art pressing. A pure lacquer (the master) can approach
80db.

Volume consumer pressings run much less.

--
Bill Vermillion - bi...@bilver.oau.org bill.ve...@oau.org
- bi...@bilver.uucp
- ..!{peora|ge-dab|tous|tarpit}!bilver!bill

Richard d Pierce

unread,

Jul 31, 1992, 3:26:09 PM7/31/92

to

In article <1992Jul31.1...@bilver.uucp> bi...@bilver.uucp (Bill Vermillion) writes:
>
>One of the engineers at Micro-Accoustics (years ago) told me
>their cartridge would respond to 1/2 micron.
>
>Now have I gotten totally screwed up (that is highly possible)
>but 1/2 micron is 1/2 of a millionth of an inch, which is a bit
>smaller than 1/2 of a millionth of a meter.
>

Whoops, Bill, the people at SI would take issue with this one. A micron is
the same as 1 uM, or 1 millionth of a meter. 1 millionth of an inch is
about 40 times smaller than that, or about 250 Angstroms, which is way the
hell up in the ultraviolet (which starts about 3500 angstroms or so).
Typical atomic diameters are in the single angstrom range, while 250 angstroms
is probably proximal to the size of the actual vinyl molecules in the record
(or so one might surmise)

John Davison

unread,

Aug 8, 1992, 2:37:01 AM8/8/92

to

In article <1992Jul28....@fid.morgan.com> se...@fid.morgan.com (Seth Breidbart) writes:

>Therefore, it is correct to say that DAT has, simultaneously, X dB
>dynamic range and Y dB snr, whenever X + Y = 96.

That is incorrect.

>Now consider a floating-point system with 16 bits, 4 bits of exponent
>and 13 bits (using the standard IEEE cheat) of mantissa. That system
>has 90 dB of dynamic range (using the exponent only), and 78 dB snr.
>Similarly to 16-bit linear, we can increase the dynamic range at the
>expense of snr (but not vice versa).

That is true; in fact, an interview in _The_Music_Machine_ (Curtis
Roads, ed.) with James Moorer (or is it F. Richard Moore? duh! duh! duh!) in
which one of those two guys proposes just such a system, with (I think) 12 bits
of significand and 4 bits of exponent.

--
NOTE: The opinions and other information contained herein are not to be
construed as representative of Northern Telecom or any of its
subsidiaries. Please direct all electronic mail to
=======> davi...@ecn.purdue.edu. <=======