DSD fact and fiction
Anonymous
The trouble with DSD
1) DSD is supposed to give better transient response and wider frequency
response. However, the bit rate of DSD is only 3Mbit/sec. The is
considered minimal by modern audio conversion standards. Why is this a
problem? Well, it means that you need a very high order of noise-shaping
in order to achieve > 100 dB of signal-to-noise ratio (at least
5th-order). This in turn means that the shaped quantization noise rises
at an extremely fast rate just above the audio spectrum. This is a
law-of-physics type of constraint and cannot be gotten around by clever
circuit design; only a higher oversampling factor could fix this, but
there is not enough room on the disc for all those bits.
So what?, you say. Well, it means that there must be a brick-wall
filter to eliminate this noise. So we are back to 20K brick wall filters
again, just like in the current standard!
Heres a little industry secret. The prototype demo systems making the
rounds these days have a very mild filter at the DAC output, so there is
a large amount of out-of-band noise at the output of the playback unit.
It just so happens that the preamp and amp used in these demo systems
are carefully selected units that will not "smoke" under these
conditions. The current industry standard for the amount of out-of-band
noise that is allowable is somewhere in the -60 dBFs region. So any
"real" consumer player will need a 20KHz lowpass filter that is quite
sharp. So much for better transient response. A straight 96KHz PCM
system would actually give a better transient response and have wider
bandwidth than a DSD system.
2) The recording format and the conversion format should be seperate!!
With current PCM standards, the chip makers have continuously improved
the quality of the A/D and D/A converters. The technology has progressed
from successive-approximation converters (early 80's) to 1-bit
delta-sigma designs (late 80's) to multi-bit noise-shaping designs (mid
90's). This was possible because one could always digitally convert the
internal signal representation used in the A/D or D/A converter to a
standard PCM word. With DSD, we are stuck with a 1-bit system that was
used in the mid-1980's, and had some serious problems with idle tones
and other strange effects. Does this mean that converter manufacturers
will go back to making converters the old-fashioned way? Not likely.
What will probably happen is that these manufacturers will continue to
use the best algorithms available, and then as a final step they will
digitally convert to a DSD stream (and suffer some loss of bandwidth and
quality during this conversion). The D/A converters will use the same
scheme; convert DSD to PCM (using a digital filter to get rid of all
that out-of-band noise) and then use whatever conversion scheme is
currently deemed the best at the time. DSD then becomes a storage format
only, with digital converters to go to and from PCM.
3) But the demos sound so good!
Sure they do. That's because the demo material has been magnificently
recorded by some of the top talent in the industry. This material would
sound good on any high-end playback system. There is no opportunity to
do a direct ABX-type comparison. Recording techniques have a lot more to
do with sound quality than the difference between a 96dB SNR playback
system and a 110 dB system.
4) Does any of this really matter?
Any system that can deliver > 100 dB of SNR and distortion over the
audio bandwidth will continue to sound wonderfull to most people most of
the time. So we are not talking about a giant step backwards in audio
quality. Its just that, compared to a true double or quad-speed PCM
system (like DVD audio-only), the DSD system is not much better than a
conventional CD system when you are simply comparing specs, whereas, at
least on paper, the DVD system offers real benefit. History has shown,
however, that the public does not part with real money unless they
perceive a real benefit. Anyone can see the difference in picture
quality between DVD and VHS, so DVD is likely to succeed. Almost no one
will hear the difference between DSD and CD, so it comes down to a test
of the marketing skill of Sony/Phillips to convince us that what we
though was perfect sound is really not so perfect after all.
Personally, I could think of lots of better uses for all those bits.
Sandman
despicable McInturd?
Anonymous <rwa...@ma.ultranet.com> wrote in message
news:3838E2...@ma.ultranet.com...
> I should probably stay out of this, but I just can't.
(clip)
Sandman
Arny Krüger
Anybody who thinks that point (4) is a little strange needs to spend
some time listening to the bit and frequency response reduced
musical selections I've posted at www.pcabx.com.
"Anonymous" <rwa...@ma.ultranet.com> wrote in message
news:3838E2...@ma.ultranet.com...
> I should probably stay out of this, but I just can't.
>
mdryoon
>
> This summarizes my findings to date quite well.
>
> Anybody who thinks that point (4) is a little strange needs to spend
> some time listening to the bit and frequency response reduced
> musical selections I've posted at www.pcabx.com.
I don't think point 4) is particarly strange. The S/N ratio and
channel separation specs of CD is obviously much better than that of
LP. But an LP in tip-top condition played through an excellent LP
playback system seems to be subjectively very close to CD in both S/N
ratio and channel separation.
IMO, the obvious improvement CD presents over LP isn't sound quality,
but convenience and long term reliability.
> "Anonymous" <rwa...@ma.ultranet.com> wrote in message
> news:3838E2...@ma.ultranet.com...
> >
> > The trouble with DSD
<snip>
> > 4) Does any of this really matter?
> >
> > Any system that can deliver > 100 dB of SNR and distortion over
> the
> > audio bandwidth will continue to sound wonderfull to most people m
> ost of
> > the time. So we are not talking about a giant step backwards in
> audio
> > quality. Its just that, compared to a true double or quad-speed
> PCM
> > system (like DVD audio-only), the DSD system is not much better
> than a
> > conventional CD system when you are simply comparing specs,
> whereas, at
> > least on paper, the DVD system offers real benefit.
<snip>
Richard Yoon (for e-mail replies, remove one "md" from "mdmdryoon.")
Markus Laun
mdryoon wrote:
>
>
> IMO, the obvious improvement CD presents over LP isn't sound quality,
> but convenience and long term reliability.
>
I saw 2 other advantages of CD's.
1) Price/Performance ratio.
2) CD's initialy forced recording studios/engineers to be better as the
CD's revealed more of their own errors then the avarage record player
did.
In the mean-time with the advent of cheap-o CD players bomm-boxes with
cd players etc. the many recording engineers are producing worse
material mixed to sound good on boom boxes.
I personaly don't want another format, I want better recorded CD's. I
have a few that are very very good, and make me cry at how poor most
are.
John Schrader
> I should probably stay out of this, but I just can't.
>
> The trouble with DSD
>
snip
> Personally, I could think of lots of better uses for all those bits.
Why are you "Anonymous"? I like to know whos "fact" I'm reading so as not
to be confused with B.S. fiction.
John Schrader
The Audio Source, Inc.
Rick Gardner
I have had the SACD-1 in my home system (BAT VK-50-SE line stage,
VK-500 amp, ESP Concert Grands spk.s, Jenna Labs wires.
At the time, (a few weeks ago), we had ALL the software currently
available.
No one from SONY anywhere around.
Oh, you never said if you had heard it, yet. I suspect you haven't and
are just blowing wind, like most people on this forum who insist they
know what it is, before any direct experience.
Try this. As it makes its way into people's homes, go hear it.
Until then, you have no idea of what you speak.
The experience was breathtaking.
* Sent from RemarQ http://www.remarq.com The Internet's Discussion Network *
The fastest and easiest way to search and participate in Usenet - Free!
Arny Krüger
"mdryoon" <mdmd...@earthlink.net> wrote in message
news:38392E0F...@earthlink.net...
> "Arny Krüger" wrote:
> >
> > This summarizes my findings to date quite well.
> >
> > Anybody who thinks that point (4) is a little strange needs to
spend
> > some time listening to the bit and frequency response reduced
> > musical selections I've posted at www.pcabx.com.
> I don't think point 4) is particarly strange. The S/N ratio and
> channel separation specs of CD is obviously much better than that
of LP.
To say the least. In my measurements the SNR of LP's rarely breaks
50 dB. LP channel separation can break 20 dB, but not at low
frequencies if there are significant levels of program material.
Then the separation has to be dramatically reduced or the cutting
stylus will make a hole in the record or lift off the surface of the
master or both.
> But an LP in tip-top condition played through an excellent LP
> playback system seems to be subjectively very close to CD in both
S/N
> ratio and channel separation.
This Might something to do with the fact that the ear has an
instaneous SNR limit of about 65 dB, and separation > 20 dB is not
heard that well either.
"Good" LP's I've measured generally have SNR's in the high 40 to low
50 dB range. If audibility weighting is applied ("A" weighting),
then LP SNR can creep into the high 50's or low 60's.
One major source of masking can be the program material itself.
Performances recorded with an audience present have SNR's well under
60 dB due to crowd noise, and therefore can be used to make LP's
sound more like CD's. Notice that a certain magazine that advocates
LP and sometimes releases the same performance on LP and CD tends to
release recordings of live performances.
Studio recordings can have SNR's in the high 60's and low 70's. LP's
rarely if ever can maintain this kind of performance.
In recent times we've seen some interesting contradictions where
people claim to hear things that exceed the general limits of the
human ear, say w/r/t amps and cables and digital precision (claiming
to "hear" 20 bits), but reverse their approach and use the known
limitations of the human ear to support their preference for LP. If
someone adds technical claims to their statements about their
preferences, then the technical claims can be scrutinized as if they
were technical claims, right?
> IMO, the obvious improvement CD presents over LP isn't sound
quality,
> but convenience and long term reliability.
That hangs on what one considers "obvious". Frankly, what bugs me
the most about LP is the audible jitter. I like piano recordings,
and between short and long term speed variations, nothing but
digital really meets my personal desires for sonic purity. Good
half-track high speed magnetic tape can be pretty good.
After the jitter that is endemic in LP's, there are a number of
pretty audible distortions due to the use of offset tonearms that I
don't like, but straight line tracking can"fix" that. Then there is
general increase in nonlinear distortion above 5-8 KHz due to the
fact that record cutters have sharp edges and styli are always
rounded to control recor wear. Well, IM makes massed instruments
sound more "together", right?
> > "Anonymous" <rwa...@ma.ultranet.com> wrote in message
> > news:3838E2...@ma.ultranet.com...
> > >
> <snip>
>
> > > 4) Does any of this really matter?
> > >
> > > Any system that can deliver > 100 dB of SNR and distortion
over the
> > > audio bandwidth will continue to sound wonderfull to most
people most of
Arny Krüger
"Markus Laun" <goos...@cdsoft.de> wrote in message
news:38397804...@cdsoft.de...
>
>
> mdryoon wrote:
> >
> >
> > IMO, the obvious improvement CD presents over LP isn't sound
quality,
> > but convenience and long term reliability.
> >
>
>
> 1) Price/Performance ratio.
> 2) CD's initialy forced recording studios/engineers to be better
as the
> CD's revealed more of their own errors then the avarage record
player
> did.
>
>
> In the mean-time with the advent of cheap-o CD players bomm-boxes
with
> cd players etc. the many recording engineers are producing worse
> material mixed to sound good on boom boxes.
Agreed. The "best" current recordings I can find don't come within
20 dB of exploiting the intrinsic capabiilty of CD technology
because of recording technique and limitations of commonly-used
venues. So here is the proposed "cure" - go to a format with
marginally better intrinsic capabilties. What's wrong with this
picture?
Arny Krüger
"Rick Gardner" <RickLGX...@aol.com.invalid> wrote in message
news:259caecb...@usw-ex0106-048.remarq.com...
> Hum . . . well . . .
>
> I have had the SACD-1 in my home system (BAT VK-50-SE line stage,
> VK-500 amp, ESP Concert Grands spk.s, Jenna Labs wires.
>
> At the time, (a few weeks ago), we had ALL the software currently
> available.
>
> No one from SONY anywhere around.
>
> Oh, you never said if you had heard it, yet. I suspect you haven't
and
> are just blowing wind, like most people on this forum who insist
they
> know what it is, before any direct experience.
>
> Try this. As it makes its way into people's homes, go hear it.
>
> Until then, you have no idea of what you speak.
>
> The experience was breathtaking.
>
And of course, the fact that all the program material was new or
remixed, and there was all this build-up and hype had nothing at all
to do with your perceptions...
Paul Bamborough
Reading the original post, a couple of questions arose in my mind about various
details: I do not know this subject well enough to answer them myself, and
wondered if anyone else could address them properly. Then this came to my aid:
Arny Krüger wrote in message...
>This summarizes my findings to date quite well.
I was delighted that Mr. Krüger has used the past two weeks so diligently, to
find out so many highly technical things about DSD. I say the last two weeks
because that is how long ago he wrote the following, in answer to a request for
a description of the subject in layman's terms:
"I would like to do that, but the technical literature that I've been able to
pull together from various sources so far lacks the detail I feel I need to
reliably do so."
and then went on describe it completely wrongly, including a claim that DSD
involved variable word-length perceptual coding and was related to HDCD. I was
grateful to Mr. Krüger at the time, because I felt that the poor person asking
the question deserved something better, and that prompted me to check out the
subject and clarify my own understanding by writing a brief non-technical
description.
Paul Frindle then discussed it much more fully and technically, politely
correcting Mr. Krüger's nonsense. (Bizarrely, rather than simply acknowledging
the truth, Mr. Krüger then tried to claim he was right in the first place, by
redefining the standard terms he had used into some meaning of his own.)
But, since then, he has found out *everything* that "Anonymous" has been
commenting on. And how graciously he makes his compliment ("summarises my
findings... quite well". Well done, Mr. Krüger: I am glad to see that your
thirst for knowledge was so great, and that you have learned so much.
Here is my question: it relates to points 1 and 2: is there any intrinsic
reason why the bitstream cannot be converted back to PCM words, and converted
conventionally? If so, what are the implications?
If you transcoded to (say) 48 or 96 KHz samples, what would their resolution be?
p
>Anybody who thinks that point (4) is a little strange needs to spend
>some time listening to the bit and frequency response reduced
>musical selections I've posted at www.pcabx.com.
>
>"Anonymous" <rwa...@ma.ultranet.com> wrote in message
>news:3838E2...@ma.ultranet.com...
>> I should probably stay out of this, but I just can't.
>> The trouble with DSD
>> 1) DSD is supposed to give better transient response and wider
>> 4) Does any of this really matter?
>> Any system that can deliver 100 dB of SNR and distortion over the
>> audio bandwidth will continue to sound wonderfull to most people
>> most of the time. So we are not talking about a giant step backwards
>> in audio quality. Its just that, compared to a true double or quad-
>> speed PCM system (like DVD audio-only), the DSD system is not much
>> better than a conventional CD system when you are simply comparing
>> specs, whereas, at least on paper, the DVD system offers real
>> benefit. History has shown, however, that the public does not part
>> with real money unless they perceive a real benefit. Anyone can see
>> the difference in picture quality between DVD and VHS, so DVD is
>> likely to succeed. Almost no one will hear the difference between
>> DSD and CD, so it comes down to a test of the marketing skill of
>> Sony/Phillips to convince us that what we though was perfect sound
>> is really not so perfect after all.
>> Personally, I could think of lots of better uses for all those bits.
Scott Dorsey
>
>No one from SONY anywhere around.
>
>Oh, you never said if you had heard it, yet. I suspect you haven't and
>are just blowing wind, like most people on this forum who insist they
>know what it is, before any direct experience.
>
>Try this. As it makes its way into people's homes, go hear it.
I have heard two demos.
The demo that Sony gave at the AES show was very disappointing. No
imaging whatsoever; all the sound came directly from the speakers.
The overall feeling wasn't natural at all.
Another demo, given by a high-end dealer, was much better, but still
didn't blow me away. It was very well-recorded material and well-reproduced,
but I wasn't jumping up and down and deciding to sell my turntable by any
means.
>The experience was breathtaking.
I'm not saying that breathtaking experiences aren't possible, just that
I haven't had one yet.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."
AudioMaven
The demo that Sony gave at the AES show was very disappointing. No
imaging whatsoever; all the sound came directly from the speakers.
The overall feeling wasn't natural at all.
Another demo, given by a high-end dealer, was much better, but still
didn't blow me away. It was very well-recorded material and well-reproduced,
but I wasn't jumping up and down and deciding to sell my turntable by any
means.
>The experience was breathtaking.
I'm not saying that breathtaking experiences aren't possible, just that
I haven't had one yet.
--scott>>
Hi Scott,
I understand what you're saying about DSD and the SCD-1. I heard the player at
four demos over the last 18 months or so--and each time I thought it was awful.
I walked away from the LA demo with a headache. I think some of the problem is
that Sony insists on using their own speakers, which tweet all the way out to
heaven.
That said, I have had the chance to live with the unit for a month in my
system. All I have to say is to keep an open mind and reserve judgement until
you can hear it in your system! I think you will be pleasantly surprised.
DSD has been covered in Positive Feedback, TAS, Stereophile (and a few others)
and we'll have our say on the new technology in the upcoming Winter issue of
Ultimate Audio. I think our reviewers feel it is a real special technology.
Myles Astor
Publisher
Ultimate Audio magazine
www.ultimateaudio.com
Vivin Oberoi
my favorite dealer's on a system that I am familiar with.
I was able to A/B compare at least one album - Miles Davis's
Kind of Blue both on CD and SACD using the same player.
The SACD came out in front, hands down. Cymbals sounded
like cymbals (none of the etchy shrilly sound).
The bass was much better defined. The imaging and
soundstaging was breathtaking. The palpability of this is old recording
increased 100%. I am not saying that the SACD is better than
DVD-Audio (since I haven't heard it yet). But there is little doubt in
my mind that what I heard was better than CD.
After a limited audition (limited by program material that I am
"very" familiar with, I was impressed. So was my wife (I trust her
ears more than I do mine). Just listening to CDs, the
player held its own against some of the more expensive
gear that I have heard.
I agree with with Myles Astor. I have seen enough mumbo jumbo
new technologies in my time that make little difference to the
bottom line - how good does it sound. As for me, I am waiting
to go hear it again after they get more SACDs in.
Vivin Oberoi.
AudioMaven <audio...@aol.com> wrote in message
news:19991122160542...@ng-fz1.aol.com...
Arny Krüger
has the time to come up with a wad of personal-attack-laden science
fiction like this.
Why would he do such a thing?
Well one of the regulars on RAO said he paid Paul to do it.
Sad.
"Paul Bamborough" <pa...@bamborough.com> wrote in message
news:81c8dp$o8h$1...@bgtnsc03.worldnet.att.net...
Andrew Thibault
"Arny Krüger" wrote:
>
> I think its really great when one of the "greats' of the industry
> has the time to come up with a wad of personal-attack-laden science
> fiction like this.
>
> Why would he do such a thing?
>
> Well one of the regulars on RAO said he paid Paul to do it.
Hook, line, sinker....
> Sad.
You forgot this: ;-(
Arny Krüger
"Andrew Thibault" <teeb...@frontiernet.net> wrote in message
news:383A9145...@frontiernet.net...
>
>
> "Arny Krüger" wrote:
> >
> > I think its really great when one of the "greats' of the
industry
> > has the time to come up with a wad of personal-attack-laden
science
> > fiction like this.
> >
> > Why would he do such a thing?
> >
> > Well one of the regulars on RAO said he paid Paul to do it.
>
Thanks for taking the bait so gracefully. ;-)
Edward M. Shain
Perhaps it's time for you to take stock and do a little
inventory taking.
The past month has seen you engaged in bitter battle with a
fairly large number of people. That group's roster incorporates people
of every stripe and persuasion on RAO. It has included audiophiles,
High End journalists, industry figures, and scientists.
None of these people have anything in common besides a love of
audio and music PLUS the experience of being in a nasty fight with
you. Are they *all* wrong? Is it *everyone* else who's at fault here?
I've looked through the threads carefully, Arnold, and I can't
find a particular objective position which is at the center of it. The
fights haven't been about the usual rumbles. Bodies aren't strewn
across the "Great Divide" of hobbyist vs. scientist. Instead, the
fights have been about tactics, about how the
discussions/arguments/battles have been fought. The only common thread
has been you, Arnold, and the way you conduct yourself.
Is it barely possible that you might be the cause of it all?
Ed
On Tue, 23 Nov 1999 12:43:15 GMT, "Arny Krüger" <ar...@flash.net>
wrote:
trotsky
"Edward M. Shain" wrote:
>
> Arnold,
>
> Perhaps it's time for you to take stock and do a little
> inventory taking.
>
> The past month has seen you engaged in bitter battle with a
> fairly large number of people. That group's roster incorporates people
> of every stripe and persuasion on RAO. It has included audiophiles,
> High End journalists, industry figures, and scientists.
>
> None of these people have anything in common besides a love of
> audio and music PLUS the experience of being in a nasty fight with
> you. Are they *all* wrong? Is it *everyone* else who's at fault here?
>
> I've looked through the threads carefully, Arnold, and I can't
> find a particular objective position which is at the center of it. The
> fights haven't been about the usual rumbles. Bodies aren't strewn
> across the "Great Divide" of hobbyist vs. scientist. Instead, the
> fights have been about tactics, about how the
> discussions/arguments/battles have been fought. The only common thread
> has been you, Arnold, and the way you conduct yourself.
>
> Is it barely possible that you might be the cause of it all?
>
> Ed
I find these posts confusing from you, Ed. I mean, it's like trying to ask
Jeffrey Dahmer why he used to eat people (when he was still alive, that is).
What do you expect Krooger to say? Do any of us expect him to admit to his
psychopathology? The fact that he claims to be a "Christian" is further
evidence of his insanity. Every single one of us knows that in reality Krooger
is just a sick fuck, so what could possibly be the point of trying to reason
with him?
Edward M. Shain
Hard to argue with that, Greg. Chalk it up to naive optimism
on my part.
Ed
George M. Middius
> I think its really great when one of the "greats' of the industry
> has the time to come up with a wad of personal-attack-laden science
> fiction like this.
>
> Why would he do such a thing?
>
> Well one of the regulars on RAO said he paid Paul to do it.
>
> Sad.
As your palborgs Steindrone and LibraryClerk are fond of
saying, "Got any facts? Didn't think so."
You're pathetic. Who are you fooling -- other than Mikey, of
course?
George M. Middius
George M. Middius
> Thanks for taking the bait so gracefully. ;-)
Sadder and sadder every time you pretend you didn't pants
yourself in your previous post. What's left for you? Will
you shoot up the shopping mall and then torch the Kroobitch
in her bed, or will you just eat your gun?
George M. Middius
George M. Middius
> The only common thread
> has been you, Arnold, and the way you conduct yourself.
> Is it barely possible that you might be the cause of it all?
Religious attack! Bigot!
George M. Middius
George M. Middius
> Every single one of us knows that in reality Krooger
> is just a sick fuck, so what could possibly be the point of trying to reason
> with him?
SOLD! to the salesman from Chicago.
George M. Middius
trotsky
"Edward M. Shain" wrote:
>
> On Tue, 23 Nov 1999 08:00:20 -0600, trotsky <gsi...@mc.net> wrote:
>
> >
> >
> Hard to argue with that, Greg. Chalk it up to naive optimism
> on my part.
>
>
You know, for once I would have to completely agree with you. Wouldn't it
completely change the tenor of the discussions around here if Krooger behaved
even *remotely* like a human being?
Andrew Thibault
"Arny Krüger" wrote:
>
> "Andrew Thibault" <teeb...@frontiernet.net> wrote in message
> news:383A9145...@frontiernet.net...
> >
> >
> > "Arny Krüger" wrote:
> > >
> > > I think its really great when one of the "greats' of the
> industry
> > > has the time to come up with a wad of personal-attack-laden
> science
> > > fiction like this.
> > >
> > > Why would he do such a thing?
> > >
> > > Well one of the regulars on RAO said he paid Paul to do it.
> >
> > Hook, line, sinker....
>
> Thanks for taking the bait so gracefully. ;-)
Nice to see an IKYABWAI done so poorly...
Mikeylikst
>Hum . . . well . . .
>
>I have had the SACD-1 in my home system (BAT VK-50-SE line stage,
>VK-500 amp, ESP Concert Grands spk.s, Jenna Labs wires.
>
>At the time, (a few weeks ago), we had ALL the software currently
>available.
>
>No one from SONY anywhere around.
>
>Oh, you never said if you had heard it, yet. I suspect you haven't and
>are just blowing wind, like most people on this forum who insist they
>know what it is, before any direct experience.
>
>Try this. As it makes its way into people's homes, go hear it.
>
>Until then, you have no idea of what you speak.
>
>The experience was breathtaking.
>
Learn how to quote, so people know what and who you are talking about.
Mike McKelvy
the love you take is equal to the love you make..............
Steve Zipser
mikey...@aol.com says...
> >RickLGX...@aol.com.invalid
>
> >Hum . . . well . . .
> >
> >I have had the SACD-1 in my home system (BAT VK-50-SE line stage,
> >VK-500 amp, ESP Concert Grands spk.s, Jenna Labs wires.
> >
> >At the time, (a few weeks ago), we had ALL the software currently
> >available.
> >
> >No one from SONY anywhere around.
> >
> >Oh, you never said if you had heard it, yet. I suspect you haven't and
> >are just blowing wind, like most people on this forum who insist they
> >know what it is, before any direct experience.
> >
> >Try this. As it makes its way into people's homes, go hear it.
> >
> >Until then, you have no idea of what you speak.
> >
> >The experience was breathtaking.
> >
>
> Learn how to quote, so people know what and who you are talking about.
That hasn't helped you at all ;-)
Arny Krüger
"Edward M. Shain" <Vizsl...@worldnet.att.net> wrote in message
news:383aa4ef...@netnews.worldnet.att.net...
> >
> >I find these posts confusing from you, Ed. I mean, it's like
trying to ask
> >Jeffrey Dahmer why he used to eat people (when he was still
alive, that is).
> >What do you expect Krooger to say? Do any of us expect him to
admit to his
> >psychopathology? The fact that he claims to be a "Christian" is
further
> >evidence of his insanity. Every single one of us knows that in
reality Krooger
> >is just a sick fuck, so what could possibly be the point of
trying to reason
> >with him?
>
> Hard to argue with that, Greg. Chalk it up to naive optimism
> on my part.
>
Or it's just Ed up to his usual run of personal attacks on me.
BTW Ed, how many audio companies do you have investments in?
Edward M. Shain
wrote:
>
>"Edward M. Shain" <Vizsl...@worldnet.att.net> wrote in message
>news:383aa4ef...@netnews.worldnet.att.net...
>> On Tue, 23 Nov 1999 08:00:20 -0600, trotsky <gsi...@mc.net> wrote:
>>
>
>> >
>> >I find these posts confusing from you, Ed. I mean, it's like
>trying to ask
>> >Jeffrey Dahmer why he used to eat people (when he was still
>alive, that is).
>> >What do you expect Krooger to say? Do any of us expect him to
>admit to his
>> >psychopathology? The fact that he claims to be a "Christian" is
>further
>> >evidence of his insanity. Every single one of us knows that in
>reality Krooger
>> >is just a sick fuck, so what could possibly be the point of
>trying to reason
>> >with him?
>>
>> Hard to argue with that, Greg. Chalk it up to naive optimism
>> on my part.
>>
>
>Or it's just Ed up to his usual run of personal attacks on me.
>
>BTW Ed, how many audio companies do you have investments in?
Lemme see now....unnnhhh........37, Arnold, I should have
only 36 but I had to take my shares in your web-site off the market.
People were asking me to pay them for taking it off my hands.
Ed
>
Paul Bamborough
Well, I really am interested to get an answer to my question: I've been reading
up a little more on DSD, and I don't feel that I understand enough to reach a
conclusion. The question is:
Is there any intrinsic reason why the bitstream cannot be
converted back to PCM words, and thence to analogue conventionally?
If so, what are the implications? If you transcoded to (say)
48 or 96 KHz samples, what would their resolution be?
Can anyone help?
With respect to Mr. Krüger's post:
>I think its really great when one of the "greats' of the industry
>has the time to come up with a wad of personal-attack-laden science
>fiction like this.
>Why would he do such a thing?
>Well one of the regulars on RAO said he paid Paul to do it.
>Sad.
No personal attacks, no science-fiction. Mr Krüger *did* write rubbish about
this subject two weeks ago. He *was* belligerent and graceless enough to argue
pointlessly when the truth was spoken. I have no idea why he would do this,
because there is no shame in not knowing things, nor in admitting that one
doesn't know them. Quite the reverse: this is how one learns.
But, the way I see it, there *is* shame in not knowing things and then
aggressive and actively attempting to pretend that you knew them all along, in
order to plug your own hobby-horse.
p
P.S. Mr. Krüger, Mr. Shain's comment about paying me was a *joke*. I believe
you may be having a problem because he forgot to signal that clearly enough with
one of those little symbols. Let me see... how does it go... %() .. no
:::()()() ... not that..
@#$) ... almost... [:::( ah, I've got it.... ;-). A joke. Not true. Not
serious. No point in harping on it.
>"Paul Bamborough" wrote...
jj, curmudgeon and tiring philalethist
Paul Bamborough <pa...@bamborough.com> wrote:
>Is there any intrinsic reason why the bitstream cannot be
>converted back to PCM words, and thence to analogue conventionally?
>If so, what are the implications? If you transcoded to (say)
>48 or 96 KHz samples, what would their resolution be?
>Can anyone help?
SACD is a kind of sigma-delta (or delta-sigma) convertor.
You could indeed transcode it to PCM of a particular
sampling rate, where you could more or less control the
sampling rate, with some concerns about noise floor.
BUT, I'm seriously unconvinced regarding SACD in general, personally.
I want to be able to do a REAL listening comparison, please.
In a quiet place, with fast switching and time alignment. NOT
on a show floor, with sequential presentation, thank you.
--
Copyright j...@research.att.com 1999, all rights reserved, except transmission
by USENET and like facilities granted. This notice must be included. Any
use by a provider charging in any way for the IP represented in and by this
article and any inclusion in print or other media are specifically prohibited.
Arny Krüger
> On Tue, 23 Nov 1999 17:34:21 GMT, "Arny Krüger" <ar...@flash.net>
> wrote:
>
> >
> >"Edward M. Shain" <Vizsl...@worldnet.att.net> wrote in message
> >news:383aa4ef...@netnews.worldnet.att.net...
> >> On Tue, 23 Nov 1999 08:00:20 -0600, trotsky <gsi...@mc.net>
wrote:
> >>
> >
> >> >
> >trying to ask
> >> >Jeffrey Dahmer why he used to eat people (when he was still
> >alive, that is).
> >> >What do you expect Krooger to say? Do any of us expect him to
> >admit to his
> >> >psychopathology? The fact that he claims to be a "Christian"
is
> >further
> >> >evidence of his insanity. Every single one of us knows that
in
> >reality Krooger
> >> >is just a sick fuck, so what could possibly be the point of
> >trying to reason
> >> >with him?
> >>
> >> on my part.
> >>
> >
> >Or it's just Ed up to his usual run of personal attacks on me.
> >
> >BTW Ed, how many audio companies do you have investments in?
>
> Lemme see now....unnnhhh........37, Arnold, I should have
> only 36 but I had to take my shares in your web-site off the
market.
> People were asking me to pay them for taking it off my hands.
>
Please post again when a serious thought enters your head.
Joe Duffy
George M. Middius <Glan...@ipo.net> wrote:
>trotsky said:
>
>> Every single one of us knows that in reality Krooger
>> is just a sick fuck, so what could possibly be the point of trying to reason
>> with him?
>
>
>
the ASF's have spoken!
in anger that the anointed
sickest-of-all will not join.
joe
Joe Duffy
Edward M. Shain <Vizsl...@worldnet.att.net> wrote:
>
> Is it barely possible that you might be the cause of it all?
>
such an almost black-and-white statement :-)
be careful who we demonize,
for next it may be you!
joe
Alvin Bloom
>such an almost black-and-white statement :-)
> be careful who we demonize,
> for next it may be you!
>
>
Very GOOD Joe! Let me hazard a guess: You ordered Chinese Take-Out and found
this dazzling little gem in a fortune cookie, right?
Thank you
Alvin Bloom
trotsky
>
> "Edward M. Shain" <Vizsl...@worldnet.att.net> wrote in message
> news:383aa4ef...@netnews.worldnet.att.net...
> > On Tue, 23 Nov 1999 08:00:20 -0600, trotsky <gsi...@mc.net> wrote:
> >
>
> > >
> trying to ask
> > >Jeffrey Dahmer why he used to eat people (when he was still
> alive, that is).
> > >What do you expect Krooger to say? Do any of us expect him to
> admit to his
> > >psychopathology? The fact that he claims to be a "Christian" is
> further
> > >evidence of his insanity. Every single one of us knows that in
> reality Krooger
> > >is just a sick fuck, so what could possibly be the point of
> trying to reason
> > >with him?
> >
> > on my part.
> >
>
> Or it's just Ed up to his usual run of personal attacks on me.
>
> BTW Ed, how many audio companies do you have investments in?
Those were MY comments that Ed was agreeing with, Kroogles. What's the
matter, don't you want a piece of me? My God, you can't even attribute
the trvth correctly.
jj, curmudgeon and tiring philalethist
Paul Bamborough <pa...@bamborough.com> wrote:
>I wasn't expressing myself particulalry clearly: the original poster had
>pointed out two problems that he perceived with DSD: the first was that there
>is so much noise just out-of-band that you have to use a fairly brutal filter,
>with all that this implies, and the second was that the means of doing DA
>conversion would be sub-optimal comapred to the best current solutions.
Well, you'd have the same problems digitally, too.
>>BUT, I'm seriously unconvinced regarding SACD in general, personally.
>>I want to be able to do a REAL listening comparison, please.
>>In a quiet place, with fast switching and time alignment. NOT
>>on a show floor, with sequential presentation, thank you.
>And with matched sources of known origin.
Ayup!
trotsky
>
> In article <383b8ec5...@netnews.worldnet.att.net>,
> Edward M. Shain <Vizsl...@worldnet.att.net> wrote:
> >
> > Is it barely possible that you might be the cause of it all?
> >
>
> be careful who we demonize,
> for next it may be you!
>
Also, joe, you should point out not to
piss on the third rail of an electric train,
because it is possible to electrocute your
dick, and that would almost be as bad as
demonizing people!
Joe Duffy
Alvin Bloom <alvin...@aol.com> wrote:
>>From: du...@bcstec.ca.boeing.com
>
>>such an almost black-and-white statement :-)
>> be careful who we demonize,
>> for next it may be you!
>>
>>
>
>this dazzling little gem in a fortune cookie, right?
>
no, i was thinking that near-genocide
was ALL one man's fault, actually.
joe
Paul Bamborough
jj, curmudgeon and tiring philalethist wrote in message ...
>In article <81eme8$613$1...@bgtnsc04.worldnet.att.net>,
>Paul Bamborough <pa...@bamborough.com> wrote:
>>Is there any intrinsic reason why the bitstream cannot be
>>converted back to PCM words, and thence to analogue conventionally?
>>If so, what are the implications? If you transcoded to (say)
>>48 or 96 KHz samples, what would their resolution be?
>
>>Can anyone help?
>SACD is a kind of sigma-delta (or delta-sigma) convertor.
>You could indeed transcode it to PCM of a particular
>sampling rate, where you could more or less control the
>sampling rate, with some concerns about noise floor.
I wasn't expressing myself particulalry clearly: the original poster had
pointed out two problems that he perceived with DSD: the first was that there
is so much noise just out-of-band that you have to use a fairly brutal filter,
with all that this implies, and the second was that the means of doing DA
conversion would be sub-optimal comapred to the best current solutions.
If these things are true (and I am not at all sure about the second), they could
presumably be mitigated by simply transcoding to PCM. Which raised the question
of what the resultant peformance looks like as compared to a conventional (say)
96/24 PCM setup.
>BUT, I'm seriously unconvinced regarding SACD in general, personally.
>I want to be able to do a REAL listening comparison, please.
>In a quiet place, with fast switching and time alignment. NOT
>on a show floor, with sequential presentation, thank you.
And with matched sources of known origin.
p
Brian L. McCarty
obe...@home.com wrote on 23/11/99 10:02:
> I heard the SCD-1 last weekend for the first time at
> my favorite dealer's on a system that I am familiar with.
> I was able to A/B compare at least one album - Miles Davis's
> Kind of Blue both on CD and SACD using the same player.
>
> The SACD came out in front, hands down. Cymbals sounded
> like cymbals (none of the etchy shrilly sound).
> The bass was much better defined. The imaging and
> soundstaging was breathtaking. The palpability of this is old recording
> increased 100%. I am not saying that the SACD is better than
> DVD-Audio (since I haven't heard it yet). But there is little doubt in
> my mind that what I heard was better than CD.
This is a completely irrelevant test. The "Kind of Blue" album, recorded in
the late 50's, was REMASTERED for SACD and you are comparing apples and
oranges.
Can you say "boost the EQ" boys and girls? I KNEW you could!
---
Steve Zipser (Sunshine Stereo) is a proven:
Zipser is a liar http://dejanews.com/=dnc/getdoc.xp?AN=369217967
Zipser is a scammer http://dejanews.com/=dnc/getdoc.xp?AN=368363274
Zipser is a cheater http://dejanews.com/=dnc/getdoc.xp?AN=374900703
Zipser is a THIEF http://dejanews.com/=dnc/getdoc.xp?AN=509980240
Gene Lyle
wrote:
>
>Or it's just Ed up to his usual run of personal attacks on me.
Arnold, for God's sake, you've got "Kick Me" written all over you.
Can't you see that?
>
>BTW Ed, how many audio companies do you have investments in?
>
Who cares.
Gene Lyle
stephen campbell
>Every single one of us knows that in
>reality Krooger is just a sick fuck, so what
>could possibly be the point of trying to
>reason with him?
Well I for one like to see him squirm and be exposed when he resorts to
weasel tactics.
Stephen
Arny Krüger
"Paul Bamborough" <pa...@bamborough.com> wrote in message
news:81eme8$613$1...@bgtnsc04.worldnet.att.net...
>
> Well, I really am interested to get an answer to my question:
I've been reading
> up a little more on DSD, and I don't feel that I understand enough
to reach a
> conclusion. The question is:
>
> Is there any intrinsic reason why the bitstream cannot be
> converted back to PCM words, and thence to analogue
conventionally?
> If so, what are the implications? If you transcoded to (say)
> 48 or 96 KHz samples, what would their resolution be?
>
> Can anyone help?
Drawings like the one at
http://home.sol.no/~espen-b/dvd/audio/index.html make it seem like
you could. BTW, I think I've seen this same drawing in English,
perhaps even in a paper from Sony, but I could not find it quickly.
As is shown, and per numerous other comments from various credible
sources, DSD technology seems to drawsheavily from Delta-Sigma
convertor technology. Therefore, it is not unreasonable to use what
is known about Delta-Sigma convertor technology to estimate the
performance of proposed conversions of the DSD data stream.
It strikes me that the oversampling factor gets lost when one tries
to pull more samples (more data words) out of a given bit stream.
http://www.rmsinst.com/dt3.htm says "Even the very simple
first-order modulator can achieve very attractive SNRs; a good 'rule
of thumb' is that for every doubling of the oversampling ratio, the
SNR will improve by approximately 9 dB." Practical D-S convertors
seem to be fourth or fifth order, suggesting that far higher rates
of return from oversampling are common. I haven't readily found a
reference that says what they are. Neverthelss this suggests that
designers are depending on a very high rate of return in terms of
SNR improvement, from oversampling.
It seems to me that one result of reconstructing more data words
(increasing sampling rate) would be a corresponding decrease in
oversampling. Therefore, as appreciably more PCM data words are
reconstructed from a given data stream, dramatic losses in SNR might
be the unhappy result.
This all seems pretty moot to me in the context of consumer audio,
as it is not that hard to find consumer digital audio equipment that
is already sonically transparent as typically used.
However, in a music business production environment, the result of
technical mistakes can be the loss of priceless artistic content,
not to mention great expense for retakes that may never capture "the
moment". Producing equipment that is practically fool-proof (i.e.,
40 dB SNR margins) for the music business production environment
seems to have a decent potential for cost-justification and
therefore makes good engineering sense if there are business people
who have quality concerns and the money to spend. I suspect that
there is quite a bit of money for capital improvements like this
among the larger producers of audio recordings, especially those who
also have to contend with technical staff whose proficiency can be
highly variable.
Arny Krüger
"trotsky" <gsi...@mc.net> wrote in message
news:383AAD75...@mc.net...
>
> You know, for once I would have to completely agree with you.
Wouldn't it
> completely change the tenor of the discussions around here if
Krooger behaved
> even *remotely* like a human being?
Why not make a big change in your posting style around here, and
practice what you preach?
George M. Middius
> >Every single one of us knows that in reality Krooger is just a sick fuck,
> >so what could possibly be the point of trying to reason with him?
> Well I for one like to see him squirm and be exposed when he resorts to
> weasel tactics.
Looks like you've lost another one, Arnii.
George M. Middius
George M. Middius
>> You know, for once I would have to completely agree with you.
>> Wouldn't it completely change the tenor of the discussions around
>> here if Krooger behaved even *remotely* like a human being?
> Why not make a big change in your posting style around here, and
> practice what you preach?
Despising, mocking, and sneering at you is the utmost in
human responses to your conveyor belt of snot and feces.
Once again, you have no way of comprehending the human point
of view, what with your Snot Module having supplanted your
erstwhile limbic region completely. Sad.;-(™
[This post reformatted by Paul Bamborough Engineering,
laboring tirelessly to de-Kroogerize Usenet]
George M. Middius
trotsky
"Arny Krüger" wrote:
>
> "trotsky" <gsi...@mc.net> wrote in message
> news:383AAD75...@mc.net...
> >
> > You know, for once I would have to completely agree with you.
> Wouldn't it
> > completely change the tenor of the discussions around here if
> Krooger behaved
> > even *remotely* like a human being?
>
> Why not make a big change in your posting style around here, and
> practice what you preach?
Does "fuck you" answer your question, turd-monster?
Dale Goudey
>Paul Bamborough wrote
>
>Well, I really am interested to get an answer to my question: I've
been reading
>up a little more on DSD, and I don't feel that I understand enough to
reach a
>conclusion. The question is:
>
>Is there any intrinsic reason why the bitstream cannot be
>converted back to PCM words, and thence to analogue conventionally?
>If so, what are the implications? If you transcoded to (say)
>48 or 96 KHz samples, what would their resolution be?
>
>Can anyone help?
>
Yes, the bitstream can be converted to lower sample rate PCM, but that
would eliminate the main argument in favor of DSD (that the brick wall
filter is eliminated). But the matter of converting to PCM to "fix"
the conversion process is more fundamentally flawed, since the high
level of DSD modulation noise in the ultrasonic frequency regime is
inherent in the information stream. ANY conversion process must
either pass it through or filter it with a sharp filter. Either
approach can be criticized. Passing the high frequency noise through
preserves the brick-wall-less nature of the processing chain (which
some believe affects sound perception). Filtering the high frequency
noise out introduces the brick wall filter, which then makes DSD a
poorly designed variation of PCM. One "problem" with DSD is that the
chip makers (who make the D/A chips) may ignore the objective to
eliminate brick wall filtering, and produce DSD D/A chips that use the
classic brick wall filter approach to achieve better technical
specifications. Many consumers may never hear what Sony has been
promoting as "a better way".
Point 2, below, is a good one. Rather than DSD, I would prefer a high
sample rate storage format that stored PCM using lossless compression.
A sample rate of 192k or 384k would be sufficient, and we would have a
good separation of storage format and conversion technology. Lossless
compression is more effective at the higher sample rates. At such
high sample rates anti-alias filters could be designed for better
pulse response. Another option would be DSD at a higher sample rate,
using an open specification for the noise shaping method. Sony has
indicated that they are working on an improved DSD with a higher
sample rate. It seems strange that Sony is promoting a consumer
storage format for ultimate quality, while at the same time they are
promising to soon introduce a higher sample rate version "for
professional use".
The "trouble with DSD" post is a very good one, with points worth
discussing.
I have heard a DSD demo. I was most impressed. Better
recorded/reproduced sound than I have heard by any other method,
though there was no attempt at a fair comparison of the various format
options so this is more a statement of the entire record/playback
chain than of the recording format itself. Still, most impressive.
If the price were right, I'd buy it! Maybe in a few years....
George M. Middius
> Does "fuck you" answer your question, turd-monster?
Careful, Greg. Boogers may actually want you to schlep over to
Michigan and give him a swift kick.
George M. Middius
trotsky
Is there trouble with the Kroobitch? Is this the only way his Kielbasa will see
some action?
greg pavlov
wrote:
>I heard the SCD-1 last weekend for the first time at
>my favorite dealer's on a system that I am familiar with.
>I was able to A/B compare at least one album - Miles Davis's
>Kind of Blue both on CD and SACD using the same player.
>
>The SACD came out in front, hands down. Cymbals sounded
>like cymbals (none of the etchy shrilly sound). ....
>
In comparisons like this, you are comparing not only
the media but a lot of other things as well, including
the time, effort, and technology that was invested in
mastering/remastering and producing the final versions
of the recording.
greg pavlov
[not affiliated with DFCI or Harvard]
**************************************************************************
For the definitive intro guide to rao, see:
http://members.aol.com/whosbest54/
**************************************************************************
Paul Frindle
modulator in a modern ADC design (this is available at the test pins of a Crystal
CS5390 ADC). Normally this is filtered (on the IC) to extract the signal in the
wanted band by filtering, which requires that the data word is increased to
accomodate the extra SNR produced. So if you replace the filter that was deleted in
the DSD ADC you will have EXACTLY the same thing as a normal PCM ADC.
Now if you want to store the unfiltered 1bit modulated information instead thats
fine - this is DSD. Obviously this constitutes much more information than the band
limited PCM but it conveniently fills a new DVD disc nicely! Of course you can play
the data back through a digital filter, increase the wordlength sufficiently and
resample at the appropriate rate for the wanted bandwidth and convert to analogue in
the conventional way. This is the best way to convert DSD into analogue - believe
me. If you do this with the proposed DSD format (i.e 64FS) you will end up with a
SNR of 120dB roughly and it will sound fine. Higher 1bit sampling rates are needed
to realise better 'audible' SNRs.
In any sampled system sampling rate and wordlengths are theoretically
interchangeable so resolution on its own is a null concept. If theory were
achievable in practice, DSD would have a higher quality than 96KHz 24bit since the
data rates are higher and more information is potentially transfered. But the
impossibility of making an extremely high order modulator (i.e one that has a brick
wall filter in its loop) means that in practice the DSD format is less data
efficient, in the audio band that we can hear, than 96KHz 24bit (hence the worse SNR
figures).
There are other practical issues to consider:
The philosophical requirement to avoid parallel data streams (and digital filters)
within the DAC means that all (or most) filtering must occur in the analogue domain.
This means that you are relying once again on opamp performance and component
tolerances etc. This means that the filtering is necessarily gentle if quality is to
be preserved. In reality most successful DSD DACs employ a kind of FIR, using
multiple DACs driven from delayed signals and mixed together resistively. This is
called a transversal filter and is the equivalent of a digital FIR filter (but don't
tell anyone), this helps to roll-off the extreme HF and make life easier for the
analogue filter, as well as reducing the increased sensitivity of the system to
clock jitter by averaging..
It is also a mistake to imagine that there is any physical reason that analogue
filtering is superior to digital filtering. A given filter response has the same
theoretical disturbance on the signal however it is produced if you ignor the error
mechanisms, (this is a mathematical reality and would apply equally even to a
mechanical system). Therefore it is very difficult to make an analogue filter that
can even approach the performance of a digital design with a 24bits or more internal
wordlength.
The phylosophical requirement for single bit conversion methods excludes the use of
advanced multibit devices recently developed by the major IC manufacturers (i.e
AKM4393 DAC, Crystal CS5396 ADC). Therefore the format itself is potentially
excluded from already existent advances in conversion art. However it would be very
possible to realise the performance of these new generation devices if the PCM they
deliver were remodulated digitally into a 1bit stream (interesting thought), and
converted back to PCM before the DAC on playback (even more interesting - what do
you have then?).
Despite all of this, with enough effort the scheme can be made to work well enough
to be apparently as transparent as existing systems, largely due to the fact that we
are less sensitive (!) to stuff above 20KHz and SNRs beyond about 80dBs are adequate
for domestic listening. The advocates of this format attribute the alleged 'higher
quality' of this format to factors that cannot be measured.
To be kind to these people I would say that this unlikely in the extreme. As ever,
the very same people do not believe that the easily measured errors that exist in
lots of comercial CD players can be heard, so they would not believe that their
absence could possibly be the cause of better sound - would they? The errors caused
by a badly designed DSD DAC will be worse than a cheap PCM DAC, they may be
different and sound different but they will still be there. It is therefore possible
that these errors are somehow better liked in the estimation of those that form
opinions at the present time? From this point of view they may just be right? In the
present climate, lack of technical superiority does not necessarily exclude consumer
desirability does it and the performance of the format has proven adequate for those
that seem to matter in consumer opinion at the moment.
Paul Bamborough
and much more besides. By the time I've mulled it, and a couple of papers I've
been sent, I may actually know enough about this to come to some useful
conclusions....
One question, for now: you say
>The philosophical requirement to avoid parallel data streams (and digital
filters)
>within the DAC means that all (or most) filtering must occur in the analogue
domain.
By 'philosophical', do you mean an *unavoidable* requirement of a DSD DAC, or a
choice that has been made?
p
Paul Frindle wrote in message <383C81A2...@free-online.net>...
Paul Dormer
>One major source of masking can be the program material itself.
>Performances recorded with an audience present have SNR's well under
>60 dB due to crowd noise, and therefore can be used to make LP's
>sound more like CD's. Notice that a certain magazine that advocates
>LP and sometimes releases the same performance on LP and CD tends to
>release recordings of live performances.
To most people the sound of a crowd is not mere noise, it is part of
the recording. I would extend this to the character and colour
imparted by audio equipment in the recording chain.. it's not always
just 'noise' it is often part and parcel of the creation.
Paul Dormer Me...@clara.net
++++++++++++++++++++++++++++++++
Sound Design, Editing, Mastering
Paul Dormer
>Agreed. The "best" current recordings I can find don't come within
>20 dB of exploiting the intrinsic capabiilty of CD technology
>because of recording technique and limitations of commonly-used
>venues. So here is the proposed "cure" - go to a format with
>marginally better intrinsic capabilties. What's wrong with this
>picture?
What is this recording?
Paul Dormer
>> You know, for once I would have to completely agree with you.
>Wouldn't it
>> completely change the tenor of the discussions around here if
>Krooger behaved
>> even *remotely* like a human being?
>
>Why not make a big change in your posting style around here
Still formatting your posts like garbage eh?
Paul Frindle
Paul Bamborough wrote:
> Thank you for all this meaty information, which gives me what I was looking for
> and much more besides. By the time I've mulled it, and a couple of papers I've
> been sent, I may actually know enough about this to come to some useful
> conclusions....
>
> One question, for now: you say
> >The philosophical requirement to avoid parallel data streams (and digital
> filters)
> >within the DAC means that all (or most) filtering must occur in the analogue
> domain.
>
> By 'philosophical', do you mean an *unavoidable* requirement of a DSD DAC, or a
> choice that has been made?
>
> p
>
It is part of the philosphy that is attributed with the 'better sound' that the
signal should stay as a single bit data format. The argument is that the filtering
in the PCM system is a possible cause of quality reduction. However this is a choice
made by the DSD people, you can just as well put back the filter and convert to PCM,
and use a good PCM DAC to get the output. In this way you can avoid the HF noise
problem entirely, avoid analogue filtering components and opamp issues and get the
maximum SNR and lowest distortion in the wanted band. This would work very nicely
and would be arguably as good as the ADC and DAC combination would have been in the
PCM domain regardless that the data has been stored as DSD signal.
RickLGX
The whole demo thing can be very frustrating both for manufacturers and
listeners.
However, I am not naieve enough to believe everyone will feel as I do. I would
still encourage you to try to hear the SACD set up properly in familiar
conditions.
Arny Krüger
"RickLGX" <ric...@aol.com> wrote in message
news:19991125155225...@ng-fu1.aol.com...
If Sony were serious about demonstrating their allegedly superior
technology fair and square, they'd stop remixing ALL the stuff they
had previously offered on CD.
Arny Krüger
"George M. Middius" <Glan...@ipo.net> wrote in message
news:Neg7OND3IQExBQ=Z=qU8GO...@4ax.com...
Greg Singh whimpered:
> > >Every single one of us knows that in reality Krooger is just a
sick fuck,
> > >so what could possibly be the point of trying to reason with
him?
> Looks like you've lost another one, Arnii.
>
Any 10 year old on my block can talk dirty like Singh does. He'd
only be impressive if he'd show proof he's six years old or less.
Arny Krüger
"Paul Dormer" <me...@clara.net> wrote in message
news:385bd63d...@news.clara.net...
> "Arny Krüger" <ar...@flash.net> wrote:
>
> >Agreed. The "best" current recordings I can find don't come
within
> >20 dB of exploiting the intrinsic capabiilty of CD technology
> >because of recording technique and limitations of commonly-used
> >venues. So here is the proposed "cure" - go to a format with
> >marginally better intrinsic capabilties. What's wrong with this
> >picture?
> What is this recording?
For months I've challenged everybody to find a recorded musical
track where the level difference between the loudest and softest
open-mic parts is more than 76 dB, fade-ins, fade-outs and the
artificially-created silence between tracks excluded. Obviously,
things like zero-crossing regions of high amplitude waves would not
count. However, short-term, 1 cycle peaks do count. The lowest-level
samples would have to last for like a few hundred milliseconds so
they would be perceptible by the listener as being quiet and not
masked by adjacent high-level samples.
I've found a fair number of CD tracks that are in the low 70's.
The Ricky Lee Jones track I was working with last year was pretty
good on this account.
Got any candidates among in-print commercial CD recordings?
Peter Corey
Hang in Arnii;
There was only one "fuck" in that post....aside from the author.
Do you suppose it might be the holiday spirit invading RAO ??
*:~._.~:*'*:~._.~:*'*:~._.~:*'*:~._.~:*'*:~._.~:*'*:~._.~:*'*:~
| ^ ^ |
[ 0 = ]
__oOOo-(_)-oOOo______________ _ _____
) )_ _ _ _(_\_____o /_/_ |
)'tis the season to be jolly! ) >-----._/_/__]>
)____________________________ ) `0 |
P.Corey@The Hi-End Haven™
http://home.att.net/~pcor/Pages/hiendhaven.html
George M. Middius
> Well I for one like to see him squirm and be exposed when he resorts to
> weasel tactics.
> Any 10 year old on my block can talk dirty like Singh does. He'd
> only be impressive if he'd show proof he's six years old or less.
Your Boogership, and just admit you like it when the humans
compare you to an animated turd?
George M. Middius
trotsky
Wouldn't 20 years be long enough for you to realize her name is "Rickie Lee Jones"?
trotsky
"Arny Krüger" wrote:
>
> "George M. Middius" <Glan...@ipo.net> wrote in message
> news:Neg7OND3IQExBQ=Z=qU8GO...@4ax.com...
>
> Greg Singh whimpered:
>
> > > >Every single one of us knows that in reality Krooger is just a
> sick fuck,
> > > >so what could possibly be the point of trying to reason with
> him?
>
> > Looks like you've lost another one, Arnii.
> >
>
> Any 10 year old on my block can talk dirty like Singh does. He'd
> only be impressive if he'd show proof he's six years old or less.
OH MY GOD! ARNII IS ADMITTING TO SPENDING 'QUALITY TIME' WITH EVERY TEN YEAR
OLD ON HIS BLOCK! AND I THOUGHT THE NAMBLA RUMORS WERE MERELY SLANCERS! AY CARAMBA!
Dale Goudey
below.
>Paul Frindle wrote
>
>
>>Paul Bamborough wrote:
>
>> Thank you for all this meaty information, which gives me what I was
looking for
>> and much more besides. By the time I've mulled it, and a couple of
papers I've
>> been sent, I may actually know enough about this to come to some
useful
>> conclusions....
>>
>> One question, for now: you say
>> >The philosophical requirement to avoid parallel data streams (and
digital
>> filters)
>> >within the DAC means that all (or most) filtering must occur in
the analogue
>> domain.
>>
>> By 'philosophical', do you mean an *unavoidable* requirement of a
DSD DAC, or a
>> choice that has been made?
>>
>> p
>>
>
>It is part of the philosphy that is attributed with the 'better
sound' that the
>signal should stay as a single bit data format. The argument is that
the filtering
>in the PCM system is a possible cause of quality reduction. However
this is a choice
>made by the DSD people, you can just as well put back the filter and
convert to PCM,
>and use a good PCM DAC to get the output. In this way you can avoid
the HF noise
>problem entirely, avoid analogue filtering components and opamp
issues and get the
>maximum SNR and lowest distortion in the wanted band. This would work
very nicely
>and would be arguably as good as the ADC and DAC combination would
have been in the
>PCM domain regardless that the data has been stored as DSD signal.
Going from DSD to PCM and then to analog does add a processing step
that produces an overall system impulse response that is different in
character from a system that goes directly from DSD to analog. The
PCM route will (typically) produce a sin(x)/x type of impulse
response, while the DSD system (in its pure form) will produce an
impulse response very much like a low-order analog filter. By
reverse-engineering the DSD process I came up with an estimate for its
impulse response. It does indeed look much like the impulse response
obtained from an analog system. If this different impulse response
characteristic affects how the ear/cochlea/neural network system
responds to sound stimulus, then the more natural (in the sense of
occurring in nature) analog filter response characteristic may be an
important thing to preserve. The neural network behavior is extremely
nonlinear, so interpretation/extrapolation of neural net response
using a linear model is not appropriate. And please note that much
neural network research does NOT pertain to genuine biological neural
networks.
References (VERY brief list):
Cochlea:
http://www.sissa.it/bp/Cochlea/
A/D, D/A converters:
Delta-Sigma converters, Theory, Design, and Simulation,
IEEE Press, 1997, ISBN 0-7803-1045-4
Neural Networks:
My references genuine biological neural networks are not recent. Does
anyone have current models/descriptions of neural networks pertaining
to sound perception?
Arny Krüger
> "Arny Krüger" <ar...@flash.net> wrote:
>
> >One major source of masking can be the program material itself.
> >Performances recorded with an audience present have SNR's well
under
> >60 dB due to crowd noise, and therefore can be used to make LP's
> >sound more like CD's. Notice that a certain magazine that
advocates
> >LP and sometimes releases the same performance on LP and CD tends
to
> >release recordings of live performances.
>
> To most people the sound of a crowd is not mere noise, it is part
of
> the recording. I would extend this to the character and colour
> imparted by audio equipment in the recording chain.. it's not
always
> just 'noise' it is often part and parcel of the creation.
I think that's a matter of personal preference.
It remains that crowd noise is pretty good at masking, and if you
want to show off a technically inferior medium, it can be quite
helpful.
Paul Frindle
The impulse response shape has nothing to do with whether the filter is
analogue or digital. The sin(x)/x (like) response you talk about is due to
the phase alignment of the filter in use. It is possible to align phase
correctly in the digital domain but difficult in analogue. But if an
analogue filter had the same phase alignment it would of course look the
same and have the same impulse response. So if you agree that aligning the
phase response at all frequencies is a good idea (you can certainly hear
it if you don't) what does this say about the impulse response of the
other unit you measured? Either that it has insufficient roll-off, or it
is not phase aligned. Both of these situations will lead to sound
degradation either due to phase shift or excess HF noise in the case of
SACD. It is an unavoidable physical reality I'm afraid.
Dale Goudey wrote:
> Again a caution about the expedient conversion from DSD to PCM. See
> below.
>
> >Paul Frindle wrote
> >
> >
> >>Paul Bamborough wrote:
> >
> >> Thank you for all this meaty information, which gives me what I was
> looking for
> >> and much more besides. By the time I've mulled it, and a couple of
> papers I've
> >> been sent, I may actually know enough about this to come to some
> useful
> >> conclusions....
> >>
> >> One question, for now: you say
> >> >The philosophical requirement to avoid parallel data streams (and
> digital
> >> filters)
> >> >within the DAC means that all (or most) filtering must occur in
> the analogue
> >> domain.
> >>
> >> By 'philosophical', do you mean an *unavoidable* requirement of a
> DSD DAC, or a
> >> choice that has been made?
> >>
> >> p
> >>
> >
Dale Goudey
due to the linear phase response. The issue I was addressing is the
"optimality criterion". Should the record/playback chain provide an
impulse response that is (very loosely speaking) non-causal while
preserving the phase response, or should the phase response be
compromised to maintain a "causal" (as found in nature) impulse
response? If we downsample to a lower sample rate we must choose such
a compromise. If we maintain a high sample rate we can minimize BOTH
errors. Both errors are small with DSD due to the high cutoff
frequency employed (100 kHz). Conventional PCM gives us the sin(x)/x
shape. Some may argue whether sin(x)/x is good or bad, I am just
raising the issue since Sony suggests that avoiding sample rate
reduction is a good thing, and converting to low sample rate PCM
essentially destroys what some folks are trying to preserve.
Another option for PCM design is to maintain a high sample rate, use a
digital filter that provides a "reasonable" impulse response, and use
lossless packing to reduce storage requirements. This would preserve
what may be "right" about DSD, while providing the convenience of PCM
for the purposes of signal manipulation (as in mastering). This is an
approach that I have a "preference" for, but only from a philosophical
point of view.
As to whether digital or analog has anything to do with the general
character of the impulse response, I'm not sure why you thought I
implied this. In fact, I used a digital filter implementation to
"model" the analog system (using a high sample rate to maintain model
fidelity).
For whatever reason, I have formed the following preferences (not at
all scientific):
1. DSD (very limited experience, a single demo)
2. 24/96 (much experience at this point, much better than Compact
Disc)
3. LP (sounds like DSD, but suffers from surface noise, etc.)
4. Compact Disc (sounds like Compact Disc)
Given this experience, I am not inclined to discredit the notion that
impulse response shape should not be governed by narrow bandwidth PCM
(though I won't file it away as fact until neural network models tell
us why it should be the case).
>Paul Frindle wrote
Paul Frindle
Thanks for the reply. Some comments in reply.
Dale Goudey wrote:
> Yes, the sin(x)/x shape (the time symmetry property, at least) arises
> due to the linear phase response. The issue I was addressing is the
> "optimality criterion". Should the record/playback chain provide an
> impulse response that is (very loosely speaking) non-causal while
> preserving the phase response, or should the phase response be
> compromised to maintain a "causal" (as found in nature) impulse
> response? If we downsample to a lower sample rate we must choose such
> a compromise. If we maintain a high sample rate we can minimize BOTH
> errors. Both errors are small with DSD due to the high cutoff
> frequency employed (100 kHz). Conventional PCM gives us the sin(x)/x
> shape. Some may argue whether sin(x)/x is good or bad, I am just
> raising the issue since Sony suggests that avoiding sample rate
> reduction is a good thing, and converting to low sample rate PCM
> essentially destroys what some folks are trying to preserve.
Sadly the decison is not possible for a 1bit system at 64fs because of the
existence of HF noise at a constant -6dB ref flat out! You really do have
to have a significant filter I'm afraid, so actually there is no choice.
You are not preserving causality by using an uncorrected filter - you are
destroying it - are you not? I.e. you are deliberately allowing the LF to
propagate before the HF. There is nothing natural about this in my
opinion.
> Another option for PCM design is to maintain a high sample rate, use a
> digital filter that provides a "reasonable" impulse response, and use
> lossless packing to reduce storage requirements. This would preserve
> what may be "right" about DSD, while providing the convenience of PCM
> for the purposes of signal manipulation (as in mastering). This is an
> approach that I have a "preference" for, but only from a philosophical
> point of view.
Yes it is a philosophical rather than proven benefit but at least it can
be done - you are correct in proposing this possibility.
Some more comments:
I am not really sure why you have a problem with impulse responses and why
people believe that the phase corrected impulse response is somehow
'unnatural'? The constant phase impulse response is natural for any
accurate and 'causally correct' band limited system when excited with
signals that have risetimes greater than the passband. The one you prefer
and refer to as 'natural' is in fact in error since it has phase
inaccuracies (i.e. delay proportional to frequency, like a simple R/C
network). Why do you think that these differential delay errors are
somehow causally correct and natural?
It is true that the amount of time a filter spends in this ringing impulse
response is proportional to the roll-off rate and the stop band
attenuation, the phase correction bit just positions the peak equally
within the ringing disturbance. It is also true that in a system where the
stop band is very very high it does not matter if the final roll-off is
phase linear or not since the bit we hear (<20KHz) is unaffected (as in
analogue systems with fast opamps). But if we imagine that for a 1bit
system we do a roll-off at 40KHz or 75KHz selectable as proposed, (100KHz
is not possible without roasting the speakers) with an uncorrected phase
filter, we risk having phase errors at <20KHz that we can hear. The 1bit
system is flawed in this respect since it requires a considerably steep
filter in order to avoid the HF noise energy problem. The noise level
rises at potentially 30dB/octave after the notional 'pass band', this can
start at 20KHz for a 64fs system! If this filter is not phase corrected,
and it produces the impulse response you prefer, there will be audible
differences in the sound quality when comparing in and out in a listening
test. We have found that single order phase errors that result in as
little as 15deg shifts at 20KHz are audible.
It is not reasonable to simply keep increasing the sampling and data rates
to remove phase shift error on uncorrected filters, just because for some
reason we don't like the shape of the natural phase corrected impulse
response. This is not simply not science and has no bearing on the issues
of sound quality. For instance in the days when mixing consoles were
analogue we fought day and night to remove accumulated phase errors due to
all the amps in the signal chain, because we could hear it (even though
the B/W was flat to 100KHz). There were 2 ways to do this; either make the
B/W of the system flat into the MHz region using RF devices and risk
instability and radio demodulation, or correct the phase with extra
networks, which added noise etc.. Of course we never succeeded and they
were never sonically transparent as a result. Thank goodness we can now
make a digital filter that does this without the noise.
As you suggested, a far far better way to produce a reduced preamble
impulse response in the filter is to sample at 96KHz and aim for a B/W of
say 24KHz. At least in this system you can make the choice between B/W and
ringing around 40KHz when you hit it with squarewaves.You can then
roll-off the filter more gently between 24KHz and 40KHz and this will
reduce the time spent in the ringing modes radically. You can then do the
phase correction without producing the philosphically disliked impulse
response you wish to avoid. It will look once again more like an analogue
system's impulse response. From an engineering point of view I would also
prefer this method since it is easier to design. However, none of this
will make the slightest difference to the sound quality of the system
PROVIDED THAT you can keep all the phase errors or filter ringing above
20KHz. And these criteria can be achieved amply at 48KHz sampling rates.
Dale Goudey
Paul Frindle wrote in message
<383F2AD7...@frindle.freeserve.co.uk>...
Happily, the decision IS possible, but yes, that HF noise is real and
can present a problem for some amplifiers. That is why Sony provides
two filter options in their players.
Your number for the HF noise level is in error (significantly).
Published data shows that the shaped noise curve is about 100 dB down
from full scale at 50 kHz in a 1/3 octave bandwidth (at this frequency
the noise power density is at its highest). The noise is real, but
not at all terrifying.
In nature, sounds are generated by vibrating matter whose motions
obey, to a fairly high degree of accuracy, ordinary or partial
differential equations. Many natural impulsive sounds can be modelled
with fair fidelity as impulses filtered through, in electrical
parlance, a lumped parameter system (ordinary differential equations).
Some sounds (like drums) also include distributed parameter behavior
(partial differential equations). That is what I meant by "occuring
in nature". There is no natural phenomenon I am aware of that produce
impulsive sounds that exhibit the sin(x)/x shape. By adding an analog
filter stage after the model for sound generation, we are simply
changing the total filter model by the added polse/zeros. If we
"distort" a natural sound by very slight changes in the model filter
parameters/filter degree, the character of the sound shouldn't change
(though the degree of change is debatable). If we "distort" a natural
sound using a sin(x)/x-like filter, then we have introduced a change
that is not well modelled by natural phenomenon. Of course, whether
this is significant is also debatable.
... Because natural sounds can be accurately modelled by such
"defective" filters. Natural sounds do not exhibit the sin(x)/x
shape. Whether this is important is open to debate, but if a
proponent of a new sound distribution format declares that the brick
wall filters are best avioded, it seems silly to add them back in the
playback chain, which is what you are suggesting.
Yes, a 100 kHz filter is too high a cutoff for playback. I am not
sure where you get your numbers, but your analysis is grossly in
error. After compensation for group delay, a five pole analog filter
with a cutoff of 40 kHz or above will exhibit less than 1 degree of
phase error at 20 kHz (elliptic or Chebyshev designs). I do not
consider this a problem, but I suppose some will argue that it is. I
have not yet assessed the phase response of the Sony SACD players (I
have the necessary analysis tools, but it takes time). I suspect that
the phase errors are much less than 15 degrees at 20k. Even after
both the record and playback equipment is included in the analysis.
>
>It is not reasonable to simply keep increasing the sampling and data
rates
>to remove phase shift error on uncorrected filters, just because for
some
>reason we don't like the shape of the natural phase corrected impulse
>response. This is not simply not science and has no bearing on the
issues
>of sound quality. For instance in the days when mixing consoles were
>analogue we fought day and night to remove accumulated phase errors
due to
>all the amps in the signal chain, because we could hear it (even
though
>the B/W was flat to 100KHz). There were 2 ways to do this; either
make the
>B/W of the system flat into the MHz region using RF devices and risk
>instability and radio demodulation, or correct the phase with extra
>networks, which added noise etc.. Of course we never succeeded and
they
>were never sonically transparent as a result. Thank goodness we can
now
>make a digital filter that does this without the noise.
I am not the one declaring that the impulse response must be a certain
way. I am simply saying that some people suggest so (we have no proof
either way), and to simply defeat the design intent of DSD and
eliminate the possible benefit seems silly. Basically, since you
don't "believe" then you are deciding for others. Is this science?
>As you suggested, a far far better way to produce a reduced preamble
>impulse response in the filter is to sample at 96KHz and aim for a
B/W of
>say 24KHz. At least in this system you can make the choice between
B/W and
>ringing around 40KHz when you hit it with squarewaves.You can then
>roll-off the filter more gently between 24KHz and 40KHz and this will
>reduce the time spent in the ringing modes radically. You can then do
the
>phase correction without producing the philosphically disliked
impulse
>response you wish to avoid. It will look once again more like an
analogue
>system's impulse response. From an engineering point of view I would
also
>prefer this method since it is easier to design. However, none of
this
>will make the slightest difference to the sound quality of the system
>PROVIDED THAT you can keep all the phase errors or filter ringing
above
>20KHz. And these criteria can be achieved amply at 48KHz sampling
rates.
Agreed. Except for the part about necessary and sufficient conditions
for ultimate sound quality. Our science is not that far advanced
regarding sound perception. Our hearing / perception apparatus is not
linear. Sounds we can't hear can (possibly) affect our perceptions of
what we can hear. We haven't proven otherwise.
A nit: I see no harm in going to a higher sample rate than 96k,
though 96k does sound very good to me.
Dale Goudey
Another nit: The name is Goudey, not Goodey.
The saga continues...
>Paul Frindle wrote in message
>Dear Mr Goodey
>
>Thanks for this interesting conversation. I would like to pick up on
some
>of your points. I am beginning to understand where you ar coming
from.
>
.....
>> Your number for the HF noise level is in error (significantly).
>> Published data shows that the shaped noise curve is about 100 dB
down
>> from full scale at 50 kHz in a 1/3 octave bandwidth (at this
frequency
>> the noise power density is at its highest). The noise is real, but
>> not at all terrifying.
>
>I'm afraid that this is not possible unless the signal is also being
>filtered significantly at this frequency. Did the spec show the
frequency
>response when this measurement was made? I cannot comment further on
this
>except to say that there would be no need to provide a filter switch
if
>the native noise from the modulator was as low as you suggest it is.
>
You seem to be confused about the signal transfer function versus the
noise transfer function. I recommend you review linear system theory
and delta-sigma ADC design theory (a bit too much for a newsgroup
post, particularly the equations and graphs needed). An analysis of
the NTF (noise transfer function) should show that the numbers I
quoted are consistent with the signal transfer function cutoff of 40
kHz. The data I referred to showed both the frequency response and
the noise spectrum, for both filter options. The correct mathematical
analysis and the measurements of the hardware are in agreement. The
noise generated on playback IS NOT WHITE, it is SHAPED by the NTF.
Hint: If the STF (signal transfer function) is G, then the NTF is
H=1/(1+G), this for the DSD architecture. G is a low pass filter. H
is a high pass filter (DC gain of G is high). I assume that dither in
introduced at the input to the quantizer (comparator).
>> In nature, sounds are generated by vibrating matter whose motions
>> obey, to a fairly high degree of accuracy, ordinary or partial
>> differential equations......
>
>You are correct in your description of what mostly happens in the
sound
>generation systems when sounds are filtered or generated by
>resilience/mass systems. But we are talking about a reproduction
system.
>We certainly do not want to impose further phase shifts on the
natural
>signal, unless we actually want to modify the sound deliberately.
Whilst
>it is difficult to find a natural equivalent of a phase corrected
filter,
>that does not mean we need to chose uncorrected filters when
reproducing a
>signal with a limited bandwidth system. This is not reasonable. It is
not
>a musical instrument.
You are correct, the record/playback chain would ideally not change
the signal at all. At issue is the set of compromises we are willing
to make (nature and degree). Implicit in my argument was that a VERY
SLIGHT increase in a phase distortion that is ALREADY PRESENT in
nature and associated audio equipment will not change the model
(mathematical description) of the waveform much. Again, less than a
degree of phase error at 20 kHz. Many audio components do not do as
well as this.
>> Yes, a 100 kHz filter is too high a cutoff for playback. I am not
>> sure where you get your numbers, but your analysis is grossly in
>> error. After compensation for group delay, a five pole analog
filter
>> with a cutoff of 40 kHz or above will exhibit less than 1 degree of
>> phase error at 20 kHz (elliptic or Chebyshev designs). I do not
>> consider this a problem, but I suppose some will argue that it is.
I
>> have not yet assessed the phase response of the Sony SACD players
(I
>> have the necessary analysis tools, but it takes time). I suspect
that
>> the phase errors are much less than 15 degrees at 20k. Even after
>> both the record and playback equipment is included in the analysis.
>
>I think your are referring to a 5 pole filter where the -3dB point is
at
>40KHz.The order of the filter at 5 poles with -3dB at 40KHz is not
enough
>to get the noise down to -100dB. You must measure a system and see
what it
>actually does, you will find that I am correct that in the lowest B/W
>setting (useable with modest amps) there will likely be more phase
shift
>than you would like. Remember that the analogue filter alone is not
the
>only filter in the system! The sometimes used transversal FIR can
correct
>for some of the phase lag of the filter as well if done properly. And
of
>course we can make a partially corrected analogue filter, but both
these
>approaches would start producing the impulse responses that are
disliked.
This is nonsense. See above comments. Again, the noise spectrum and
frequency response were measured, confirming current day delta-sigma
design methods as applied to the DSD design.
>
>> I am not the one declaring that the impulse response must be a
certain
>> way. I am simply saying that some people suggest so (we have no
proof
>> either way), and to simply defeat the design intent of DSD and
>> eliminate the possible benefit seems silly. Basically, since you
>> don't "believe" then you are deciding for others. Is this science?
>
>No it is not a question of belief at all, it is fact I'm afraid. The
>impulse response you prefer is an error in reproduction systems, the
phase
>corrected one is correct. The maths works and the sound tests prove
it.
>I've been at this a very long time.
So, you are saying that less than 1 degree of phase error is just
plain WRONG. Just one degree. My goodness. Totally unacceptable.
Well, at least not ideal, we both agree. I believe I understand your
point, at least in principle.
>> . However, none of
>> this
>> >will make the slightest difference to the sound quality of the
system
>> >PROVIDED THAT you can keep all the phase errors or filter ringing
>> above
>> >20KHz. And these criteria can be achieved amply at 48KHz sampling
>> rates.
>>
>> Agreed. Except for the part about necessary and sufficient
conditions
>> for ultimate sound quality. Our science is not that far advanced
>> regarding sound perception. Our hearing / perception apparatus is
not
>> linear. Sounds we can't hear can (possibly) affect our perceptions
of
>> what we can hear. We haven't proven otherwise.
>
>The problem is that we haven't proven it either. The idea that sounds
>above what we can hear can affect our perceptions is a compelling one
I
>agree. And I agree that the ear is a non-linear system followed by a
>corrective processor in the brain. I have tried out these theories
under
>strict controls because I once thought this was possible too. However
I
>have found that the ear itself does indeed cut off and is totally
>insensitive to the presence of stuff above audible range even if it
is
>very loud indeed, to the point where speakers burn out without being
>heard, (as has apparently happened on some early DSD demonstrations).
>The mechanism by which you hear a difference in the presence of
extreme HF
> >20KHz is due to speaker non-linearities which cause harmonics of
the
>frequencies below 20KHz to alias with those above. In other words the
>stuff above 20KHz is exacerbating errors in the reproduction system
and
>making the response in the audible range worse.
>
I am sure I can't hear a thing above 16.5 kHz. I am not certain that
ultrasonic sounds have no affect on me. I understand (now) that you
believe that ultrasonic frequencies do not affect sound perception. I
do not take this as proven. That we disagree yet can converse makes
for an interesting discussion, providing us with an opportunity to
revisit our beliefs, update our technical knowledge, and generally
strive for a better understanding of ourselves and others.
An interesting note on sound perception appeared in the December 1999
issue of Audiophile (page 27). Good entertaining reading. This
controversey will continue for awhile.
>> A nit: I see no harm in going to a higher sample rate than 96k,
>> though 96k does sound very good to me.
>
>As a designer I still reckon that sampling at 96KHz and filtering
gently
>from around 24KHz will give the best situation, in that such a system
>would be much less sensitive to cheap consumer design short cuts
without
>being a complete waste of data bandwidth. An optimal compromise if
you
>will.
24/96 does sound good, I agree. To me it is significantly better than
16/44.1, though specifically WHY this is so is not well understood.
This fact makes me inclined to apply a bit of overkill if possible to
"be certain" that we haven't compromised the sound playback
experience. Proof of optimality or lack thereof can be deferred.
Regards, Dale Goudey
(please note the spelling)
Paul Frindle
Thanks for this interesting conversation. I would like to pick up on some
of your points. I am beginning to understand where you ar coming from.
Dale Goudey wrote:
> Mr Frindle, the discussion is still alive.
>
> Paul Frindle wrote in message
> <383F2AD7...@frindle.freeserve.co.uk>...
> >Dear Mr Goodey.
> >Thanks for the reply. Some comments in reply.
> >
> >Dale Goudey wrote:
> >
> Happily, the decision IS possible, but yes, that HF noise is real and
> can present a problem for some amplifiers. That is why Sony provides
> two filter options in their players.
> Your number for the HF noise level is in error (significantly).
> Published data shows that the shaped noise curve is about 100 dB down
> from full scale at 50 kHz in a 1/3 octave bandwidth (at this frequency
> the noise power density is at its highest). The noise is real, but
> not at all terrifying.
I'm afraid that this is not possible unless the signal is also being
filtered significantly at this frequency. Did the spec show the frequency
response when this measurement was made? I cannot comment further on this
except to say that there would be no need to provide a filter switch if
the native noise from the modulator was as low as you suggest it is.
> In nature, sounds are generated by vibrating matter whose motions
> obey, to a fairly high degree of accuracy, ordinary or partial
> differential equations. Many natural impulsive sounds can be modelled
> with fair fidelity as impulses filtered through, in electrical
> parlance, a lumped parameter system (ordinary differential equations).
> Some sounds (like drums) also include distributed parameter behavior
> (partial differential equations). That is what I meant by "occuring
> in nature". There is no natural phenomenon I am aware of that produce
> impulsive sounds that exhibit the sin(x)/x shape. By adding an analog
> filter stage after the model for sound generation, we are simply
> changing the total filter model by the added polse/zeros. If we
> "distort" a natural sound by very slight changes in the model filter
> parameters/filter degree, the character of the sound shouldn't change
> (though the degree of change is debatable). If we "distort" a natural
> sound using a sin(x)/x-like filter, then we have introduced a change
> that is not well modelled by natural phenomenon. Of course, whether
> this is significant is also debatable.
>
You are correct in your description of what mostly happens in the sound
generation systems when sounds are filtered or generated by
resilience/mass systems. But we are talking about a reproduction system.
We certainly do not want to impose further phase shifts on the natural
signal, unless we actually want to modify the sound deliberately. Whilst
it is difficult to find a natural equivalent of a phase corrected filter,
that does not mean we need to chose uncorrected filters when reproducing a
signal with a limited bandwidth system. This is not reasonable. It is not
a musical instrument.
> Why do you think that these differential delay errors are
> >somehow causally correct and natural?
>
> ... Because natural sounds can be accurately modelled by such
> "defective" filters. Natural sounds do not exhibit the sin(x)/x
> shape. Whether this is important is open to debate, but if a
> proponent of a new sound distribution format declares that the brick
> wall filters are best avioded, it seems silly to add them back in the
> playback chain, which is what you are suggesting.
As my last comment, there is no reason why a reproduction system should
take on the impulse response of a natural sound generation system. We are
trying very hard to NOT make a sound changing system. This is not a sound
modelling system (I hope). Rather than accept a view of things because it
seems a natural and appealing argument, we should look at the science.
There are many situations in the technical world where concepts abound
that are not in any way connected to natural events around us. This does
not exclude their validity or usefulness amongst engineers.
> Yes, a 100 kHz filter is too high a cutoff for playback. I am not
> sure where you get your numbers, but your analysis is grossly in
> error. After compensation for group delay, a five pole analog filter
> with a cutoff of 40 kHz or above will exhibit less than 1 degree of
> phase error at 20 kHz (elliptic or Chebyshev designs). I do not
> consider this a problem, but I suppose some will argue that it is. I
> have not yet assessed the phase response of the Sony SACD players (I
> have the necessary analysis tools, but it takes time). I suspect that
> the phase errors are much less than 15 degrees at 20k. Even after
> both the record and playback equipment is included in the analysis.
I think your are referring to a 5 pole filter where the -3dB point is at
40KHz.The order of the filter at 5 poles with -3dB at 40KHz is not enough
to get the noise down to -100dB. You must measure a system and see what it
actually does, you will find that I am correct that in the lowest B/W
setting (useable with modest amps) there will likely be more phase shift
than you would like. Remember that the analogue filter alone is not the
only filter in the system! The sometimes used transversal FIR can correct
for some of the phase lag of the filter as well if done properly. And of
course we can make a partially corrected analogue filter, but both these
approaches would start producing the impulse responses that are disliked.
> I am not the one declaring that the impulse response must be a certain
> way. I am simply saying that some people suggest so (we have no proof
> either way), and to simply defeat the design intent of DSD and
> eliminate the possible benefit seems silly. Basically, since you
> don't "believe" then you are deciding for others. Is this science?
No it is not a question of belief at all, it is fact I'm afraid. The
impulse response you prefer is an error in reproduction systems, the phase
corrected one is correct. The maths works and the sound tests prove it.
I've been at this a very long time.
> . However, none of
> this
> >will make the slightest difference to the sound quality of the system
> >PROVIDED THAT you can keep all the phase errors or filter ringing
> above
> >20KHz. And these criteria can be achieved amply at 48KHz sampling
> rates.
>
> Agreed. Except for the part about necessary and sufficient conditions
> for ultimate sound quality. Our science is not that far advanced
> regarding sound perception. Our hearing / perception apparatus is not
> linear. Sounds we can't hear can (possibly) affect our perceptions of
> what we can hear. We haven't proven otherwise.
The problem is that we haven't proven it either. The idea that sounds
above what we can hear can affect our perceptions is a compelling one I
agree. And I agree that the ear is a non-linear system followed by a
corrective processor in the brain. I have tried out these theories under
strict controls because I once thought this was possible too. However I
have found that the ear itself does indeed cut off and is totally
insensitive to the presence of stuff above audible range even if it is
very loud indeed, to the point where speakers burn out without being
heard, (as has apparently happened on some early DSD demonstrations).
The mechanism by which you hear a difference in the presence of extreme HF
>20KHz is due to speaker non-linearities which cause harmonics of the
frequencies below 20KHz to alias with those above. In other words the
stuff above 20KHz is exacerbating errors in the reproduction system and
making the response in the audible range worse.
> A nit: I see no harm in going to a higher sample rate than 96k,
> though 96k does sound very good to me.
As a designer I still reckon that sampling at 96KHz and filtering gently
from around 24KHz will give the best situation, in that such a system
would be much less sensitive to cheap consumer design short cuts without
being a complete waste of data bandwidth. An optimal compromise if you
will.
Many thanks
Arny Krüger
"Dale Goudey" <gou...@earthlink.net> wrote in message
news:81p5ou$3ah$1...@ash.prod.itd.earthlink.net...
> Agreed. Except for the part about necessary and sufficient
conditions
> for ultimate sound quality.
> Our science is not that far advanced regarding sound perception.
OSAF
>Our hearing / perception apparatus is not
> linear.
Agreed.
> Sounds we can't hear can (possibly) affect our perceptions of
> what we can hear. We haven't proven otherwise.
It's a negative hypothesis, so we never will.
Why not mosey over to my freeware web site www.pcabx.com and
download some of the samples that are brick-wall filtered at 18 KHz
(but for which the reference versions you should compare them to
have significant content above 18 KHz). Then just show us how you
can hear the difference between the same program material filtered
and unfiltered, with a reliable listening test using the ABX
Comparator software also freely downloadable from that source.
George M. Middius
>> Agreed. Except for the part about necessary and sufficient
>> conditions for ultimate sound quality.
>> Our science is not that far advanced regarding sound perception.
> OSAF
You're such a dimwit, Arnii. How much humiliation do you
have to endure before you realize you're only a slug
compared to any real engineer?
>> Sounds we can't hear can (possibly) affect our perceptions of what
>> we can hear. We haven't proven otherwise.
> It's a negative hypothesis, so we never will.
Put your dick back in your knickers, you sad sack of shit.
> Why not mosey over to my snotware web site
WARNING to all humans! Don't load that page if you harbor
any esthetic sense at all. It's the worst disaster of a Web
site ever designed by an "adult."
I'll bet mental health professionals throughout the English-
speaking world are contemplating "web design" as a therapy
for their confined patients. Arnii, you may yet make a
worthwhile contribution to science. ;-)
[This post reformatted by Paul Bamborough Engineering,
laboring tirelessly to de-Kroogerize Usenet]
George M. Middius
Paul Frindle
Many apologies for my typing error regarding your name. This was
unintentional and in no way implies any disrespect. Thanks again for
another interesting post. I will try to answer some of your points if I
can:
Dale Goudey wrote:
Yes the noise is shaped this is why it is called a noise shaping loop. The
operation of the noise shaping loop is basically an unstable modulator
which oscillates in a random pattern with a 1 bit digital output that has
a SNR of 6dB. The low pass filter in the loop provides high gain at low
frequencies which pushes the modulation noise into the frequencies above
the turnover of the filter. Therefore we get a noise spectrum where there
is a band in the LF region where there is very much less noise energy.
This is the passband (wanted frequencies) of the system, In current
designs this can be as low as -108dB or so up to 20KHz with a 64fs clock
speed. When all frequencies above 20KHz are filtered out, as is the case
in a normal ADC like the Crystal 5390, we can get a SNR of around -108dB
etc..
Now if we look at the spectrum of the idling noise in the 1bit signal
(i.e. DSD) we see that the rise of the noise above the loop filter
turnover is roughly at the rate of the order of the filter. In other words
the filter's efficiency at reducing noise in the loop falls of at a rate
defined by the filter order. In a 5th order loop therefore the noise will
rise with a 5th order slope, i.e 30dB/octave.
2 things to consider here, If we have a modulator in which the loop filter
is set higher than 20KHz the passband noise will increase since the total
noise of a 1bit signal can only legally be at -6dB to maintain stability
so there is less bandwidth available to accomodate the passband signal.
Therefore we would expect at least another 3dB of passband noise (actually
I think 4.5dB is the correct figure, 3dB for more passband +1.5dB for
loss of total noise bandwidth ). So the SNR of the system even if
perfectly filtered could only be at around -104dB or so. There is a trade
off therefore between the loop filter turnover frequency and order,
passband noise and out of band noise. This is why the IC manufacturers are
now employing multibit loop modulators in the latest 96KHz designs
(incidentally this concept was invented by Bob Adams in the 1980s who
produced the first audio DSM ADC using a mulitbit loop).
In a 1bit system the aim is to keep the noise out of band down to levels
that do not destroy speakers etc. As we can see that even with a 5th order
loop filter set at 40KHz the 80KHz noise will be at around -70dB or so
best case even if we had an 80KHz brickwall filter. .
Of course a DSD system does not use a brickwall filter at any frequency
since this is counter the philosophy. So in reality we use a 20KHz noise
shaping loop filter for the ADC to maximise SNR. This gives noise at
around -78dB at 40KHz and -48dB at 80KHz. Employ a DAC filter of around 8
poles (not all at the same frequency to minimise phase and opamp slew rate
errors, using a VSVR design to minimise roll-off at 20KHz) set at around
60KHz for -3dB response. This in practice keeps the noise energy down to
around -60dB peaking at around 55KHz or so.
Under this scheme it is clearly the case that -100dB SNR at 100KHz is out
of the question, even for a 40KHz target response.
> Hint: If the STF (signal transfer function) is G, then the NTF is
> H=1/(1+G), this for the DSD architecture. G is a low pass filter. H
> is a high pass filter (DC gain of G is high). I assume that dither in
> introduced at the input to the quantizer (comparator).
See above. I submit that your model is incomplete.
Again get a practical DSD system and measure it. For instance see my
former example of a practical system and see what the total phase shift
would be in an uncorrected system.
I am sorry you think this is nonsense! You have personally measured such a
system I understand, so I can only stand corrected? I would be interested
in knowing the manufacturer and model reference number of the units you
measured?
> >
> >> I am not the one declaring that the impulse response must be a
> certain
> >> way. I am simply saying that some people suggest so (we have no
> proof
> >> either way), and to simply defeat the design intent of DSD and
> >> eliminate the possible benefit seems silly. Basically, since you
> >> don't "believe" then you are deciding for others. Is this science?
> >
> >No it is not a question of belief at all, it is fact I'm afraid. The
> >impulse response you prefer is an error in reproduction systems, the
> phase
> >corrected one is correct. The maths works and the sound tests prove
> it.
> >I've been at this a very long time.
>
> So, you are saying that less than 1 degree of phase error is just
> plain WRONG. Just one degree. My goodness. Totally unacceptable.
> Well, at least not ideal, we both agree. I believe I understand your
> point, at least in principle.
No, I said that 15deg seemed to be the audible threshold, if you remember.
My assertion is that the 1deg figure you quoted for an SACD player without
phase correction is entirely impossible. With phase correction, the
impulse response will begin to resemble the sin(x)/x function you dislike.
I am also saying that the designs employ a transversal filter (i.e
multiple delayed and combined DACs) that 'could' be used as a phase
pre-corrector if the impulse response stopped being an issue. This last
statement is a hint.
Yes I agree it would be wrong to present our (however) extensive tests as
rigorous scientific proof, I would never be that presumptuous. I am as
ever open to the possibility that energy above 20KHz may affect our
perception of sound even if we can't hear it. It's just that I can find no
mechanism that illustrates this under controlled conditions except in the
presence of considerable speaker non-linearities as I stated. I have tried
everything I can think of with music and deliberately generated signals
with various phase relationships and so far I can find no other effect
other than LS (and in some case amplifier) errors producing frequencies
that were not there in the original programme. Yes sounding different, but
no - not part of the input signal.
I would be very interested in any information that might inspire tests I
haven't done or thought of yet?.
> An interesting note on sound perception appeared in the December 1999
> issue of Audiophile (page 27). Good entertaining reading. This
> controversey will continue for awhile.
I will try to find this article - thanks.
> >> A nit: I see no harm in going to a higher sample rate than 96k,
> >> though 96k does sound very good to me.
> >
> >As a designer I still reckon that sampling at 96KHz and filtering
> gently
> >from around 24KHz will give the best situation, in that such a system
> >would be much less sensitive to cheap consumer design short cuts
> without
> >being a complete waste of data bandwidth. An optimal compromise if
> you
> >will.
>
> 24/96 does sound good, I agree. To me it is significantly better than
> 16/44.1, though specifically WHY this is so is not well understood.
> This fact makes me inclined to apply a bit of overkill if possible to
> "be certain" that we haven't compromised the sound playback
> experience. Proof of optimality or lack thereof can be deferred.
Having designed, listened and measured such systems for more than 12 years
now, to this day (hint!!). My humble opinion is that you are hearing
errors in the converters at 44.1Khz which are possibly reduced at 88.2 and
96KHz since they are variously moved out of band.
A list of possible causes of audible error in systems I have witnessed:
Absent or incorrect dithering in truncation systems (i.e between 24bit
processing and 16bit media, happens most often in mastering situations).
Differential non-linearity distortion (especially in parallel converters).
Crosstalk between digital circuits and the analogue summing and output
stages.
Insufficient stopband attenuation causing aliassing, sometimes referred to
as timing quantisation (paritcularly in cheap systems using parallel
DACs).
Timing jitter at 50/60Hz line, programme or display multiplexing
frequencies (seen this in some very expensive players).
Uncorrected phase reponses (often due to analogue post filtering in
oversampled DACs).
Frequency response ripple or not flat to within +/- 0.2dB.
Noise in the presence of and modulated by signal (particularly in so
called bit stream or 1bit PCM DACs).
Differential delay between L and R channels, Not necessarily in whole
samples (can occur in oversampled designs).
Signal modulated converter idle tones (particularly in bit stream PCM
designs).
All these can be heard and measured when they occur, so nothing magic
about it at all. There are of course more less common ones.
> Regards, Dale Goudey
> (please note the spelling)
Duly noted - my apologies. I hope you will understand, that if I cannot go
into even more detail about these matters, it is not because the
conversation is in any way uninteresting!
Thanks
Paul Dormer
>> >Agreed. The "best" current recordings I can find don't come
>within
>> >20 dB of exploiting the intrinsic capabiilty of CD technology
>> >because of recording technique and limitations of commonly-used
>> >venues. So here is the proposed "cure" - go to a format with
>> >marginally better intrinsic capabilties. What's wrong with this
>> >picture?
>
>> What is this recording?
>
>For months I've challenged everybody to find a recorded musical
>track where the level difference between the loudest and softest
>open-mic parts is more than 76 dB, fade-ins, fade-outs and the
>artificially-created silence between tracks excluded. Obviously,
>things like zero-crossing regions of high amplitude waves would not
>count. However, short-term, 1 cycle peaks do count. The lowest-level
>samples would have to last for like a few hundred milliseconds so
>they would be perceptible by the listener as being quiet and not
>masked by adjacent high-level samples.
In other words you're trying to seek out an extremely obscure type of
recording? No wonder you're having difficulty!
As I've said before, I don't think it's fair to say dynamic range is
lower due to microphone noise, ambience, crowd noise etc.
Technically, maybe.. but these noises are part and parcel of the
recording itself, they are colourations which constitute aesthetic
elements that engineers and producers have arrived at through
decisions during the recording process itself. If one microphone is
chosen over another during recording, because of it's characteristic
self-noise or euphonic distortion, then that self-noise or distortion
should be reproducable as part of the artistic creation. Even then,
there always the POTENTIAL for recordings to be made with full CD
dynamic range specification, should anyone desire to listen to bursts
of silence.. and personally I have made recordings which don't involve
microphones at all, and do meet the full dynamic range specification
of CD so your point is moot. One track I worked on has a break of a
few hundred milliseconds.. digital silence.
>I've found a fair number of CD tracks that are in the low 70's.
>
>The Ricky Lee Jones track I was working with last year was pretty
>good on this account.
Probably used a noise gate.
>Got any candidates among in-print commercial CD recordings?
I doubt I possess any recordings that use the sort of minimalist
microphone techniques that are necessary to arrive at such high
*measurable* dynamic ranges, and that also contain obvious pauses in
the material with the microphone channels left wide open.. that really
is a bizarre requirement. The music I enjoy typically has high levels
of colouration, distortion or noise built in.. eg 70's dub recordings.
I want this noise to be reproduced as faithfully as possible..
therefore I require media and equipment of high dynamic range and
signal to noise ratio. Get it?
I suggest you investigate measuring recordings of John Cages "4'33"..
if that's what gives you a hard on.
Arny Krüger
No, you are missing my point, and that is that the effective dynamic
range of commercial recordings is about 20 dB worse than what CD can
do, technically.
> As I've said before, I don't think it's fair to say dynamic range
is
> lower due to microphone noise, ambience, crowd noise etc.
That's your preference. Enjoy!
My preference is to take a more analytical and critical approach.
My experiments show that is that there is no need to reproduce noise
with lots of bits. Bit-reduced noise sounds an awful lot like
non-bit-reduced noise until the bit reduction becomes extreme, and
even then reliably hearing a difference can be tough. In fact, of
all the kinds of "program material" there is, noise seems to be the
most tolerant of being reproduced with a minimial number of bits.
Furthermore when we are talking noise that is 70 dB down, its not
like the listener's ear is at maximum acuity.
> Technically, maybe..
Thank you!
> but these noises are part and parcel of the
> recording itself, they are colourations which constitute aesthetic
> elements that engineers and producers have arrived at through
> decisions during the recording process itself.
In a few cases, yes. In general, background noise at about -60 dB
and below is just something that "is" and it is usually accepted. A
recording with 58 dB SNR will be perceived as being "noisy" and the
background noise, whether natural or artificial can be distracting.
I agree that noise that is like 60 dB down is going to be audible
and reproducing it with some reasonable degree of faithfulness is a
good goal. Now let's see what might be an acceptable SNR for
reproducing -58 or -60 dB dB noise. Well, some people seem to like
vinyl and that often has a basic SNR of about 38 dB. So lets go for
a system noise floor that is about 38 dB below -58 dB. Why that
is -96 dB and here we are back at 16 bits!
A recording with 72 dB SNR will generally be perceived as being
"clean" and the background noise will be generally easy to ignore
and/or will hardly be perceived at all. Any who doubt this need only
listen to the bit-reduced samples that are available for free
download at www.pcabx.com. Thusfar, I've not heard anybody claim
reliable detection below 14 bits.
> If one microphone is
> chosen over another during recording, because of it's
characteristic
> self-noise or euphonic distortion, then that self-noise or
distortion
> should be reproducable as part of the artistic creation.
OK, so we reproduce it (-72 dB noise) with >20 dB to the noise floor
of the basic 16/44 system. Where is the beef?
> Even then, there always the POTENTIAL for recordings to be made
with full CD
> dynamic range specification, should anyone desire to listen to
bursts
> of silence.. and personally I have made recordings which don't
involve
> microphones at all, and do meet the full dynamic range
specification
> of CD so your point is moot. One track I worked on has a break of
a
> few hundred milliseconds.. digital silence.
16 bits reproduces digital silence as digital silence - that means
no output at all. Some of my analytical tools call it "-999 dB" OK,
there is some noise in the follow-on analog stages. But that's not
digital noise, it is analog noise.
> >I've found a fair number of CD tracks that are in the low 70's.
> >
> >The Ricky Lee Jones track I was working with last year was pretty
> >good on this account.
> Probably used a noise gate.
I dunno. If that was done, why is the noise as high as it turns out
to be?
> >Got any candidates among in-print commercial CD recordings?
> I doubt I possess any recordings that use the sort of minimalist
> microphone techniques that are necessary to arrive at such high
> *measurable* dynamic ranges, and that also contain obvious pauses
in
> the material with the microphone channels left wide open.. that
really
> is a bizarre requirement.
It just goes to show that normal musical recordings aren't even
going to have 70 dB of dynamic range. Works for me!
>The music I enjoy typically has high levels
> of colouration, distortion or noise built in.. eg 70's dub
recordings.
"Rickie Lee James" was recorded in analog in 1976, I believe. A
really good 2-track high speed analog tape system can perform this
well.
> I want this noise to be reproduced as faithfully as possible..
So how many bits does it take to provide audibly indistinguishable
reproduction of noise that is 72 dB down below peak levels. Remember
that the instantaneous SNR capacity of the human ear is about -65
dB. Much of the time -72 dB noise won't be reliably perceived at
all.
> therefore I require media and equipment of high dynamic range and
> signal to noise ratio. Get it?
If 72 dB is as you claim a "bizarre" requirement, lets go for say, a
basic system noise floor that is 20 dB below that. So now we are 20
dB below "bizarre". Seems like it -92 dB noise should be enough,
right?
> I suggest you investigate measuring recordings of John Cages
"4'33"..
> if that's what gives you a hard on.
Are you trying to be a little obtuse here?
You succeeded. ;-(
Dale Goudey
Our thread has been interesting, but we now seem to be not quite
addressing the issues I at first tried to raise or address. I will
restrict comments to just a few points this time (and edit out some).
Just clarifications or requests for such this time.
>Paul Frindle wrote in message
>Dear Mr Goudey.
>
>...
>> >Paul Frindle wrote in message
>> >...
>> >Thanks for this interesting conversation. I would like to pick up
on
>> some
>> >of your points. I am beginning to understand where you ar coming
>> from.
>> >
>> .....
>> >> Your number for the HF noise level is in error (significantly).
>> >> Published data shows that the shaped noise curve is about 100 dB
>> down
>> >> from full scale at 50 kHz in a 1/3 octave bandwidth (at this
>> frequency
>> >> the noise power density is at its highest). The noise is real,
but
>> >> not at all terrifying.
Well, the above quoted number is for "digital silence". Introducing a
signal can increase the ultrasonic noise power density considerably.
Yes, that isn't a good thing. A compromise that can be inappropriate
for some systems (which may need a sharper high frequency cutoff).
Yes, SACD is making some non-mainstream compromises. I wouldn't
design a system like that (but it sure sounds good).
>.............
>Now if we look at the spectrum of the idling noise in the 1bit signal
>(i.e. DSD) we see that the rise of the noise above the loop filter
>turnover is roughly at the rate of the order of the filter. In other
words
>the filter's efficiency at reducing noise in the loop falls of at a
rate
>defined by the filter order. In a 5th order loop therefore the noise
will
>rise with a 5th order slope, i.e 30dB/octave.
>
>................
>Under this scheme it is clearly the case that -100dB SNR at 100KHz is
out
>of the question, even for a 40KHz target response.
Check out the measurements/graphs in the November 1999 issue of
Stereophile. The ultrasonic power spectral density changes with
signal content (significantly). Does your analysis show this? Does
your understanding fit their measurement results?
>
>> Hint: If the STF (signal transfer function) is G, then the NTF is
>> H=1/(1+G), this for the DSD architecture. G is a low pass filter.
H
>> is a high pass filter (DC gain of G is high). I assume that dither
in
>> introduced at the input to the quantizer (comparator).
>
>See above. I submit that your model is incomplete.
>
Well, actually I did have an error. The term G in the above is the
loop gain. The STF is then G/(1+G). And yes, the STF/NTF formulas
are useless (almost) by themselves. Simulation and hardware testing
is required. The formulas are for insight only (as is most
simulation).
>>......
>
>Again get a practical DSD system and measure it. For instance see my
>former example of a practical system and see what the total phase
shift
>would be in an uncorrected system.
>
It would be nice to see a comprehensive set of measurements of the
SACD system, along with a simluation that matches well with the
measurements. I may eventually get around to some of that (simulation
to match a small set of test results).
.....................
I am very puzzled by your statement that the phase error for a 5-pole
filter with a 40 kHz cutoff cannot be on the order of one degree
(without phase compensation). Or are you referring to the SACD player
specifically? A 5-pole Chebyshev filter, for example, with very small
passband ripple and a 40 kHz cutoff will give us a phase error of
about 1 degree. Since Chebyshev filter design and analysis is so
elementary, this should not be difficult to confirm. (of course, if
group delay is not accounted for then the phase error would seem to be
more like 70 degrees, maybe that is what you are referring to?)
>..............
Yes, higher sample rates will give us better sound one way or another.
Actually, there are not many A/D converters that obey the nyquist
limit (the specs look better if the nyquist limit is violated, I
suppose). Theory and practice are quite diverged here. The higher
sample rates will suffer less from the aliasing (aliased signals being
placed "beyond audibility").
Regards,
Dale Goudey
Paul Frindle
Thanks for this post. I will just pick up on some points but I must be
brief.
Dale Goudey wrote:
> Well, the above quoted number is for "digital silence". Introducing a
> signal can increase the ultrasonic noise power density considerably.
> Yes, that isn't a good thing. A compromise that can be inappropriate
> for some systems (which may need a sharper high frequency cutoff).
> Yes, SACD is making some non-mainstream compromises. I wouldn't
> design a system like that (but it sure sounds good).
This is interesting since there is no real concept of digital silence in a
1bit system. If you have a modulator connected that is producing scarlet
book DSD then the noise is a permanent -6dB ref flat out to nyquist all
the time.Now, there is a bodge (in my opinion) that exists that tries to
say that a DSD silence can consist of a generated signal that repeats 1's
and 0's in order to provide a signal that averages to mid point DC. This
of course gives a much much better answer for a DAC with whatever filter
in is use. In my view this is discreditable as this is not real
measurement and is akin to the old trick of specifying the noise of a PCM
converter with no changing codes at the input.
> >.............
> >Now if we look at the spectrum of the idling noise in the 1bit signal
> >(i.e. DSD) we see that the rise of the noise above the loop filter
> >turnover is roughly at the rate of the order of the filter. In other
> words
> >the filter's efficiency at reducing noise in the loop falls of at a
> rate
> >defined by the filter order. In a 5th order loop therefore the noise
> will
> >rise with a 5th order slope, i.e 30dB/octave.
> >
> >................
> >Under this scheme it is clearly the case that -100dB SNR at 100KHz is
> out
> >of the question, even for a 40KHz target response.
>
> Check out the measurements/graphs in the November 1999 issue of
> Stereophile. The ultrasonic power spectral density changes with
> signal content (significantly). Does your analysis show this? Does
> your understanding fit their measurement results?
No, this should not happen to any great degree in theory unless we are
seeing the harmonic distortion products of the converter aliassing with
the noise above the passband. But then that is probably what we ARE
seeing? This is another potential error mode of a 1bit converter whether
DSD or PCM.
> >> Hint: If the STF (signal transfer function) is G, then the NTF is
> >> H=1/(1+G), this for the DSD architecture. G is a low pass filter.
> H
> >> is a high pass filter (DC gain of G is high). I assume that dither
> in
> >> introduced at the input to the quantizer (comparator).
> >
> >See above. I submit that your model is incomplete.
> >
>
> Well, actually I did have an error. The term G in the above is the
> loop gain. The STF is then G/(1+G). And yes, the STF/NTF formulas
> are useless (almost) by themselves. Simulation and hardware testing
> is required. The formulas are for insight only (as is most
> simulation).
>
> >>......
> >
> >Again get a practical DSD system and measure it. For instance see my
> >former example of a practical system and see what the total phase
> shift
> >would be in an uncorrected system.
> >
>
> It would be nice to see a comprehensive set of measurements of the
> SACD system, along with a simluation that matches well with the
> measurements. I may eventually get around to some of that (simulation
> to match a small set of test results).
>
Yes I agree that this would be the best idea. As you may have gathered, I
have had the privilege of measuring such devices. But that is not to say
that others are not better than those I have had access to.
> I am very puzzled by your statement that the phase error for a 5-pole
> filter with a 40 kHz cutoff cannot be on the order of one degree
> (without phase compensation). Or are you referring to the SACD player
> specifically? A 5-pole Chebyshev filter, for example, with very small
> passband ripple and a 40 kHz cutoff will give us a phase error of
> about 1 degree. Since Chebyshev filter design and analysis is so
> elementary, this should not be difficult to confirm. (of course, if
> group delay is not accounted for then the phase error would seem to be
> more like 70 degrees, maybe that is what you are referring to?)
I am referring to the phase difference between the lowest and the highest
frequencies of interest. In a non-sampled filter the lowest frequency
would have virtually no delay. Therefore the existence of any relative
delay in the HF would indicate uncorrected phase shift. But yes I am aware
that you can make a filter with 5 poles that rolls off at 40KHz with only
1deg error at 20KHz. I am saying that this is not necessarily what is in
the players if they produce no vestige of the sin(x)/x impulse response.
> >
> >A list of possible causes of audible error in systems I have
> witnessed:
> >
> >..............
>
> Yes, higher sample rates will give us better sound one way or another.
> Actually, there are not many A/D converters that obey the nyquist
> limit (the specs look better if the nyquist limit is violated, I
> suppose). Theory and practice are quite diverged here. The higher
> sample rates will suffer less from the aliasing (aliased signals being
> placed "beyond audibility").
>
> Regards,
> Dale Goudey
I am glad that we agree on this point, at least leaving open the
possibility that it may not be specifically the increase in the sampling
rate that gives us theoretically better sound. That is the crux of my
whole argument and applies equally to DVD-A etc..
Regards
Paul
Paul Dormer
>> >For months I've challenged everybody to find a recorded musical
>> >track where the level difference between the loudest and softest
>> >open-mic parts is more than 76 dB, fade-ins, fade-outs and the
>> >artificially-created silence between tracks excluded. Obviously,
>> >things like zero-crossing regions of high amplitude waves would
>not
>> >count. However, short-term, 1 cycle peaks do count. The
>lowest-level
>> >samples would have to last for like a few hundred milliseconds so
>> >they would be perceptible by the listener as being quiet and not
>> >masked by adjacent high-level samples.
>
>> In other words you're trying to seek out an extremely obscure type
>of
>> recording? No wonder you're having difficulty!
>
>No, you are missing my point, and that is that the effective dynamic
>range of commercial recordings is about 20 dB worse than what CD can
>do, technically.
>
>> As I've said before, I don't think it's fair to say dynamic range
>is
>> lower due to microphone noise, ambience, crowd noise etc.
>
>That's your preference. Enjoy!
I don't agree, crowd noise and ambience are fundamental aspects to a
signal.
>My preference is to take a more analytical and critical approach.
Surprise me.
>My experiments show that is that there is no need to reproduce noise
>with lots of bits. Bit-reduced noise sounds an awful lot like
>non-bit-reduced noise until the bit reduction becomes extreme, and
>even then reliably hearing a difference can be tough. In fact, of
>all the kinds of "program material" there is, noise seems to be the
>most tolerant of being reproduced with a minimial number of bits.
>Furthermore when we are talking noise that is 70 dB down, its not
>like the listener's ear is at maximum acuity.
Hmm.. so you say. Noise is obviously less distinguishable than more
clearly defined sounds.. but not to the extent you suggest. What
about dither?
>> Technically, maybe..
>
>Thank you!
OK.
>> but these noises are part and parcel of the
>> recording itself, they are colourations which constitute aesthetic
>> elements that engineers and producers have arrived at through
>> decisions during the recording process itself.
>
>In a few cases, yes.
More often than that. Recording persons get very attached to certain
pieces of equipment due to their perceived colourations (or lack of
them). Choosing the right mic pre-amp or desk can be a big deal.
>In general, background noise at about -60 dB
>and below is just something that "is" and it is usually accepted.
Who are you speaking on behalf of here? I find noise at -60dB
questionable, generally speaking.
>A
>recording with 58 dB SNR will be perceived as being "noisy" and the
>background noise, whether natural or artificial can be distracting.
Well, it can depend somewhat on how the noise manifests itself.
>I agree that noise that is like 60 dB down is going to be audible
>and reproducing it with some reasonable degree of faithfulness is a
>good goal. Now let's see what might be an acceptable SNR for
>reproducing -58 or -60 dB dB noise. Well, some people seem to like
>vinyl and that often has a basic SNR of about 38 dB. So lets go for
>a system noise floor that is about 38 dB below -58 dB. Why that
>is -96 dB and here we are back at 16 bits!
Huh? What are you trying to say? You're not making sense.
>A recording with 72 dB SNR will generally be perceived as being
>"clean" and the background noise will be generally easy to ignore
>and/or will hardly be perceived at all. Any who doubt this need only
>listen to the bit-reduced samples that are available for free
>download at www.pcabx.com. Thusfar, I've not heard anybody claim
>reliable detection below 14 bits.
Above, you mean. I've managed 15 bits, but not always reliably. Can
people hear the difference between 14 or 15 bits and 20? Using decent
equipment? That's a more interesting question!
>> If one microphone is
>> chosen over another during recording, because of it's
>characteristic
>> self-noise or euphonic distortion, then that self-noise or
>distortion
>> should be reproducable as part of the artistic creation.
>
>OK, so we reproduce it (-72 dB noise) with >20 dB to the noise floor
>of the basic 16/44 system. Where is the beef?
The claim that you regularly make, that because YOU cannot find
commercial material that is technically better than 75dB or so in
dynamic range, the specification of CD is not being exploited. I'm
suggesting that your way of justifying this is more than a little
bizarre.. if you widen your selection process you will find material
that does exploit the specification.
Arny Krüger
> "Arny Krüger" <ar...@flash.net> wrote:
>
> >A recording with 72 dB SNR will generally be perceived as being
> >"clean" and the background noise will be generally easy to ignore
> >and/or will hardly be perceived at all. Any who doubt this need
only
> >listen to the bit-reduced samples that are available for free
> >download at www.pcabx.com. Thusfar, I've not heard anybody claim
> >reliable detection below 14 bits.
>
Can
> people hear the difference between 14 or 15 bits and 20? Using
decent
> equipment? That's a more interesting question!
Letsee, the difference between 14 bits and 16 bits is what, 64 times
as big as the difference between 16 bits and 20. Do you seriously
think that improving the reference with 1/64 more accuracy would
make that much of a difference? We are talking 2% distortion of the
LSB, right?
> >> If one microphone is
> >> chosen over another during recording, because of it's
> >characteristic
> >> self-noise or euphonic distortion, then that self-noise or
> >distortion
> >> should be reproducable as part of the artistic creation.
> >
> >OK, so we reproduce it (-72 dB noise) with >20 dB to the noise
floor
> >of the basic 16/44 system. Where is the beef?
> The claim that you regularly make, that because YOU cannot find
> commercial material that is technically better than 75dB or so in
> dynamic range, the specification of CD is not being exploited.
Surprise me. Find a commerical recording with better than 75 dB.
More specifically, find a recording where the quotent of the
amplitude of the largest single peak is more than 75 dB greater than
the amplitude of any natural, unattenuated sound that has duration
of 10 mSec or one full cycle of whatever frequency the fundamental
of the sound is, whichever is greater.
>I'm
> suggesting that your way of justifying this is more than a little
> bizarre.. if you widen your selection process you will find
material
> that does exploit the specification.
So you say. But of course we know that you don't always tell the
whole truth.
jj, curmudgeon and tiring philalethist
Paul Dormer <me...@clara.net> wrote:
>The claim that you regularly make, that because YOU cannot find
>commercial material that is technically better than 75dB or so in
>dynamic range, the specification of CD is not being exploited.
Well, I'll chime in and say I agree with that, most commercial
material doesn't come close, and is usually badly mixed, badly scaled,
and badly processed to boot.
>if you widen your selection process you will find material
>that does exploit the specification.
Um, that isn't very common. Yes, I have some, but it's not all
that common, nor is it where I expected to find it.
--
Copyright j...@research.att.com 1999, all rights reserved, except transmission
by USENET and like facilities granted. This notice must be included. Any
use by a provider charging in any way for the IP represented in and by this
article and any inclusion in print or other media are specifically prohibited.
Paul Dormer
>> >A recording with 72 dB SNR will generally be perceived as being
>> >"clean" and the background noise will be generally easy to ignore
>> >and/or will hardly be perceived at all. Any who doubt this need
>only
>> >listen to the bit-reduced samples that are available for free
>> >download at www.pcabx.com. Thusfar, I've not heard anybody claim
>> >reliable detection below 14 bits.
>>
>> Above, you mean. I've managed 15 bits, but not always reliably.
>Can
>> people hear the difference between 14 or 15 bits and 20? Using
>decent
>> equipment? That's a more interesting question!
>
>Letsee, the difference between 14 bits and 16 bits is what, 64 times
>as big as the difference between 16 bits and 20. Do you seriously
>think that improving the reference with 1/64 more accuracy would
>make that much of a difference? We are talking 2% distortion of the
>LSB, right?
Whilst I accept that there is some logic to what you say it's by no
means a foregone conclusion, in my opinion. I am not happy to draw
conclusions about >16bit material based on experiments revolving
around 16bit material. You have already decided, it seems, that 14
bits is about all we can hear.. so no doubt you see no point in going
any further. We have recently heard from Glenn that 15 bits is
achievable.. I have passed tests at 15 bits myself, albeit not
everytime. So I say it's worth comparing >16bits with 16 or lower
truncations.
>> >> If one microphone is
>> >> chosen over another during recording, because of it's
>> >characteristic
>> >> self-noise or euphonic distortion, then that self-noise or
>> >distortion
>> >> should be reproducable as part of the artistic creation.
>> >
>> >OK, so we reproduce it (-72 dB noise) with >20 dB to the noise
>floor
>> >of the basic 16/44 system. Where is the beef?
>
>> The claim that you regularly make, that because YOU cannot find
>> commercial material that is technically better than 75dB or so in
>> dynamic range, the specification of CD is not being exploited.
>
>Surprise me. Find a commerical recording with better than 75 dB.
>More specifically, find a recording where the quotent of the
>amplitude of the largest single peak is more than 75 dB greater than
>the amplitude of any natural, unattenuated sound that has duration
>of 10 mSec or one full cycle of whatever frequency the fundamental
>of the sound is, whichever is greater.
Without the open mic channel requirement?
>>I'm
>> suggesting that your way of justifying this is more than a little
>> bizarre.. if you widen your selection process you will find
>material
>> that does exploit the specification.
>
>So you say. But of course we know that you don't always tell the
>whole truth.
No, YOU are the liar Arnold.. try and stay in context.
Paul Dormer
>>The claim that you regularly make, that because YOU cannot find
>>commercial material that is technically better than 75dB or so in
>>dynamic range, the specification of CD is not being exploited.
>
>Well, I'll chime in and say I agree with that, most commercial
>material doesn't come close, and is usually badly mixed, badly scaled,
>and badly processed to boot.
I agree with this, but Arny implies that there is NO material that
meets the dynamic range spec of CD.. based on his funky research
requirements. I might check out a few CDs from a prestigous sample
library, recorded using state of the art equipment. I'm pretty sure
some of these samples will blow his sweeping statements away.
>>if you widen your selection process you will find material
>>that does exploit the specification.
>
>Um, that isn't very common. Yes, I have some, but it's not all
>that common, nor is it where I expected to find it.
Care to elaborate on that? Cite some material?
jj, curmudgeon and tiring philalethist
Arny Krüger <ar...@flash.net> wrote:
>Letsee, the difference between 14 bits and 16 bits is what, 64 times
>as big as the difference between 16 bits and 20.
Setting 1 lsb error at 1, the error at 14 bits is 4, and the error
at 20 bits is 1/16.
I don't think you meant what you said, rather, I think you meant that
the ratio of error between the 14 and 20 bit samples was 1/64.
However, the difference between 1 LSB error at 16 and 1 LSB error at
14, using these units, is 4-1 = 3.
The difference between 16 and 20 bits is 1-1/16, or 15/16. So, the
difference in error is 45/16. Of course, what I think you intended was
in fact the relevant point.
The metric you specified, in fact, goes down if you go from 20 to 24
bits, which shows the conditioning error in what you said.
STILL, the point holds. If someone can't hear the change from 16 to 14 bits,
they aren't likely to hear a change 1/64 of that.
jj, curmudgeon and tiring philalethist
jj, curmudgeon and tiring philalethist <j...@research.att.com> wrote:
>STILL, the point holds. If someone can't hear the change from 16 to 14 bits,
>they aren't likely to hear a change 1/64 of that.
Now that was clear as mud, wasn't it? If one can't hear the
noise at 14 bits, rather, one is unlikely to hear the noise when
it's 1/64 of that, or 20 bits.
The point is, of course, if you can't hear the difference between
14 and 16 bits, then you don't hear the noise floor due to quantization yet.
Then, of course, you're not likely to hear something that is 1/64 of what
you can't hear already.
David Wareing
and tiring philalethist) wrote:
Not much better.... unless you hear in fractions of amplitude.
Paul Dormer
>STILL, the point holds. If someone can't hear the change from 16 to 14 bits,
>they aren't likely to hear a change 1/64 of that.
Glenn says he knows of people that can hear 15 bits vs 16 bits.. I've
managed it too, hard though it was. Where do we go from there?
Arny Krüger
> >>commercial material that is technically better than 75dB or so
in
> >>dynamic range, the specification of CD is not being exploited.
>
> >Well, I'll chime in and say I agree with that, most commercial
> >material doesn't come close, and is usually badly mixed, badly
scaled,
> >and badly processed to boot.
> I agree with this, but Arny implies that there is NO material that
> meets the dynamic range spec of CD.. based on his funky research
> requirements.
Where are the Bamborough stuffin'-words--in-my-mouth police when a
guy needs them? First off, all I've said is that I've looked and
found none, in accordance with certain criteria.
My requirements are perceptually-based. (actually way too lax)
> I might check out a few CDs from a prestigous sample
> library, recorded using state of the art equipment. I'm pretty
sure
> some of these samples will blow his sweeping statements away.
No sweeping statements, just observations, Paul.
> >>if you widen your selection process you will find material
> >>that does exploit the specification.
> >Um, that isn't very common. Yes, I have some, but it's not all
> >that common, nor is it where I expected to find it.
> Care to elaborate on that? Cite some material?
Yes, JJ, I'm all ears.
Arny Krüger
> In article <yRr14.11873$Lm2.2...@news.rdc2.mi.home.com>,
> Arny Krüger <ar...@flash.net> wrote:
> >Letsee, the difference between 14 bits and 16 bits is what, 64
times
> >as big as the difference between 16 bits and 20.
> Hmm, is it?
>
> Setting 1 lsb error at 1, the error at 14 bits is 4, and the error
> at 20 bits is 1/16.
>
> I don't think you meant what you said, rather, I think you meant
that
> the ratio of error between the 14 and 20 bit samples was 1/64.
Letsee, let's take bit 14 as being 1.000
the 15th bit is 0.5
the 16th bit is 0.25
the 17 th bit is 0.125
the 18 th bit is 0.0625
the 19th bit is 0.03125
the 20 th bit is 0.015625
1.00 / 0.015625 is 1/64
That is what I meant.
> However, the difference between 1 LSB error at 16 and 1 LSB error
at
> 14, using these units, is 4-1 = 3.
> The difference between 16 and 20 bits is 1-1/16, or 15/16. So, the
> difference in error is 45/16. Of course, what I think you
intended was
> in fact the relevant point.
Isn't the maximum error in 'x bit" digital information just a tad
1 tad = a true mathematical "delta") under the value of the LSB?
> The metric you specified, in fact, goes down if you go from 20 to
24
> bits, which shows the conditioning error in what you said.
> STILL, the point holds. If someone can't hear the change from 16
to 14 bits,
> they aren't likely to hear a change 1/64 of that.
I'd be willing to live with that, but...
the 21 st bit is 0.0078125
the 22 nd bit is 0.00390625
the 23 rd bit is 0.001953125
the 24th bit is 0.0009765625
1/ 0.0009765625 = 1024
Arny Krüger
to 14 bits,
> >they aren't likely to hear a change 1/64 of that.
>
I've
> managed it too, hard though it was. Where do we go from there?
(1) what does "can hear" mean? A DBT?
(2) If you take something and cut it by 40 dB and then amplify it by
40 dB then you have done something like throw away 7 bits. You can
definitely hear the difference between 7 bits and 8 bits in a DBT.
In fact, most can pretty readily hear the difference between 11 and
12 bits, as many who have visited www.pcabx.com now know from their
own experience.
So, if your evaluation (which may be done for legitimate engineering
reasons) includes cutting by 40 dB, throwing away the 16th bit, and
then boosting by 40 dB then most should be able to hear the
difference between 15 and 16 bit processing of the attenuated
signal.
The legitimate engineering reasons that come to mind for this kind
of processing at all points of a home stereo (i.e., justification to
just go 24 bit wall-to-wall) seems to me to have nothing to do with
playing CD's on a properly set up system. But if such a reason
exists, then I'd like to hear about it.
And there is another caveat here, and that is that digital filters
that process 16 bit data often need to have > 16 bit precision to
have adequately smooth frequency response. 20 bits seems to do the
job for most current applications, but I can hypothesize the need
for a few more bits. No matter, cheap DSPs are now 24 bit.
jj, curmudgeon and tiring philalethist
David Wareing <war...@dircon.co.uk> wrote:
>Not much better.... unless you hear in fractions of amplitude.
After the cochlear filtering, that's basically what deltaL/L is all about.
jj, curmudgeon and tiring philalethist
Paul Dormer <me...@clara.net> wrote:
>Glenn says he knows of people that can hear 15 bits vs 16 bits.. I've
>managed it too, hard though it was. Where do we go from there?
You can do that in my listening room. It is possible. However, what you loose
is a sense of ambience when you go to 16 bits, generally, i.e. the noise floor
goes down.
If you hear anything else, it's time to check someone's dithering apparatus,
or low-level linearity requirements. My system setup is very careful to
get decent low-level linearity, btw.
Arny Krüger
> In article <3849ed7d...@news.clara.net>,
> Paul Dormer <me...@clara.net> wrote:
> >Glenn says he knows of people that can hear 15 bits vs 16 bits..
I've
> >managed it too, hard though it was. Where do we go from there?
>
> You can do that in my listening room. It is possible. However,
what you loose
> is a sense of ambience when you go to 16 bits, generally, i.e. the
noise floor
> goes down.
What is the average level of the program material when such things
are noticed ( 1 second moving average).
David Wareing
and tiring philalethist) wrote:
>In article <3846e1b7...@news.dircon.co.uk>,
>David Wareing <war...@dircon.co.uk> wrote:
>>Not much better.... unless you hear in fractions of amplitude.
>??? Of course, you do, in one very real sense.
>After the cochlear filtering, that's basically what deltaL/L is all about.
>
>--
>Copyright j...@research.att.com 1999, all rights reserved, except transmission
>by USENET and like facilities granted. This notice must be included. Any
>use by a provider charging in any way for the IP represented in and by this
>article and any inclusion in print or other media are specifically prohibited.
It does not seem like a very big number when it is not squared
(power). I guess it isn't huge.....now what does (1/16 of not much)
sound like.....hum.....Never mind!
jj, curmudgeon and tiring philalethist
In the good listening room, bear in mind. NC8, remember?
No, it wasn't ABX, it was blind ABC/hr. But 10/10 on the
first try is sorta definitive, you know, listen, click, listen
click score ...
>(2) If you take something and cut it by 40 dB and then amplify it by
>40 dB then you have done something like throw away 7 bits. You can
>definitely hear the difference between 7 bits and 8 bits in a DBT.
Irrelevant.
>And there is another caveat here, and that is that digital filters
>that process 16 bit data often need to have > 16 bit precision to
>have adequately smooth frequency response. 20 bits seems to do the
>job for most current applications, but I can hypothesize the need
>for a few more bits. No matter, cheap DSPs are now 24 bit.
It depends on the ACCUMULATED NOISE.
For many 24 bits is NOT ENOUGH.
Arny Krüger
> >(1) what does "can hear" mean? A DBT?
> Pretty easily heard in normal material wth some quiet spots.
> In the good listening room, bear in mind. NC8, remember?
Yeah, I'd like to have one of those!
> No, it wasn't ABX, it was blind ABC/hr. But 10/10 on the
> first try is sorta definitive, you know, listen, click, listen
> click score ...
Works for me
> >(2) If you take something and cut it by 40 dB and then amplify it
by
> >40 dB then you have done something like throw away 7 bits. You
can
> >definitely hear the difference between 7 bits and 8 bits in a
DBT.
> Irrelevant.
That was my point. But irrelevant to what? If I was going to design
a recording console that was fool proof, then I might define fool
proof as being able to tolerate cutting something down 40 dB too
far, and then jacking it up by 40 dB someplace else without audible
degradation.
> >And there is another caveat here, and that is that digital
filters
> >that process 16 bit data often need to have > 16 bit precision to
> >have adequately smooth frequency response. 20 bits seems to do
the
> >job for most current applications, but I can hypothesize the need
> >for a few more bits. No matter, cheap DSPs are now 24 bit.
> It depends on the ACCUMULATED NOISE.
> For many 24 bits is NOT ENOUGH.
Many what? People? Complex processing applications?