I think what hasn't been stressed about the issue of digital
distortion is that it is not a traditional phase distortion we are
talking about, whereby the phase anomalies are related to frequency in
a linear way, such as would happen in a cheap equalizer.

digital high frequency distortion is more destructive to music because
it makes the waveform fit into a time grid.

on a 100 Hz sine wave this is not so much of a problem because there
are hundreds of samples of one complete wave cycle therefore the wave
is reproduced fairly accurately.

a 5 Khz waveform only has around 10 or so samples for the whole
waveform 10 Khz has half of that and 20 half of half etc.

raising the sample rate does raise bandwith, but the human ear cannot
hear above 20KHz (most people stop at 15-16Khz).

High sample rates are useful because they increase resolution for
high frequencies.

5khz at 192K sampling rate will have more samples to re-build the
waveform with.

reconstruction filters only fill in the holes. If something happens
inbetween the samples it simply isn't recorded in the first place.

Filters just 'connect the dots'


If it were possible to sample at an extremely high sample rate from
the beginning cost-wise (keeping within a consumer price range) it
would already have been done.


remember that 15 years ago digital recorders costed a fortune even
though they only went up to 48khz.

Now the price barrier isn't there anymore.

In article , maxdm
wrote:

I think what hasn't been stressed about the issue of digital
distortion is that it is not a traditional phase distortion we are
talking about, whereby the phase anomalies are related to frequency in
a linear way, such as would happen in a cheap equalizer.

digital high frequency distortion is more destructive to music because
it makes the waveform fit into a time grid.

on a 100 Hz sine wave this is not so much of a problem because there
are hundreds of samples of one complete wave cycle therefore the wave
is reproduced fairly accurately.

a 5 Khz waveform only has around 10 or so samples for the whole
waveform 10 Khz has half of that and 20 half of half etc.

raising the sample rate does raise bandwith, but the human ear cannot
hear above 20KHz (most people stop at 15-16Khz).

High sample rates are useful because they increase resolution for
high frequencies.

5khz at 192K sampling rate will have more samples to re-build the
waveform with.

reconstruction filters only fill in the holes. If something happens
inbetween the samples it simply isn't recorded in the first place.

There isn't anything "between the samples" that is missed because the
input to the ADC MUST be band limited for the system to work correctly.
If the input to the ADC is correctly band limited, then the
reconstruction filter EXACTLY reproduces the input. Do not trust your
intuition on this. If you have enough of a math background, look into
the math behind discrete time sampling. If you do not have enough math
background to follow the math you will just have to take it on faith.
However, this is proven in the same way that 1+1=2 is proven. It's not
a "theory" - it's a theorem.


Filters just 'connect the dots'


If it were possible to sample at an extremely high sample rate from
the beginning cost-wise (keeping within a consumer price range) it
would already have been done.


If you can't hear above 20 kHz, there is no reason to sample at more
than 40 kHz plus a little extra for the filter transition. It won't
give you any more information below 20 kHz if you sample at 96 kHz. If
new systems are being introduced with higher sample rates it can only
be because 1) it doesn't cost much and it's good marketing fodder, 2)
someone doesn't understand how digital works, or 3) someone believes
that people can hear to 40 kHz. My money is on 1.

Marc Foster


remember that 15 years ago digital recorders costed a fortune even
though they only went up to 48khz.

Now the price barrier isn't there anymore.

inbetween the samples it simply isn't recorded in the first place.

There isn't anything "between the samples" that is missed because the
input to the ADC MUST be band limited for the system to work correctly.
If the input to the ADC is correctly band limited, then the
reconstruction filter EXACTLY reproduces the input. Do not trust your
intuition on this. If you have enough of a math background, look into
the math behind discrete time sampling. If you do not have enough math
background to follow the math you will just have to take it on faith.
However, this is proven in the same way that 1+1=2 is proven. It's not
a "theory" - it's a theorem.


This is true if you only consider a waveform as a simple helmholtz
model.
although I do not have a math backround and it sounds like you do, the
issue at hand is a lot simpler:

I know it's hard to talk about something which is difficult to hear
for some people. If the distortion was more evident (in the mid
frequencies) then perhaps more 'complete' theorems to describe what
is happening would have gained popularity, I feel.

musical sound is not only made of harmonic content, or frequency
dependent information.
the best music reproduction systems can accurately reproduce
transients in waveforms in a natural way.

if you were to take an analog lc filter and limit bandwith at 6 KHz
the effect would be vastly different than lowering the sampling
frequency to 6Khz bandwidth on a digital system.

even if you filter the signal going into the adc so it has no harmonic
content over 6 KHz and then filter all the 'steps' created by the dac
out with another filter that limits the bandwith at 6KHz in output the
sound is grainy. Towards the upper end of the spectrum distortion
which is mathematically related to the sampling frequency occurs.

if the waveform manifests a small yet significant waveform
irregularity between samples, the best that a digital system can do is
take the output of the lowpass filter at the next sample and, since
the lowpass filter is supposed (theoretically) to filter out any
irrelevant information (above 6KHz) it is assumed that this is an
accurate representation of what is being recorded.

In practice this is not so. Information gets lost, and as we near the
bandwith limit the system becomes even more unstable because any error
due to jitter or less than perfect brickwall filters/converters.

It is more probable that a digital system operates within it's
theoretical ideal at a point well below it's sampling frequency.

I am sure that math apart, most people who have the capacity to listen
critically to sound will notice a difference in 192Khz audio compared
to 44.1 even if the speakers or headphones or amplifiers are frequency
limited to20 Khz.

Anyone out there with listening experience ?

"maxdm" wrote in message
...
Filters just 'connect the dots'

This is sufficient for perfect reproduction.

An excellent analogy, used in a recently posted link is to imagine a circle.
Only three points on the circumference are required to fully define it, and
any more are an unnecessary waste of bandwidth.

Marc Foster wrote:
In article , maxdm
wrote:
reconstruction filters only fill in the holes. If something happens
inbetween the samples it simply isn't recorded in the first place.

There isn't anything "between the samples" that is missed because the
input to the ADC MUST be band limited for the system to work correctly.
If the input to the ADC is correctly band limited, then the
reconstruction filter EXACTLY reproduces the input. Do not trust your
intuition on this. If you have enough of a math background, look into
the math behind discrete time sampling. If you do not have enough math
background to follow the math you will just have to take it on faith.

Alternatively, there is a good graphical ilustration of this counterintuitive
property at:
http://www.lavryengineering.com/docu...ing_Theory.pdf

See in pages 23-25 the example of a 17 KHz sine wave sampled at 44.1 KHz.

However, this is proven in the same way that 1+1=2 is proven. It's not
a "theory" - it's a theorem.


--
http://www.mat.uc.pt/~rps/

..pt is Portugal| `Whom the gods love die young'-Menander (342-292 BC)
Europe | Villeneuve 50-82, Toivonen 56-86, Senna 60-94

maxdm wrote:
inbetween the samples it simply isn't recorded in the first place.

You have to understand that if the signal is band-limited to 20KHz,
there is no additional information gained by sampling at higher than
40KHz. Sampling a 20 KHz band-limited signal at 192 KHz results in no
additional information over sampling at 44.1 KHz.

Please note that we are not talking about only sinewaves as input
signals. We are talking about music, speech, or whatever time-varying
waveforms that are band-limited. Music with transients are band-limited.
You have a problem with redbook CD sampling only if you can hear above
20 KHz.


There isn't anything "between the samples" that is missed because the
input to the ADC MUST be band limited for the system to work correctly.
If the input to the ADC is correctly band limited, then the
reconstruction filter EXACTLY reproduces the input. Do not trust your
intuition on this. If you have enough of a math background, look into
the math behind discrete time sampling. If you do not have enough math
background to follow the math you will just have to take it on faith.
However, this is proven in the same way that 1+1=2 is proven. It's not
a "theory" - it's a theorem.


This is true if you only consider a waveform as a simple helmholtz
model.

No. The key thing you have to understand is that the input waveform is
band-limited, i.e., all its energy are inside a frequency band. For CD,
it is approx. DC to 20 KHz.

Helmholtz models are irrelevant in this discussion.

although I do not have a math backround and it sounds like you do, the
issue at hand is a lot simpler:

Sampling theorem may not be intuitive. Your intuition in this case is
wrong. You don't need a strong math background, just basic calculus.

So far you seem to prefer to be led by your intuition instead of doing
the work to understand sampling.

I know it's hard to talk about something which is difficult to hear
for some people. If the distortion was more evident (in the mid
frequencies) then perhaps more 'complete' theorems to describe what
is happening would have gained popularity, I feel.

Sampling does not introduce any distortion, if the input signal is
band-limited. This may not be intuitive.


musical sound is not only made of harmonic content, or frequency
dependent information.
the best music reproduction systems can accurately reproduce
transients in waveforms in a natural way.

Music is band-limited. The best music production systems are also band
limited.


if you were to take an analog lc filter and limit bandwith at 6 KHz
the effect would be vastly different than lowering the sampling
frequency to 6Khz bandwidth on a digital system.

This is highly irrelevant to the discussion. Sampling is not like analog
filtering.


even if you filter the signal going into the adc so it has no harmonic
content over 6 KHz and then filter all the 'steps' created by the dac
out with another filter that limits the bandwith at 6KHz in output the
sound is grainy. Towards the upper end of the spectrum distortion
which is mathematically related to the sampling frequency occurs.

if the waveform manifests a small yet significant waveform
irregularity between samples,

It cannot, if the waveform is band-limited. There is only one possible
waveform that can be represented by those samples. This is a key to
understand sampling theorem. There cannot be an arbitrary number of
band-limited waveforms with the same samples. It is *not* like
connecting dots.

the best that a digital system can do is
take the output of the lowpass filter at the next sample and, since
the lowpass filter is supposed (theoretically) to filter out any
irrelevant information (above 6KHz) it is assumed that this is an
accurate representation of what is being recorded.

Again, you do not understand sampling.

In practice this is not so. Information gets lost, and as we near the
bandwith limit the system becomes even more unstable because any error
due to jitter or less than perfect brickwall filters/converters.

It is more probable that a digital system operates within it's
theoretical ideal at a point well below it's sampling frequency.

I am sure that math apart, most people who have the capacity to listen
critically to sound will notice a difference in 192Khz audio compared
to 44.1 even if the speakers or headphones or amplifiers are frequency
limited to20 Khz.

Any difference is due to implementation differences, and not a result of
the sampling rate difference, if the input signal is band-limited to 20 KHz.


Anyone out there with listening experience ?

maxdm wrote:
inbetween the samples it simply isn't recorded in the first place.

There isn't anything "between the samples" that is missed because the
input to the ADC MUST be band limited for the system to work correctly.
If the input to the ADC is correctly band limited, then the
reconstruction filter EXACTLY reproduces the input. Do not trust your
intuition on this. If you have enough of a math background, look into
the math behind discrete time sampling. If you do not have enough math
background to follow the math you will just have to take it on faith.
However, this is proven in the same way that 1+1=2 is proven. It's not
a "theory" - it's a theorem.


This is true if you only consider a waveform as a simple helmholtz
model.
although I do not have a math backround and it sounds like you do, the
issue at hand is a lot simpler:

I know it's hard to talk about something which is difficult to hear
for some people. If the distortion was more evident (in the mid
frequencies) then perhaps more 'complete' theorems to describe what
is happening would have gained popularity, I feel.

musical sound is not only made of harmonic content, or frequency
dependent information.
the best music reproduction systems can accurately reproduce
transients in waveforms in a natural way.

if you were to take an analog lc filter and limit bandwith at 6 KHz
the effect would be vastly different than lowering the sampling
frequency to 6Khz bandwidth on a digital system.

even if you filter the signal going into the adc so it has no harmonic
content over 6 KHz and then filter all the 'steps' created by the dac
out with another filter that limits the bandwith at 6KHz in output the
sound is grainy. Towards the upper end of the spectrum distortion
which is mathematically related to the sampling frequency occurs.

if the waveform manifests a small yet significant waveform
irregularity between samples, the best that a digital system can do is
take the output of the lowpass filter at the next sample and, since
the lowpass filter is supposed (theoretically) to filter out any
irrelevant information (above 6KHz) it is assumed that this is an
accurate representation of what is being recorded.

In practice this is not so. Information gets lost, and as we near the
bandwith limit the system becomes even more unstable because any error
due to jitter or less than perfect brickwall filters/converters.

It is more probable that a digital system operates within it's
theoretical ideal at a point well below it's sampling frequency.

I am sure that math apart, most people who have the capacity to listen
critically to sound will notice a difference in 192Khz audio compared
to 44.1 even if the speakers or headphones or amplifiers are frequency
limited to20 Khz.

Anyone out there with listening experience ?

How about yourself? You're *sure* that the difference is audible to
a critical listener. I must presume you count youself as one.
So, try a comparison under blind conditions. If you have the
capability, record some 196Khz audio to redbook. Surely you'll pass
with flying colors.

If you want other data points, you might also want to post
your questionn over on rec.audio.tech, and perhaps to
George Massenburg;s Mastering web board

http://webbd.nls.net:8080/~mastering/login


--

-S.
Why don't you just admit that you hate music and leave people alone. --
spiffy

In article , maxdm
wrote:

I think what hasn't been stressed about the issue of digital
distortion is that it is not a traditional phase distortion we are
talking about, whereby the phase anomalies are related to frequency in
a linear way, such as would happen in a cheap equalizer.
digital high frequency distortion is more destructive to music because
it makes the waveform fit into a time grid.

The reason it is not stressed, except in the high-end audio realm,
is that your claimm is, to put it as simply and as bluntly as possible,
just plain wrong. I mean no disrespect in saying this, and please
bear with me if I sound impatient, but while your "technical"
description seems to make sense, aevent a lot of intuitive sense,
intuition, in this case, is absolutely wrong. Only in the high
end audio business does this quite incorrect view of how sampling
works just keep hanging on. It is technically wrong.

on a 100 Hz sine wave this is not so much of a problem because there
are hundreds of samples of one complete wave cycle therefore the wave
is reproduced fairly accurately.

a 5 Khz waveform only has around 10 or so samples for the whole
waveform 10 Khz has half of that and 20 half of half etc.

And, as long as it is more than 2 per cycle, that's ALL that's
needed to FULLY encode the waveform with NO distortion. It may
not make intuitive sense, but it's because intuition is wrong.

raising the sample rate does raise bandwith, but the human ear cannot
hear above 20KHz (most people stop at 15-16Khz).

High sample rates are useful because they increase resolution for
high frequencies.

Absolutely, 100% false, I am afraid. Once you sample at more than
twice the highest signal frequency, sampling any more WILL NOT
inclrease the resolution one bit.

5khz at 192K sampling rate will have more samples to re-build the
waveform with.

reconstruction filters only fill in the holes. If something happens
inbetween the samples it simply isn't recorded in the first place.

If there is something "in between" the samples, then, by definition,
the signal CANNOT be at 5 kHz. WHat you must learn to understand
is that if the bandwidth is limited to 20 kHz, THERE CANNOT BE ANYTHING
IN BETWEEN THE SAMPLES THAT IS NOT FULLY, ENTIRELY, FAITHFULLY AND
WITHOUT ERROR OR DISTORTION CAPTURED BY A PROPERLY IMPLEMENTED SYSTEM
THAT SAMPLES AT ANY FREQUENCY ABOVE 40 kHZ.

Sampling ALL waveforms contained within a 20 kHz bandwidth IS NOT,
indeed, CANNOT by done ANY more accurately, with ANY more resolution
at a sample rate of 96 kHz than it is at a sample rate of 48 kHz.
Or 44.1 kHz. Or, for that matter, 40.0001 kHz.

This is not a matter of specualtion, or intuition, it is a matter
of a well founded, rigorously proven THEOREM (not "theory," but
"theorem," please note the difference). Please refer to the work
of Claude Shannon on the topic from more than a half century ago,

Filters just 'connect the dots'

No, they do not. You would do well to, in fact, read up onj the topic
before making such a claim. If the orginal material was properly band
limited to less than 1/2 the sample rate, there IS NOT ANYTHING
'between the dots' that's missing, that the filter is 'incorrectly'
connecting. What exists "between the dots" is FULLY and UNAMBIGUOUSLY
determined by the fact that the bandwidth of the signal is limited.

If it were possible to sample at an extremely high sample rate from
the beginning cost-wise (keeping within a consumer price range) it
would already have been done.

remember that 15 years ago digital recorders costed a fortune even
though they only went up to 48khz.

Now the price barrier isn't there anymore.

Excuse me, but pretty decent DAT recorders could be had for $2000
back then.

+---------------------------------------+
| Dick Pierce |
| Professional Audio Development |
| (1) 781/826-4953 Voice and FAX |
| |
+---------------------------------------+

In article wyhAc.46425$2i5.22508@attbi_s52, maxdm
wrote:

inbetween the samples it simply isn't recorded in the first place.

There isn't anything "between the samples" that is missed because the
input to the ADC MUST be band limited for the system to work correctly.
If the input to the ADC is correctly band limited, then the
reconstruction filter EXACTLY reproduces the input. Do not trust your
intuition on this. If you have enough of a math background, look into
the math behind discrete time sampling. If you do not have enough math
background to follow the math you will just have to take it on faith.
However, this is proven in the same way that 1+1=2 is proven. It's not
a "theory" - it's a theorem.


This is true if you only consider a waveform as a simple helmholtz
model.
although I do not have a math backround and it sounds like you do, the
issue at hand is a lot simpler:

I know it's hard to talk about something which is difficult to hear
for some people. If the distortion was more evident (in the mid
frequencies) then perhaps more 'complete' theorems to describe what
is happening would have gained popularity, I feel.

musical sound is not only made of harmonic content, or frequency
dependent information.
the best music reproduction systems can accurately reproduce
transients in waveforms in a natural way.

if you were to take an analog lc filter and limit bandwith at 6 KHz
the effect would be vastly different than lowering the sampling
frequency to 6Khz bandwidth on a digital system.

No, it wouldn't. The digital system would have to be analog filtered to
6 kHz bandwidth before being sampled. At this point (the input to the
ADC) the signal is identical to the analog waveform. If it is sampled
at the proper rate and the correct reconstruction filter is
implemented, then the reconstructed signal at the output of the digital
system is IDENTICAL to the band limited analog waveform.


even if you filter the signal going into the adc so it has no harmonic
content over 6 KHz and then filter all the 'steps' created by the dac
out with another filter that limits the bandwith at 6KHz in output the
sound is grainy. Towards the upper end of the spectrum distortion
which is mathematically related to the sampling frequency occurs.

No it doesn't. The use of dither (another REQUIRED part of a
functioning digital system) prevents signal or sample rate correlated
errors from appearing at the output.


if the waveform manifests a small yet significant waveform
irregularity between samples, the best that a digital system can do is
take the output of the lowpass filter at the next sample and, since
the lowpass filter is supposed (theoretically) to filter out any
irrelevant information (above 6KHz) it is assumed that this is an
accurate representation of what is being recorded.

If there is a "small yet significant waveform irregularity between
samples" then the signal has not been properly bandlimited prior to the
ADC. If the signal is properly bandlimited then there is no information
in the waveform that is not captured in the sampled data. Once again,
this is well established fact. It was old news when I took digital
signal processing classes in the 1970's.


In practice this is not so. Information gets lost, and as we near the
bandwith limit the system becomes even more unstable because any error
due to jitter or less than perfect brickwall filters/converters.

It is more probable that a digital system operates within it's
theoretical ideal at a point well below it's sampling frequency.

I am sure that math apart, most people who have the capacity to listen
critically to sound will notice a difference in 192Khz audio compared
to 44.1 even if the speakers or headphones or amplifiers are frequency
limited to20 Khz.

If the only difference between the two recordings is the sample rate
then one of the two systems must be broken to produce a real
difference.


Anyone out there with listening experience ?

Christopher Key wrote:

"maxdm" wrote in message
...
Filters just 'connect the dots'

This is sufficient for perfect reproduction.

An excellent analogy, used in a recently posted link is to imagine a circle.
Only three points on the circumference are required to fully define it, and
any more are an unnecessary waste of bandwidth.

True enough, but you also must know it's a circle. The
information content of that knowledge is nontrivial, and
without it, an infinity of shapes could be defined by three
points.

So maybe that's not such a great analogy to discribe
reproduction of an audio waveform.

Mike Prager
North Carolina, USA

"Mike Prager" wrote in message
news:z9tAc.66738$HG.63024@attbi_s53...
Christopher Key wrote:

"maxdm" wrote in message
...
Filters just 'connect the dots'

This is sufficient for perfect reproduction.

An excellent analogy, used in a recently posted link is to imagine a
circle.
Only three points on the circumference are required to fully define it,
and
any more are an unnecessary waste of bandwidth.

True enough, but you also must know it's a circle. The
information content of that knowledge is nontrivial, and
without it, an infinity of shapes could be defined by three
points.

Good point, although the fact that the signal is band limited means that you
do know a certain amount about the signal, ie that it can be expressed as
the sum of a set of sinusoids up to a given frequency.

So maybe that's not such a great analogy to discribe
reproduction of an audio waveform.


I do accept that it isn't a precise analogy, but thought nonetheless that it
was pretty good for showing people of a non mathematical background
(specifically the OP) roughly what was going on.

Chris Key

Dick Pierce wrotE:

....snips to content ......

In article , maxdm
wrote:

Filters just 'connect the dots'

No, they do not. You would do well to, in fact, read up onj the topic
before making such a claim. If the orginal material was properly band
limited to less than 1/2 the sample rate, there IS NOT ANYTHING
'between the dots' that's missing, that the filter is 'incorrectly'
connecting. What exists "between the dots" is FULLY and UNAMBIGUOUSLY
determined by the fact that the bandwidth of the signal is limited.

IMO the "sine wave vs the stairstep analogy" has been the leading problem with
a simple description of digital processing. People thinkof a scope trace of a
sound as being the "signal" and a stairstep of samples as 'missing' the points
between the samples.

They forget (or didn't know) that the scope trace is NOT the sound ( or the
signal) but simply an analog representation of it. It's just a picture of the
sound that has its own deficiencies. It is "not" the signal.

Mike Prager wrote in message news:z9tAc.66738$HG.63024@attbi_s53...
Christopher Key wrote:

"maxdm" wrote in message
...
Filters just 'connect the dots'

This is sufficient for perfect reproduction.

An excellent analogy, used in a recently posted link is to imagine a circle.
Only three points on the circumference are required to fully define it, and
any more are an unnecessary waste of bandwidth.

True enough, but you also must know it's a circle.

In exactly the same way that in a properly implemented sampled
system, you KNOW, a priori, that the bandwidth is limited to
less than 1/2 the sample rate.

The
information content of that knowledge is nontrivial, and
without it, an infinity of shapes could be defined by three
points.

You don't need to know it's a cricle, you need to know simply
that it is a closed continuous function of the form ax + by = c
or something similar.

Everyone, absolutely EVERYONE is simply ignoring the basic tenets
of sampling when making all these handwaving objections about why
it can't work without having the slightest understanding of how
it does.

So maybe that's not such a great analogy to discribe
reproduction of an audio waveform.

It's a perfetcly good analogy, FAR more accurate than nonsense about
"connecting the dots" and "stuff between the samples" and the rest
of the high-end hooey being spouted.

Hello Mike,

Mike Prager wrote in
news:z9tAc.66738$HG.63024@attbi_s53:
....
True enough, but you also must know it's a circle. The
....
So maybe that's not such a great analogy to discribe
reproduction of an audio waveform.

isn't this pretty much the same as the need to know that the original
signal is bandlimited? I think that from this point of view the analogy can
be considered a pretty good one.

Bye,

--
Denis Sbragion
InfoTecna
Tel: +39 0362 805396, Fax: +39 0362 805404
URL: http://www.infotecna.it

Nousaine wrote:
... If the orginal material was properly band
limited to less than 1/2 the sample rate, there IS NOT ANYTHING
'between the dots' that's missing, that the filter is 'incorrectly'
connecting. What exists "between the dots" is FULLY and UNAMBIGUOUSLY
determined by the fact that the bandwidth of the signal is limited.

IMO the "sine wave vs the stairstep analogy" has been the leading
problem with a simple description of digital processing. People
thinkof a scope trace of a sound as being the "signal" and a
stairstep of samples as 'missing' the points between the samples.

They forget (or didn't know) that the scope trace is NOT the sound (
or the signal) but simply an analog representation of it. It's just a
picture of the sound that has its own deficiencies. It is "not" the
signal.

The output filter is as much part of the D/A converter as the swiched
current sources or the I/U-converter. You just cannot judge the sound by
looking at intermediate stages, but only at the final output to the
amplifier, and there the signal is identical to the original bandlimited
input.
The filtering to bandlimit will certainly affect the signal, but those high
frequency parts that are removed do not make any difference to our
perception, because they are outside our hearing range.
So if somebody argues against digital, he should not mention the
quantisation and time discretion, but the input filter stages at the coding,
which will make a difference when done improperly, like in the early 80s.
Generally I think most people have difficulties understanding the sampling
theorem, it is anti-intuitive and you need a certain level of education and
abstraction to grasp its implications.
--
ciao Ban
Bordighera, Italy

And, as long as it is more than 2 per cycle, that's ALL that's
needed to FULLY encode the waveform with NO distortion. It may
not make intuitive sense, but it's because intuition is wrong.

raising the sample rate does raise bandwith, but the human ear cannot
hear above 20KHz (most people stop at 15-16Khz).

High sample rates are useful because they increase resolution for
high frequencies.

Absolutely, 100% false, I am afraid. Once you sample at more than
twice the highest signal frequency, sampling any more WILL NOT
inclrease the resolution one bit.


O.K.

Since you appear to have studied the matter in depth, could you please
explain or comment on the following points:
let's imagine a digital system with a very low sampling rate.

If we record a sine wave that is not perfect and has a 'glitch'
(energy in a higher spectrum) that occurs inbetween samples, the
lowpass filter will limit the slew rate and therefore if the sine wave
is near the bandwith limit of the said brickwall filter this 'glitch'
should be filtered out of the signal right?
is it possible that a waveform that has a short sharp transient
inbetween samples could charge a capacitor in the brickwall filter or
in the ADC and therefore cause mistracking?
if the waveform has a very short but definite 'lump' lasting a
fraction of a millisecond the filter would eliminate the lump, but the
lump would charge the filter with a positive or negative voltage
therefore offsetting the signal read by the adc on the next samples.

If we have a system of any kind meant to reproduce sound, from a
practical engineering standpoint, which is meant to be mass produced
and not a one-off costing huge sums of money, doesn't common sense
dictate that the system should, if possible cost-wise, never operate
near its theoretical limits?
what you say about analog digital conversion makes sense in a perfect
system that behaves in a perfect way.

My experience is that the filters in most home and pro AD converters
do not limit bandwidth 100%, and when they come close there are side
effects.

raising the bandwith takes the brickwall frequency out of the audio
spectrum and any imperfections in the filter due to it being an
electronic device and not a mathematical model.

Why do experts in the recording and mastering field, who have been
used to making the records you would play on your home system claim
that the resolution on analog tape machines and laquer discs is
superior to analog?

I have recorded on both, and have gone to the trouble of modifying my
ADC's which use the akm chips to make sure what I was hearing was not
due to the analog stages, and analog is more distorted from a thd and
low frequency standpoint but it has imaging and depth that digital
recording has not achieved.
you hear this effect more in multitrack recording where the
side-effects are summed because of the summing of many discrete
sounds.
Higher sampling rates sound better.

Excuse me, but pretty decent DAT recorders could be had for $2000
back then.

Digital has been available since the early seventies and was very
expensive.

Mitsubishi multitrack digital recorders used for making records costed
a fortune.

Perhaps, then what is at issue is how the original analog signal gets
bandwidth limited? Is it possible that certain methods of bandwidth limitimg
are responsible for the nastiness present in the higher frequencies,
particularly in older CDs?

"Dick Pierce" wrote in message
...
In article , maxdm
wrote:

I think what hasn't been stressed about the issue of digital
distortion is that it is not a traditional phase distortion we are
talking about, whereby the phase anomalies are related to frequency in
a linear way, such as would happen in a cheap equalizer.
digital high frequency distortion is more destructive to music because
it makes the waveform fit into a time grid.

The reason it is not stressed, except in the high-end audio realm,
is that your claimm is, to put it as simply and as bluntly as possible,
just plain wrong. I mean no disrespect in saying this, and please
bear with me if I sound impatient, but while your "technical"
description seems to make sense, aevent a lot of intuitive sense,
intuition, in this case, is absolutely wrong. Only in the high
end audio business does this quite incorrect view of how sampling
works just keep hanging on. It is technically wrong.

on a 100 Hz sine wave this is not so much of a problem because there
are hundreds of samples of one complete wave cycle therefore the wave
is reproduced fairly accurately.

a 5 Khz waveform only has around 10 or so samples for the whole
waveform 10 Khz has half of that and 20 half of half etc.

And, as long as it is more than 2 per cycle, that's ALL that's
needed to FULLY encode the waveform with NO distortion. It may
not make intuitive sense, but it's because intuition is wrong.

raising the sample rate does raise bandwith, but the human ear cannot
hear above 20KHz (most people stop at 15-16Khz).

High sample rates are useful because they increase resolution for
high frequencies.

Absolutely, 100% false, I am afraid. Once you sample at more than
twice the highest signal frequency, sampling any more WILL NOT
inclrease the resolution one bit.

5khz at 192K sampling rate will have more samples to re-build the
waveform with.

reconstruction filters only fill in the holes. If something happens
inbetween the samples it simply isn't recorded in the first place.

If there is something "in between" the samples, then, by definition,
the signal CANNOT be at 5 kHz. WHat you must learn to understand
is that if the bandwidth is limited to 20 kHz, THERE CANNOT BE ANYTHING
IN BETWEEN THE SAMPLES THAT IS NOT FULLY, ENTIRELY, FAITHFULLY AND
WITHOUT ERROR OR DISTORTION CAPTURED BY A PROPERLY IMPLEMENTED SYSTEM
THAT SAMPLES AT ANY FREQUENCY ABOVE 40 kHZ.

Sampling ALL waveforms contained within a 20 kHz bandwidth IS NOT,
indeed, CANNOT by done ANY more accurately, with ANY more resolution
at a sample rate of 96 kHz than it is at a sample rate of 48 kHz.
Or 44.1 kHz. Or, for that matter, 40.0001 kHz.

This is not a matter of specualtion, or intuition, it is a matter
of a well founded, rigorously proven THEOREM (not "theory," but
"theorem," please note the difference). Please refer to the work
of Claude Shannon on the topic from more than a half century ago,

Filters just 'connect the dots'

No, they do not. You would do well to, in fact, read up onj the topic
before making such a claim. If the orginal material was properly band
limited to less than 1/2 the sample rate, there IS NOT ANYTHING
'between the dots' that's missing, that the filter is 'incorrectly'
connecting. What exists "between the dots" is FULLY and UNAMBIGUOUSLY
determined by the fact that the bandwidth of the signal is limited.

If it were possible to sample at an extremely high sample rate from
the beginning cost-wise (keeping within a consumer price range) it
would already have been done.

remember that 15 years ago digital recorders costed a fortune even
though they only went up to 48khz.

Now the price barrier isn't there anymore.

Excuse me, but pretty decent DAT recorders could be had for $2000
back then.

+---------------------------------------+
| Dick Pierce |
| Professional Audio Development |
| (1) 781/826-4953 Voice and FAX |
| |
+---------------------------------------+

Dick Pierce wrote:

You don't need to know it's a cricle, you need to know simply
that it is a closed continuous function of the form ax + by = c
or something similar.

Seems to me that more than one ellipse will fit through the
same three points.

Everyone, absolutely EVERYONE is simply ignoring the basic tenets
of sampling when making all these handwaving objections about why
it can't work without having the slightest understanding of how
it does.

Not everyone.

So maybe that's not such a great analogy to discribe
reproduction of an audio waveform.

It's a perfetcly good analogy,

Still disagree with that, but I never disputed sampling
theory.

FAR more accurate than nonsense about
"connecting the dots" and "stuff between the samples" and the rest
of the high-end hooey being spouted.

There's plenty of hooey, of different flavors, to go around.

Mike Prager
North Carolina, USA

(Dick Pierce) wrote in
news:vaDAc.133992$Ly.97088@attbi_s01:

Mike Prager wrote in message
news:z9tAc.66738$HG.63024@attbi_s53...
Christopher Key wrote:

"maxdm" wrote in message
...
Filters just 'connect the dots'

This is sufficient for perfect reproduction.

An excellent analogy, used in a recently posted link is to imagine a
circle. Only three points on the circumference are required to fully
define it, and any more are an unnecessary waste of bandwidth.

True enough, but you also must know it's a circle.

In exactly the same way that in a properly implemented sampled
system, you KNOW, a priori, that the bandwidth is limited to
less than 1/2 the sample rate.

The
information content of that knowledge is nontrivial, and
without it, an infinity of shapes could be defined by three
points.

You don't need to know it's a cricle, you need to know simply
that it is a closed continuous function of the form ax + by = c
or something similar.

Everyone, absolutely EVERYONE is simply ignoring the basic tenets
of sampling when making all these handwaving objections about why
it can't work without having the slightest understanding of how
it does.

So maybe that's not such a great analogy to discribe
reproduction of an audio waveform.

It's a perfetcly good analogy, FAR more accurate than nonsense about
"connecting the dots" and "stuff between the samples" and the rest
of the high-end hooey being spouted.


Dick,

Putting aside such malarky like "stuff between the samples" etc., I have
heard a definite difference between CD players. I call it digital grunge.
Is the better sound experienced with well engineered, pricier products
due to better D-As, better DSP processing or ????

What happened between 1990 and 2002 as far as CDPs are concerned? Things
are better AFAIC.

thanks

r

--
Nothing beats the bandwidth of a station wagon filled with DLT tapes.

jw wrote:

Perhaps, then what is at issue is how the original analog signal gets
bandwidth limited? Is it possible that certain methods of bandwidth limitimg
are responsible for the nastiness present in the higher frequencies,
particularly in older CDs?


Any technology can be poorly implemented, so it is possible. However,
there is really no excuse for having poor anti-aliasing filters in front
of the A-D converter, since that filter can be built with tight
tolerances because cost is not as big an issue as in the players.

In any event, mastering plays a much bigger role when it comes to adding
nastiness to the end result. Of course, there are many of us who do not
think that as a group the older CD's sound nasty at all in the high
frequencies, so your assumption may not be valid in general.

Nowadays, oversampling ADC's and higher sampling rates have made that
filter much easier to design. Do you still find the newer CD's nasty in
the higher frequencies?

On 6/20/04 1:41 PM, in article E%jBc.62919$2i5.57513@attbi_s52, "chung"
wrote:

jw wrote:

Perhaps, then what is at issue is how the original analog signal gets
bandwidth limited? Is it possible that certain methods of bandwidth limitimg
are responsible for the nastiness present in the higher frequencies,
particularly in older CDs?


Any technology can be poorly implemented, so it is possible. However,
there is really no excuse for having poor anti-aliasing filters in front
of the A-D converter, since that filter can be built with tight
tolerances because cost is not as big an issue as in the players.

While I would agree - how good is good? There are any number of different
ADC's on the market - from inexpensive sound cards to machines costing
$10,000 and more. Some of the things are the number of channels and the
types of convenience features - but some are the technical specifications.

Given the downmixing and so on, as well as for posterity - it is in
everyone's interest to record at a much, much, much higher fidelity than
would be practically affordable for playback - but that is not always done.

In any event, mastering plays a much bigger role when it comes to adding
nastiness to the end result. Of course, there are many of us who do not
think that as a group the older CD's sound nasty at all in the high
frequencies, so your assumption may not be valid in general.

Whew. I suppose it dpeends upon what you like to listen to. I really like
Elvis Costello and also like Led Zeppelin. IN the former it was recorded so
poorly that it is all but unremasterable (despite valiant efforts by Rhino
Records) - and Led Zeppelin sounded much better on LP than the CD's I had
(and the remasters are even worse!).

I guess all I am saying - it "as a group" springs to mind as "which group
would that be?" I have some real clinkers most of them are c. 1985...

Nowadays, oversampling ADC's and higher sampling rates have made that
filter much easier to design. Do you still find the newer CD's nasty in
the higher frequencies?

Myself - I find that the discs claiming to be better sounding generally are
- and the remasters of older less compentantly mastered CD's sound better.

Tori Amos - The originals sounded great - and the main difference in the
remasters is that noise floor seems to be a touch lower bringing some piano
overtones to the fore much better than before - as one example.

jw wrote:
Perhaps, then what is at issue is how the original analog
signal gets bandwidth limited? Is it possible that certain
methods of bandwidth limitimg are responsible for the
nastiness present in the higher frequencies, particularly in
older CDs?

Yes, it's possible to screw up the design of either the anti-
aliasing filter (used in the A-D stage) or the anti-imaging
filter (used in the D-A conversion), but to do so would be in
total ignorance of techniques that have been in play for a very
long time.

It's tough, for example, to implement an appropriate steep
filter with minimal effects in the audio band using analog
design techniques. And,m if you ever COULD implement such a
filter, you'd be at the mercy of impossible-to-achieve component
tolerances.

And that's PRECISELY the reason why nobody does it that way.
That's the WHOLE point of oversampling. You oversample your
imcoming stream at, say, 64x times the base sample rate and
implement your brickwall filter in the digital domain. You have
enormous felxibility in the design and are not subject at all
to issues like component tolerances and such. Instead of having
to worry about aliasing and imaging components just above 20 kHz,
you only have to worry about them at just above 64 X 20 kHz, or
1.3 MEGAhertz.

But, this is old stuff: I don't know of a single 44.1 kHz product
on the market that does NOT use oversampling.

More to the point about the falacy of the original poster's
assertions: I would challenge him and anyone else with similar
assertions to, fact, present the problems being asserted. The
original poster talks about "digital high frequency distortion."
If the problem exists as asserted, this distortion should be
TRIVIAL to measure. It should be ENORMOUS. Where, then, are the
distortion measurements that are so huge? Where are the pictures
of the stair-steps, of the connect-the-dot waveforms, of the
wandering and utterly imprecise and quantized phase problems?

(hint: they aren't there!)

+---------------------------------------+
| Dick Pierce |
| Professional Audio Development |
| (1) 781/826-4953 Voice and FAX |
| |
+---------------------------------------+

On 6/20/04 8:27 PM, in article 7YpBc.64321$2i5.24596@attbi_s52, "Dick
Pierce" wrote:

And that's PRECISELY the reason why nobody does it that way.
That's the WHOLE point of oversampling. You oversample your
imcoming stream at, say, 64x times the base sample rate and
implement your brickwall filter in the digital domain. You have
enormous felxibility in the design and are not subject at all
to issues like component tolerances and such. Instead of having
to worry about aliasing and imaging components just above 20 kHz,
you only have to worry about them at just above 64 X 20 kHz, or
1.3 MEGAhertz.

The trade off, as there is one, is the more you over-sample the less
tolerant your system is to timing errors/jitter in the bit stream.

That is why after some brief forays into truly huge oversampling (64x and so
on)- the more moderate 4-8x won out.

On Sun, 20 Jun 2004 16:47:20 GMT, "Rich.Andrews"

..... some stuff deleted.......
Dick,

Putting aside such malarky like "stuff between the samples" etc., I have
heard a definite difference between CD players. I call it digital grunge.
Is the better sound experienced with well engineered, pricier products
due to better D-As, better DSP processing or ????

What happened between 1990 and 2002 as far as CDPs are concerned? Things
are better AFAIC.

My experience is quite limited, but I was able to compare two CD
players for about a month. One was an old NAD player, from about 1985,
and the other is a Linn Numerik/Karik combo, with the new power
supplies etc. One sold for about $300, the other about $4000. I set
them both up so that playback levels were within 0.1 db, put the same
CD's into each (starting at the same time), and listened for hours,
switched back and forth etc. I even had a random switcher so I
couldn't tellwhich was playing, and had to figure it out myself, and
then compare with the (hidden) indicator. My guesses were essentially
random, I could not determine which was playing (50-100 tests,
whenever I wanted, whatever music, for whatever time interval).
I then used my equipment to see what the differences were,
especially for things like jitter. The differences were beyond the
ability of my HP 3581A wave analyzer (90 db dynamic range), and only
by doing some special circuitry that would reach down to -100 to -110
db was I able to see much difference.
The extra "stuff" that was different was very close to the tested
frequencies, so audio masking would have made it impossible for me to
discern.
Before I did any serious testing I DID hear differences. Afterwards
I DIDN'T! My expectations changed my perceptions of sound.
I have noticed that the audio quality (subjective) and overall
effect of listening to a CD changes dramatically from one listening
session to another. This is probably heresy for this newsgroup, but it
is my opinion that my physical, psychological and spiritual state have
many orders of magnitude (that's POWERS of 10) more effect on sound
than what you could possible hear between REASONABLY designed CD
players. On the other hand, things like phono cartridges, and speakers
have a very big difference. Although you might think I have tin ears,
I am very picky about those electromechanical thingies like
cartridges.
Because of the influence of my mental state, I tend to trust the
equipment more than my ears, especially when I know what kind of
measured "junk" causes my ears grief.
So my question to you is how do you know (RELIABLY) that your
preference for one type of equipment over another is not the result of
mental state, expectations, hype, magic (any high technology is
indistinguishable from magic), coffee, stress, the cool look of the
equipment, etc. Could you get more than random guesses if you set up
your comparison as above? It's damned hard to make a good testing
setup! It has to be very well executed, or you'll believe the test
setup has confounded the test itself.

-Paul

.................................................. ..............
Paul Guy
Somewhere in the Nova Scotia fog

Paul Guy wrote in :

On Sun, 20 Jun 2004 16:47:20 GMT, "Rich.Andrews"

..... some stuff deleted.......
Dick,

Putting aside such malarky like "stuff between the samples" etc., I have
heard a definite difference between CD players. I call it digital
grunge.
Is the better sound experienced with well engineered, pricier products
due to better D-As, better DSP processing or ????

What happened between 1990 and 2002 as far as CDPs are concerned?
Things are better AFAIC.

My experience is quite limited, but I was able to compare two CD
players for about a month. One was an old NAD player, from about 1985,
and the other is a Linn Numerik/Karik combo, with the new power
supplies etc. One sold for about $300, the other about $4000. I set
them both up so that playback levels were within 0.1 db, put the same
CD's into each (starting at the same time), and listened for hours,
switched back and forth etc. I even had a random switcher so I
couldn't tellwhich was playing, and had to figure it out myself, and
then compare with the (hidden) indicator. My guesses were essentially
random, I could not determine which was playing (50-100 tests,
whenever I wanted, whatever music, for whatever time interval).
I then used my equipment to see what the differences were,
especially for things like jitter. The differences were beyond the
ability of my HP 3581A wave analyzer (90 db dynamic range), and only
by doing some special circuitry that would reach down to -100 to -110
db was I able to see much difference.
The extra "stuff" that was different was very close to the tested
frequencies, so audio masking would have made it impossible for me to
discern.
Before I did any serious testing I DID hear differences. Afterwards
I DIDN'T! My expectations changed my perceptions of sound.
I have noticed that the audio quality (subjective) and overall
effect of listening to a CD changes dramatically from one listening
session to another. This is probably heresy for this newsgroup, but it
is my opinion that my physical, psychological and spiritual state have
many orders of magnitude (that's POWERS of 10) more effect on sound
than what you could possible hear between REASONABLY designed CD
players. On the other hand, things like phono cartridges, and speakers
have a very big difference. Although you might think I have tin ears,
I am very picky about those electromechanical thingies like
cartridges.
Because of the influence of my mental state, I tend to trust the
equipment more than my ears, especially when I know what kind of
measured "junk" causes my ears grief.
So my question to you is how do you know (RELIABLY) that your
preference for one type of equipment over another is not the result of
mental state, expectations, hype, magic (any high technology is
indistinguishable from magic), coffee, stress, the cool look of the
equipment, etc. Could you get more than random guesses if you set up
your comparison as above? It's damned hard to make a good testing
setup! It has to be very well executed, or you'll believe the test
setup has confounded the test itself.

-Paul

.................................................. .............
Paul Guy
Somewhere in the Nova Scotia fog


Paul,

I own two Denon DCD1520 players and a McIntosh MCD7008. When I bought the
MCD7008, I wasn't expecting anything. I was interested in the changer.
The audible differences between the Denon and the McIntosh are not subtle.
A friend of mine compared a cheap 1 year old Sony to his MCD7007 player
and heard significant differences. He then purchased a new McIntosh CDP
to replace the Sony.

The audible differences are probably most pronounced with good orchestral
program material generally found on labels like Telarc, EMI, and DG.

r

--
Nothing beats the bandwidth of a station wagon filled with DLT tapes.

"Dick Pierce"
But, this is old stuff: I don't know of a single 44.1 kHz
product
on the market that does NOT use oversampling.


Richard, I am not arguing with you, simply looking for a better
understanding, like so many others.
Can you please explain this/these then:
John
http://www.sakurasystems.com/articles/Kusunoki.html
http://www.audionote.co.jp/digital/essay.htm

Well, don't we know it's a sine wave? Anything else has a higher frequency
component.


"Christopher Key" wrote in message
news:X5DAc.50592$Hg2.42246@attbi_s04...
"Mike Prager" wrote in message
news:z9tAc.66738$HG.63024@attbi_s53...
Christopher Key wrote:

"maxdm" wrote in message
...
Filters just 'connect the dots'

This is sufficient for perfect reproduction.

An excellent analogy, used in a recently posted link is to imagine a
circle.
Only three points on the circumference are required to fully define
it,
and
any more are an unnecessary waste of bandwidth.

True enough, but you also must know it's a circle. The
information content of that knowledge is nontrivial, and
without it, an infinity of shapes could be defined by three
points.

Good point, although the fact that the signal is band limited means that
you
do know a certain amount about the signal, ie that it can be expressed as
the sum of a set of sinusoids up to a given frequency.

So maybe that's not such a great analogy to discribe
reproduction of an audio waveform.


I do accept that it isn't a precise analogy, but thought nonetheless that
it
was pretty good for showing people of a non mathematical background
(specifically the OP) roughly what was going on.

Chris Key

On 6/20/04 11:07 PM, in article , "Paul Guy"
wrote:

So my question to you is how do you know (RELIABLY) that your
preference for one type of equipment over another is not the result of
mental state, expectations, hype, magic (any high technology is
indistinguishable from magic), coffee, stress, the cool look of the
equipment, etc. Could you get more than random guesses if you set up
your comparison as above? It's damned hard to make a good testing
setup! It has to be very well executed, or you'll believe the test
setup has confounded the test itself.

This is why the only valid testing is long term testing. If I have had a
stressful day, it takes awhile before music - no matter what playback method
or if it is a live concert - will move me. Though if in the right mood, it
might move me to become teary.

My "subjective" method is that there are some recordings that will move me
emotionally if I play them - and if the playback stack doesn't mangle them
too badly. The ones that help with this, stay.

I lost faith in simple ABX testing because it didn't seem to match how I
would actually USE the equipment.

In article KGXBc.160040$Ly.55785@attbi_s01,
"Midlant" wrote:

"Dick Pierce"
But, this is old stuff: I don't know of a single 44.1 kHz
product
on the market that does NOT use oversampling.


Richard, I am not arguing with you, simply looking for a better
understanding, like so many others.
Can you please explain this/these then:
John
http://www.sakurasystems.com/articles/Kusunoki.html
http://www.audionote.co.jp/digital/essay.htm


You can add Audio Research to the list.

http://www.audioresearch.com/CD3.htm

"While using the latest 24/192-capable Crystal DAC, the CD3 does not
upsample, because our empirical research shows sonic compromise is
unavoidable due to sample rate manipulation and approximating errors."

On 6/22/04 10:50 AM, in article KGXBc.160040$Ly.55785@attbi_s01, "Midlant"
wrote:

"Dick Pierce"
But, this is old stuff: I don't know of a single 44.1 kHz
product
on the market that does NOT use oversampling.


Richard, I am not arguing with you, simply looking for a better
understanding, like so many others.
Can you please explain this/these then:
John
http://www.sakurasystems.com/articles/Kusunoki.html
http://www.audionote.co.jp/digital/essay.htm


Has anyone used this stuff? It seems to spring out of the Japanese
"minimalist" esthetic - apparently some really like it - and since one is
not supposed to be able to hear above 22.05kHz anyway - why even filter it,
eh? :-)

On 21 Jun 2004 03:05:33 GMT, Bromo wrote:

On 6/20/04 8:27 PM, in article 7YpBc.64321$2i5.24596@attbi_s52, "Dick
Pierce" wrote:

And that's PRECISELY the reason why nobody does it that way.
That's the WHOLE point of oversampling. You oversample your
imcoming stream at, say, 64x times the base sample rate and
implement your brickwall filter in the digital domain. You have
enormous felxibility in the design and are not subject at all
to issues like component tolerances and such. Instead of having
to worry about aliasing and imaging components just above 20 kHz,
you only have to worry about them at just above 64 X 20 kHz, or
1.3 MEGAhertz.

The trade off, as there is one, is the more you over-sample the less
tolerant your system is to timing errors/jitter in the bit stream.

That is why after some brief forays into truly huge oversampling (64x and so
on)- the more moderate 4-8x won out.

Actually, the most popular current systems use *at least* 128x
oversampling - the so-called 'single-bit' or 'delta-sigma' systems.
You are thinking of multi-bit systems, which have *never* exceeded
16x, due to basic technical limitations.

--

Stewart Pinkerton | Music is Art - Audio is Engineering

jw wrote:

Perhaps, then what is at issue is how the original analog signal gets
bandwidth limited? Is it possible that certain methods of bandwidth limitimg
are responsible for the nastiness present in the higher frequencies,
particularly in older CDs?

That's a pretty sweeping generalisation. If you can get hold of a 1983
copy of Dire Straits 'Love Over Gold' (one of the first CDs ever
made), you'll find that the sound quality will easily match any modern
production. It was *always* possible to 'do it right' with CD, but
just as was the case with vinyl the '70s, it didn't happen all that
often in the mainstream.....................
--

Stewart Pinkerton | Music is Art - Audio is Engineering

On Thu, 17 Jun 2004 14:05:17 GMT, (maxdm)
wrote:

I am sure that math apart, most people who have the capacity to listen
critically to sound will notice a difference in 192Khz audio compared
to 44.1 even if the speakers or headphones or amplifiers are frequency
limited to20 Khz.

Anyone out there with listening experience ?

Yup, loads of us - and there is as yet *zero* evidence that 24/192
sounds different from 16/44 - when that is the *only* difference.

The real test is to compare 24/192 to 16/48, since that is a very
simple mathematical transform, and will introduce no peculiar
artifacts. As noted, there is as yet *zero* evidence that these sound
different.
--

Stewart Pinkerton | Music is Art - Audio is Engineering

Bromo wrote in message news:Nq9Cc.91780$eu.10945@attbi_s02...
On 6/22/04 10:50 AM, in article KGXBc.160040$Ly.55785@attbi_s01, "Midlant"
wrote:

"Dick Pierce"
But, this is old stuff: I don't know of a single 44.1 kHz
product
on the market that does NOT use oversampling.


I should have satated: there are no competent products...

Richard, I am not arguing with you, simply looking for a better
understanding, like so many others.
Can you please explain this/these then:
John
http://www.sakurasystems.com/articles/Kusunoki.html
http://www.audionote.co.jp/digital/essay.htm


Has anyone used this stuff? It seems to spring out of the Japanese
"minimalist" esthetic - apparently some really like it - and since one is
not supposed to be able to hear above 22.05kHz anyway - why even filter it,
eh? :-)

Because it's a real bad idea?

One of the overarching principles of sampled data is that the data
stream contains the baseband audio and EVERY image of that baseband
out to infinity, and the entire point to anti-imaging filtering is
to remove all the images of the original, and leave only the original.

The problem with NOT filtering them out is that every image contains
the same total energy as the baseband original. That means that if you
can imagine the spectrum of the music contained in the 0-20kHz band,
there is a mirror image of that in the 24-44 kHz band, another non-
mirror image in the 44-64 kHz band, a second mirror image in the 68-88
kHz band, a second non-mirro4ed image in the 88-108 kHz band, and so on
all the way up to infinity (ignoring practical bandwidth limitations).
Every one of those 20 kHz bands has exactly the same total energy as
the original 20 kHz band.

(now, before someone comes running in waving there hands claiming
I said these things have infinite power output because the images
go out to infinity, look at what I said: their bandwidth is
ultimately limited by practical bandwidth limitations.
Mathematically, the images DO go out to infinity, and if the
equipment had infinite bandwidth but finite power output, then
each image, and the original, would have infinitesimal power)

Now, imagine all your downstream equipment being fed all this energy,
much of which is up in the RF reqion. That excess energy is wasting
power, interfering with the proper operation of the electronics,
combining with other sources of energy and getting modulated down
wherte it can be heard, heating tweeter voice coils and a whole
lot more.

And, because you DO have all the images present, which you WILL get
if you omit the antiimaging reconstruction filter, only then will
you see the "classic" stair-step output that audiophilic digiphobes
around the globe have been pointing as the source of all digital
evil.

So, in effect, what this equipment says is "if you don't like your
preconceptions of digital, try this, which is precisely your worst
nightmare."

Only in high end audio to you see this sort of insanity.

In article iy8Cc.78427$Hg2.37239@attbi_s04,
MINe 109 wrote:

"Dick Pierce"
But, this is old stuff: I don't know of a single 44.1 kHz
product
on the market that does NOT use oversampling.


Richard, I am not arguing with you, simply looking for a better
understanding, like so many others.
Can you please explain this/these then:
John
http://www.sakurasystems.com/articles/Kusunoki.html
http://www.audionote.co.jp/digital/essay.htm


You can add Audio Research to the list.

http://www.audioresearch.com/CD3.htm

"While using the latest 24/192-capable Crystal DAC, the CD3 does not
upsample, because our empirical research shows sonic compromise is
unavoidable due to sample rate manipulation and approximating errors."

The Crystal Semiconductor DACs are all (as best as I can tell) of the
delta-sigma design. This type of DAC architecture generates output
(analog) samples at a much higher rate than the incoming digital
samples.

Audio Research's statement indicates that they do not implement their
own oversampling/upsampling logic _before_ the signal is fed into the
actual DAC chip. They don't need to, as the DAC chip used in the CD3
does the oversampling/upsampling internally, as part of the
delta-sigma modulation process.

--
Dave Platt AE6EO
Hosting the Jade Warrior home page: http://www.radagast.org/jade-warrior
I do _not_ wish to receive unsolicited commercial email, and I will
boycott any company which has the gall to send me such ads!

Bromo wrote:



On 6/20/04 11:07 PM, in article , "Paul Guy"
wrote:

So my question to you is how do you know (RELIABLY) that your
preference for one type of equipment over another is not the result of
mental state, expectations, hype, magic (any high technology is
indistinguishable from magic), coffee, stress, the cool look of the
equipment, etc. Could you get more than random guesses if you set up
your comparison as above? It's damned hard to make a good testing
setup! It has to be very well executed, or you'll believe the test
setup has confounded the test itself.

This is why the only valid testing is long term testing. If I have had a
stressful day, it takes awhile before music - no matter what playback method
or if it is a live concert - will move me. Though if in the right mood, it
might move me to become teary.

My "subjective" method is that there are some recordings that will move me
emotionally if I play them - and if the playback stack doesn't mangle them
too badly. The ones that help with this, stay.

I lost faith in simple ABX testing because it didn't seem to match how I
would actually USE the equipment.

You mean you don't plan on listening to it? That's the beauty of any type of
bias controlled listening (ABX or otherwise); it requires that you "listen"
with only your ears leaving out all the personal, social and written middlemen.

Bromo wrote:

This is why the only valid testing is long term testing.* If I have had a
stressful day, it takes awhile before music - no matter what playback
method
or if it is a live concert - will move me.* Though if in the right mood, it
might move me to become teary.

My "subjective" method is that there are some recordings that will move me
emotionally if I play them - and if the playback stack doesn't mangle them
too badly.* The ones that help with this, stay.

I lost faith in simple ABX testing because it didn't seem to match how I
would actually USE the equipment.

You are confusing two very different things:

1) your optimal mental state for enjoying/appreciating music, and

2) the conditions which maximize your ability to discern subtle sonic
differences.

Relaxed, long-term listening may well be best for music enjoyment. But for
discerning subtle sonic differences, you are simply factually wrong. As any
basic psychoacoustics textbook will tell you, a quick-switching comparison
will be more sensitive to such differences than a long-term comparison. To
argue otherwise is the equivalent of claiming that 1+1=2.1.

bob

__________________________________________________ _______________
Get fast, reliable Internet access with MSN 9 Dial-up – now 3 months FREE!
http://join.msn.click-url.com/go/onm...ave/direct/01/

Bromo wrote:

On 6/20/04 11:07 PM, in article , "Paul Guy"
wrote:

So my question to you is how do you know (RELIABLY) that your
preference for one type of equipment over another is not the result of
mental state, expectations, hype, magic (any high technology is
indistinguishable from magic), coffee, stress, the cool look of the
equipment, etc. Could you get more than random guesses if you set up
your comparison as above? It's damned hard to make a good testing
setup! It has to be very well executed, or you'll believe the test
setup has confounded the test itself.

This is why the only valid testing is long term testing. If I have had a
stressful day, it takes awhile before music - no matter what playback method
or if it is a live concert - will move me. Though if in the right mood, it
might move me to become teary.

My "subjective" method is that there are some recordings that will move me
emotionally if I play them - and if the playback stack doesn't mangle them
too badly. The ones that help with this, stay.

I lost faith in simple ABX testing because it didn't seem to match how I
would actually USE the equipment.

So, as an RF engineer, have you lost faith in all measuring equipment
because that was not how you would use the devices under test? For
example, you would lose faith in a distortion analyzer because no one
listens to sine waves? Or you would lose faith in sqaure wave responses
because these devices do not recieve square waves in real life?

I find this statement by an alleged engineer utterly astonishing.

On Sat, 19 Jun 2004 16:50:43 GMT, (maxdm)
wrote:

Since you appear to have studied the matter in depth, could you please
explain or comment on the following points:
let's imagine a digital system with a very low sampling rate.

Irrelevant, but let's carry on...................

If we record a sine wave that is not perfect and has a 'glitch'
(energy in a higher spectrum) that occurs inbetween samples, the
lowpass filter will limit the slew rate and therefore if the sine wave
is near the bandwith limit of the said brickwall filter this 'glitch'
should be filtered out of the signal right?

Nothing to do with slew rate, but the higher frequency components (aka
'the glitch') will indeed be removed. Correctly so, as a sine wave *by
definition* has no glitches.

is it possible that a waveform that has a short sharp transient
inbetween samples could charge a capacitor in the brickwall filter or
in the ADC and therefore cause mistracking?

No, if it's a decently implemented ADC.

if the waveform has a very short but definite 'lump' lasting a
fraction of a millisecond the filter would eliminate the lump, but the
lump would charge the filter with a positive or negative voltage
therefore offsetting the signal read by the adc on the next samples.

No, it wouldn't.

If we have a system of any kind meant to reproduce sound, from a
practical engineering standpoint, which is meant to be mass produced
and not a one-off costing huge sums of money, doesn't common sense
dictate that the system should, if possible cost-wise, never operate
near its theoretical limits?

Not if you can use techniques like oversampling to retain the
precision stuff in the digital domain.

what you say about analog digital conversion makes sense in a perfect
system that behaves in a perfect way.

My experience is that the filters in most home and pro AD converters
do not limit bandwidth 100%, and when they come close there are side
effects.

If that is true, then you have only experience of poor products....

raising the bandwith takes the brickwall frequency out of the audio
spectrum and any imperfections in the filter due to it being an
electronic device and not a mathematical model.

That's the nice thing about digital filters - they *are* perfect - or
at least they are *exactly* the same in *every* production unit.

Why do experts in the recording and mastering field, who have been
used to making the records you would play on your home system claim
that the resolution on analog tape machines and laquer discs is
superior to analog?

Because they are :

a) (deleted at the request of the moderators)......... :-)

b) a tiny but very vocal minority of such 'experts', who have a
commercial axe to grind

c) technically ignorant

d) trying to to justify their preference for the sound of the added
*distortions* of analogue techniques

Choose from any of the above................

I have recorded on both, and have gone to the trouble of modifying my
ADC's which use the akm chips to make sure what I was hearing was not
due to the analog stages, and analog is more distorted from a thd and
low frequency standpoint but it has imaging and depth that digital
recording has not achieved.

Due to *added* artifacts.................

you hear this effect more in multitrack recording where the
side-effects are summed because of the summing of many discrete
sounds.
Higher sampling rates sound better.

Prove it.
--

Stewart Pinkerton | Music is Art - Audio is Engineering

On 6/23/04 10:57 PM, in article AqrCc.97729$Sw.3328@attbi_s51, "chung"
wrote:

I lost faith in simple ABX testing because it didn't seem to match how I
would actually USE the equipment.

So, as an RF engineer, have you lost faith in all measuring equipment
because that was not how you would use the devices under test? For
example, you would lose faith in a distortion analyzer because no one
listens to sine waves? Or you would lose faith in sqaure wave responses
because these devices do not recieve square waves in real life?

I find this statement by an alleged engineer utterly astonishing.

No, sir, but we as a group are forgetting that simulation (whether that be
on a computer or using sine waves to test a piece of equipment that is
designed to reproduce music) is NOT how the piece of equipment is *used.*

I have not lost faith in measuring equipment, but I do know that putting
full faith in tests that are not fully representative of actual practice, is
a mistake. It may indicate something - but it may not be accurate.

For instance in the 1920's there were ABX demonstrations where an orchestra
played music, stopped and a 78 was played - and the audience didn't notice
in the rapid ABX fashion.

Does this mean that 78's sounded the same as live music? Of course not, but
it may reveal a weakness in a test.

IN this manner, I will take the data from a test, but think about what it
does mean, and what it does not.

To do less would be to be sloppy.