"Karl Uppiano" wrote in message
...

"Randy Yates" wrote in message
...
"Karl Uppiano" writes:

"Arny Krueger" wrote in message
...
"Karl Uppiano" wrote in message

"François Yves Le Gal" wrote in message
...
On Sun, 01 Feb 2004 12:53:18 -0800, chung
wrote:

One, DSD/SACD proponents
claim the much wider bandwidth over CD's, and filtering will
reduce
significantly that claimed advantage.

Having a gentle low pass filter at 60 or 100 KHz doesn't
significantly reduce SACD's bandwith. You get more than PCM 96 or
192 in both cases!

I would start a 3 dB/octave rolloff at 20kHz or so.

Ironically, most if not all people can't hear the difference a
brickwall
filter at 16 KHz makes, if the filter is well-designed.

Don't believe me?

Listen for yourself at
http://www.pcabx.com/technical/low_pass/index.htm .

I believe you. I realize it isn't considered state of the art anymore,
but I
wonder how many people realize that FM stereo uses a 38 kHz sample
rate?

Actually, the L-R signal is AM-DSB-SC-modulated on a 38 kHz subcarrier
and
a 19 kHz pilot tone added. This, along with the L+R baseband signal, is
then FM modulated onto the carrier.

The balanced modulator is mathematically equivalent to a sampled system in
which the left and right channels are alternately sampled at a 38 kHz rate
and used to directly modulate the carrier. The 19 kHz pilot is derived
from
the 38 kHz sample frequency and injected onto the carrier at 9%
modulation.
Even without the math, it makes sense if you think about it: Alternately
sampling a mono signal (L = R) will give you no subcarrier (L - R = 0),
which is exactly what happens with AM-DSB-SC. The analysis is a little
more
complicated when a stereo signal is encoded, but the results are identical
using either approach.

The balanced modulator approach was used in FM stereo generators until the
mid 1970's or so, when the alternately sampled approach became possible
with
the advent of TTL and FET analog switches. The switched stereo generators
require less maintenance. The phase-lock-loop demodulators in most
receivers
is the same system in reverse.

Nyquist's sampling theorem applies to FM stereo. Therefore, 19 kHz is the
highest audio frequency theoretically possible, but you need a guard band
for the pilot, so 15 kHz is typical, with some extremely good stereo
generators yielding 16 to 18 kHz.

By the way, the transition band for the analog anti-aliasing filters
(required for any FM stereo generator) need to go from 0 dB at 15 kHz to -60
dB at 19 kHz. Stereo decoders in most receivers usually don't bother to
control the high frequency energy very well, which is why many cassette
decks with Dolby noise reduction have "MPX filters" to prevent this energy
from messing up the Dolby encoding.

chung wrote in message news:ec705$401ebdd9$c247604



If you take into account tolerances of the 5 caps, you would have a
noticeable ripple in the passband, as well as mismatches between L/R.
Plus the cost of the 10 capacitors is not insignificant. And more if you
have 5 channels.

The much better way is simply apply digital filtering, or use better
dithering schemes, but then we are back to LPCM, and not DSD .


I agree. If you look at the data you will see that with perfect
component values, the flatness is: -0.1/-0.2 which 0.1 dB p-p. Using
real componet will increase the ripple. A pratical best flatness spec.
(for real world components) would be +0.1/-0.3 dB.

Using 1% tolerance components, I ran a Monte Carlo analysis (on
100,000 units). I got the following results:

+0.20/-0.40 dB flatness 100% yield

+0.15/-0.35 dB flatness 99.91% yield

+0.10/-0.30 dB flatness 98.80% yield


It would be fairly expensive to use 1% tolerance components. Probably
add $3 or $4 to the cost of manufacture. For an SACD player that sells
for $500 to $1500, that would be acceptable.

However, I don't know if a flatness spec of +/- 0.2dB would be
acceptable to high-end buyers?

Bob Stanton

chung wrote in message news:ec705$401ebdd9$c247604



If you take into account tolerances of the 5 caps, you would have a
noticeable ripple in the passband, as well as mismatches between L/R.
Plus the cost of the 10 capacitors is not insignificant. And more if you
have 5 channels.

The much better way is simply apply digital filtering, or use better
dithering schemes, but then we are back to LPCM, and not DSD .


I agree. If you look at the data you will see that with perfect
component values, the flatness is: -0.1/-0.2 which 0.1 dB p-p. Using
real componet will increase the ripple. A pratical best flatness spec.
(for real world components) would be +0.1/-0.3 dB.

Using 1% tolerance components, I ran a Monte Carlo analysis (on
100,000 units). I got the following results:

+0.20/-0.40 dB flatness 100% yield

+0.15/-0.35 dB flatness 99.91% yield

+0.10/-0.30 dB flatness 98.80% yield


It would be fairly expensive to use 1% tolerance components. Probably
add $3 or $4 to the cost of manufacture. For an SACD player that sells
for $500 to $1500, that would be acceptable.

However, I don't know if a flatness spec of +/- 0.2dB would be
acceptable to high-end buyers?

Bob Stanton

chung wrote in message news:ec705$401ebdd9$c247604



If you take into account tolerances of the 5 caps, you would have a
noticeable ripple in the passband, as well as mismatches between L/R.
Plus the cost of the 10 capacitors is not insignificant. And more if you
have 5 channels.

The much better way is simply apply digital filtering, or use better
dithering schemes, but then we are back to LPCM, and not DSD .


I agree. If you look at the data you will see that with perfect
component values, the flatness is: -0.1/-0.2 which 0.1 dB p-p. Using
real componet will increase the ripple. A pratical best flatness spec.
(for real world components) would be +0.1/-0.3 dB.

Using 1% tolerance components, I ran a Monte Carlo analysis (on
100,000 units). I got the following results:

+0.20/-0.40 dB flatness 100% yield

+0.15/-0.35 dB flatness 99.91% yield

+0.10/-0.30 dB flatness 98.80% yield


It would be fairly expensive to use 1% tolerance components. Probably
add $3 or $4 to the cost of manufacture. For an SACD player that sells
for $500 to $1500, that would be acceptable.

However, I don't know if a flatness spec of +/- 0.2dB would be
acceptable to high-end buyers?

Bob Stanton

chung wrote in message news:ec705$401ebdd9$c247604



If you take into account tolerances of the 5 caps, you would have a
noticeable ripple in the passband, as well as mismatches between L/R.
Plus the cost of the 10 capacitors is not insignificant. And more if you
have 5 channels.

The much better way is simply apply digital filtering, or use better
dithering schemes, but then we are back to LPCM, and not DSD .


I agree. If you look at the data you will see that with perfect
component values, the flatness is: -0.1/-0.2 which 0.1 dB p-p. Using
real componet will increase the ripple. A pratical best flatness spec.
(for real world components) would be +0.1/-0.3 dB.

Using 1% tolerance components, I ran a Monte Carlo analysis (on
100,000 units). I got the following results:

+0.20/-0.40 dB flatness 100% yield

+0.15/-0.35 dB flatness 99.91% yield

+0.10/-0.30 dB flatness 98.80% yield


It would be fairly expensive to use 1% tolerance components. Probably
add $3 or $4 to the cost of manufacture. For an SACD player that sells
for $500 to $1500, that would be acceptable.

However, I don't know if a flatness spec of +/- 0.2dB would be
acceptable to high-end buyers?

Bob Stanton

"François Yves Le Gal" wrote in message

On Tue, 03 Feb 2004 04:21:57 GMT, "Karl Uppiano"
wrote:

Nope, FM stereo is a sampled system.

Well, it's not according to my book, as the 38 KHz carrier isn't a
sampling frequency.

Just another page in the big book of things you're ignorant about, François.

The FM stereo system is generally demodulated these days by means of a
circuit that could easily pass for an analog demultiplexer. It well known
among technical folk who have studied it with sufficient depth, that there
is an interesting duality about FM stereo.

FM Stereo is as Randy says:

"Actually, the L-R signal is AM-DSB-SC-modulated on a 38 kHz subcarrier and
a 19 kHz pilot tone added. This, along with the L+R baseband signal, is
then FM modulated onto the carrier."

and it can be demodulated accordingly.

But, FM Stereo can also be viewed as Karl says:

"Left and right sampled alternately at 38 kHz is the same as L+R and
L-R fed into a balanced modulator with a suppressed carrier. The balanced
modulator approach was used in FM stereo generators until the mid 1970's or
so, when the alternately sampled approach became possible with the advent of
TTL and FET analog switches. Both systems also inject a 19 kHz pilot tone at
a 9% modulation level to turn on the "stereo" light, and to give
phase-lock-loop stereo decoders (also sampled systems) something to lock
onto."

Now I know from years of experience that in the posturing-ridden one-track
mind of the anonymous internet troll that posts as "François Yves Le Gal"
both of these statements can't be true at the same time. But, in the real
world, they are. The reasons why can be instructive to those with the mental
powers and sufficient training and experience to properly contemplate them.
Sorry to leave you out, "François Yves Le Gal" or whoever you really are.

"François Yves Le Gal" wrote in message

On Tue, 03 Feb 2004 04:21:57 GMT, "Karl Uppiano"
wrote:

Nope, FM stereo is a sampled system.

Well, it's not according to my book, as the 38 KHz carrier isn't a
sampling frequency.

Just another page in the big book of things you're ignorant about, François.

The FM stereo system is generally demodulated these days by means of a
circuit that could easily pass for an analog demultiplexer. It well known
among technical folk who have studied it with sufficient depth, that there
is an interesting duality about FM stereo.

FM Stereo is as Randy says:

"Actually, the L-R signal is AM-DSB-SC-modulated on a 38 kHz subcarrier and
a 19 kHz pilot tone added. This, along with the L+R baseband signal, is
then FM modulated onto the carrier."

and it can be demodulated accordingly.

But, FM Stereo can also be viewed as Karl says:

"Left and right sampled alternately at 38 kHz is the same as L+R and
L-R fed into a balanced modulator with a suppressed carrier. The balanced
modulator approach was used in FM stereo generators until the mid 1970's or
so, when the alternately sampled approach became possible with the advent of
TTL and FET analog switches. Both systems also inject a 19 kHz pilot tone at
a 9% modulation level to turn on the "stereo" light, and to give
phase-lock-loop stereo decoders (also sampled systems) something to lock
onto."

Now I know from years of experience that in the posturing-ridden one-track
mind of the anonymous internet troll that posts as "François Yves Le Gal"
both of these statements can't be true at the same time. But, in the real
world, they are. The reasons why can be instructive to those with the mental
powers and sufficient training and experience to properly contemplate them.
Sorry to leave you out, "François Yves Le Gal" or whoever you really are.

"François Yves Le Gal" wrote in message

On Tue, 03 Feb 2004 04:21:57 GMT, "Karl Uppiano"
wrote:

Nope, FM stereo is a sampled system.

Well, it's not according to my book, as the 38 KHz carrier isn't a
sampling frequency.

Just another page in the big book of things you're ignorant about, François.

The FM stereo system is generally demodulated these days by means of a
circuit that could easily pass for an analog demultiplexer. It well known
among technical folk who have studied it with sufficient depth, that there
is an interesting duality about FM stereo.

FM Stereo is as Randy says:

"Actually, the L-R signal is AM-DSB-SC-modulated on a 38 kHz subcarrier and
a 19 kHz pilot tone added. This, along with the L+R baseband signal, is
then FM modulated onto the carrier."

and it can be demodulated accordingly.

But, FM Stereo can also be viewed as Karl says:

"Left and right sampled alternately at 38 kHz is the same as L+R and
L-R fed into a balanced modulator with a suppressed carrier. The balanced
modulator approach was used in FM stereo generators until the mid 1970's or
so, when the alternately sampled approach became possible with the advent of
TTL and FET analog switches. Both systems also inject a 19 kHz pilot tone at
a 9% modulation level to turn on the "stereo" light, and to give
phase-lock-loop stereo decoders (also sampled systems) something to lock
onto."

Now I know from years of experience that in the posturing-ridden one-track
mind of the anonymous internet troll that posts as "François Yves Le Gal"
both of these statements can't be true at the same time. But, in the real
world, they are. The reasons why can be instructive to those with the mental
powers and sufficient training and experience to properly contemplate them.
Sorry to leave you out, "François Yves Le Gal" or whoever you really are.

"François Yves Le Gal" wrote in message

On Tue, 03 Feb 2004 04:21:57 GMT, "Karl Uppiano"
wrote:

Nope, FM stereo is a sampled system.

Well, it's not according to my book, as the 38 KHz carrier isn't a
sampling frequency.

Just another page in the big book of things you're ignorant about, François.

The FM stereo system is generally demodulated these days by means of a
circuit that could easily pass for an analog demultiplexer. It well known
among technical folk who have studied it with sufficient depth, that there
is an interesting duality about FM stereo.

FM Stereo is as Randy says:

"Actually, the L-R signal is AM-DSB-SC-modulated on a 38 kHz subcarrier and
a 19 kHz pilot tone added. This, along with the L+R baseband signal, is
then FM modulated onto the carrier."

and it can be demodulated accordingly.

But, FM Stereo can also be viewed as Karl says:

"Left and right sampled alternately at 38 kHz is the same as L+R and
L-R fed into a balanced modulator with a suppressed carrier. The balanced
modulator approach was used in FM stereo generators until the mid 1970's or
so, when the alternately sampled approach became possible with the advent of
TTL and FET analog switches. Both systems also inject a 19 kHz pilot tone at
a 9% modulation level to turn on the "stereo" light, and to give
phase-lock-loop stereo decoders (also sampled systems) something to lock
onto."

Now I know from years of experience that in the posturing-ridden one-track
mind of the anonymous internet troll that posts as "François Yves Le Gal"
both of these statements can't be true at the same time. But, in the real
world, they are. The reasons why can be instructive to those with the mental
powers and sufficient training and experience to properly contemplate them.
Sorry to leave you out, "François Yves Le Gal" or whoever you really are.

"chung" wrote in message
vers.com
Robert Stanton wrote:

"Arny Krueger" wrote in message
...
"Robert Stanton" wrote in message
m
chung wrote in message
...
Harry Lavo wrote:



Isn't it interesting that we "subjectivists" here are always
whipsawed by the "objectivists" for thinking that extended
frequency response is a benefit "because it can't be heard".
Then, in defense of DVD-A, the increased noise in the ultrasonic
range is bandied about as making DSD/SACD "inferior".


Actually it's interesting that subjectivists said that they need
the ultrasonic bandwidth to fully perceive music, and yet they
like SACD's despite the much higher (by orders of magnitude)
ultrasonic noise inherent in the SACD format.


I often see ultrasonic noise mentioned as a problem for SACD
players. If ultrasonic noise were really a problem, it could be
easily eliminated with a small, active lowpass filter.

Why wouldn't the manufactures of "high end" SACD players, just
filter it out?

Wouldn't the filter affect the overtones of the music just as much
as it affects the noise?

Yes, it will chop off all overtones above 30kHz. But, we humans can't
hear above 25kHz, so we won't hear the difference.

You and I may agree on that, but one of the touted features of SACD is
the much broader bandwidth, higher than 25 KHz. If you limit it to
25KHz, it could not compete against the other hi-rez formats, such as
24/96 or 24/192, which are flat up to close to half the sampling
frequency.

Agreed.

"chung" wrote in message
vers.com
Robert Stanton wrote:

"Arny Krueger" wrote in message
...
"Robert Stanton" wrote in message
m
chung wrote in message
...
Harry Lavo wrote:



Isn't it interesting that we "subjectivists" here are always
whipsawed by the "objectivists" for thinking that extended
frequency response is a benefit "because it can't be heard".
Then, in defense of DVD-A, the increased noise in the ultrasonic
range is bandied about as making DSD/SACD "inferior".


Actually it's interesting that subjectivists said that they need
the ultrasonic bandwidth to fully perceive music, and yet they
like SACD's despite the much higher (by orders of magnitude)
ultrasonic noise inherent in the SACD format.


I often see ultrasonic noise mentioned as a problem for SACD
players. If ultrasonic noise were really a problem, it could be
easily eliminated with a small, active lowpass filter.

Why wouldn't the manufactures of "high end" SACD players, just
filter it out?

Wouldn't the filter affect the overtones of the music just as much
as it affects the noise?

Yes, it will chop off all overtones above 30kHz. But, we humans can't
hear above 25kHz, so we won't hear the difference.

You and I may agree on that, but one of the touted features of SACD is
the much broader bandwidth, higher than 25 KHz. If you limit it to
25KHz, it could not compete against the other hi-rez formats, such as
24/96 or 24/192, which are flat up to close to half the sampling
frequency.

Agreed.

"chung" wrote in message
vers.com
Robert Stanton wrote:

"Arny Krueger" wrote in message
...
"Robert Stanton" wrote in message
m
chung wrote in message
...
Harry Lavo wrote:



Isn't it interesting that we "subjectivists" here are always
whipsawed by the "objectivists" for thinking that extended
frequency response is a benefit "because it can't be heard".
Then, in defense of DVD-A, the increased noise in the ultrasonic
range is bandied about as making DSD/SACD "inferior".


Actually it's interesting that subjectivists said that they need
the ultrasonic bandwidth to fully perceive music, and yet they
like SACD's despite the much higher (by orders of magnitude)
ultrasonic noise inherent in the SACD format.


I often see ultrasonic noise mentioned as a problem for SACD
players. If ultrasonic noise were really a problem, it could be
easily eliminated with a small, active lowpass filter.

Why wouldn't the manufactures of "high end" SACD players, just
filter it out?

Wouldn't the filter affect the overtones of the music just as much
as it affects the noise?

Yes, it will chop off all overtones above 30kHz. But, we humans can't
hear above 25kHz, so we won't hear the difference.

You and I may agree on that, but one of the touted features of SACD is
the much broader bandwidth, higher than 25 KHz. If you limit it to
25KHz, it could not compete against the other hi-rez formats, such as
24/96 or 24/192, which are flat up to close to half the sampling
frequency.

Agreed.

"chung" wrote in message
vers.com
Robert Stanton wrote:

"Arny Krueger" wrote in message
...
"Robert Stanton" wrote in message
m
chung wrote in message
...
Harry Lavo wrote:



Isn't it interesting that we "subjectivists" here are always
whipsawed by the "objectivists" for thinking that extended
frequency response is a benefit "because it can't be heard".
Then, in defense of DVD-A, the increased noise in the ultrasonic
range is bandied about as making DSD/SACD "inferior".


Actually it's interesting that subjectivists said that they need
the ultrasonic bandwidth to fully perceive music, and yet they
like SACD's despite the much higher (by orders of magnitude)
ultrasonic noise inherent in the SACD format.


I often see ultrasonic noise mentioned as a problem for SACD
players. If ultrasonic noise were really a problem, it could be
easily eliminated with a small, active lowpass filter.

Why wouldn't the manufactures of "high end" SACD players, just
filter it out?

Wouldn't the filter affect the overtones of the music just as much
as it affects the noise?

Yes, it will chop off all overtones above 30kHz. But, we humans can't
hear above 25kHz, so we won't hear the difference.

You and I may agree on that, but one of the touted features of SACD is
the much broader bandwidth, higher than 25 KHz. If you limit it to
25KHz, it could not compete against the other hi-rez formats, such as
24/96 or 24/192, which are flat up to close to half the sampling
frequency.

Agreed.

"François Yves Le Gal" wrote in message
...
On Tue, 03 Feb 2004 04:21:57 GMT, "Karl Uppiano"

wrote:

Nope, FM stereo is a sampled system.

Well, it's not according to my book, as the 38 Khz carrier isn't a
sampling
frequency.

You're in denial, then. Switching generators are "the way FM stereo was
done" right up until DSP processors came along that generate the signals
numerically. All three systems are in daily use now. It doesn't matter
anyway, 19 kHz is the Nyquist frequency regardless of the mechanism used to
generate the signal. The spectra are identical. My original point was this:
All stereo generators require brick wall anti-aliasing filters to keep the
baseband and the subcarrier sidebands from overlapping (aliasing), in
addition to a guard band to protect the 19 kHz pilot. So FM stereo is
strictly limited in bandwidth to about 15 kHz to 18 kHz best case. FWIW, TV
stereo uses a similar system, except it's tied to the horizontal sweep
frequency, which is even lower (15.734 kHz doubled to a sampling rate of
31.468 kHz). Meaning that TV stereo is limited to about 14 kHz. Same type
brick wall AA filters are required.

"François Yves Le Gal" wrote in message
...
On Tue, 03 Feb 2004 04:21:57 GMT, "Karl Uppiano"

wrote:

Nope, FM stereo is a sampled system.

Well, it's not according to my book, as the 38 Khz carrier isn't a
sampling
frequency.

You're in denial, then. Switching generators are "the way FM stereo was
done" right up until DSP processors came along that generate the signals
numerically. All three systems are in daily use now. It doesn't matter
anyway, 19 kHz is the Nyquist frequency regardless of the mechanism used to
generate the signal. The spectra are identical. My original point was this:
All stereo generators require brick wall anti-aliasing filters to keep the
baseband and the subcarrier sidebands from overlapping (aliasing), in
addition to a guard band to protect the 19 kHz pilot. So FM stereo is
strictly limited in bandwidth to about 15 kHz to 18 kHz best case. FWIW, TV
stereo uses a similar system, except it's tied to the horizontal sweep
frequency, which is even lower (15.734 kHz doubled to a sampling rate of
31.468 kHz). Meaning that TV stereo is limited to about 14 kHz. Same type
brick wall AA filters are required.

"François Yves Le Gal" wrote in message
...
On Tue, 03 Feb 2004 04:21:57 GMT, "Karl Uppiano"

wrote:

Nope, FM stereo is a sampled system.

Well, it's not according to my book, as the 38 Khz carrier isn't a
sampling
frequency.

You're in denial, then. Switching generators are "the way FM stereo was
done" right up until DSP processors came along that generate the signals
numerically. All three systems are in daily use now. It doesn't matter
anyway, 19 kHz is the Nyquist frequency regardless of the mechanism used to
generate the signal. The spectra are identical. My original point was this:
All stereo generators require brick wall anti-aliasing filters to keep the
baseband and the subcarrier sidebands from overlapping (aliasing), in
addition to a guard band to protect the 19 kHz pilot. So FM stereo is
strictly limited in bandwidth to about 15 kHz to 18 kHz best case. FWIW, TV
stereo uses a similar system, except it's tied to the horizontal sweep
frequency, which is even lower (15.734 kHz doubled to a sampling rate of
31.468 kHz). Meaning that TV stereo is limited to about 14 kHz. Same type
brick wall AA filters are required.

"François Yves Le Gal" wrote in message
...
On Tue, 03 Feb 2004 04:21:57 GMT, "Karl Uppiano"

wrote:

Nope, FM stereo is a sampled system.

Well, it's not according to my book, as the 38 Khz carrier isn't a
sampling
frequency.

You're in denial, then. Switching generators are "the way FM stereo was
done" right up until DSP processors came along that generate the signals
numerically. All three systems are in daily use now. It doesn't matter
anyway, 19 kHz is the Nyquist frequency regardless of the mechanism used to
generate the signal. The spectra are identical. My original point was this:
All stereo generators require brick wall anti-aliasing filters to keep the
baseband and the subcarrier sidebands from overlapping (aliasing), in
addition to a guard band to protect the 19 kHz pilot. So FM stereo is
strictly limited in bandwidth to about 15 kHz to 18 kHz best case. FWIW, TV
stereo uses a similar system, except it's tied to the horizontal sweep
frequency, which is even lower (15.734 kHz doubled to a sampling rate of
31.468 kHz). Meaning that TV stereo is limited to about 14 kHz. Same type
brick wall AA filters are required.

Robert Stanton wrote:

chung wrote in message news:ec705$401ebdd9$c247604



If you take into account tolerances of the 5 caps, you would have a
noticeable ripple in the passband, as well as mismatches between L/R.
Plus the cost of the 10 capacitors is not insignificant. And more if you
have 5 channels.

The much better way is simply apply digital filtering, or use better
dithering schemes, but then we are back to LPCM, and not DSD .


I agree. If you look at the data you will see that with perfect
component values, the flatness is: -0.1/-0.2 which 0.1 dB p-p. Using
real componet will increase the ripple. A pratical best flatness spec.
(for real world components) would be +0.1/-0.3 dB.

Using 1% tolerance components, I ran a Monte Carlo analysis (on
100,000 units). I got the following results:

+0.20/-0.40 dB flatness 100% yield

+0.15/-0.35 dB flatness 99.91% yield

+0.10/-0.30 dB flatness 98.80% yield


It would be fairly expensive to use 1% tolerance components. Probably
add $3 or $4 to the cost of manufacture. For an SACD player that sells
for $500 to $1500, that would be acceptable.

1% capacitors are very expensive. Maybe Sony can get them cheaper, but
SACD is not that huge a market, yet. And a 98.8% yield for a filter is
absolutely unacceptable for companies like Sony. (It's way too low.) I
think you have to get at least a 6-sigma yield for a small circuit like
a filter. That's why digital filtering is so wonderful.


However, I don't know if a flatness spec of +/- 0.2dB would be
acceptable to high-end buyers?

It looks worse than good redbook CD players. The bigger problem, of
course, is that you can't reduce the BW to 25KHz and still claim wide-band.

Robert Stanton wrote:

chung wrote in message news:ec705$401ebdd9$c247604



If you take into account tolerances of the 5 caps, you would have a
noticeable ripple in the passband, as well as mismatches between L/R.
Plus the cost of the 10 capacitors is not insignificant. And more if you
have 5 channels.

The much better way is simply apply digital filtering, or use better
dithering schemes, but then we are back to LPCM, and not DSD .


I agree. If you look at the data you will see that with perfect
component values, the flatness is: -0.1/-0.2 which 0.1 dB p-p. Using
real componet will increase the ripple. A pratical best flatness spec.
(for real world components) would be +0.1/-0.3 dB.

Using 1% tolerance components, I ran a Monte Carlo analysis (on
100,000 units). I got the following results:

+0.20/-0.40 dB flatness 100% yield

+0.15/-0.35 dB flatness 99.91% yield

+0.10/-0.30 dB flatness 98.80% yield


It would be fairly expensive to use 1% tolerance components. Probably
add $3 or $4 to the cost of manufacture. For an SACD player that sells
for $500 to $1500, that would be acceptable.

1% capacitors are very expensive. Maybe Sony can get them cheaper, but
SACD is not that huge a market, yet. And a 98.8% yield for a filter is
absolutely unacceptable for companies like Sony. (It's way too low.) I
think you have to get at least a 6-sigma yield for a small circuit like
a filter. That's why digital filtering is so wonderful.


However, I don't know if a flatness spec of +/- 0.2dB would be
acceptable to high-end buyers?

It looks worse than good redbook CD players. The bigger problem, of
course, is that you can't reduce the BW to 25KHz and still claim wide-band.

Robert Stanton wrote:

chung wrote in message news:ec705$401ebdd9$c247604



If you take into account tolerances of the 5 caps, you would have a
noticeable ripple in the passband, as well as mismatches between L/R.
Plus the cost of the 10 capacitors is not insignificant. And more if you
have 5 channels.

The much better way is simply apply digital filtering, or use better
dithering schemes, but then we are back to LPCM, and not DSD .


I agree. If you look at the data you will see that with perfect
component values, the flatness is: -0.1/-0.2 which 0.1 dB p-p. Using
real componet will increase the ripple. A pratical best flatness spec.
(for real world components) would be +0.1/-0.3 dB.

Using 1% tolerance components, I ran a Monte Carlo analysis (on
100,000 units). I got the following results:

+0.20/-0.40 dB flatness 100% yield

+0.15/-0.35 dB flatness 99.91% yield

+0.10/-0.30 dB flatness 98.80% yield


It would be fairly expensive to use 1% tolerance components. Probably
add $3 or $4 to the cost of manufacture. For an SACD player that sells
for $500 to $1500, that would be acceptable.

1% capacitors are very expensive. Maybe Sony can get them cheaper, but
SACD is not that huge a market, yet. And a 98.8% yield for a filter is
absolutely unacceptable for companies like Sony. (It's way too low.) I
think you have to get at least a 6-sigma yield for a small circuit like
a filter. That's why digital filtering is so wonderful.


However, I don't know if a flatness spec of +/- 0.2dB would be
acceptable to high-end buyers?

It looks worse than good redbook CD players. The bigger problem, of
course, is that you can't reduce the BW to 25KHz and still claim wide-band.

Robert Stanton wrote:

chung wrote in message news:ec705$401ebdd9$c247604



If you take into account tolerances of the 5 caps, you would have a
noticeable ripple in the passband, as well as mismatches between L/R.
Plus the cost of the 10 capacitors is not insignificant. And more if you
have 5 channels.

The much better way is simply apply digital filtering, or use better
dithering schemes, but then we are back to LPCM, and not DSD .


I agree. If you look at the data you will see that with perfect
component values, the flatness is: -0.1/-0.2 which 0.1 dB p-p. Using
real componet will increase the ripple. A pratical best flatness spec.
(for real world components) would be +0.1/-0.3 dB.

Using 1% tolerance components, I ran a Monte Carlo analysis (on
100,000 units). I got the following results:

+0.20/-0.40 dB flatness 100% yield

+0.15/-0.35 dB flatness 99.91% yield

+0.10/-0.30 dB flatness 98.80% yield


It would be fairly expensive to use 1% tolerance components. Probably
add $3 or $4 to the cost of manufacture. For an SACD player that sells
for $500 to $1500, that would be acceptable.

1% capacitors are very expensive. Maybe Sony can get them cheaper, but
SACD is not that huge a market, yet. And a 98.8% yield for a filter is
absolutely unacceptable for companies like Sony. (It's way too low.) I
think you have to get at least a 6-sigma yield for a small circuit like
a filter. That's why digital filtering is so wonderful.


However, I don't know if a flatness spec of +/- 0.2dB would be
acceptable to high-end buyers?

It looks worse than good redbook CD players. The bigger problem, of
course, is that you can't reduce the BW to 25KHz and still claim wide-band.

Fran=E7ois Yves Le Gal wrote:

On Tue, 03 Feb 2004 16:53:36 GMT, "Karl Uppiano" karl_uppiano@verizon.=
net
wrote:
=20
You're in denial, then.=20
=20
I'm not, as FM is still transmitted in analog - not digital - form, eve=
n if
the wave is the result of digital manipulations.

Sampling can be done in the analog as well as digital domains. Your=20
problem is that you cannot visualize analog sampling.

You can't properly state
that 38 KHz is the sampling frequency -=20

Why not? You are sequentially "looking at" left and right channels.

and what about SCA side channels at
76 KHz, BTW ?
=20
"Frequency modulation (FM): Modulation in which the instantaneous frequ=
ency
of a sine wave carrier is caused to depart from the center frequency by=
an
amount proportional to the instantaneous value of the modulating signal=
=2E=20

Non-sequitors.

Go get a textbook on communications, and look at the mathematical=20
descriptions of the FM multiplex signal.
=20
Note 1: In FM, the carrier frequency is called the center frequency.
"
=20

The sampling frequency has nothing to do with the final carrier frequency=
=2E


Federal Standard 1037C
=20

Fran=E7ois Yves Le Gal wrote:

On Tue, 03 Feb 2004 16:53:36 GMT, "Karl Uppiano" karl_uppiano@verizon.=
net
wrote:
=20
You're in denial, then.=20
=20
I'm not, as FM is still transmitted in analog - not digital - form, eve=
n if
the wave is the result of digital manipulations.

Sampling can be done in the analog as well as digital domains. Your=20
problem is that you cannot visualize analog sampling.

You can't properly state
that 38 KHz is the sampling frequency -=20

Why not? You are sequentially "looking at" left and right channels.

and what about SCA side channels at
76 KHz, BTW ?
=20
"Frequency modulation (FM): Modulation in which the instantaneous frequ=
ency
of a sine wave carrier is caused to depart from the center frequency by=
an
amount proportional to the instantaneous value of the modulating signal=
=2E=20

Non-sequitors.

Go get a textbook on communications, and look at the mathematical=20
descriptions of the FM multiplex signal.
=20
Note 1: In FM, the carrier frequency is called the center frequency.
"
=20

The sampling frequency has nothing to do with the final carrier frequency=
=2E


Federal Standard 1037C
=20

Fran=E7ois Yves Le Gal wrote:

On Tue, 03 Feb 2004 16:53:36 GMT, "Karl Uppiano" karl_uppiano@verizon.=
net
wrote:
=20
You're in denial, then.=20
=20
I'm not, as FM is still transmitted in analog - not digital - form, eve=
n if
the wave is the result of digital manipulations.

Sampling can be done in the analog as well as digital domains. Your=20
problem is that you cannot visualize analog sampling.

You can't properly state
that 38 KHz is the sampling frequency -=20

Why not? You are sequentially "looking at" left and right channels.

and what about SCA side channels at
76 KHz, BTW ?
=20
"Frequency modulation (FM): Modulation in which the instantaneous frequ=
ency
of a sine wave carrier is caused to depart from the center frequency by=
an
amount proportional to the instantaneous value of the modulating signal=
=2E=20

Non-sequitors.

Go get a textbook on communications, and look at the mathematical=20
descriptions of the FM multiplex signal.
=20
Note 1: In FM, the carrier frequency is called the center frequency.
"
=20

The sampling frequency has nothing to do with the final carrier frequency=
=2E


Federal Standard 1037C
=20

Fran=E7ois Yves Le Gal wrote:

On Tue, 03 Feb 2004 16:53:36 GMT, "Karl Uppiano" karl_uppiano@verizon.=
net
wrote:
=20
You're in denial, then.=20
=20
I'm not, as FM is still transmitted in analog - not digital - form, eve=
n if
the wave is the result of digital manipulations.

Sampling can be done in the analog as well as digital domains. Your=20
problem is that you cannot visualize analog sampling.

You can't properly state
that 38 KHz is the sampling frequency -=20

Why not? You are sequentially "looking at" left and right channels.

and what about SCA side channels at
76 KHz, BTW ?
=20
"Frequency modulation (FM): Modulation in which the instantaneous frequ=
ency
of a sine wave carrier is caused to depart from the center frequency by=
an
amount proportional to the instantaneous value of the modulating signal=
=2E=20

Non-sequitors.

Go get a textbook on communications, and look at the mathematical=20
descriptions of the FM multiplex signal.
=20
Note 1: In FM, the carrier frequency is called the center frequency.
"
=20

The sampling frequency has nothing to do with the final carrier frequency=
=2E


Federal Standard 1037C
=20

"François Yves Le Gal" wrote in message

On Tue, 03 Feb 2004 16:53:36 GMT, "Karl Uppiano"
wrote:

You're in denial, then.

I'm not, as FM is still transmitted in analog - not digital - form,
even if the wave is the result of digital manipulations. You can't
properly state that 38 KHz is the sampling frequency

Sure we can.

- and what about SCA side channels at 76 KHz, BTW ?

Irrelevant.

"Frequency modulation (FM): Modulation in which the instantaneous
frequency of a sine wave carrier is caused to depart from the center
frequency by an amount proportional to the instantaneous value of the
modulating signal.

Note 1: In FM, the carrier frequency is called the center frequency.
"

Federal Standard 1037C

So what?

"François Yves Le Gal" wrote in message

On Tue, 03 Feb 2004 16:53:36 GMT, "Karl Uppiano"
wrote:

You're in denial, then.

I'm not, as FM is still transmitted in analog - not digital - form,
even if the wave is the result of digital manipulations. You can't
properly state that 38 KHz is the sampling frequency

Sure we can.

- and what about SCA side channels at 76 KHz, BTW ?

Irrelevant.

"Frequency modulation (FM): Modulation in which the instantaneous
frequency of a sine wave carrier is caused to depart from the center
frequency by an amount proportional to the instantaneous value of the
modulating signal.

Note 1: In FM, the carrier frequency is called the center frequency.
"

Federal Standard 1037C

So what?

"François Yves Le Gal" wrote in message

On Tue, 03 Feb 2004 16:53:36 GMT, "Karl Uppiano"
wrote:

You're in denial, then.

I'm not, as FM is still transmitted in analog - not digital - form,
even if the wave is the result of digital manipulations. You can't
properly state that 38 KHz is the sampling frequency

Sure we can.

- and what about SCA side channels at 76 KHz, BTW ?

Irrelevant.

"Frequency modulation (FM): Modulation in which the instantaneous
frequency of a sine wave carrier is caused to depart from the center
frequency by an amount proportional to the instantaneous value of the
modulating signal.

Note 1: In FM, the carrier frequency is called the center frequency.
"

Federal Standard 1037C

So what?

"François Yves Le Gal" wrote in message

On Tue, 03 Feb 2004 16:53:36 GMT, "Karl Uppiano"
wrote:

You're in denial, then.

I'm not, as FM is still transmitted in analog - not digital - form,
even if the wave is the result of digital manipulations. You can't
properly state that 38 KHz is the sampling frequency

Sure we can.

- and what about SCA side channels at 76 KHz, BTW ?

Irrelevant.

"Frequency modulation (FM): Modulation in which the instantaneous
frequency of a sine wave carrier is caused to depart from the center
frequency by an amount proportional to the instantaneous value of the
modulating signal.

Note 1: In FM, the carrier frequency is called the center frequency.
"

Federal Standard 1037C

So what?

chung wrote in message news:936b4$401fe986$

1% capacitors are very expensive. Maybe Sony can get them cheaper, but
SACD is not that huge a market, yet. And a 98.8% yield for a filter is
absolutely unacceptable for companies like Sony. (It's way too low.) I
think you have to get at least a 6-sigma yield for a small circuit like
a filter. That's why digital filtering is so wonderful.


That explains why SACD manufacturers allow ultrasonic noise. Analog
active filters are a little too expensive.

chung wrote in message news:936b4$401fe986$

1% capacitors are very expensive. Maybe Sony can get them cheaper, but
SACD is not that huge a market, yet. And a 98.8% yield for a filter is
absolutely unacceptable for companies like Sony. (It's way too low.) I
think you have to get at least a 6-sigma yield for a small circuit like
a filter. That's why digital filtering is so wonderful.


That explains why SACD manufacturers allow ultrasonic noise. Analog
active filters are a little too expensive.

chung wrote in message news:936b4$401fe986$

1% capacitors are very expensive. Maybe Sony can get them cheaper, but
SACD is not that huge a market, yet. And a 98.8% yield for a filter is
absolutely unacceptable for companies like Sony. (It's way too low.) I
think you have to get at least a 6-sigma yield for a small circuit like
a filter. That's why digital filtering is so wonderful.


That explains why SACD manufacturers allow ultrasonic noise. Analog
active filters are a little too expensive.

Robert Stanton wrote:
chung wrote in message news:936b4$401fe986$

1% capacitors are very expensive. Maybe Sony can get them cheaper, but
SACD is not that huge a market, yet. And a 98.8% yield for a filter is
absolutely unacceptable for companies like Sony. (It's way too low.) I
think you have to get at least a 6-sigma yield for a small circuit like
a filter. That's why digital filtering is so wonderful.


That explains why SACD manufacturers allow ultrasonic noise. Analog
active filters are a little too expensive.

It's also non-trivial to design an active filter that preserves the
120dB linearity that is specified in these hi-rez formats.

Robert Stanton wrote:
chung wrote in message news:936b4$401fe986$

1% capacitors are very expensive. Maybe Sony can get them cheaper, but
SACD is not that huge a market, yet. And a 98.8% yield for a filter is
absolutely unacceptable for companies like Sony. (It's way too low.) I
think you have to get at least a 6-sigma yield for a small circuit like
a filter. That's why digital filtering is so wonderful.


That explains why SACD manufacturers allow ultrasonic noise. Analog
active filters are a little too expensive.

It's also non-trivial to design an active filter that preserves the
120dB linearity that is specified in these hi-rez formats.

Robert Stanton wrote:
chung wrote in message news:936b4$401fe986$

1% capacitors are very expensive. Maybe Sony can get them cheaper, but
SACD is not that huge a market, yet. And a 98.8% yield for a filter is
absolutely unacceptable for companies like Sony. (It's way too low.) I
think you have to get at least a 6-sigma yield for a small circuit like
a filter. That's why digital filtering is so wonderful.


That explains why SACD manufacturers allow ultrasonic noise. Analog
active filters are a little too expensive.

It's also non-trivial to design an active filter that preserves the
120dB linearity that is specified in these hi-rez formats.

"Karl Uppiano" writes:

"Randy Yates" wrote in message
...
"Karl Uppiano" writes:

"Arny Krueger" wrote in message
...
"Karl Uppiano" wrote in message

"François Yves Le Gal" wrote in message
...
On Sun, 01 Feb 2004 12:53:18 -0800, chung
wrote:

One, DSD/SACD proponents
claim the much wider bandwidth over CD's, and filtering will reduce
significantly that claimed advantage.

Having a gentle low pass filter at 60 or 100 KHz doesn't
significantly reduce SACD's bandwith. You get more than PCM 96 or
192 in both cases!

I would start a 3 dB/octave rolloff at 20kHz or so.

Ironically, most if not all people can't hear the difference a
brickwall
filter at 16 KHz makes, if the filter is well-designed.

Don't believe me?

Listen for yourself at
http://www.pcabx.com/technical/low_pass/index.htm .

I believe you. I realize it isn't considered state of the art anymore,
but I
wonder how many people realize that FM stereo uses a 38 kHz sample rate?

Actually, the L-R signal is AM-DSB-SC-modulated on a 38 kHz subcarrier and
a 19 kHz pilot tone added. This, along with the L+R baseband signal, is
then FM modulated onto the carrier.

The balanced modulator is mathematically equivalent to a sampled system in
which the left and right channels are alternately sampled at a 38 kHz rate
and used to directly modulate the carrier.

Yeah, I think I see that. I did a simple initial system analysis and it
looks like it may indeed work out that way. There will be harmonics at
n*38 kHz, too, though, but I guess you just filter those out?

Thanks for the new perspective.
--
% Randy Yates % "Rollin' and riding and slippin' and
%% Fuquay-Varina, NC % sliding, it's magic."
%%% 919-577-9882 %
%%%% % 'Living' Thing', *A New World Record*, ELO
http://home.earthlink.net/~yatescr

"Karl Uppiano" writes:

"Randy Yates" wrote in message
...
"Karl Uppiano" writes:

"Arny Krueger" wrote in message
...
"Karl Uppiano" wrote in message

"François Yves Le Gal" wrote in message
...
On Sun, 01 Feb 2004 12:53:18 -0800, chung
wrote:

One, DSD/SACD proponents
claim the much wider bandwidth over CD's, and filtering will reduce
significantly that claimed advantage.

Having a gentle low pass filter at 60 or 100 KHz doesn't
significantly reduce SACD's bandwith. You get more than PCM 96 or
192 in both cases!

I would start a 3 dB/octave rolloff at 20kHz or so.

Ironically, most if not all people can't hear the difference a
brickwall
filter at 16 KHz makes, if the filter is well-designed.

Don't believe me?

Listen for yourself at
http://www.pcabx.com/technical/low_pass/index.htm .

I believe you. I realize it isn't considered state of the art anymore,
but I
wonder how many people realize that FM stereo uses a 38 kHz sample rate?

Actually, the L-R signal is AM-DSB-SC-modulated on a 38 kHz subcarrier and
a 19 kHz pilot tone added. This, along with the L+R baseband signal, is
then FM modulated onto the carrier.

The balanced modulator is mathematically equivalent to a sampled system in
which the left and right channels are alternately sampled at a 38 kHz rate
and used to directly modulate the carrier.

Yeah, I think I see that. I did a simple initial system analysis and it
looks like it may indeed work out that way. There will be harmonics at
n*38 kHz, too, though, but I guess you just filter those out?

Thanks for the new perspective.
--
% Randy Yates % "Rollin' and riding and slippin' and
%% Fuquay-Varina, NC % sliding, it's magic."
%%% 919-577-9882 %
%%%% % 'Living' Thing', *A New World Record*, ELO
http://home.earthlink.net/~yatescr

"Karl Uppiano" writes:

"Randy Yates" wrote in message
...
"Karl Uppiano" writes:

"Arny Krueger" wrote in message
...
"Karl Uppiano" wrote in message

"François Yves Le Gal" wrote in message
...
On Sun, 01 Feb 2004 12:53:18 -0800, chung
wrote:

One, DSD/SACD proponents
claim the much wider bandwidth over CD's, and filtering will reduce
significantly that claimed advantage.

Having a gentle low pass filter at 60 or 100 KHz doesn't
significantly reduce SACD's bandwith. You get more than PCM 96 or
192 in both cases!

I would start a 3 dB/octave rolloff at 20kHz or so.

Ironically, most if not all people can't hear the difference a
brickwall
filter at 16 KHz makes, if the filter is well-designed.

Don't believe me?

Listen for yourself at
http://www.pcabx.com/technical/low_pass/index.htm .

I believe you. I realize it isn't considered state of the art anymore,
but I
wonder how many people realize that FM stereo uses a 38 kHz sample rate?

Actually, the L-R signal is AM-DSB-SC-modulated on a 38 kHz subcarrier and
a 19 kHz pilot tone added. This, along with the L+R baseband signal, is
then FM modulated onto the carrier.

The balanced modulator is mathematically equivalent to a sampled system in
which the left and right channels are alternately sampled at a 38 kHz rate
and used to directly modulate the carrier.

Yeah, I think I see that. I did a simple initial system analysis and it
looks like it may indeed work out that way. There will be harmonics at
n*38 kHz, too, though, but I guess you just filter those out?

Thanks for the new perspective.
--
% Randy Yates % "Rollin' and riding and slippin' and
%% Fuquay-Varina, NC % sliding, it's magic."
%%% 919-577-9882 %
%%%% % 'Living' Thing', *A New World Record*, ELO
http://home.earthlink.net/~yatescr

"François Yves Le Gal" wrote in message
...
On Tue, 03 Feb 2004 16:53:36 GMT, "Karl Uppiano"

wrote:

You're in denial, then.

I'm not, as FM is still transmitted in analog - not digital - form, even
if
the wave is the result of digital manipulations. You can't properly state
that 38 KHz is the sampling frequency - and what about SCA side channels
at
76 KHz, BTW ?

The application of sampling theorems is not limited to digital applications.
There are lots of ways to multiplex analog signals, and periodic sampling is
one of them. It's called time-division multiplexing. 38 kHz is the sampling
frequency for the FM stereo service. SCA is a different service completely
unrelated to FM stereo.

Please refer to http://transmitters.tripod.com/stereo.htm for more
information. It isn't the best writeup I've seen, but it's the only one I
could find on short notice.

"Frequency modulation (FM): Modulation in which the instantaneous
frequency
of a sine wave carrier is caused to depart from the center frequency by an
amount proportional to the instantaneous value of the modulating signal.

The modulating signal... which is provided by a stereo generator that
samples the left and right audio channels at 38 kHz, summed with a 19 kHz
pilot, and optionally, other subcarrier services.

Note 1: In FM, the carrier frequency is called the center frequency.
"

Federal Standard 1037C

"François Yves Le Gal" wrote in message
...
On Tue, 03 Feb 2004 16:53:36 GMT, "Karl Uppiano"

wrote:

You're in denial, then.

I'm not, as FM is still transmitted in analog - not digital - form, even
if
the wave is the result of digital manipulations. You can't properly state
that 38 KHz is the sampling frequency - and what about SCA side channels
at
76 KHz, BTW ?

The application of sampling theorems is not limited to digital applications.
There are lots of ways to multiplex analog signals, and periodic sampling is
one of them. It's called time-division multiplexing. 38 kHz is the sampling
frequency for the FM stereo service. SCA is a different service completely
unrelated to FM stereo.

Please refer to http://transmitters.tripod.com/stereo.htm for more
information. It isn't the best writeup I've seen, but it's the only one I
could find on short notice.

"Frequency modulation (FM): Modulation in which the instantaneous
frequency
of a sine wave carrier is caused to depart from the center frequency by an
amount proportional to the instantaneous value of the modulating signal.

The modulating signal... which is provided by a stereo generator that
samples the left and right audio channels at 38 kHz, summed with a 19 kHz
pilot, and optionally, other subcarrier services.

Note 1: In FM, the carrier frequency is called the center frequency.
"

Federal Standard 1037C