[Anahata]

On Tue, 13 Oct 2009 05:47:26 -0700, William Sommerwerck wrote:

The PCM on the disk, and a series of PWM pulses quantized to the values
of the PCM are fungible. They represent exactly the same data. What is
so hard to understand about this?

That's only true in an imaginary world where there is no noise and the
surface of the disk is perfectly free of distortion, dirt and other
surface defects, and you have a very sensor and quantizer to measure the
pit length. You have to be able to measure each pulse length (a few
microns on the disk) with an accuracy of better than 1 part in 65536,
assuming you want to match the 16 bit resolution of conventional CD.

In practice those defects will push the measured pulse width of your PWM
version many quantized steps away from the correct value, so you'll end
up choosing some exact quantized value, but the wrong one.

With PCM you only have to distingush between a pit and the absence of a
pit, so far higher noise and dirt levels can be tolerated. You have to do
it 16 times instead of once per sample, but the chances are still much
higher of getting all of those 16 bits right.

--
Anahata
==//== 01638 720444
http://www.treewind.co.uk ==//== http://www.myspace.com/maryanahata

[Don Pearce[_3_]]

On Tue, 13 Oct 2009 18:02:32 -0700, "William Sommerwerck"
wrote:

One other thing I forgot...

Have you ever read a magazine article -- or even a book -- on a subject you
were knowledgeable about, and found it loaded with errors?

You won't answer that, because the answer is "yes", you have -- and you
won't have the guts to admit it.


Of course, plenty.

And I always have the guts to admit my errors, viz. that other thread
about sampling and quantizing not being inextricably linked. I
answered you without thinking, I got it wrong and I owned up in the
very next post. That is how you do it.

d

[William Sommerwerck]

The PCM on the disk, and a series of PWM pulses quantized to
the values of the PCM are fungible. They represent exactly the
same data. What is so hard to understand about this?

That's only true in an imaginary world where there is no noise and
the surface of the disk is perfectly free of distortion, dirt and other
surface defects, and you have a very sensor and quantizer to measure
the pit length. You have to be able to measure each pulse length (a few
microns on the disk) with an accuracy of better than 1 part in 65536,
assuming you want to match the 16-bit resolution of conventional CD.

I'm reminded of the line in "The Day the Earth Stood Still" where Klaatu
corrects Professor Barnhardt's (ie, Einstein's) equations. Barnhardt
protests that the solution is not exact. * Klaatu agrees, but says that it
works well enough to get him from planet to planet.

The original question was whether CDs could be made with PWM encoding that
could be fed directly to a switching amplifier. The answer is yes, in
principle, but there are practical problems that would keep it from working
well.

As someone pointed out, a digital PWM CD could represent exactly the same
data as a PCM CD. The data on both are fungible.

* Edmund North, the screenwriter, apparently had a decent understanding of
math. Or the sense to ask someone.


In practice those defects will push the measured pulse width of your
PWM version many quantized steps away from the correct value, so
you'll end up choosing some exact quantized value, but the wrong one.

I already made this point, when I said that PCM is more-robust than PWM.

You are confusing the principle with its implementation. We're not talking
about how well something does or does not work in practice, but the
principle involved.

I should point out that LVs use analog PWM, and it works pretty well. Of
course, the "dynamic range" required is not nearly so wide.


With PCM you only have to distingush between a pit and the absence
of a pit, so far higher noise and dirt levels can be tolerated. You have
to do it 16 times instead of once per sample, but the chances are still
much higher of getting all of those 16 bits right.

Absolutely correct, but that has nothing to do with the issue involved.

[Wecan do it]

"William Sommerwerck" wrote in
message ...

The original question was whether CDs could be made with PWM
encoding that could be fed directly to a switching amplifier.
The answer is yes, in principle, but there are practical
problems that would keep it from working well.

As someone pointed out, a digital PWM CD could represent
exactly the same data as a PCM CD. The data on both are
fungible.


Considering that a red book CD is two channels of audio
interleaved into one signal decoded by one receiver, how would
you put stereo on this PWM signal? You are talking about
putting the precise signal that would drive the class D amp
directly right?

Next the reason you cited for your question was , in my own
words, to eliminate unnecessary conversions. Since when is
converting digital data such a cost prohibitive process. Gonna
save a two dollar microprocessor converter are we?

peace
dawg

*

[William Sommerwerck]

The original question was whether CDs could be made with PWM
encoding that could be fed directly to a switching amplifier.
The answer is yes, in principle, but there are practical
problems that would keep it from working well.

As someone pointed out, a digital PWM CD could represent
exactly the same data as a PCM CD. The data on both are
fungible.


Considering that a red book CD is two channels of audio
interleaved into one signal decoded by one receiver, how would
you put stereo on this PWM signal? You are talking about
putting the precise signal that would drive the class D amp
directly right?

Next the reason you cited for your question was , in my own
words, to eliminate unnecessary conversions. Since when is
converting digital data such a cost prohibitive process. Gonna
save a two dollar microprocessor converter are we?

I didn't raise the original question. Someone else did.

My response was that it was doable in principle, but there are practical
problems that are worse than the issue of how you implement multiple
channels in a PWM system.

[Anahata]

On Wed, 14 Oct 2009 05:25:04 -0700, William Sommerwerck wrote:

The original question was whether CDs could be made with PWM encoding
that could be fed directly to a switching amplifier. The answer is yes,
in principle, but there are practical problems that would keep it from
working well.

Quite right.
My objection was (I thought clearly enough) to your earlier assertion
that:

if you quantized the pulse width -- that
is, digitized it to a finite number of states (widths) -- these problems
would be reduced.

This is simply not true under any circumstances.
If any noise introduced in the process of reading the pulse widths
exceeds the quantization step size, then quantization won't make it
better, and if the noise is less than that, the quantization process
itself simply added some noise. Neither case "reduces the problem".

--
Anahata
-+- http://www.treewind.co.uk
Home: 01638 720444 Mob: 07976 263827

[Don Pearce[_3_]]

On Wed, 14 Oct 2009 11:50:12 -0500, Anahata
wrote:

On Wed, 14 Oct 2009 05:25:04 -0700, William Sommerwerck wrote:

The original question was whether CDs could be made with PWM encoding
that could be fed directly to a switching amplifier. The answer is yes,
in principle, but there are practical problems that would keep it from
working well.

Quite right.
My objection was (I thought clearly enough) to your earlier assertion
that:

if you quantized the pulse width -- that
is, digitized it to a finite number of states (widths) -- these problems
would be reduced.

This is simply not true under any circumstances.
If any noise introduced in the process of reading the pulse widths
exceeds the quantization step size, then quantization won't make it
better, and if the noise is less than that, the quantization process
itself simply added some noise. Neither case "reduces the problem".

You are wasting your time. I have already walked him through all this
stuff, but he just doesn't get it.

d

[Anahata]

On Wed, 14 Oct 2009 17:50:02 +0000, Don Pearce wrote:

You are wasting your time. I have already walked him through all this
stuff, but he just doesn't get it.

Actually I'm beginning to suspect a troll. I shan't post again on this
thread either.

--
Anahata
-+- http://www.treewind.co.uk
Home: 01638 720444 Mob: 07976 263827

[William Sommerwerck]

if you quantized the pulse width -- that
is, digitized it to a finite number of states (widths) -- these problems
would be reduced.

This is simply not true under any circumstances.
If any noise introduced in the process of reading the pulse widths
exceeds the quantization step size, then quantization won't make it
better, and if the noise is less than that, the quantization process
itself simply added some noise. Neither case "reduces the problem".

To paraphrase Dr. Morbius... "If if if if if if if."

The quantization levels would have to be selected taking into account the
phase and amplitude noise of the transmission system.

[Don Pearce]

On Wed, 14 Oct 2009 13:23:07 -0500, Anahata
wrote:

On Wed, 14 Oct 2009 17:50:02 +0000, Don Pearce wrote:

You are wasting your time. I have already walked him through all this
stuff, but he just doesn't get it.

Actually I'm beginning to suspect a troll. I shan't post again on this
thread either.

Likewise.

BTW it is nice to hear the cello played without heavy vibrato for a
change.

d

[William Sommerwerck]

This is simply not true under any circumstances.
If any noise introduced in the process of reading the pulse widths
exceeds the quantization step size, then quantization won't make it
better, and if the noise is less than that, the quantization process
itself simply added some noise. Neither case "reduces the problem".

You are wasting your time. I have already walked him through all this
stuff, but he just doesn't get it.

I get it very well.

The statement that "This is simply not true under any circumstances." is
absurd. See my previous post.

[Ben Bradley[_2_]]

On Sat, 10 Oct 2009 15:13:33 GMT, (Don Pearce)
wrote:

On Thu, 8 Oct 2009 13:12:28 -0700, "William Sommerwerck"
wrote:

No, digitization is representing those quantized levels
by numbers (these are the digits). Quantization is just
quantization, nothing more. It is a jaggy waveform.

Not so. Think about the _meaning_ of digital.

Sampling and digitization are separate and independent
phenomena. Perhaps you are unaware of CCD aka.
"bucket-brigade" which is most certainly *sampling*
but also most certainly ANALOG and NOT digital.

Of course. That wasn't what I was talking about. I was talking about
amplitude quantization -- not temporal quantization.


You're going to have to explain how you can have one without the
other.

There's of course BBD's for discrete time with continuous
amplitude, but for generating discrete amplitude with continuous time
try National's LM3914. Actually the logarithmically-responding LM3916
is probably more common, and is used for LED level meters.

d