[William Sommerwerck]

Well, you said it... "Discrete". If it's discrete, it's digital.

Chiming in late here... You really fundamentally misunderstand the
difference between analog and digital. The entire discussion about
dither and whatnot misses the point. The output signal of a DAC is
analog, totally regardless of quantization errors, "discrete" steps or
whatever. Why? Because when you feed this signal to an amp, and into
speakers or headphones, the resulting changes of air pressure can be
heard and recognized as audio that is more or less similar (aka
analog) to the original. A digital signal can not be heard/recognized
by your ears. If you feed an SPDIF signal to your speakers, it will
not sound anywhere even remotely similar to the original signal...

And if you feed the 10.7MHz output of a tuner's IF strip to your speakers --
an analog signal -- what will you hear?

You -- and most other people -- are confusing signals with data. So much
college educations these days.


I used to tear my hair out (figuratively) when people said that
phone-line
modems converted the digital data from your computer into an analog form
so
they could be transmitted over a conventional phone line. Wrong. When you
represent data as a finite number of qunatized values -- whether by
amplitude, phase, and/or frequency -- the representation is digital.

Not a good analogy.

It's a perfect analogy.


What does happen, however, is that the modem signal
needs to be in a frequency range that can be transmitted by a phone
line, which doesn't really reach up to 20k...
But yes, this is indeed a digital signal.

If you understand THAT, how can you not understand it with respect to audio?
It is EXACTLY the same thing.

Continuously varying data is analog. Quantized data is digital.

[William Sommerwerck]

Chiming in late here... You really fundamentally misunderstand the
difference between analog and digital. The entire discussion about
dither and whatnot misses the point. The output signal of a DAC is
analog, totally regardless of quantization errors, "discrete" steps or
whatever. Why? Because when you feed this signal to an amp, and into
speakers or headphones, the resulting changes of air pressure can be
heard and recognized as audio that is more or less similar (aka
analog) to the original. A digital signal can not be heard/recognized
by your ears. If you feed an SPDIF signal to your speakers, it will
not sound anywhere even remotely similar to the original signal...

Now wait a minute, by this definition, wouldn't PWM be analog?

PWM can be analog or digital, depending on whether the pulse width is
quantized.

Most switching amplifers use some form of pulse-width modulation -- and most
are analog amplifiers, because the pulse widths are nto quantized.

Duh...

[Laurence Payne[_2_]]

On Sun, 25 Oct 2009 02:57:54 +0100, Daniel Fuchs
wrote:

William Sommerwerck wrote:


Well, you said it... "Discrete". If it's discrete, it's digital.


Chiming in late here... You really fundamentally misunderstand the
difference between analog and digital.

You have to understand that William shares Humpty Dumpty's attitude to
words. We wasted a lot of time in this thread before realising he had
his own private definition of "digital". Once this became clear we
had the choice of joining a willie-waving party on esoteric audio
theory or saying "yeah, whatever" :-)

[Daniel Fuchs[_3_]]

William Sommerwerck wrote:


And if you feed the 10.7MHz output of a tuner's IF strip to your speakers --
an analog signal -- what will you hear?


Nothing - I can't hear 10 Mhz...
Besides that, AM and FM are entirely analog - but the signal is encoded
in a way and needs to be decoded to be heard. The carrier signal is just
that, a carrier. Like an analog magnetic tape or an LP record's groove.
Many of these formats alter the original signal for storage, and apply
the reverse effect for playback, e.g. RIAA equalization or Dolby
(remember?).

So again, a failed analogy.


You -- and most other people -- are confusing signals with data. So much
college educations these days.

The output of a DAC is indeed an audio signal, and not data. So?


It's a perfect analogy.

No, the modem analogy has nothing to do with a CD player. The signal you
hear out of the modem's little speaker would represent the SPDIF output
of the CD player, not the DAC.

BTW, said SPDIF output is of course digital. So the answer to your
original question is "it can be both"... But you were referring to the
DAC's output.


If you understand THAT, how can you not understand it with respect to audio?
It is EXACTLY the same thing.

No, it is not. See above.


Continuously varying data is analog. Quantized data is digital.

The DAC's output signal is _not_ quantized, with or without dither etc.
There can be no "steps" in a waveform, just like there is no perfect
square wave. You realize that, don't you?

And the audible waveform that you hear out of the modem is actually
continuous, too (look mom, no steps). It becomes a digital
representation of encapsulated data only because the receiver at the
other end knows what to pay attention to and what to ignore, and how to
deal with it to turn it into a fax or something else. Your ears can not
reassemble the fax or email from the modem signal, can they? But they
can hear the music that comes out of the DAC. Which means the signal is
analog, i.e. similar to the original changes in air pressure that
reached one or several microphones. Or similar to the changes in voltage
at the master output of a mixer. Or whatever else there was initially.

BTW, did you know that on recordable DVD media, storage is not digital,
but in fact analog (i.e. continuous)? This signal has to be read off
the disk, then AD converted and turned into digital video or whatever
again, and then gets converted into analog again...

D.

[Daniel Fuchs[_3_]]

William Sommerwerck wrote:


But yes, this is indeed a digital signal.

If you understand THAT, how can you not understand it with respect to audio?
It is EXACTLY the same thing.


PS: What your ears are hearing out of a modem speaker is of course _not_
the actual digital signal. It's the (frequency modulated) carrier
signal, which happens to be in an audible frequency range. If you were
to transpose an FM carrier signal from 100 MHz or so down to 1k, it
might sound a bit similar, but the changes in frequency would be
continuous (because the encoding is analog, not digital). A bit
simplified here, but maybe, just maybe, you will get the idea.

If not, I guess it's really "Yeah, whatever..."


D.

[William Sommerwerck]

Continuously varying data is analog. Quantized data is digital.

The DAC's output signal is _not_ quantized, with or without dither etc.

Hello? Worst-case example -- no dither. The output is quantized. And it
doesn't matter whether it's gone through a low-pass filter or not.

If the output were not quantized, there would be no quantization error, and
no quantization noise/distortion. But there is.


There can be no "steps" in a waveform, just like there is no perfect
square wave. You realize that, don't you?

You are confusing the fact that no signal has infinite bandwidth with the
principle involved. You are confusing the waveform with the information it
contains.


And the audible waveform that you hear out of the modem is actually
continuous, too (look mom, no steps).

DUH! It is quantized! The waveforms -- which ARE NOT the data -- have
discrete, quantized values of ampliftude, frequency, and polarirty.


This is amazing. It's like my presenting all the evidence about the
roundness of the Earth, and being told it's flat.

[William Sommerwerck]

"Laurence Payne" wrote in message
...
On Sun, 25 Oct 2009 02:57:54 +0100, Daniel Fuchs
wrote:

William Sommerwerck wrote:

Well, you said it... "Discrete". If it's discrete, it's digital.

Chiming in late here... You really fundamentally misunderstand the
difference between analog and digital.

You have to understand that William shares Humpty Dumpty's
attitude to words. We wasted a lot of time in this thread before
realising he had his own private definition of "digital". Once this
became clear we
had the choice of joining a willie-waving party on esoteric audio
theory or saying "yeah, whatever" :-)

My "private" definition of digital is the correct one. Yours (meaning
everyone else here) is wrong.

Digital has nothing to do with "numbers". It has to do with reducing the
data from a continuous representation to one with a finite quantity of
discrete values.

How can you NOT understand that?

Ken Pohlman's book will arriving in a few days. I'm curious as to what he
will say...

[Anahata]

On Sun, 25 Oct 2009 02:57:54 +0100, Daniel Fuchs wrote:

[much snippage]
A DAC's output signal is not
digital.

We all (except one) know that.
Don't get sucked into the madness!

(I gave up when he told me "numbers" doesn't mean "digital")

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

[William Sommerwerck]

I gave up when he told me "numbers" doesn't mean "digital".

No, I said the opposite. I said that "digital" does not mean "numbers".

I understand something -- which to me is simple, straightforward common
sense, that requires no particular intelligence or insight -- and no one
else does. (In this group, anyway.)

This bothers me a great deal, because of what it says about the general
intelligence of people. I'm one of the folk who reads gum wrappers -- and I
don't even chew gum.

[Don Pearce[_3_]]

On Sun, 25 Oct 2009 05:07:19 -0700, "William Sommerwerck"
wrote:

No, I said the opposite. I said that "digital" does not mean "numbers".

Do you even know what a digit is? I'm guessing not.

d

[William Sommerwerck]

Perhaps the best way to get through to you is to ask for an answer to the
following question...

I assume you have a rough idea of how an animal's sensory system works. What
your brain "sees" is a pulse train from receptors in your eye. The stronger
the stimulus, the more pulses are sent to the brain each second.

So, you tell me... Is the information conveyed by that pulse train analog or
digital? And more to the point... WHY is it analog or digital?

[Don Pearce[_3_]]

On Sun, 25 Oct 2009 05:11:28 -0700, "William Sommerwerck"
wrote:

Perhaps the best way to get through to you is to ask for an answer to the
following question...

I assume you have a rough idea of how an animal's sensory system works. What
your brain "sees" is a pulse train from receptors in your eye. The stronger
the stimulus, the more pulses are sent to the brain each second.

So, you tell me... Is the information conveyed by that pulse train analog or
digital? And more to the point... WHY is it analog or digital?


You are making the error of the excluded middle here when you limit
this to analogue and digital. You also make the error of thinking that
once this is all within the nerves it is somehow transmitting an
equivalent of the original stimulus that somehow requires fidelity.
That isn't how the brain and nervous system operate. You need to find
another analogy. Better yet, forget analogy - stick to the real thing.
Do you know what a digit is? Tell is.

d

[Daniel Fuchs[_3_]]

Hello? Worst-case example -- no dither. The output is quantized.

No. It is not. It is continuous. Always, with or without dither. It can
not contain those infamous "steps" that we often see in graphs that are
supposed to show "digital" waveforms.
Show me a _perfect_ square wave, i.e. with perfect right angles. Then
I'll believe you.... Everything else would be continuous.


If the output were not quantized, there would be no quantization error, and
no quantization noise/distortion. But there is.

Huh? The existence of a quantization error (that would have occured
during
ADC already) is supposed to prove that the resulting waveform after DAC
is also
"quantized"? Are you mad?


You are confusing the fact that no signal has infinite bandwidth with the
principle involved. You are confusing the waveform with the information it
contains.

No. The waveform is the waveform (be it alectrical or air pressure). The
information it contains is decoded in your brain. It is mostly
meaningless to a bird or a horse. The information they will gather from
other signals is meaningless to us.


DUH! It is quantized! The waveforms -- which ARE NOT the data -- have
discrete, quantized values of ampliftude, frequency, and polarirty.

No. They do not. Not after the DAC. And "frequency" is totally
irrelevant here anyhow.

This is amazing. It's like my presenting all the evidence about the
roundness of the Earth, and being told it's flat.

Other way round. It's been an established fact for a very long time that
it's round. You are trying to convince us that is, after all, flat...


D.

[William Sommerwerck]

"Don Pearce" wrote in message
...
On Sun, 25 Oct 2009 05:11:28 -0700, "William Sommerwerck"
wrote:

Perhaps the best way to get through to you is to ask for an answer to the
following question...

I assume you have a rough idea of how an animal's sensory system works.
What
your brain "sees" is a pulse train from receptors in your eye. The
stronger
the stimulus, the more pulses are sent to the brain each second.

So, you tell me... Is the information conveyed by that pulse train analog
or
digital? And more to the point... WHY is it analog or digital?


You are making the error of the excluded middle here when you limit
this to analogue and digital.

And there are other forms of data representation? Such as?


You also make the error of thinking that
once this is all within the nerves it is somehow transmitting an
equivalent of the original stimulus

Well, it is!

that somehow requires fidelity.

Shouldn't it? On some level?


That isn't how the brain and nervous system operate.

It is! Rush right down to your library in a reading frenzy!

Do you remember the Popular Electronics article from the early 60s about an
"electronic nerve cell" built at Bell Labs? It closely mimicked to behavior
of a real nerve cell. I built one for a course project. It worked just the
way it was suppsosed to -- the stronger the stimulus, the faster it output
pulses. (It also mimicked the recovery period between pulses.) I don't know
what happened to it -- the transistors cost several dollars apiece, and I
wouldn't have knowingly thrown it out.


You need to find another analogy. Better yet, forget analogy -
stick to the real thing. Do you know what a digit is? Tell is.

Nothing like presenting factual information, asking a reasonable question
about it, then having the facts dismissed so the question can be ignored.
Amazing!

[Daniel Fuchs[_3_]]

William Sommerwerck wrote:

I understand something -- which to me is simple, straightforward common
sense, that requires no particular intelligence or insight -- and no one
else does. (In this group, anyway.)


You'd say the same thing about "everybody else" if you were driving in
the wrong direction on the freeway, I guess...

:-/

[Don Pearce[_3_]]

On Sun, 25 Oct 2009 13:32:24 +0100, Daniel Fuchs
wrote:



William Sommerwerck wrote:

I understand something -- which to me is simple, straightforward common
sense, that requires no particular intelligence or insight -- and no one
else does. (In this group, anyway.)


You'd say the same thing about "everybody else" if you were driving in
the wrong direction on the freeway, I guess...

:-/

He always reminds me of an old Punch cartoon. To proud parents were
watching their retarded son at a military passing-out parade. Mother
said to father "Oh, look dear, everybody is out of step except for our
Sydney".

d

[William Sommerwerck]

I understand something -- which to me is simple, straightforward
common sense, that requires no particular intelligence or insight --
and no one else does. (In this group, anyway.)

You'd say the same thing about "everybody else" if you were driving
in the wrong direction on the freeway, I guess...

But what if they were?

I really don't want to quote Swift, so I won't. But you know the quote.

[William Sommerwerck]

If the output were not quantized, there would be no quantization error,
and no quantization noise/distortion. But there is.

Huh? The existence of a quantization error (that would have occured
during ADC already) is supposed to prove that the resulting waveform
after DAC is also "quantized"? Are you mad?

No, I'm very sober, and am stating a fact. If the side of a barn is blue,
one can reasonably assume that it was painted with blue paint.

Find a test disk with an undithered sweep tone, and give it a listen. The
quantization noise (more properly, distortion) is plainly audible at and
near the frequencies that are submultiples of the sampling rate.

Now... How can you have a signal with audible quantization distortion -- and
say that it isn't digital (ie, quantized)?


For those desiring "proof by authority", I refer you to Wannamaker's thesis.
(I don't know if everything in this thesis is correct -- it probably is --
but I respect his intelligence, if only because he writes in a roughly
conversational style, and tries to directly engage the reader.)

www.digitalsignallabs.com/phd.pdf

Please turn to page 2, paragraph 1.1. He sez...

"Analogue-to-digital conversion is customarily decomposed into two separate
processes: time sampling of the input analogue waveform and amplitude
quantization of the signal values in order that the samples may be
represented by binary words of a prescribed length."

Then go to page 5 and study the "archetypal" quantizing systems in the
drawings. Note, please, what is missing from these drawings. And then think
about it.

[Scott Dorsey]

Don P. wrote:
Daniel Fuchs scribbled:

Chiming in late here... You really fundamentally misunderstand the
difference between analog and digital. The entire discussion about
dither and whatnot misses the point. The output signal of a DAC is
analog, totally regardless of quantization errors, "discrete" steps or
whatever. Why? Because when you feed this signal to an amp, and into
speakers or headphones, the resulting changes of air pressure can be
heard and recognized as audio that is more or less similar (aka
analog) to the original. A digital signal can not be heard/recognized
by your ears. If you feed an SPDIF signal to your speakers, it will
not sound anywhere even remotely similar to the original signal...

Now wait a minute, by this definition, wouldn't PWM be analog?

Yes. PWM is analogue.
--scott

--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Scott Dorsey]

Don Pearce wrote:
On Sun, 25 Oct 2009 05:07:19 -0700, "William Sommerwerck"
wrote:

No, I said the opposite. I said that "digital" does not mean "numbers".

Do you even know what a digit is? I'm guessing not.

It's the body part I use to press the K key and add this subject to my killfile.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Daniel Fuchs[_3_]]

Find a test disk with an undithered sweep tone, and give it a listen. The
quantization noise (more properly, distortion) is plainly audible at and
near the frequencies that are submultiples of the sampling rate.

Oh boy...
a) quantization noise is practically inaudible. Converters are pretty
good these days.
b) the sampling rate has nothing to do with this, really. Also,
distortion at e.g. 22k would also be inaudible.

What's the point?


Now... How can you have a signal with audible quantization distortion -- and
say that it isn't digital (ie, quantized)?

Yeah, whatever.
This is completely absurd, but it's been done to death here before. You
really don't understand the difference between analog and digital...



"Analogue-to-digital conversion is customarily decomposed into two separate
processes: time sampling of the input analogue waveform and amplitude
quantization of the signal values in order that the samples may be
represented by binary words of a prescribed length."

So?


Then go to page 5 and study the "archetypal" quantizing systems in the
drawings. Note, please, what is missing from these drawings. And then think
about it.

Where is the DA conversion?

Bye.

D.

[jwvm]

On Oct 25, 7:38*am, "William Sommerwerck"
wrote:
Continuously varying data is analog. Quantized data is digital.
The DAC's output signal is _not_ quantized, with or without dither etc.

Hello? Worst-case example -- no dither. The output is quantized. And it
doesn't matter whether it's gone through a low-pass filter or not.


Wow, things are moving backwards again. Let's take the worst case
situation here, a 1 bit DAC with no oversampling and no dither. The
quantization errors will be horrible I suspect you will agree. When
the output of the DAC is low-pass filtered such that the signal is
bandwidth limited to less than half of the sampling frequency, the
resulting signal can have any level over some voltage range. In fact,
if the only frequency present is relatively close to the sampling rate
and of sufficiently high amplitude, the reconstructed signal will be a
sinusoidal waveform. The amplitude will very like be off by a
considerable amount but it will still be sinusoidal. You would be
hard pressed to convince any competent electrical engineer that this
result is a digital signal.

[George's Pro Sound Co.]

"William Sommerwerck" wrote in message
...
I gave up when he told me "numbers" doesn't mean "digital".

No, I said the opposite. I said that "digital" does not mean "numbers".

I understand something -- which to me is simple, straightforward common
sense, that requires no particular intelligence or insight -- and no one
else does. (In this group, anyway.)

This bothers me a great deal, because of what it says about the general
intelligence of people. I'm one of the folk who reads gum wrappers -- and
I
don't even chew gum.

I LOVED Bazooka Joe!!!
george

[William Sommerwerck]

Then go to page 5 and study the "archetypal" quantizing
systems in the drawings. Note, please, what is missing
from these drawings. And then think about it.

Where is the D/A conversion?

Exactly, exactly.

The D/A conversion occurs when the signal is quantized.

To put it a slightly different way... Wannamaker's analysis would not be
changed in the least if the "channel" (which isn't described, because it
doesn't need to be) produced a PCM signal.

You can't quantize a continuously varying value and they say the result is
still analog! It's digital. It can be represented by an integer, and that's
all that matters.

Why is this so difficult to understand? Any quantized data representation is
digital.

Data * cannot vary continuously and be quantized at the same time -- it
cannot be simultaneously analog and digital.

* I said data, not waveforms. A waveform is not, per se, data.

[William Sommerwerck]

Then go to page 5 and study the "archetypal" quantizing
systems in the drawings. Note, please, what is missing
from these drawings. And then think about it.

Where is the D/A conversion?

Whoops! The brain was not engaged when the fingers touched the keyboard.
Very, very sorry.

The "missing" item is whatever it is that converts the quantized signal into
PCM -- or any other transmission system. The analysis doesn't require it.

To put it a slightly different way... Wannamaker's analysis would not be
changed in the least if the "channel" (which isn't described, because it
doesn't need to be) produced a PCM signal.

You can't quantize a continuously varying value and they say the result is
still analog! It's digital. It can be represented by an integer, and
that's
all that matters.

Why is this so difficult to understand? Any quantized data representation
is
digital.

Data * cannot vary continuously and be quantized at the same time -- it
cannot be simultaneously analog and digital.

* I said data, not waveforms. A waveform is not, per se, data.

[Anahata]

He told me "numbers" doesn't mean "digital".

No, I said the opposite. I said that "digital" does not mean "numbers".

Whatever. You are wrong either way.

It is an absolutely fundamental and axiomatic definiton of a digital
communication or storage medium that numerical values fed into one end
can be reproduced exactly (as numerical values) at the other end.

The output of a CD player, for many reasons including some that have been
gone through ad nauseam in this thread, cannot successfully be part of
such a system, because the original 16 bit PCM numerical values cannot be
retrieved from it.

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

[Laurence Payne[_2_]]

On Sun, 25 Oct 2009 04:43:17 -0700, "William Sommerwerck"
wrote:

My "private" definition of digital is the correct one. Yours (meaning
everyone else here) is wrong.

So pick a different word. It will annoy people less, and make you
look less ignorant.


Digital has nothing to do with "numbers". It has to do with reducing the
data from a continuous representation to one with a finite quantity of
discrete values.

You really don't understand the reconstruction filter, do you? Maybe
it's because you can't accept what it does that you object to its
name?

[Daniel Fuchs[_3_]]

William Sommerwerck wrote:


The D/A conversion occurs when the signal is quantized.

(...)

You can't quantize a continuously varying value and they say the result is
still analog! It's digital. It can be represented by an integer, and that's
all that matters.


What you're saying would make some sort of sense - if it referred to AD
conversion, not DA...


D.

[William Sommerwerck]

The D/A conversion occurs when the signal is quantized.

(...)

You can't quantize a continuously varying value and they say the result
is
still analog! It's digital. It can be represented by an integer, and
that's
all that matters.

What you're saying would make some sort of sense - if it referred to AD
conversion, not DA...

Well, it makes perfect sense. As I hope you saw, I goofed and had to rewrite
what I miswrote. It was supposed to say A/D conversion.

[William Sommerwerck]

Digital has nothing to do with "numbers". It has to do with reducing
the data from a continuous representation to one with a finite quantity
of discrete values.

You really don't understand the reconstruction filter, do you? Maybe
it's because you can't accept what it does that you object to its
name?

All the "reconstruction" filter does is remove the ultrasonic components so
they won't cause trouble. The temporally-continuous signal is always
present, and needs no reconstruction (qv, sampling theory).

Now, once upon a time some designers experimented with oversampling in
playback using interpolated values -- the intent being to more-accurately
"reconstruct" the original waveform by inserting samples that mimicked the
waveforms of "real" sounds. This could legitimately be called a
"reconstruction" filter (or process).

Reviewers claimed that it resulted various changes in the sound, which is
plausible. However, the math of this is not immediately obvious to me, and I
don't feel competent to discuss it (though I have theories).

[Daniel Fuchs[_3_]]

Well, it makes perfect sense. As I hope you saw, I goofed and had to rewrite
what I miswrote. It was supposed to say A/D conversion.

And what does this do to your theory of the CD player's output signal
being digital?


D.