[William Sommerwerck]

"Arny Krueger" wrote in message
...
"William Sommerwerck" wrote in message
...

I have many failings, but I know how to read, and how to understand
what I read. Most people do not.
The operative language is "regardless of level" (see above), which
you and everyone else conveniently ignore.

No ignorance, just an appreciation for the weakness of your argument.
You need more qualifiers than just level to build a working hole in the
logic of self-dithering.

I am not arguing against it (or for it). I am simply saying I don't see it,
and would like a more-detailed explanation/derivation. Is that too much to
ask?

[Arny Krueger[_4_]]

"William Sommerwerck" wrote in message
...
"Arny Krueger" wrote in message
...
"William Sommerwerck" wrote in message
...

I have many failings, but I know how to read, and how to understand
what I read. Most people do not.
The operative language is "regardless of level" (see above), which
you and everyone else conveniently ignore.

No ignorance, just an appreciation for the weakness of your argument.
You need more qualifiers than just level to build a working hole in the
logic of self-dithering.

I am not arguing against it (or for it). I am simply saying I don't see
it,
and would like a more-detailed explanation/derivation. Is that too much to
ask?


Given what you have already received and err, disused...

[William Sommerwerck]

I have many failings, but I know how to read, and how to
understand what I read. Most people do not.
The operative language is "regardless of level" (see above),
which you and everyone else conveniently ignore.

No ignorance, just an appreciation for the weakness of your
argument. You need more qualifiers than just level to build
a working hole in the logic of self-dithering.

I am not arguing against it (or for it). I am simply saying I don't
see it, and would like a more-detailed explanation/derivation.
Is that too much to ask?

Given what you have already received and err [ur], disused... [discussed?]

Arny, are you saying that you always understand everything the first time?
That you have never any difficulty whatever understanding new things?

When someone says that any level of noise above the minimum required will
properly dither a signal, I don't see it. I'm not saying it's wrong, I'm
saying I don't immediately understand it. I would like an explanation or a
reference -- not hand-waving. "Oh, it's so obvious. You must be a moron for
not seeing it."

I am not an intellectual coward. If I don't understand something, I say so.

[William Sommerwerck]

I don't know if the following is correct, but I'm not the only one who's
ever questioned the "principle" or efficacy of self-dithering...

http://www.gearslutz.com/board/maste...lf-dither.html

[[email protected]]

On Sunday, March 25, 2012 12:41:32 PM UTC-4, Soundhaspriority wrote:
Ethan, I didn't use the word "most." How can this be discussed if you
attribute to me a statement that I did not make, as in "that dither makes a
difference for most music" ?

I never said "most" either. I just read my post again, and here's what I actually wrote: "if you can show that dither is audible on any pop music type mix recorded at a sensible level, I promise I will change my opinion immediately." Note the word "any."

The lack of dither should be audible on solo flute.

So post a recording of a flute solo with and without dither to let everyone hear the difference. Better, name the files simply "A" and "B" and ask people to guess which is which. If you are correct, this will prove your point conclusively!

--Ethan

[Don Pearce[_3_]]

On Mon, 26 Mar 2012 08:58:04 -0700, "William Sommerwerck"
wrote:

I don't know if the following is correct, but I'm not the only one who's
ever questioned the "principle" or efficacy of self-dithering...

http://www.gearslutz.com/board/maste...lf-dither.html


There are not so much errors as red herrings in there - particularly
the bit about fades. Fades MUST be dithered. But we are talking about
the dithering quality of excess noise, I believe. The problem with
fades is that the noise is reduced at the same rate as the signal, so
the original optimum dither is instantly too small when you fade. The
same goes for any amplitude change in the digital domain.

Will you accept that to properly dither an analogue signal, an
uncorrelated noise signal must be present that bridges the gap between
adjacent ADC levels? If not, then we must retreat yet further into
theory.

If that is accepted, then the question is - does an oversized noise
signal contain levels that bridge the gap between adjacent ADC levels?
The answer to that is yes - a noise signal contains ALL levels. So an
oversize noise signal will dither properly, and in addition add
unwanted noise.

d

[Don Pearce[_3_]]

On Mon, 26 Mar 2012 09:23:10 -0700 (PDT), wrote:

On Sunday, March 25, 2012 12:41:32 PM UTC-4, Soundhaspriority wrote:
Ethan, I didn't use the word "most." How can this be discussed if you
attribute to me a statement that I did not make, as in "that dither makes a
difference for most music" ?

I never said "most" either. I just read my post again, and here's what I actually wrote: "if you can show that dither is audible on any pop music type mix recorded at a sensible level, I promise I will change my opinion immediately." Note the word "any."

The lack of dither should be audible on solo flute.

So post a recording of a flute solo with and without dither to let everyone hear the difference. Better, name the files simply "A" and "B" and ask people to guess which is which. If you are correct, this will prove your point conclusively!

--Ethan

Ethan. Any chance you can attend to the settings in your news reader?
It is not putting any line breaks in.

d

[[email protected]]

On Sunday, March 25, 2012 5:59:20 PM UTC-4, Peter Larsen wrote:
Please define "showing", I may be lame-brained tonight, but I don't quite
understand how to meet your requirements. What is it you ask for?

All I ask for is a pair of 16-bit files - one reduced from 24 bits using dither, and the other simply truncated. Then post the files here for everyone to hear. Even better is to name them anonymously as I suggested to Robert, such as "File A" and "File B." Then people here can state which file they think is which, to see if people really can tell a difference. That's what I did in my Dither article:

http://www.ethanwiner.com/dither.html

So you should do something similar, using any source you believe shows dither making a difference. But please don't use a sine wave recorded at -50 DBMS, or any other such contrived example. What I've said repeatedly is that dither makes no audible difference on typical pop music recorded at sensible levels. So that's what you or someone else needs to disprove.

--Ethan

[William Sommerwerck]

"Don Pearce" wrote in message
...

Will you accept that to properly dither an analogue signal, an
uncorrelated noise signal must be present that bridges the gap
between adjacent ADC levels? If not, then we must retreat yet
further into theory.

Yes, yes, yes. Of course! We're not arguing that.


If that is accepted, then the question is -- does an oversized noise
signal contain levels that bridge the gap between adjacent ADC
levels? The answer to that is yes -- a noise signal contains ALL
levels. So an oversize noise signal will dither properly, and in
addition add unwanted noise.

So -- if that's correct -- the only thing potentially "wrong" with
self-dithering is that the spectrum might not be optimum, or the noise
insufficiently random. Okay.

I still don't fully "see" this. I'll have to chew on it a bit. (ar, ar)

I have completed my mastication. Now I understand. The noise, riding on top
of the signal, introduces random amplitude variations that decorrelate the
quantization error from the sampling rate, /regardless of the amplitude of
the noise/.

[[email protected]]

Sorry, I meant -50 dBFS. :-)

[Don Pearce[_3_]]

On Mon, 26 Mar 2012 09:44:09 -0700, "William Sommerwerck"
wrote:

"Don Pearce" wrote in message
...

Will you accept that to properly dither an analogue signal, an
uncorrelated noise signal must be present that bridges the gap
between adjacent ADC levels? If not, then we must retreat yet
further into theory.

Yes, yes, yes. Of course! We're not arguing that.


If that is accepted, then the question is -- does an oversized noise
signal contain levels that bridge the gap between adjacent ADC
levels? The answer to that is yes -- a noise signal contains ALL
levels. So an oversize noise signal will dither properly, and in
addition add unwanted noise.

So -- if that's correct -- the only thing potentially "wrong" with
self-dithering is that the spectrum might not be optimum, or the noise
insufficiently random. Okay.

I still don't fully "see" this. I'll have to chew on it a bit. (ar, ar)

I have completed my mastication. Now I understand. The noise, riding on top
of the signal, introduces random amplitude variations that decorrelate the
quantization error from the sampling rate, /regardless of the amplitude of
the noise/.


Yes, just provided there is enough to stretch between two adjacent
levels.

When it come to spectrum - you get what you are given. You don't get
to choose the spectrum of self-dither noise. It is whatever is there.
There is actually a lot of bull talked about optimum spectrum. Sure
you can see the difference with a pure tone on an analyser, but I
would challenge anybody to hear it blind.

d

[Scott Dorsey]

Les Cargill wrote:
Scott Dorsey wrote:

No. Try it yourself, with your own file. Start out with uncompressed
clean audio so you know where you're beginning. Then when you get it
back, subtract it from the original. It's going to take you some time
to get the two lined up perfectly so they subtract at all,

There is usually quiet at the first, so add a single-sample pulse.
Use that as a slate to line 'em up.

if there's no quiet, add a second or so of zero samples.

Add the pulse, then encode, then use the pulse to
line them up. It may be delayed a fraction of a sample
depending on the codec.

Or it can be shredded horribly, to the point where it's no longer a
single-sample pulse at all.
--scott

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

[Les Cargill[_4_]]

PStamler wrote:
On Mar 26, 6:21 am, "Arny wrote:

I think that Mac 075s were under $200 in their day. 275s were under $400 if
memory serves.

But $400 in 1960 was the equivalent of $2,911.23 in 2010 (the latest
year the Inflation Calculator at Westegg.com covers). That's high-end
territory for sure.

Peace,
Paul


Technology has had a faster deflator than other goods. it's
hard to say. Fender MI gear that's comparable is generally
the same as its 1960 price asjusted for inflation with the
regular deflator ) so...

--
Les Cargill

[Les Cargill[_4_]]

Scott Dorsey wrote:
Les wrote:
Scott Dorsey wrote:

No. Try it yourself, with your own file. Start out with uncompressed
clean audio so you know where you're beginning. Then when you get it
back, subtract it from the original. It's going to take you some time
to get the two lined up perfectly so they subtract at all,

There is usually quiet at the first, so add a single-sample pulse.
Use that as a slate to line 'em up.

if there's no quiet, add a second or so of zero samples.

Add the pulse, then encode, then use the pulse to
line them up. It may be delayed a fraction of a sample
depending on the codec.

Or it can be shredded horribly, to the point where it's no longer a
single-sample pulse at all.
--scott



In general, the impulse response of MP3 encoders is *SPLAT*.

--
Les Cargill

[MarkK]

Some of the answers to the question about the importance of dither being
applied at an exact LEVEL are addressed in this book and other similar books
by John Watkinson

Introduction to Digital Audio
http://books.google.com/books/about/..._audio.html?id
=l-DF7RTArnMC


and this one
Intoduction to Digital Video
http://books.google.com/books/about/..._video.html?id
=CvwO8zm3SqoC

and this one
http://books.google.com/books?id=8JH...28&dq=optimum+
dither+amplitude&source=bl&ots=V6o6WuwGqH&sig=F9fh P6fQSY07Frr311C85ao8OqY&hl
=en&sa=X&ei=qOtwT_uGOILx0gHzoPjqBg&ved=0CFgQ6AEwCQ #v=onepage&q&f=false


There are three areas of optimaztion of dither
1) it should complelty linearize the quantizing distortion
2) it should add the least amount of excess noise
3) there should be no noise modulation (where the signal modulates the
noise)

There are 3 common probability densities of dither commonly used
(Q means 1 quantizationlevel)


RECTANGULAR PROBABILITY DENSITY
The amplitude RPD is critical. Linearity is perfect for dither = 1 Q p-p
and all integer multiples. RPD dither causes noise modulation.

TRIANGUALR PROBABILITY DENSITY
optimum linearity occurs at dither = 2 Q p-p.
I'm not sure what happens to the linearity with excess triangular dither
No noise modulation

GAUSSIAN PROBABILOITY DENSITY
Linearity is achieved with dither = 1/2 Q RMS and noise modulation is
negligible.
I'm not sure what happens to the linearity with excess Gaussian dither.

The orignal paper is by
Vanderkooy and Lip****z

http://www.drewdaniels.com/dither.pdf



There are a lot of articles about the advantages of dither but not that many
about optimization of the dither amplitude.



Search terms to read more are "dither amplitude"


This is an interesting question academic question.

In practice, my work is with 16 bit systems, I turn the dither on and don't
worry about it, I don't hear the difference.

Mark

[Arny Krueger[_4_]]

"PStamler" wrote in message
...
On Mar 26, 6:21 am, "Arny Krueger" wrote:

I think that Mac 075s were under $200 in their day. 275s were under $400
if
memory serves.

But $400 in 1960 was the equivalent of $2,911.23 in 2010 (the latest
year the Inflation Calculator at Westegg.com covers). That's high-end
territory for sure.

I wasn't far that I have been reviewing the 1955 Radio Shack catalog, and
the MC-60 was indeed $198.50, while the K-107 (200 watts) was 494.50. The
H.H. Scott 70 watt 265-A was 195.95. The Fisher 50 wass 50-AZ was $159.50.
Audio was much more of a rich man's hobby in those days.

[Arny Krueger[_4_]]

"William Sommerwerck" wrote in message
...
I don't know if the following is correct, but I'm not the only one who's
ever questioned the "principle" or efficacy of self-dithering...

http://www.gearslutz.com/board/maste...lf-dither.html


Yeah, but Gearslutz has never made its reputation based on theoretical
treatises.

They get to be wrong, because self-dither works out in actual use.

However, they make one good point - fade ins and fade outs and other
examples of massive attenuation can move the built in noise down so low that
there is no self-dither.

It's my policy to dither everything regardless, and the DAW software I use
gives me no choice - it dithers everything except file format conversions
automatically and in general you have no choice about it.

[Arny Krueger[_4_]]

"MarkK" wrote in message
...
Some of the answers to the question about the importance of dither being
applied at an exact LEVEL are addressed in this book and other similar
books
by John Watkinson

Introduction to Digital Audio
http://books.google.com/books/about/..._audio.html?id
=l-DF7RTArnMC

On page 120 it says "the amplitude of the dither is not critcal.". It backs
up my statement that if noise provides its own dither, its amplitude is not
reduced.

I don't have the time to do the research on all of these references,
especially without any quotes or page numbers.

I've personally discussed this matter with Vandekooy and Lip****z.

[Scott Dorsey]

Arny Krueger wrote:

I wasn't far that I have been reviewing the 1955 Radio Shack catalog, and
the MC-60 was indeed $198.50, while the K-107 (200 watts) was 494.50. The
H.H. Scott 70 watt 265-A was 195.95. The Fisher 50 wass 50-AZ was $159.50.
Audio was much more of a rich man's hobby in those days.

And that's why so many people were buying Eico and Dynaco amps, and
building their own from articles in Popular Electronics that used
6V6s and universal output transformers.

There was a _lot_ of homebrewing going on, in part because the good
equipment was so expensive, and in part because it was expensive due to
assembly labour.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Arny Krueger[_4_]]

"Les Cargill" wrote in message
...
Scott Dorsey wrote:
Les wrote:
Scott Dorsey wrote:

No. Try it yourself, with your own file. Start out with uncompressed
clean audio so you know where you're beginning. Then when you get it
back, subtract it from the original. It's going to take you some time
to get the two lined up perfectly so they subtract at all,

There is usually quiet at the first, so add a single-sample pulse.
Use that as a slate to line 'em up.

if there's no quiet, add a second or so of zero samples.

Add the pulse, then encode, then use the pulse to
line them up. It may be delayed a fraction of a sample
depending on the codec.

Or it can be shredded horribly, to the point where it's no longer a
single-sample pulse at all.
--scott



In general, the impulse response of MP3 encoders is *SPLAT*.

True to an extent. Carefully selected impulse-like sounds in music are one
area where otherwise good coders expose themselves.

MP3 only works because the impulse response of the ear isn't that good,
either.

[Arny Krueger[_4_]]

"Scott Dorsey" wrote in message
...
Arny Krueger wrote:

I wasn't far that I have been reviewing the 1955 Radio Shack catalog, and
the MC-60 was indeed $198.50, while the K-107 (200 watts) was 494.50. The
H.H. Scott 70 watt 265-A was 195.95. The Fisher 50 wass 50-AZ was $159.50.
Audio was much more of a rich man's hobby in those days.

And that's why so many people were buying Eico and Dynaco amps, and
building their own from articles in Popular Electronics that used
6V6s and universal output transformers.

Hmm, by 1960 the price of the Macintosh equipment was generally unchanged.

RS shows only one Mac stereo component, the C-20 at $225. The MC-30 was
143.50.

The Dyna PAS-2 was $59.95 kit. The Dyna Stereo 70 was 99.95 kit and
$129.95 wired.

There was a _lot_ of homebrewing going on, in part because the good
equipment was so expensive, and in part because it was expensive due to
assembly labour.

Been there, done that. The Macintosh equipment of the day was generally
point-to-point wired with parts on terminal strips, and did not contain PC
boards. Dyna made heavy use of PC boards which were not always robust when
subject to the heat of tubed equipment.

[Arny Krueger[_4_]]

"Arny Krueger" wrote in message
...

Hmm, by 1960 the price of the Macintosh equipment was generally unchanged.

RS shows only one Mac stereo component, the C-20 at $225. The MC-30 was
143.50.

The Dyna PAS-2 was $59.95 kit. The Dyna Stereo 70 was 99.95 kit and
$129.95 wired.

It needs to be pointed out that the MC-30 was more like a 45 wpc amp, if one
took a Dyna ST-70 for a 35 wpc amp.

If the tubes were not brand new the ST-70 struggled to be a 35 wpc
amplifier, 30 watts was more like it.

[Trevor]

"William Sommerwerck" wrote in message
...
The operative language is "regardless of level" (see above), which you and
everyone else conveniently ignore. How can you have possibly missed this?
How can it be misunderstood? Why is it necessary to have to explain things
that have been clearly stated?

OF COURSE it doesn't matter "where" the dither is applied. THAT ISN'T THE
ISSUE. It's the question of the correct or appropriate level.


Too much will work just as well as the "optimum" amount. And many analog
sources, including all tape derived ones, have too much already. Of course
adding a little more will go completely unnoticed as well, which is why most
people do it, "to be sure, to be sure" :-)

Trevor.

[PStamler]

On Mar 26, 8:11*pm, "Arny Krueger" wrote:
"Arny Krueger" wrote in message

...

Hmm, by 1960 the price of the Macintosh equipment was generally unchanged.

RS shows only one Mac stereo component, the C-20 at $225. *The MC-30 was
143.50.

The Dyna PAS-2 was $59.95 kit. The *Dyna Stereo 70 *was 99.95 kit and
$129.95 wired.

It needs to be pointed out that the MC-30 was more like a 45 wpc amp, if one
took a Dyna ST-70 for a 35 wpc amp.

If the tubes were not brand new the ST-70 struggled to be a 35 wpc
amplifier, 30 watts was more like it.

And if you tried to run both channels full power at once, it was
something like 30W/ch even with new tubes, 25W/ch with not-so-new
tubes.

Peace,
Paul

[Scott Dorsey]

Ty Ford wrote:
On Mon, 26 Mar 2012 12:34:10 -0400, Don Pearce wrote

Ethan. Any chance you can attend to the settings in your news reader?
It is not putting any line breaks in.


Odd. I'm seeing them here.

That's because your newsreader is adding them.
--scott


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

[Mike Rivers]

On 3/27/2012 12:02 PM, Scott Dorsey wrote:

On Mon, 26 Mar 2012 12:34:10 -0400, Don Pearce wrote
Ethan. Any chance you can attend to the settings in your news reader?
It is not putting any line breaks in.

Odd. I'm seeing them here.

That's because your newsreader is adding them.

Maybe Don needs a new newsreader. Every once in a while I'll
see a message without line breaks or too-long lines, but I
have a button that makes those perfectly readable.


--
"Today's production equipment is IT based and cannot be
operated without a passing knowledge of computing, although
it seems that it can be operated without a passing knowledge
of audio." - John Watkinson

http://mikeriversaudio.wordpress.com - useful and
interesting audio stuff

[Don Pearce[_3_]]

On Tue, 27 Mar 2012 12:39:08 -0400, Mike Rivers
wrote:

On 3/27/2012 12:02 PM, Scott Dorsey wrote:

On Mon, 26 Mar 2012 12:34:10 -0400, Don Pearce wrote
Ethan. Any chance you can attend to the settings in your news reader?
It is not putting any line breaks in.

Odd. I'm seeing them here.

That's because your newsreader is adding them.

Maybe Don needs a new newsreader. Every once in a while I'll
see a message without line breaks or too-long lines, but I
have a button that makes those perfectly readable.

I'm using the same one I have used for years - Agent. Never had a
problem with it before.

d

[William Sommerwerck]

If you're running Outlook Express, choose Tools / Options / Send. The Plain
Text Settings let you select the line length.

Other editors probably have something similar.

[Don Pearce[_3_]]

On Wed, 28 Mar 2012 09:04:21 -0700 (PDT), wrote:

I have no idea what settings I could change to affect word wrap. I type as I always do, and in Firefox via Google Groups what I write looks fine once it appears.

--Ethan

It's very strange. I can only think that your (and other) news readers
fix the line length of incoming posts.

d

[[email protected]]

Years ago I used Outlook Express, but now I read and post to this group via Google Groups using Firefox. So there are no settings.

[Don Pearce[_3_]]

On Tue, 3 Apr 2012 08:40:00 -0700 (PDT), wrote:

Years ago I used Outlook Express, but now I read and post to this group via Google Groups using Firefox. So there are no settings.

That is probably it. I've been trying to work out whose posts look
wrong, and they all post the way you do. Good old Google Groups
strikes again.

d

[Mike Rivers]

On Apr 3, 11:51*am, (Don Pearce) wrote:

That is probably it. I've been trying to work out whose posts look
wrong, and they all post the way you do. Good old Google Groups
strikes again.

Sorry, I couldn't resist replying to this from Google Groups. If you
can't read this, you're probably not interested anyway.

I sometimes use Google Groups to read this newsgroup (it's the only
one I read) when Eternal September is down or constipated. I'm glad
that they're here. I'm willing to forgive the occasional e-mail that
starts out "I read your post on Google" because I was once there
myself.

[Don Pearce[_3_]]

On Tue, 3 Apr 2012 12:00:34 -0700 (PDT), Mike Rivers
wrote:

On Apr 3, 11:51*am, (Don Pearce) wrote:

That is probably it. I've been trying to work out whose posts look
wrong, and they all post the way you do. Good old Google Groups
strikes again.

Sorry, I couldn't resist replying to this from Google Groups. If you
can't read this, you're probably not interested anyway.

I sometimes use Google Groups to read this newsgroup (it's the only
one I read) when Eternal September is down or constipated. I'm glad
that they're here. I'm willing to forgive the occasional e-mail that
starts out "I read your post on Google" because I was once there
myself.

Damn. This post has line breaks in it. That blows my theory out of the
water.

d

[Peter Larsen[_3_]]

wrote:

On Sunday, March 25, 2012 5:59:20 PM UTC-4, Peter Larsen wrote:

Please define "showing", I may be lame-brained tonight, but I don't
quite understand how to meet your requirements. What is it you ask
for?

All I ask for is a pair of 16-bit files - one reduced from 24 bits
using dither, and the other simply truncated. Then post the files
here for everyone to hear.

Aha.

Even better is to name them anonymously as
I suggested to Robert, such as "File A" and "File B." Then people
here can state which file they think is which, to see if people
really can tell a difference. That's what I did in my Dither article:

http://www.ethanwiner.com/dither.html

I'll look, ie. listen, into this.

So you should do something similar, using any source you believe
shows dither making a difference. But please don't use a sine wave
recorded at -50 DBMS, or any other such contrived example.

Good sense.

What I've
said repeatedly is that dither makes no audible difference on typical
pop music recorded at sensible levels. So that's what you or someone
else needs to disprove.

This is different from what you ask for above and quite easy to agree with.

--Ethan

Kind regards

Peter Larsen