[bill]

Paul Stamler wrote:
"bill" wrote in message
t...

As I just mentioned in another post on this what happens if you sample
your 20KHz sine wave with a 40KHz sample rate that just happens to
sample the zero crossings and not the peaks??? I think you know.

Which is why you don't do that. Period. Congratulations; you've just
discovered aliasing.

Peace,
Paul


Sorry again, I am not talking anti-aliasing. Get out some paper and
drink some coffee, since it appears this concept will take you some time.

--
Bill (Sleepless biker) Baka

[bill]

Laurence Payne wrote:
On Wed, 25 Oct 2006 17:59:52 GMT, bill
wrote:

As I just mentioned in another post on this what happens if you sample
your 20KHz sine wave with a 40KHz sample rate that just happens to
sample the zero crossings and not the peaks??? I think you know.

Yeah. We "know". But, in this case, I'm told (by people who know
their stuff) that we're wrong. NYquist isn't very intuitive. But it
works.

It works because nobody can really hear the distortion caused at that
high a frequency, and a pure sine wave used for testing is not what a
musical instrument may produce at that frequency. A lower frequency
example is a drum, which is a decaying oscillation, so just think about
building up a good representation of that with only 2.2 samples per
sine. Just when you think you have it the amplitude has changed.
Why is something so simple going over some people's heads?

--
Bill (Sleepless biker) Baka

[bill]

APR wrote:
It is amusing how many posters here have to attack the person all the time.
It is surely a sign of something amiss when there has to be personal attacks
rather the discussion of the relevant points. I think we all know very well
qualified people who don't demonstrate the level of common sense and
intelligence the qualifications would suggest, and it appears to me that
that hypothesis is reinforced by the posters regularly resorting to personal
attack.

Of topic here, but I have had guys with PhD's working for me that were
useless because they could not think for themselves and certainly could
not think 'outside the box' they were taught not to question. 8 years of
college crams you full of 'facts' but not to question those 'facts'. You
might as well program a computer with all those 'facts' if there will
not be any creative thought to verify them.



--
Bill (Sleepless biker) Baka

[bill]

David Morgan (MAMS) wrote:
"APR" wrote in message

It is amusing how many posters here have to attack the person all the time.

No one who is *truthfully* regular attacks anyone unless it's political in nature.
The rest of it is people just passing through who think they know it all anyway.


Or, in my case, have experienced some of the shortcomings of relying
totally on a theory that applies to one thing but not to what I am
working on.

--
Bill (Sleepless biker) Baka

[bill]

William Sommerwerck wrote:
If "the scientific method" were the definition of science, then the best
scientists would be those who where the best at running scientific
experiments. But despite the need to carefully desingn, run, and interpret
experiments, the best scientists are those with the best insight, the best
imaginations.

The first hypothesis has to come from -- and it doesn't come from
experiment. It comes from imagination.


AMEN!!!

--
Bill (Sleepless biker) Baka

[bill]

Chevdo wrote:
In article ,
says...
If "the scientific method" were the definition of science, then the best
scientists would be those who where the best at running scientific
experiments. But despite the need to carefully desingn, run, and interpret
experiments, the best scientists are those with the best insight, the best
imaginations.

The first hypothesis has to come from -- and it doesn't come from
experiment. It comes from imagination.



Doesn't matter whether those people are practicing science or not, you're
talking about superior people, not superior 'scientists'. Science without the
scientific method is 'woo-woo', aka 'flummery'. It's how we got crap like
chiropracty and homeopathy that we for some reason now can't get rid of. The
fact that a genius needs to be in charge of the lab janitors to generate
exception results, and that even the lab janitors have to be cognitively
superior to, say, the fader janitors who call themselves 'audio engineers',
and that even those fader janitors have to be cognitively superior to the
original mop & broom janitors, is superfulous. However, if geniuses simply
relied on their geniusness, rather than the scientific method, we'd be
absolutely NOWHERE, as we were for thousands of years until we finally realized
that anything not proven by scientific method is WORTHLESS.

It was the Catholic church who stifled invention and burned people at
the stake for heresy in questioning 'God'. That put us back an easy
1,000 years.


There were plenty
of geniuses who were around before the scientific method, and they
unfortunately didn't get very far.

Church again. Read about what happened to some of the inventors back
around Galileo's time.

If Aristotle had been around in the
16th century, he'd probably rival or exceed Newton, and if either of them had
been in the 20th century they'd probably rival or exceed Einstein.

Aristotle and Pythagoras were true geniuses before their time but the
Roman empire (conquer, rape, and burn) set things back a bit, not to
mention all the warring for no good reason. Ever wonder what was in the
library of Alexandria that has been lost forever?

Einstein
had the benefit of the access to the entirety of Newton, Aristotle, and many
other geniuses at his disposal, without which even he wouldn't have got very
far. The point is, research was totally unreliable before the scientific
method. Now it is far more reliable (though not totally, as evidenced by the
amount of flummery that still gets published in peer-reviewed science
journals), and the various political machinations to shape and direct
research by pharmacutical companies, and over-zealous and/or ideologically
biased academics.

Let's see. There's primitives in office, like Bush declaring stem cell
research is immoral and against God's wishes. Ring a bell?


--
Bill (Sleepless biker) Baka

[Laurence Payne]

On Wed, 25 Oct 2006 18:43:54 -0700, "William Sommerwerck"
wrote:

If "the scientific method" were the definition of science, then the best
scientists would be those who where the best at running scientific
experiments. But despite the need to carefully desingn, run, and interpret
experiments, the best scientists are those with the best insight, the best
imaginations.

The first hypothesis has to come from -- and it doesn't come from
experiment. It comes from imagination.

Science is defined by the scientific method. If you want to re-define
it, we can't stop you. But you will only cause confusion.

In the scientific method, the hypothesis comes from observation. I
guess we can stretch that to include unconscious observation and
correlation.

[William Sommerwerck]

The first hypothesis has to come from somewhere -- and
it doesn't come from experiment. It comes from imagination.

I don't quite know how to evaluate your use of the term "best."
Could you quantify that a bit so we're all on the same page?

To me, this is all very obvious. But...

Who are the greatest scientists? The ones who perform the best experiments?
Or the ones who do the most-creative thinking?

[William Sommerwerck]

Why can't you see that the imagination and intuition you speak of are
integral to the method which defines science both at the level of
theory and at the level of experiment?

Because they aren't.


The experiment without a hypothesis is blind...

What you don't seem to understand is that such experiments can be useful to
learn about something you don't understand. It's precisely because people
refuse to perform such experiments that we don't have reliable, accurate
listening tests.

About 20 years ago, the late, lamented Hewlett-Packard ran a series of ads
that went something like...

"An engineer asks 'How can I?'. A scientist asks 'What if...?' We never stop
asking 'What if...?'.

More than anything, science is about asking good questions.

[William Sommerwerck]

The experiment without a hypothesis is blind

That is the WRONG term. a 'blind' experiment is a BETTER experiment. A
'double-blind' experiment is even better. 'blind' experiments are those
in
which the experimenters are not able to influence the outcome by
restricting
their awareness of the constituents of the experiment. It does NOT mean
that a
hypothesis is not being tested.

You've quite misunderstood what he meant.

[Ron Capik]

William Sommerwerck wrote:

The first hypothesis has to come from somewhere -- and
it doesn't come from experiment. It comes from imagination.

I don't quite know how to evaluate your use of the term "best."
Could you quantify that a bit so we're all on the same page?

To me, this is all very obvious. But...

Who are the greatest scientists? The ones who perform the best experiments?
Or the ones who do the most-creative thinking?

I don't quite know how to evaluate your use of the term "best."
Could you quantify that a bit so we're all on the same page?


Later...

Ron Capik
--
Will the circle
be unbroken
by and by...

[Scott Dorsey]

bill "just use comcast", "you know the drill for no spam bots." wrote:

It works because nobody can really hear the distortion caused at that
high a frequency, and a pure sine wave used for testing is not what a
musical instrument may produce at that frequency. A lower frequency
example is a drum, which is a decaying oscillation, so just think about
building up a good representation of that with only 2.2 samples per
sine. Just when you think you have it the amplitude has changed.
Why is something so simple going over some people's heads?

Because it doesn't exist.

It ONLY takes 2 samples per wave to reconstruct the original signal. THAT
is what Nyquist says. It does not matter what the phase is, it does not
matter what the frequency is. 2 samples per cycle will do it. 2.2 is more
than enough.

If you don't believe it, try it with a typical modern A/D and D/A box,
a scope and a signal generator. What goes in is what comes out, right
up close to the filter cutoff.
--scott

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

[William Sommerwerck]

It ONLY takes 2 samples per wave to reconstruct
the original signal. THAT is what Nyquist says.

Strictly speaking, it's 2 samples, over multiple cycles, because the
sampling rate has to be than twice the highest frequency to be sampled.

This makes sense if you understand the mathematical transformation between
the time and frequency domains.

[bill]

Scott Dorsey wrote:
bill "just use comcast", "you know the drill for no spam bots." wrote:
It works because nobody can really hear the distortion caused at that
high a frequency, and a pure sine wave used for testing is not what a
musical instrument may produce at that frequency. A lower frequency
example is a drum, which is a decaying oscillation, so just think about
building up a good representation of that with only 2.2 samples per
sine. Just when you think you have it the amplitude has changed.
Why is something so simple going over some people's heads?

Because it doesn't exist.

It ONLY takes 2 samples per wave to reconstruct the original signal. THAT
is what Nyquist says. It does not matter what the phase is, it does not
matter what the frequency is. 2 samples per cycle will do it. 2.2 is more
than enough.

If you don't believe it, try it with a typical modern A/D and D/A box,
a scope and a signal generator. What goes in is what comes out, right
up close to the filter cutoff.
--scott

I don't give a flying crap what Nyquist says because it is aimed at
steady state sine waves at that kind of a sample rate. If you can't
understand that, then I am arguing a point with a rock. Didn't anybody
ever teach you to think for yourself?

--
Bill (Sleepless biker) Baka

[Brian Running]

But despite the need to carefully desingn, run, and interpret
experiments, the best scientists are those with the best insight, the best
imaginations.

The first hypothesis has to come from -- and it doesn't come from
experiment. It comes from imagination.

This directly contradicts what you wrote a couple days ago, when you
said that good scientists begin their research without a hypothesis. I
don't follow you.

[Chewy Papadopoulous]

Chevdo wrote:

that anything not proven by scientific method is WORTHLESS.


Chevy, any high school science student knows that the scientific method
doesn't "prove" anything other than what tested variables don't support
a given hypothesis. It is a process of elimination, which you and Gutter
Butt ought to both appreciate. Know your terms, THEN go out and preach.

Love the "Brave New Worldly" intellectual dominance hierarchy you set up
in your sermon, though. Wonder where you fit in...

Never mind.

Chewy

[Romeo Rondeau]

David Morgan (MAMS) wrote:
"APR" wrote in message

It is amusing how many posters here have to attack the person all the time.

No one who is *truthfully* regular attacks anyone unless it's political in nature.
The rest of it is people just passing through who think they know it all anyway.



Totally not true, Dave. There are several regulars on here who attack
people personally. Once a difference of opinion, taste or technique
develops, the norm in this group is attack. This is usually followed by
questioning the person's credentials, experience and then moving into
statements about their ears or brains not working properly and ending in
them either being called a Nazi or worse (!) a Republican :-) And heaven
forbid if you don't use a tape machine :-)

[Scott Dorsey]

bill "just use comcast", "you know the drill for no spam bots." wrote:

I don't give a flying crap what Nyquist says because it is aimed at
steady state sine waves at that kind of a sample rate.

Every signal can be represented as the sum of steady state sine waves.
Fourier showed this in the 17th century. The Shannon Sampling Theorem
therefore is valid for any arbitrary waveform once it's proven for a
single sine wave, AS LONG AS that waveform is bandlimited. Discussion
of this can be found in Shannon's original BSTJ article.

If you can't
understand that, then I am arguing a point with a rock. Didn't anybody
ever teach you to think for yourself?

I understand it, but it's not relevant. If you think it IS relevant, you
haven't done the math. I strongly recommend a good introduction to DSP
which will explain the Sampling Theorem.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Ron Capik]

Scott Dorsey wrote:

bill "just use comcast", "you know the drill for no spam bots." wrote:

I don't give a flying crap what Nyquist says because it is aimed at
steady state sine waves at that kind of a sample rate.

Every signal can be represented as the sum of steady state sine waves.
Fourier showed this in the 17th century. The Shannon Sampling Theorem
therefore is valid for any arbitrary waveform once it's proven for a
single sine wave, AS LONG AS that waveform is bandlimited. Discussion
of this can be found in Shannon's original BSTJ article.

Here's a pointer to Shannon's paper:

http://cm.bell-labs.com/cm/ms/what/s...day/paper.html

Bob Lucky gave a great talk at the Shannon Day conference
addressing the impact of latency on communication theory as
applied to modern media. But that's way, way OT.

Also, there was a long sampling theory discussion in rec.audio.tech a few
months ago, [ Subject: 10 metres audio cable going into PC = too long? ]
that may be enlightening.

Later...

Ron Capik
--

[bill]

Scott Dorsey wrote:
bill "just use comcast", "you know the drill for no spam bots." wrote:
I don't give a flying crap what Nyquist says because it is aimed at
steady state sine waves at that kind of a sample rate.

Every signal can be represented as the sum of steady state sine waves.

Every STEADY state signal.

Fourier showed this in the 17th century.

I don't think he was concerned with audio. His math works nicely with
sine waves that have enough repetition to build up an approximation. He
had no clue that there would be computers doing things he never even
dreamed of.

The Shannon Sampling Theorem
therefore is valid for any arbitrary waveform once it's proven for a
single sine wave, AS LONG AS that waveform is bandlimited. Discussion
of this can be found in Shannon's original BSTJ article.

Duh! An arbitrary waveform can be a 5KHz sine wave with 'up to' 20KHz
components superimposed on it. The 5 KHz will be properly digitized but
the 20KHz parts won't be.

If you can't
understand that, then I am arguing a point with a rock. Didn't anybody
ever teach you to think for yourself?

I understand it, but it's not relevant. If you think it IS relevant, you
haven't done the math. I strongly recommend a good introduction to DSP
which will explain the Sampling Theorem.
--scott

Stubborn, aren't you. I'm an engineer and I did the math (Fourier and
LaPlace)in the early 70's, so I understand the sampling theorem AND it's
limitations. Find a book that actually mentions the limitations instead
of pretending the math is foolproof and any exceptions can be
compensated for by filtering. Anti-aliasing filters are not for waveform
purity but to make sure the converter isn't fooled into creating a
totally bogus waveform. Play with a digital oscilloscope sometime and
you will see how bad they can screw up a simple sine wave at the wrong
sweep and sample rate. They make a good case for me keeping my personal
analog scope to do sanity checks, although I know from experience what a
mis-sampled waveform looks like.
You need some humbling experience, not my typing back and forth.
Someone else can teach you, hopefully before you embarrass yourself by
preaching in a meeting of more experienced peers.
I'm done with you.


--
Bill (Sleepless biker) Baka

[Laurence Payne]

On Thu, 26 Oct 2006 04:18:38 -0700, "William Sommerwerck"
wrote:

What you don't seem to understand is that such experiments can be useful to
learn about something you don't understand. It's precisely because people
refuse to perform such experiments that we don't have reliable, accurate
listening tests.

"Experiment" has a particular meaning in the context of the Scientific
Method.
While messing around, or as a side-effect of another experiment, you
may notice something interesting and form a hypothesis. Then you
design an experiment to test it.

You're merely demonstrating that you don't understand the terminology.

[bill]

Ron Capik wrote:
Scott Dorsey wrote:

bill "just use comcast", "you know the drill for no spam bots." wrote:
I don't give a flying crap what Nyquist says because it is aimed at
steady state sine waves at that kind of a sample rate.
Every signal can be represented as the sum of steady state sine waves.
Fourier showed this in the 17th century. The Shannon Sampling Theorem
therefore is valid for any arbitrary waveform once it's proven for a
single sine wave, AS LONG AS that waveform is bandlimited. Discussion
of this can be found in Shannon's original BSTJ article.

Here's a pointer to Shannon's paper:

http://cm.bell-labs.com/cm/ms/what/s...day/paper.html

Bob Lucky gave a great talk at the Shannon Day conference
addressing the impact of latency on communication theory as
applied to modern media. But that's way, way OT.

Also, there was a long sampling theory discussion in rec.audio.tech a few
months ago, [ Subject: 10 metres audio cable going into PC = too long? ]
that may be enlightening.

I give up on you guys on sampling theory versus reality.
On the cable, 10 metres = about 33 feet which means a propagation delay
of about 50 nS at the speed electricity travels in a conductor within a
dielectric. 1/50nS = about 20 MHz , not KHz, speaking in terms of
wavelength. I probably would have just gotten a good laugh out of that
discussion.

Later...

Ron Capik
--




--
Bill (Sleepless biker) Baka

[bill]

Laurence Payne wrote:
On Thu, 26 Oct 2006 04:18:38 -0700, "William Sommerwerck"
wrote:

What you don't seem to understand is that such experiments can be useful to
learn about something you don't understand. It's precisely because people
refuse to perform such experiments that we don't have reliable, accurate
listening tests.

"Experiment" has a particular meaning in the context of the Scientific
Method.
While messing around, or as a side-effect of another experiment, you
may notice something interesting and form a hypothesis. Then you
design an experiment to test it.

You're merely demonstrating that you don't understand the terminology.

I think you are on to what I have been telling the "I was taught this in
school so it is absolutely the truth no matter what you have seen."
crowd. Education just doesn't seem to be teaching the 'question
everything' method anymore. It's more like memorize it and never
question anything. That's why we experiment.

--
Bill (Sleepless biker) Baka

[Scott Dorsey]

bill "just use comcast", "you know the drill for no spam bots." wrote:
Scott Dorsey wrote:
bill "just use comcast", "you know the drill for no spam bots." wrote:
I don't give a flying crap what Nyquist says because it is aimed at
steady state sine waves at that kind of a sample rate.

Every signal can be represented as the sum of steady state sine waves.

Every STEADY state signal.

No. It does not have to be steady state.

Fourier showed this in the 17th century.

I don't think he was concerned with audio. His math works nicely with
sine waves that have enough repetition to build up an approximation. He
had no clue that there would be computers doing things he never even
dreamed of.

That's the wonderful thing about mathematics! It works the same way,
no matter what the application is.

The Shannon Sampling Theorem
therefore is valid for any arbitrary waveform once it's proven for a
single sine wave, AS LONG AS that waveform is bandlimited. Discussion
of this can be found in Shannon's original BSTJ article.

Duh! An arbitrary waveform can be a 5KHz sine wave with 'up to' 20KHz
components superimposed on it. The 5 KHz will be properly digitized but
the 20KHz parts won't be.

They will be, just as long as the sampling rate is over 40 Ksamp/sec and
the anti-aliasing filter is properly set up.

Aperiodical signals? Square waves? Impulses? They can ALL be properly
reproduced as long as they are bandlimited; the only distortion that occurs
is the bandlimiting itself.

Stubborn, aren't you. I'm an engineer and I did the math (Fourier and
LaPlace)in the early 70's, so I understand the sampling theorem AND it's
limitations. Find a book that actually mentions the limitations instead
of pretending the math is foolproof and any exceptions can be
compensated for by filtering.

I don't think you did do the math because if you had read the original
Shannon paper, you'd understand just how incredibly elegant the whole
process is. It really is ingenious as hell.

Anti-aliasing filters are not for waveform
purity but to make sure the converter isn't fooled into creating a
totally bogus waveform.

These are the same thing.

Play with a digital oscilloscope sometime and
you will see how bad they can screw up a simple sine wave at the wrong
sweep and sample rate. They make a good case for me keeping my personal
analog scope to do sanity checks, although I know from experience what a
mis-sampled waveform looks like.

Right. You can also put a 10 MHz square wave into a 10 MHz analogue scope
and have it look like a sine wave, because of the bandlimiting. The
artifacts you see are due to bandlimiting.

You need some humbling experience, not my typing back and forth.
Someone else can teach you, hopefully before you embarrass yourself by
preaching in a meeting of more experienced peers.
I'm done with you.

I strongly, strongly want to suggest you sit down with the original Shannon
paper and read it, because once the idea hits you how the thing works, you
will be surprised. It really is pretty amazing.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Scott Dorsey]

bill "just use comcast", "you know the drill for no spam bots." wrote:

I give up on you guys on sampling theory versus reality.

Don't give up! Get the paper and read it!

Really, I read it when I was an undergrad, and it took me several times
through to really figure it out, but it blew my mind when I finally did.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Ron Capik]

bill wrote:

Ron Capik wrote:

...snip...

Also, there was a long sampling theory discussion in rec.audio.tech a few
months ago, [ Subject: 10 metres audio cable going into PC = too long? ]
that may be enlightening.

I give up on you guys on sampling theory versus reality.
On the cable, 10 metres = about 33 feet which means a propagation delay
of about 50 nS at the speed electricity travels in a conductor within a
dielectric. 1/50nS = about 20 MHz , not KHz, speaking in terms of
wavelength. I probably would have just gotten a good laugh out of that
discussion.

--
Bill (Sleepless biker) Baka

The communications theory discussion was part of the OT thread
drift much as this sub-thread has drifted from the topic, ...and yes,
sadly some of that discussion was humorous.


Later...

Ron Capik
--

[Bob Cain]

bill wrote:

I don't give a flying crap what Nyquist says because it is aimed at
steady state sine waves at that kind of a sample rate. If you can't
understand that, then I am arguing a point with a rock. Didn't anybody
ever teach you to think for yourself?

You shouldn't try that until you understand the fundamentals yourself.

So long as you are sampling a signal with a bandwidth strictly less
than twice the sampling rate you can reconstruct what was sampled
_exactly_ with the use of a filter with an impulse response of a
sampled sinc() function. That function is infinite in length in both
directions from its center which means it gathers information from the
entire signal to fill the gaps between the samples. Of course
infinity isn't physically realizable but windowed sinc() filters with
a finite number of coeficients can approximate that ideal
reconstruction with arbitrary accuracy.

This is called the sampling theorem and is one of the first things
taught and proved in any digital signal processing book or first course.

Your reliance on intuition gained from pictures is leading you astray.


Bob
--

"Things should be described as simply as possible, but no simpler."

A. Einstein

[bill]

Scott Dorsey wrote:
bill "just use comcast", "you know the drill for no spam bots." wrote:
Scott Dorsey wrote:
bill "just use comcast", "you know the drill for no spam bots." wrote:
I don't give a flying crap what Nyquist says because it is aimed at
steady state sine waves at that kind of a sample rate.
Every signal can be represented as the sum of steady state sine waves.
Every STEADY state signal.

No. It does not have to be steady state.

Fourier showed this in the 17th century.
I don't think he was concerned with audio. His math works nicely with
sine waves that have enough repetition to build up an approximation. He
had no clue that there would be computers doing things he never even
dreamed of.

That's the wonderful thing about mathematics! It works the same way,
no matter what the application is.

The Shannon Sampling Theorem
therefore is valid for any arbitrary waveform once it's proven for a
single sine wave, AS LONG AS that waveform is bandlimited. Discussion
of this can be found in Shannon's original BSTJ article.
Duh! An arbitrary waveform can be a 5KHz sine wave with 'up to' 20KHz
components superimposed on it. The 5 KHz will be properly digitized but
the 20KHz parts won't be.

They will be, just as long as the sampling rate is over 40 Ksamp/sec and
the anti-aliasing filter is properly set up.

Aperiodical signals? Square waves? Impulses? They can ALL be properly
reproduced as long as they are bandlimited; the only distortion that occurs
is the bandlimiting itself.

Stubborn, aren't you. I'm an engineer and I did the math (Fourier and
LaPlace)in the early 70's, so I understand the sampling theorem AND it's
limitations. Find a book that actually mentions the limitations instead
of pretending the math is foolproof and any exceptions can be
compensated for by filtering.

I don't think you did do the math because if you had read the original
Shannon paper, you'd understand just how incredibly elegant the whole
process is. It really is ingenious as hell.

Anti-aliasing filters are not for waveform
purity but to make sure the converter isn't fooled into creating a
totally bogus waveform.

These are the same thing.

Play with a digital oscilloscope sometime and
you will see how bad they can screw up a simple sine wave at the wrong
sweep and sample rate. They make a good case for me keeping my personal
analog scope to do sanity checks, although I know from experience what a
mis-sampled waveform looks like.

Right. You can also put a 10 MHz square wave into a 10 MHz analogue scope
and have it look like a sine wave, because of the bandlimiting. The
artifacts you see are due to bandlimiting.

You need some humbling experience, not my typing back and forth.
Someone else can teach you, hopefully before you embarrass yourself by
preaching in a meeting of more experienced peers.
I'm done with you.

I strongly, strongly want to suggest you sit down with the original Shannon
paper and read it, because once the idea hits you how the thing works, you
will be surprised. It really is pretty amazing.
--scott

I strongly advise you to sit down with some graph paper, so you feel
like you are 'enginering' and draw yourself a picture.
You can live with your 'blind' faith.
Later.

--
Bill (Sleepless biker) Baka

[bill]

Ron Capik wrote:
bill wrote:

Ron Capik wrote:

...snip...
Also, there was a long sampling theory discussion in rec.audio.tech a few
months ago, [ Subject: 10 metres audio cable going into PC = too long? ]
that may be enlightening.
I give up on you guys on sampling theory versus reality.
On the cable, 10 metres = about 33 feet which means a propagation delay
of about 50 nS at the speed electricity travels in a conductor within a
dielectric. 1/50nS = about 20 MHz , not KHz, speaking in terms of
wavelength. I probably would have just gotten a good laugh out of that
discussion.
--
Bill (Sleepless biker) Baka

The communications theory discussion was part of the OT thread
drift much as this sub-thread has drifted from the topic, ...and yes,
sadly some of that discussion was humorous.


Later...

Ron Capik
--


I think this was already off topic when I discovered this group about a
week ago, totally by accident when somebody cross posted to the bicycle
group I was on.
Hate cross-posters with nothing intelligent to say.

--
Bill (Sleepless biker) Baka

[bill]

Bob Cain wrote:
bill wrote:

I don't give a flying crap what Nyquist says because it is aimed at
steady state sine waves at that kind of a sample rate. If you can't
understand that, then I am arguing a point with a rock. Didn't anybody
ever teach you to think for yourself?

You shouldn't try that until you understand the fundamentals yourself.

I do and they never failed me with R.F. and video work where there was a
reasonably steady signal, unlike audio.

So long as you are sampling a signal with a bandwidth strictly less
than twice the sampling rate you can reconstruct what was sampled
_exactly_ with the use of a filter with an impulse response of a
sampled sinc() function. That function is infinite in length in both
directions from its center which means it gathers information from the
entire signal to fill the gaps between the samples. Of course
infinity isn't physically realizable but windowed sinc() filters with
a finite number of coeficients can approximate that ideal
reconstruction with arbitrary accuracy.

This is called the sampling theorem and is one of the first things
taught and proved in any digital signal processing book or first course.

Your reliance on intuition gained from pictures is leading you astray.

I hate to have to keep telling you that while the math was good with a
steady amplitude sine wave it does not work with a decaying sine wave,
like Cymbols (sp?) clanging, unless the decay rate is slow enough to
build up a number of samples. The proof of this would be to look at
distortion and sidebands generated by even a 15 KHz signal. While it may
be reconstructed within reason, a side by side comparison of the
original analog and the digitized output will show differences,
distortion, phase delay, and some sideband spikes a KHz or so on either
side of the desired waveform. This depends on the sampling rate, the
filter, and what you consider tolerable. If the sideband spurs are 80 dB
down with a noise floor of -120 dB then they will show up on a spectrum
analyzer but I don't think anybody could hear them. This can even be an
artifact of a clocked switched capacitor filter, which while nice, is
not perfect.
Have you ever seen anything man made that was totally flawless?


Bob


--
Bill (Sleepless biker) Baka

[Scott Dorsey]

bill "just use comcast", "you know the drill for no spam bots." wrote:

I strongly advise you to sit down with some graph paper, so you feel
like you are 'enginering' and draw yourself a picture.
You can live with your 'blind' faith.

It's not blind, because:

1. I have done the math. You ought to, if only because it's really
neat. If you don't like the original Shannon paper, get Stephen Harris'
application note from the Crystal Semi databook which goes through a
simplified version.

2. I have sat down with a scope, a signal generator, and converters. I
have seen converters that did an excellent job of reproducing a waveform.
I have seen some (like the PCM F-1) that did not, and then spent some
time investigating why they did not.

If the math says it should work, and actual bench testing shows it works,
then to my mind, it works.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Scott Dorsey]

bill "just use comcast", "you know the drill for no spam bots." wrote:
I hate to have to keep telling you that while the math was good with a
steady amplitude sine wave it does not work with a decaying sine wave,
like Cymbols (sp?) clanging, unless the decay rate is slow enough to
build up a number of samples.

Okay, here is an intuitive way of thinking about this: every aperiodic
waveform can be thought of as a small section of a longer repeating
waveform. Therefore any process that applies to a periodic waveform
can also apply to an aperiodic one over a finite time period.

The proof of this would be to look at
distortion and sidebands generated by even a 15 KHz signal. While it may
be reconstructed within reason, a side by side comparison of the
original analog and the digitized output will show differences,
distortion, phase delay, and some sideband spikes a KHz or so on either
side of the desired waveform.

But, if you actually measure it, these things do not happen. The only reason
phase shift occurs is because of the bandlimiting required as part of the
sampling system, and it is an implementation issue that has been overcome.

The sideband spikes are the result of timing error on the sampling clock
and we call it "jitter." It was a big issue back in the eighties but is
pretty much dealt with today.

Distortion can be caused by all kinds of things, including nonmonotonic
converters. Again, an implementation issue that in no way invalidates
the sampling theorem, and an issue that was dealt with a decade ago.

This depends on the sampling rate, the
filter, and what you consider tolerable. If the sideband spurs are 80 dB
down with a noise floor of -120 dB then they will show up on a spectrum
analyzer but I don't think anybody could hear them. This can even be an
artifact of a clocked switched capacitor filter, which while nice, is
not perfect.

Sideband spurs are the result of clock timing errors. There is a canonical
JAES paper from Stephen Harris on the subject, but the best intuitive
description I ever heard was from alice!jj at an IEEE thing a few years ago.
If you mail him he'll probably send you a copy.

Have you ever seen anything man made that was totally flawless?

Only mathematical models. That's what's so nice about them.
--scott

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

[bill]

Scott Dorsey wrote:
bill "just use comcast", "you know the drill for no spam bots." wrote:
I strongly advise you to sit down with some graph paper, so you feel
like you are 'enginering' and draw yourself a picture.
You can live with your 'blind' faith.

It's not blind, because:

1. I have done the math. You ought to, if only because it's really
neat. If you don't like the original Shannon paper, get Stephen Harris'
application note from the Crystal Semi databook which goes through a
simplified version.

2. I have sat down with a scope, a signal generator, and converters. I
have seen converters that did an excellent job of reproducing a waveform.
I have seen some (like the PCM F-1) that did not, and then spent some
time investigating why they did not.

If the math says it should work, and actual bench testing shows it works,
then to my mind, it works.
--scott

I would rather be reading something from Stephen Hawking or converting
Einstein's theories of relativity into a C++ program, but I guess I can
read some trivial paper and then return to this, ahhh, debate.

--
Bill (Sleepless biker) Baka

[William Sommerwerck]

I don't give a flying crap what Nyquist says because it is
aimed at steady state sine waves at that kind of a sample rate.

No, it's not. The Nyquist criterion applies to band-limited signals.

You don't seem to understand "superposition", either.

[bill]

Scott Dorsey wrote:
bill "just use comcast", "you know the drill for no spam bots." wrote:
I hate to have to keep telling you that while the math was good with a
steady amplitude sine wave it does not work with a decaying sine wave,
like Cymbols (sp?) clanging, unless the decay rate is slow enough to
build up a number of samples.

Okay, here is an intuitive way of thinking about this: every aperiodic
waveform can be thought of as a small section of a longer repeating
waveform. Therefore any process that applies to a periodic waveform
can also apply to an aperiodic one over a finite time period.

Uh-huh,
If you watch a slow, say 4.42 Hz sample of a single cycle of a 2 Hz sine
wave with random synchronization, and blip the sample points with a
scope Z-modulation on a digital scope, then look at the single sampled
reconstruction of that waveform, it will be different on every pass,
assuming the digitizer is reset before each new wave. After you see the
good in/garbage out, then try it with a 1 Hz wave, and while it will be
better, it will still not be completely accurate. Don't forget your 2 Hz
brick wall filter, which for this you can just write in C or C++.
Then tell me about how the math doesn't lie to you.
I have had the distinct pain in the ass of having to initiate PhD's into
the real world, and had to fire one guy with a Masters from Cal Poly
because he had a huge ego and nothing but a piece of paper to back it up.

The proof of this would be to look at
distortion and sidebands generated by even a 15 KHz signal. While it may
be reconstructed within reason, a side by side comparison of the
original analog and the digitized output will show differences,
distortion, phase delay, and some sideband spikes a KHz or so on either
side of the desired waveform.

But, if you actually measure it, these things do not happen. The only reason
phase shift occurs is because of the bandlimiting required as part of the
sampling system, and it is an implementation issue that has been overcome.

The sideband spikes are the result of timing error on the sampling clock
and we call it "jitter." It was a big issue back in the eighties but is
pretty much dealt with today.

I just looked at a graph heavy site on the subject a few days ago and
with a 15KHz signal on a modern system there were spikes about 30 dB
above the noise floor about every 1 KHz on every side. It wasn't a
manufacturer's ad but a site that did independent analysis, and now i
regret not bookmarking it, but it was just before I tripped into this
quagmire. They seemed to be all about discrediting the technology, so I
don't know how real the spectrum analyzer pictures were. The spikes were
post digital filter so it could have been due to the switched capacitor
clock getting through and causing strange things. I have a Tek scope but
not an audio spectrum analyzer so I can't verify what I saw on a
possibly bogus web site.

Distortion can be caused by all kinds of things, including nonmonotonic
converters. Again, an implementation issue that in no way invalidates
the sampling theorem, and an issue that was dealt with a decade ago.

This depends on the sampling rate, the
filter, and what you consider tolerable. If the sideband spurs are 80 dB
down with a noise floor of -120 dB then they will show up on a spectrum
analyzer but I don't think anybody could hear them. This can even be an
artifact of a clocked switched capacitor filter, which while nice, is
not perfect.

Sideband spurs are the result of clock timing errors. There is a canonical
JAES paper from Stephen Harris on the subject, but the best intuitive
description I ever heard was from alice!jj at an IEEE thing a few years ago.
If you mail him he'll probably send you a copy.

I can probably Google it. Everything else seems to be in their database.

Have you ever seen anything man made that was totally flawless?

Only mathematical models. That's what's so nice about them.

To borrow a quote from Bob Pease again "Just because the schematic
simulates in Spice does not mean the layout will work.". That one I can
really attest to.

--scott



--
Bill (Sleepless biker) Baka

[bill]

William Sommerwerck wrote:
I don't give a flying crap what Nyquist says because it is
aimed at steady state sine waves at that kind of a sample rate.

No, it's not. The Nyquist criterion applies to band-limited signals.

You don't seem to understand "superposition", either.


Too damned many self-proclaimed experts here. Read up on SDR, or
Software Defined Radio, and see the tricks they pull with intentional
undersampling. You might cringe at that trend.

--
Bill (Sleepless biker) Baka

[Scott Dorsey]

bill "just use comcast", "you know the drill for no spam bots." wrote:
After you see the
good in/garbage out, then try it with a 1 Hz wave, and while it will be
better, it will still not be completely accurate. Don't forget your 2 Hz
brick wall filter, which for this you can just write in C or C++.

NO. The filtering comes BEFORE the conversion. Not after. You cannot do
the filtering in the digital domain. YOU MUST BE SAMPLING A BANDLIMITED
SIGNAL, otherwise you get aliasing.

Many of the artifacts you do see on digital scopes at high speeds, though
have to do with the lack of anti-aliasing. That's an unfortunate side effect
of pushing wideband stuff out to the limits of technology; dropping the anti
aliasing filter is sometimes less of a problem than keeping it there. You will
see the same thing on datalogger systems.

Then tell me about how the math doesn't lie to you.

The math does not lie. The implementation might very well lie, and lots do.

I have had the distinct pain in the ass of having to initiate PhD's into
the real world, and had to fire one guy with a Masters from Cal Poly
because he had a huge ego and nothing but a piece of paper to back it up.

and we call it "jitter." It was a big issue back in the eighties but is

I just looked at a graph heavy site on the subject a few days ago and
with a 15KHz signal on a modern system there were spikes about 30 dB
above the noise floor about every 1 KHz on every side. It wasn't a
manufacturer's ad but a site that did independent analysis, and now i
regret not bookmarking it, but it was just before I tripped into this
quagmire.

That would seem about typical for something of the PCM 1610 level of
technology. What is even scarier is that you will see lots of noise
spikes well above the noise floor even with no modulation on some of
the earlier converters. Some of this was due to digital stuff leaking
into analogue lines, some of it due to idle tones on poorly-implemented
sigma delta systems (which is how they got called "Bitscream" converters
back then).

Things are a whole hell of a lot better than they were back then. And
back in the PCM 1610 days, you could hear all kinds of nastiness too.

I'd be curious what the site was, though. I'd bet some folks out there
like Prism and Lavry, though, might be willing to show real plots from
a AP testset on real converters. Hell, even the crappy soundcards on
Arny's site perform orders of magnitude better than anything available
in the early eighties.

They seemed to be all about discrediting the technology, so I
don't know how real the spectrum analyzer pictures were. The spikes were
post digital filter so it could have been due to the switched capacitor
clock getting through and causing strange things. I have a Tek scope but
not an audio spectrum analyzer so I can't verify what I saw on a
possibly bogus web site.

Switched capacitor filters are bad news for audio systems for the most
part, because of those modulation effects. Earlier converters tended to
use simple Bessel filters with dozens of stages chained one after the
other to make the filter sharp. Modern converters use oversampling or
sigma-delta stuff so the only filter required is a first order rolloff.

Sideband spurs are the result of clock timing errors. There is a canonical
JAES paper from Stephen Harris on the subject, but the best intuitive
description I ever heard was from alice!jj at an IEEE thing a few years ago.
If you mail him he'll probably send you a copy.

I can probably Google it. Everything else seems to be in their database.

JJ is a good guy that is willing to sit down with someone at a conference,
and answer stupid questions with good intuitive answers. He has helped me
a whole lot over the years.

Have you ever seen anything man made that was totally flawless?

Only mathematical models. That's what's so nice about them.

To borrow a quote from Bob Pease again "Just because the schematic
simulates in Spice does not mean the layout will work.". That one I can
really attest to.

I have a personal hatred of Spice, mostly due to trying to get undergrads
to understand that the Spice model doesn't encompass all the actual effects
in the real world, and that if it did, it would be harder to design than
the circuit itself.

But the CONCEPT of Spice... I love the concept.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Chris Hornbeck]

On Fri, 27 Oct 2006 00:28:58 GMT, bill
wrote:

Too damned many self-proclaimed experts here.

I think it may be more a case of everyone here having been
passed through the eye of this same needle at some time
past. Some more publicly (like me and you) and others
more privately. It's non-intuitive, so there's an eye
to be passed through.

But public or private, the Revealed Truth is very cool
and well worth the effort. It's almost, but not quite,
as cool as learning about dither, possibly the coolest
concept in technology.

Anywho, all good fortune, and glad to hear your voice here,

Chris Hornbeck

[Chevdo]

In article ,
says...

Chevdo wrote:
In article ,
says...
The experiment without a hypothesis is blind

That is the WRONG term.

No, just a different use of it than yours. Your mind is
insufficiently subtle to see the difference.

No it is the wrong term to use in this context since the term is used to mean
something else, something very specific and important, in this context. It's
unfortunate you're too immature of a usenet baby to accept that.


Hasn't school started yet?



Why, are you hoping to go molest some kids on the playground?

[Chevdo]

In article ,
says...

Chevdo wrote:
In article ,
says...
If "the scientific method" were the definition of science, then the best
scientists would be those who where the best at running scientific
experiments. But despite the need to carefully desingn, run, and interpret
experiments, the best scientists are those with the best insight, the best
imaginations.

The first hypothesis has to come from -- and it doesn't come from
experiment. It comes from imagination.



Doesn't matter whether those people are practicing science or not, you're
talking about superior people, not superior 'scientists'. Science without
the
scientific method is 'woo-woo', aka 'flummery'. It's how we got crap like
chiropracty and homeopathy that we for some reason now can't get rid of.
The
fact that a genius needs to be in charge of the lab janitors to generate
exception results, and that even the lab janitors have to be cognitively
superior to, say, the fader janitors who call themselves 'audio engineers',
and that even those fader janitors have to be cognitively superior to the
original mop & broom janitors, is superfulous. However, if geniuses simply
relied on their geniusness, rather than the scientific method, we'd be
absolutely NOWHERE, as we were for thousands of years until we finally
realized
that anything not proven by scientific method is WORTHLESS.

It was the Catholic church who stifled invention and burned people at
the stake for heresy in questioning 'God'. That put us back an easy
1,000 years.

So?



There were plenty
of geniuses who were around before the scientific method, and they
unfortunately didn't get very far.

Church again. Read about what happened to some of the inventors back
around Galileo's time.


I already know all about that. Does the name Bruno 'ring a bell'?


If Aristotle had been around in the
16th century, he'd probably rival or exceed Newton, and if either of them
had
been in the 20th century they'd probably rival or exceed Einstein.

Aristotle and Pythagoras were true geniuses before their time but the
Roman empire (conquer, rape, and burn) set things back a bit, not to
mention all the warring for no good reason. Ever wonder what was in the
library of Alexandria that has been lost forever?


Now you've ventured into erroneous territory. The Romans modelled themselves
after the ancient Greeks, they certainly didn't 'set things back'. Obviously
you didn't put as much effort into learning about Roman geniuses as Greek ones.
Anyway, it was the catholic church, again, who purged much of the ancient
Greek manuscripts in the middle ages. Most of those we have access to today
were hidden by byzantine monks, who often wrote over them with boring christian
prayer text so that they would be saved for future use (the Archamedes
manuscript, for example, was written over with christian prayer by a monk
looking to protect it, so he wrote perpendicular to the original text, making
it much easier for modern scholars to decypher what was originally there..)


Einstein
had the benefit of the access to the entirety of Newton, Aristotle, and many
other geniuses at his disposal, without which even he wouldn't have got very
far. The point is, research was totally unreliable before the scientific
method. Now it is far more reliable (though not totally, as evidenced by
the
amount of flummery that still gets published in peer-reviewed science
journals), and the various political machinations to shape and direct
research by pharmacutical companies, and over-zealous and/or ideologically
biased academics.

Let's see. There's primitives in office, like Bush declaring stem cell
research is immoral and against God's wishes. Ring a bell?


Yes, which I think I covered in the sentence you're responding to, but I also
don't think that the 'anti-science' crusade the Bush Administration is on is as
successful as those who lament it most passionately seem to think. They can't
even push creationism into public school in Kansas. Somehow I don't think
they're going to achieve any massive dumbing-down of the population at large.