Sign in with
Sign up | Sign in
Your question

"reconstruction" filters (so-called)

Last response: in Home Audio
Share
Anonymous
September 5, 2005 8:53:04 PM

Archived from groups: rec.audio.pro (More info?)

For those willing to crack a book, I recommend looking at Lahti, "An
Introduction to Random Signals and Communication Theory", which used to be a
standard college textbook on the subject. (Maybe it still is.)

Page 47 has a graphical representation of the sampling theorem. Look at it.
Think for a second or two. Then utter "Ahhh..." as you UNDERSTAND why there
is no such thing as a "reconstruction" filter. You cannot reconstruct what
is already present...
Anonymous
September 6, 2005 4:09:23 AM

Archived from groups: rec.audio.pro (More info?)

On 9/5/05 7:53 PM, in article KPKdnXvOrb0oR4HeRVn-pA@comcast.com, "William
Sommerwerck" <gizzledgeezer@comcast.net> wrote:

> For those willing to crack a book, I recommend looking at Lahti, "An
> Introduction to Random Signals and Communication Theory", which used to be a
> standard college textbook on the subject. (Maybe it still is.)
>
> Page 47 has a graphical representation of the sampling theorem. Look at it.
> Think for a second or two. Then utter "Ahhh..." as you UNDERSTAND why there
> is no such thing as a "reconstruction" filter. You cannot reconstruct what
> is already present...

OK Bill, let's try the semantic-garbage angle.
Taking a term commonly-defined and agreed-on in both lay folks AND academia,
and redefining it DIFFERENTLY when YOU talk is NOT about honest
communication. The context in ALL these texts, as well as the conversation
being attempted here, and understood and happily used by everybody else, is
that of reconstructing the -original waveform- from the sampling-donated
added information that was NOT part of the original waveform.

It ain;t ever been (except as you forcibly redefined it every time for your
own amusement) about 'reconstructing something not present' or 'something
lost' or 'something gained' or 'something borrowed' or 'something blue' ...
Anonymous
September 6, 2005 5:35:28 AM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck wrote:

> For those willing to crack a book, I recommend looking at Lahti, "An
> Introduction to Random Signals and Communication Theory", which used to be a
> standard college textbook on the subject. (Maybe it still is.)
>
> Page 47 has a graphical representation of the sampling theorem. Look at it.
> Think for a second or two. Then utter "Ahhh..." as you UNDERSTAND why there
> is no such thing as a "reconstruction" filter. You cannot reconstruct what
> is already present...

In the same way that an ADC requires an anti-aliasing filter, isn't the DAC
output filter classicly called an anti-imaging filter ?

I think the term reconstruction filter was an audio specific term for same.

Graham
Anonymous
September 6, 2005 5:35:29 AM

Archived from groups: rec.audio.pro (More info?)

"Pooh Bear" <rabbitsfriendsandrelations@hotmail.com> wrote
in message news:431CE450.B1FF2BA4@hotmail.com
> William Sommerwerck wrote:
>
>> For those willing to crack a book, I recommend looking
>> at Lahti, "An Introduction to Random Signals and
>> Communication Theory", which used to be a standard
>> college textbook on the subject. (Maybe it still is.)
>>
>> Page 47 has a graphical representation of the sampling
>> theorem. Look at it. Think for a second or two. Then
>> utter "Ahhh..." as you UNDERSTAND why there is no such
>> thing as a "reconstruction" filter. You cannot
>> reconstruct what is already present...
>
> In the same way that an ADC requires an anti-aliasing
> filter, isn't the DAC output filter classicly called an
> anti-imaging filter ?

Right. The spectrum of the sampled signal prior to low-pass
filtering following the DAC is the digitized signal
convolved with the sampling frequency:

http://www.web-ee.com/primers/files/AN-236.pdf page 6

"A final and complex consideration to understand is the
effects
of sampling. When a signal is sampled the end effect
is the multiplication of the signal by a unit sampling pulse
train as recalled from Figure 3a, c and e. The resultant
waveform has a spectrum that is the convolution of the
signal
spectrum and the spectrum of the unit sample pulse
train, i.e. Figure 3b, d, and f. If the unit sample pulse
has the
classical sin X/X spectrum5 of a rectangular pulse, see
Figure
13, then the convolution of the pulse spectrum with the
signal spectrum would produce the non-ideal sampled signal
spectrum shown in Figure 10a, b, and c."


> I think the term reconstruction filter was an audio
> specific term for same.

A rose by another name? ;-)

Here, they call it a smoothing filter:

"As a result of the D/A converter step function
response it is apparent that a large amount of undesirable
high frequency energy is present. To eliminate this the D/A
converter is usually followed by a smoothing filter, having
a
cutoff frequency no greater than half the sampling
frequency.
As its name suggests the filter output produces a
smoothed version of the D/A converter output which in fact
is a convolved function."

(reference, page 10)
Anonymous
September 6, 2005 7:43:39 AM

Archived from groups: rec.audio.pro (More info?)

>> For those willing to crack a book, I recommend looking at Lahti, "An
>> Introduction to Random Signals and Communication Theory", which used to
be a
>> standard college textbook on the subject. (Maybe it still is.)

>> Page 47 has a graphical representation of the sampling theorem. Look at
it.
>> Think for a second or two. Then utter "Ahhh..." as you UNDERSTAND why
there
>> is no such thing as a "reconstruction" filter. You cannot reconstruct
what
>> is already present...

> OK Bill, let's try the semantic-garbage angle.
> Taking a term commonly-defined and agreed-on in both lay folks AND
academia,
> and redefining it DIFFERENTLY when YOU talk is NOT about honest
> communication. The context in ALL these texts, as well as the conversation
> being attempted here, and understood and happily used by everybody else,
is
> that of reconstructing the -original waveform- from the sampling-donated
> added information that was NOT part of the original waveform.

Well, Humpty Dumpty, if a word means what _you_ want it to mean... Go ahead.

Calling a low-pass filter a "reconstruction filter" leads people to believe
things that simply aren't true.

You're like most human beings. You believe what you want to believe.
Anonymous
September 6, 2005 6:46:00 PM

Archived from groups: rec.audio.pro (More info?)

I use a 'reconstruction' filter algo to upsample and interpolate low
resolution waveforms for display.


"William Sommerwerck" <gizzledgeezer@comcast.net> wrote in message
news:KPKdnXvOrb0oR4HeRVn-pA@comcast.com...
> For those willing to crack a book, I recommend looking at Lahti, "An
> Introduction to Random Signals and Communication Theory", which used to be
a
> standard college textbook on the subject. (Maybe it still is.)
>
> Page 47 has a graphical representation of the sampling theorem. Look at
it.
> Think for a second or two. Then utter "Ahhh..." as you UNDERSTAND why
there
> is no such thing as a "reconstruction" filter. You cannot reconstruct what
> is already present...
>
>
Anonymous
September 6, 2005 7:36:16 PM

Archived from groups: rec.audio.pro (More info?)

On Tue, 6 Sep 2005 03:43:39 -0700, "William Sommerwerck"
<gizzledgeezer@comcast.net> wrote:

>>> For those willing to crack a book, I recommend looking at Lahti, "An
>>> Introduction to Random Signals and Communication Theory", which used to
>be a
>>> standard college textbook on the subject. (Maybe it still is.)

Offhand I see 16 copies for sale (bookfinder.com), all from 1968.
Dunno what happened to the author, maybe he went on to make more money
than being a professor/textbook author pays. Publishers generally
insist that authors "revise" textbooks and publish new editions every
few years to revitalize sales after too many used copies get
circulating.
Is there an equivalent or similar figure online, perhaps in the
online book at http://www.dspguide.com ?

I presume there is an earlier thread that prompted you to post
this, could you give me a reference for that?


>
>>> Page 47 has a graphical representation of the sampling theorem. Look at
>it.
>>> Think for a second or two. Then utter "Ahhh..." as you UNDERSTAND why
>there
>>> is no such thing as a "reconstruction" filter. You cannot reconstruct
>what
>>> is already present...

Certainly in the frequency domain, the filter REMOVES frequencies
that are not part of the original.

>
>> OK Bill, let's try the semantic-garbage angle.
>> Taking a term commonly-defined and agreed-on in both lay folks AND
>academia,
>> and redefining it DIFFERENTLY when YOU talk is NOT about honest
>> communication. The context in ALL these texts, as well as the conversation
>> being attempted here, and understood and happily used by everybody else,
>is
>> that of reconstructing the -original waveform- from the sampling-donated
>> added information that was NOT part of the original waveform.
>
>Well, Humpty Dumpty, if a word means what _you_ want it to mean... Go ahead.

I've been around people and groups who redefine words/overload the
language (without even knowing they're doing it), and I don't see that
happening here.

>
>Calling a low-pass filter a "reconstruction filter" leads people to believe
>things that simply aren't true.

It reconstructs the original band-limited waveform that goes into
the ADC.
This wording is quite different from "it removes the band of
frequencies above 1/2 Nyquist" but it is equivalent.

Even so, words used as technical terms often DO mean specific
things that don't correspond well to their meanings in general use.
Such differences often result in threads like this, and pat phrases
such as "It's only a theory."

>You're like most human beings. You believe what you want to believe.

I want to believe the truth. :) 
Anonymous
September 7, 2005 5:39:42 AM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck wrote:

> Page 47 has a graphical representation of the sampling theorem. Look at it.
> Think for a second or two. Then utter "Ahhh..." as you UNDERSTAND why there
> is no such thing as a "reconstruction" filter. You cannot reconstruct what
> is already present...

Sigh. It isn't present in the samples. The reconstruction
filter substitutes a sinc() function for each sample scaled
by the sample value. Where these overlap (because a sinc()
is extended in time) the results are summed. After that you
have what was present in the signal that was originally sampled.

There most certainly is such a thing as a reconstruction
filter. It reconstructs the signal from the samples.


Bob
--

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

A. Einstein
Anonymous
September 7, 2005 5:45:14 AM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck wrote:

> Calling a low-pass filter a "reconstruction filter" leads people to believe
> things that simply aren't true.

You are wrong. It isn't any old low-pass filter. A
reconstruction filter is one whose impulse response is a
sinc() function of time (or a windowed approximation to
one.) There are an infinite number of filters that can be
called a low-pass. There is only one low-pass with a sinc()
response.

> You're like most human beings. You believe what you want to believe.

It doesn't really matter what you believe about this, what
matters is the truth and that truth is unique and provable.


Bob
--

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

A. Einstein
Anonymous
September 7, 2005 5:53:25 AM

Archived from groups: rec.audio.pro (More info?)

Jona Vark wrote:
> I use a 'reconstruction' filter algo to upsample and interpolate low
> resolution waveforms for display.

This is a good way to really see what is going on. Cool
Edit Pro does the same for its display so that it can show
an accurate smooth line between samples which shows the
reconstructed continuous signal.

If you create a segment with a single sample in it that
isn't zero what you will see on the display (when zoomed in
so that a small number of samples fill the display) is the
sinc() function. The height of it be that of the sample.
If you start editing other samples to be non-zero, the
sinc()s from each are all added together to give the
reconstructed smooth signal.

What you see is the output of a sinc() reconstruction filter
in response to those samples going into it.


Bob
--

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

A. Einstein
Anonymous
September 7, 2005 6:53:31 AM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck wrote:

>>>For those willing to crack a book, I recommend looking at Lahti, "An
>>>Introduction to Random Signals and Communication Theory", which used to
>
> be a
>
>>>standard college textbook on the subject. (Maybe it still is.)
>
>
>>>Page 47 has a graphical representation of the sampling theorem. Look at
>
> it.
>
>>>Think for a second or two. Then utter "Ahhh..." as you UNDERSTAND why
>
> there
>
>>>is no such thing as a "reconstruction" filter. You cannot reconstruct
>
> what
>
>>>is already present...
>
>
>>OK Bill, let's try the semantic-garbage angle.
>>Taking a term commonly-defined and agreed-on in both lay folks AND
>
> academia,
>
>>and redefining it DIFFERENTLY when YOU talk is NOT about honest
>>communication. The context in ALL these texts, as well as the conversation
>>being attempted here, and understood and happily used by everybody else,
>
> is
>
>>that of reconstructing the -original waveform- from the sampling-donated
>>added information that was NOT part of the original waveform.
>
>
> Well, Humpty Dumpty, if a word means what _you_ want it to mean... Go ahead.
>
> Calling a low-pass filter a "reconstruction filter" leads people to believe
> things that simply aren't true.
>
> You're like most human beings. You believe what you want to believe.
>
>


Naw. It's the First Law of Communications: an error, once made,
must be propagated at all cost.

--
Les Cargill
Anonymous
September 7, 2005 9:03:21 AM

Archived from groups: rec.audio.pro (More info?)

> William Sommerwerck wrote:
>
> > Page 47 has a graphical representation of the sampling theorem. Look at
it.
> > Think for a second or two. Then utter "Ahhh..." as you UNDERSTAND why
there
> > is no such thing as a "reconstruction" filter. You cannot reconstruct
what
> > is already present...
>
> Sigh. It isn't present in the samples. The reconstruction
> filter substitutes a sinc() function for each sample scaled
> by the sample value. Where these overlap (because a sinc()
> is extended in time) the results are summed. After that you
> have what was present in the signal that was originally sampled.
>
> There most certainly is such a thing as a reconstruction
> filter. It reconstructs the signal from the samples.

Bob, I'm going to work and I don't have time to discuss this -- and I know
how futile discussing things with you can be.
Anonymous
September 7, 2005 5:55:06 PM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck wrote:

> Bob, I'm going to work and I don't have time to discuss this -- and I know
> how futile discussing things with you can be.

Yes, it can be quite futile to argue with someone who knows
of what they speak.

Make your original statement in comp.dsp and see what ensues.


Bob
--

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

A. Einstein
September 7, 2005 11:38:28 PM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck wrote:
> > You're arguing semantics, and Bob is talking about
> > what the filter does. Think about it.
>
> In a very strict, narrow sense, you're right. The filter (obviously!)
> changes the waveform to make it look more like the original waveform.
>
> The problem, though, in calling the filter a "reconstruction" filter, is
> that the term implies that the signal itself is being reconstructed. Nothing
> could be farther from the truth.

In the frequency domain, the original signal exists and the
reconstruction simply filter removes all the high frequency images. In
the time domain the reconstruction filter changes the waveform by
"connecting the dots". They are both the same thing being viewed from
two views, the frequency domain or the time domain.

tastes great and less filling

Mark
Anonymous
September 8, 2005 1:19:59 AM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck wrote:

> Bob, I'm going to work and I don't have time to discuss this -- and I know
> how futile discussing things with you can be.

You're arguing semantics, and Bob is talking about what the filter does.
Think about it.

--
ha
Anonymous
September 8, 2005 1:20:00 AM

Archived from groups: rec.audio.pro (More info?)

> You're arguing semantics, and Bob is talking about
> what the filter does. Think about it.

In a very strict, narrow sense, you're right. The filter (obviously!)
changes the waveform to make it look more like the original waveform.

The problem, though, in calling the filter a "reconstruction" filter, is
that the term implies that the signal itself is being reconstructed. Nothing
could be farther from the truth.
Anonymous
September 8, 2005 1:22:52 AM

Archived from groups: rec.audio.pro (More info?)

"Bob Cain" <arcane@arcanemethods.com> wrote in message
news:D fnk3a0emm@enews2.newsguy.com...
>
>
> William Sommerwerck wrote:
>
> > Bob, I'm going to work and I don't have time to discuss this -- and I
know
> > how futile discussing things with you can be.
>
> Yes, it can be quite futile to argue with someone who knows
> of what they speak.
>
> Make your original statement in comp.dsp and see what ensues.
>


ahh.. comp.dsp. When I am feeling like I know everything.. a trip there will
put me in my place.
Anonymous
September 8, 2005 1:59:08 AM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck wrote:
>>You're arguing semantics, and Bob is talking about
>>what the filter does. Think about it.
>
>
> In a very strict, narrow sense, you're right. The filter (obviously!)
> changes the waveform to make it look more like the original waveform.
>
> The problem, though, in calling the filter a "reconstruction" filter, is
> that the term implies that the signal itself is being reconstructed. Nothing
> could be farther from the truth.

The reason it is called a reconstruction filter is that it
reconstructs, from the discrete samples, the continuous
signal that was originally sampled.

Stuff is removed (stuff in between the samples that is
informationally redundant if the signal is band limited) by
the process of sampling and the process of reconstruction
puts the stuff back to recover the original continuous
signal. This is done by, of all things, a reconstruction
filter.

Does that make it any clearer?


Bob
--

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

A. Einstein
Anonymous
September 8, 2005 6:49:54 AM

Archived from groups: rec.audio.pro (More info?)

On 7 Sep 2005 19:38:28 -0700, "Mark" <makolber@yahoo.com> wrote:

>In the frequency domain, the original signal exists and the
>reconstruction simply filter removes all the high frequency images. In
>the time domain the reconstruction filter changes the waveform by
>"connecting the dots". They are both the same thing being viewed from
>two views, the frequency domain or the time domain.

Yer no fun. This thread could have run for days.

But seriously, thanks, as always,

Chris Hornbeck
Anonymous
September 8, 2005 8:15:02 AM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck wrote:

> > You're arguing semantics, and Bob is talking about
> > what the filter does. Think about it.
>
> In a very strict, narrow sense, you're right. The filter (obviously!)
> changes the waveform to make it look more like the original waveform.
>
> The problem, though, in calling the filter a "reconstruction" filter, is
> that the term implies that the signal itself is being reconstructed. Nothing
> could be farther from the truth.

Does the term imply, or do you infer?

--
ha
Anonymous
September 8, 2005 8:21:02 AM

Archived from groups: rec.audio.pro (More info?)

> In the frequency domain, the original signal exists and the
> reconstruction simply filter removes all the high frequency images. In
> the time domain the reconstruction filter changes the waveform by
> "connecting the dots". They are both the same thing being viewed from
> two views, the frequency domain or the time domain.

Correct.

One other point... As you cannot build a real-world filter that "completely"
removes the images, you can never "perfectly" reconstruct the original
waveform. But this is of no practical importance, because the image
components are above the range of hearing. So why talk about
"reconstructing" something that doesn't need reconstruction?
Anonymous
September 8, 2005 8:24:29 AM

Archived from groups: rec.audio.pro (More info?)

> > In a very strict, narrow sense, you're right. The filter (obviously!)
> > changes the waveform to make it look more like the original waveform.
> >
> > The problem, though, in calling the filter a "reconstruction" filter, is
> > that the term implies that the signal itself is being reconstructed.
Nothing
> > could be farther from the truth.
>
> The reason it is called a reconstruction filter is that it
> reconstructs, from the discrete samples, the continuous
> signal that was originally sampled.
>
> Stuff is removed (stuff in between the samples that is
> informationally redundant if the signal is band limited) by
> the process of sampling and the process of reconstruction
> puts the stuff back to recover the original continuous
> signal. This is done by, of all things, a reconstruction filter.

> Does that make it any clearer?

No, because you haven't stated anything you or anyone else in this group
doesn't already understand.

The term "reconstruction filter" was coined by engineers who didn't stop to
think about what they were saying, its menaing or implications.

Furthermore, the waveform is _not_ "the original", because you cannot
practically remove the images completely. How "original" is original?
Anonymous
September 8, 2005 10:27:22 AM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck kirjoitti:
> Page 47 has a graphical representation of the sampling theorem. Look at it.
> Think for a second or two. Then utter "Ahhh..." as you UNDERSTAND why there
> is no such thing as a "reconstruction" filter. You cannot reconstruct what
> is already present...

Strangely enough, I still feel it *does* reconstruct.

I guess a sculptor doesn't sculpt, either, in your opinion, because
the clay or wood or stone or whatever was already there?

Look, sampling theorem isn't limited to digital or even PCM.
Reconstruction filter is PCM specific - not sampling theorem
specific. Sampling theory isn't about real world - PCM is.

Timo
Anonymous
September 8, 2005 10:27:23 AM

Archived from groups: rec.audio.pro (More info?)

> > Page 47 has a graphical representation of the sampling theorem. Look at
it.
> > Think for a second or two. Then utter "Ahhh..." as you UNDERSTAND why
there
> > is no such thing as a "reconstruction" filter. You cannot reconstruct
what
> > is already present...


> Strangely enough, I still feel it *does* reconstruct.

> I guess a sculptor doesn't sculpt, either, in your opinion, because
> the clay or wood or stone or whatever was already there?

It is, in the sense that the sculptor removes all the material that isn't
needed. But that's an invalid comparison.


> Look, sampling theorem isn't limited to digital or even PCM.
> Reconstruction filter is PCM specific - not sampling theorem
> specific. Sampling theory isn't about real world - PCM is.

What?

The sampling theorem is not limited to digital. Aren't you aware that you
can have sampled analog systems?
Anonymous
September 8, 2005 4:12:16 PM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck wrote:

> The term "reconstruction filter" was coined by engineers who didn't stop to
> think about what they were saying, its menaing or implications.

I really don't understand your point. A reconstruction
filter reconstructs the signal that was originally sampled.
What is wrong, exactly, with that statement?

Sure a practical implementation is an approximation, and a
damned good one these days, but so is everything to one
degree or another. Platonic forms do not exist other than
in the mind.


Bob
--

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

A. Einstein
Anonymous
September 8, 2005 4:50:50 PM

Archived from groups: rec.audio.pro (More info?)

On Thu, 8 Sep 2005 04:24:29 -0700, William Sommerwerck wrote:

>>> In a very strict, narrow sense, you're right. The filter (obviously!)
>>> changes the waveform to make it look more like the original waveform.
>>>
>>> The problem, though, in calling the filter a "reconstruction" filter, is
>>> that the term implies that the signal itself is being reconstructed.
> Nothing
>>> could be farther from the truth.
>>
>> The reason it is called a reconstruction filter is that it
>> reconstructs, from the discrete samples, the continuous
>> signal that was originally sampled.
>>
>> Stuff is removed (stuff in between the samples that is
>> informationally redundant if the signal is band limited) by
>> the process of sampling and the process of reconstruction
>> puts the stuff back to recover the original continuous
>> signal. This is done by, of all things, a reconstruction filter.
>
>> Does that make it any clearer?
>
> No, because you haven't stated anything you or anyone else in this group
> doesn't already understand.
>
> The term "reconstruction filter" was coined by engineers who didn't stop to
> think about what they were saying, its menaing or implications.
>
> Furthermore, the waveform is _not_ "the original", because you cannot
> practically remove the images completely. How "original" is original?

OK, how does this sound for an explanation?

The first part of the A/D process is a very gentle analogue anti-alias
filter. Next comes the oversampling A/D going at n * 44.1kHz. Next comes
the software filter, which chops the signal hard at 20kHz, with minimal
phase error. This generates a trajectory at the oversampled rate which we
will call waveform "A".

Because of the software filtering, we can decimate the samples back to
44.1kHz without generating any alias products; this we store on a CD.

Now, to the reproduction stage. The CD is read, and a piece of software
oversamples again. This software exactly reconstructs waveform "A" from the
decimated samples, which can now be anti-aliased by another gentle analogue
filter.

That waveform reconstruction is done by a process that amounts to a low
pass filter - it is called a reconstruction filter for that reason.

d
Anonymous
September 8, 2005 7:45:16 PM

Archived from groups: rec.audio.pro (More info?)

>> The term "reconstruction filter" was coined by engineers who didn't
>> stop to think about what they were saying, its meaning or implications.

> I really don't understand your point. A reconstruction
> filter reconstructs the signal that was originally sampled.
> What is wrong, exactly, with that statement?

You just made my point for me -- you said signal instead of waveform.

In a certain sense, it does indeed "reconstruct" the waveform -- but it does
not reconstruct the original signal. That is the issue.
Anonymous
September 8, 2005 9:27:41 PM

Archived from groups: rec.audio.pro (More info?)

On Thu, 8 Sep 2005 04:24:29 -0700, "William Sommerwerck"
<gizzledgeezer@comcast.net> wrote:


>> { Bob Cain's completely reasonable response snipped. }

>> Does that make it any clearer?
>
>No, because you haven't stated anything you or anyone else in this group
>doesn't already understand.
>
>The term "reconstruction filter" was coined by engineers who didn't stop to
>think about what they were saying, its menaing or implications.

Are you also posting something similar on an armed services
newsgroup about the phrase "military intelligence?"

>Furthermore, the waveform is _not_ "the original", because you cannot
>practically remove the images completely. How "original" is original?

How many hundred db down do the images have to be before you
consider them 'practically' removed? Whatever it is, We Have The
Technology...
Anonymous
September 9, 2005 2:10:05 AM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck wrote:
>>>The term "reconstruction filter" was coined by engineers who didn't
>>>stop to think about what they were saying, its meaning or implications.
>
>
>>I really don't understand your point. A reconstruction
>>filter reconstructs the signal that was originally sampled.
>>What is wrong, exactly, with that statement?
>
>
> You just made my point for me -- you said signal instead of waveform.
>
> In a certain sense, it does indeed "reconstruct" the waveform -- but it does
> not reconstruct the original signal. That is the issue.

In what way, exactly and of any practical signifigance, does
it fail to reconstruct the sampled signal? Technical
details, please.


Bob
--

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

A. Einstein
Anonymous
September 9, 2005 4:12:45 AM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck wrote:

>>>The term "reconstruction filter" was coined by engineers who didn't
>>>stop to think about what they were saying, its meaning or implications.
>
>
>>I really don't understand your point. A reconstruction
>>filter reconstructs the signal that was originally sampled.
>>What is wrong, exactly, with that statement?
>
>
> You just made my point for me -- you said signal instead of waveform.
>

Isn't that a distinction without a difference ( what a Hasidic
friedn refers to as "pilpul" )?

> In a certain sense, it does indeed "reconstruct" the waveform -- but it does
> not reconstruct the original signal. That is the issue.
>
>

In what, a "you can't stick your hand in the same river twice"
sense? Delay is the whole point of re-cordin'.

--
Les Cargill
Anonymous
September 9, 2005 8:03:13 AM

Archived from groups: rec.audio.pro (More info?)

>> In a certain sense, it does indeed "reconstruct" the waveform --
>> but it does not reconstruct the original signal. That is the issue.

> In what way, exactly and of any practical signifigance, does
> it fail to reconstruct the sampled signal? Technical details,
> please.

Because the original signal is always present. Sampling does not change,
alter, or remove the original signal.

Sampling is not the process of removing, it is a process of adding. The
sampled signal is the original signal, plus the images.
Anonymous
September 9, 2005 10:08:07 PM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck kirjoitti:
>>Look, sampling theorem isn't limited to digital or even PCM.
> What?
> The sampling theorem is not limited to digital. Aren't you aware that you
> can have sampled analog systems?

Isn't that what I just said?

Timo
Anonymous
September 9, 2005 10:08:08 PM

Archived from groups: rec.audio.pro (More info?)

> >>Look, sampling theorem isn't limited to digital or even PCM.
> > What?
> > The sampling theorem is not limited to digital. Aren't you aware
> > that you can have sampled analog systems?

> Isn't that what I just said?

Not the way I interpreted it. You said something quite different.
Anonymous
September 10, 2005 2:57:10 AM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck wrote:
>>>In a certain sense, it does indeed "reconstruct" the waveform --
>>>but it does not reconstruct the original signal. That is the issue.
>
>
>>In what way, exactly and of any practical signifigance, does
>>it fail to reconstruct the sampled signal? Technical details,
>>please.
>
>
> Because the original signal is always present. Sampling does not change,
> alter, or remove the original signal.
>
> Sampling is not the process of removing, it is a process of adding. The
> sampled signal is the original signal, plus the images.

I give up.


Bob
--

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

A. Einstein
Anonymous
September 10, 2005 8:16:11 AM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck wrote:

> > >>Look, sampling theorem isn't limited to digital or even PCM.
> > > What?
> > > The sampling theorem is not limited to digital. Aren't you aware
> > > that you can have sampled analog systems?
>
> > Isn't that what I just said?
>
> Not the way I interpreted it. You said something quite different.

Might want to re-peruse it... <g>

--
ha
Anonymous
September 10, 2005 8:47:12 AM

Archived from groups: rec.audio.pro (More info?)

>>>>In a certain sense, it does indeed "reconstruct" the waveform --
> >>>but it does not reconstruct the original signal. That is the issue.

>>>In what way, exactly and of any practical signifigance, does
>>>it fail to reconstruct the sampled signal? Technical details, please.

>> Because the original signal is always present. Sampling does not
>> change, alter, or remove the original signal.

>> Sampling is not the process of removing, it is a process of adding.
>> The sampled signal is the original signal, plus the images.

> I give up.


As I do. I cannot understand why you don't "get" such a simple mathematical
principle.

I mean no personal offense, Bob, but (as we saw with the Doppler discussion)
you don't seem to have the ability to see things in terms of general
principles. I refer you to Dr. Land's observation about people who are
educated, but lack imagination.

My discussions on various UseNet groups have revealed two important
things...

1. I'm sometimes wrong about particular topics. Sometimes it's due to
simple ignorance, but it's most often because I haven't thought them through
carefully.

2. I'm considerably sharper and more insightful than a lot of the people on
these groups, many of whom have "educations". This is truly, horribly
frightening.

Regardless, I've said what I had to say. These are my last comments on the
subject.
Anonymous
September 10, 2005 2:52:50 PM

Archived from groups: rec.audio.pro (More info?)

=?ISO-8859-1?Q?Timo_Haanp=E4=E4?= <thaanpaa@sci.fi> wrote:
>William Sommerwerck kirjoitti:
>>>Look, sampling theorem isn't limited to digital or even PCM.
>> What?
>> The sampling theorem is not limited to digital. Aren't you aware that you
>> can have sampled analog systems?
>
>Isn't that what I just said?

Hardly anyone does, though, now that bucket brigade delay lines are gone.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."
Anonymous
September 10, 2005 7:40:54 PM

Archived from groups: rec.audio.pro (More info?)

"William Sommerwerck" <gizzledgeezer@comcast.net> writes:

>>> In a certain sense, it does indeed "reconstruct" the waveform --
>>> but it does not reconstruct the original signal. That is the issue.
>
>> In what way, exactly and of any practical signifigance, does
>> it fail to reconstruct the sampled signal? Technical details,
>> please.
>
> Because the original signal is always present. Sampling does not change,
> alter, or remove the original signal.

Stretching the limits of semantics here, how do you define a signal? Is
it the voltage that varies on a SPECIFIC wire (for example) at a specific
set of time values? In this sense, one can never copy, reconstruct, or
otherwise replicate an original signal since, in the physical world,
the reproduction will always be at later point in time. Thus, even if
the pattern of the replica matched exactly the pattern of the original,
the time origins of the two signals would differ.

No, sampling does not "destroy" the original signal in this sense. So
if you want to stretch semantics, you can certainly say that you can't
"reconstruct" what hasn't yet been destroyed. It goes without saying
(for most people) that what is being reconstructed is a copy of the
original signal.
--
% Randy Yates % "Bird, on the wing,
%% Fuquay-Varina, NC % goes floating by
%%% 919-577-9882 % but there's a teardrop in his eye..."
%%%% <yates@ieee.org> % 'One Summer Dream', *Face The Music*, ELO
http://home.earthlink.net/~yatescr
Anonymous
September 10, 2005 7:40:55 PM

Archived from groups: rec.audio.pro (More info?)

> No, sampling does not "destroy" the original signal in this sense. So
> if you want to stretch semantics, you can certainly say that you can't
> "reconstruct" what hasn't yet been destroyed. It goes without saying
> (for most people) that what is being reconstructed is a copy of the
> original signal.

Correct. You "get it".
Anonymous
September 11, 2005 2:40:20 AM

Archived from groups: rec.audio.pro (More info?)

On Sat, 10 Sep 2005 08:49:11 -0700, "William Sommerwerck"
<gizzledgeezer@comcast.net> wrote:

>> No, sampling does not "destroy" the original signal in this sense. So
>> if you want to stretch semantics, you can certainly say that you can't
>> "reconstruct" what hasn't yet been destroyed. It goes without saying
>> (for most people) that what is being reconstructed is a copy of the
>> original signal.
>
>Correct. You "get it".
>

I suppose I now "get it" too. Your problem isn't with the word
"reconstruct" (or perhaps you do have a problem with it in addition to
this aspect), it's that you can't make the ORIGINAL signal, you can
only make a copy, even if it's exact, or at least has unmeasurably low
error.
You have indeed stretched semantics here, and you've misled me, Bob
Cain and whoever else followed this thread, with no good reason.
Anonymous
September 11, 2005 2:40:21 AM

Archived from groups: rec.audio.pro (More info?)

"Ben Bradley" <ben_nospam_bradley@frontiernet.net> wrote in message
news:ivn6i1hcofr54esimd5mugeg4dr3g3rm4h@4ax.com...
> On Sat, 10 Sep 2005 08:49:11 -0700, "William Sommerwerck"
> <gizzledgeezer@comcast.net> wrote:
>
> >> No, sampling does not "destroy" the original signal in this sense. So
> >> if you want to stretch semantics, you can certainly say that you can't
> >> "reconstruct" what hasn't yet been destroyed. It goes without saying
> >> (for most people) that what is being reconstructed is a copy of the
> >> original signal.
> >
> >Correct. You "get it".

> I suppose I now "get it" too. Your problem isn't with the word
> "reconstruct" (or perhaps you do have a problem with it in addition to
> this aspect), it's that you can't make the ORIGINAL signal, you can
> only make a copy, even if it's exact, or at least has unmeasurably low
> error.

No, no, no, no, no. You're really reading something into this that I never
put there.


> You have indeed stretched semantics here, and you've misled me, Bob
> Cain and whoever else followed this thread, with no good reason.

No, Ben, I never made any such semantic distinction. I simply commented that
the person who made that post "got it" -- "you can certainly say that you
can't "reconstruct" what hasn't yet been destroyed." I was about to remove
his last sentence from the posting and decided not to. I should have.

I have mislead no one about anything. "You people" simply refuse to even
_consider_ a point of view that differs in the slightest from what you
already know to be "true".
Anonymous
September 11, 2005 5:10:00 AM

Archived from groups: rec.audio.pro (More info?)

"William Sommerwerck" <gizzledgeezer@comcast.net> writes:
> [...]
> "You people" simply refuse to even _consider_ a point of view that
> differs in the slightest from what you already know to be "true".

I think you are falsely accusing. The problem with your point of view is that it
is based on an extremely narrow use of words. In order for information to flow,
it requires both parties to work at trying to "get" the other one. You seem
to like to sit on your throne and have the world revolve around your meanings
and definitions.

And by the way, your narrow view of the term "reconstruction" doesn't
change my belief that its use is proper in the slightest. I.e., it IS
still true that it reconstructs in the sense intended by everyone else
in the world but you.
--
% Randy Yates % "My Shangri-la has gone away, fading like
%% Fuquay-Varina, NC % the Beatles on 'Hey Jude'"
%%% 919-577-9882 %
%%%% <yates@ieee.org> % 'Shangri-La', *A New World Record*, ELO
http://home.earthlink.net/~yatescr
Anonymous
September 11, 2005 5:46:47 AM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck wrote:

> I claim (without proof) that I can tell which people are talking sense, and
> are spouting drivel, by the "fact" that the former talk in terms of
> principles -- that when you read their work, you react with an "Aha! So
> that's it." Your writing is remarkably lacking in "Ahas!" And if you don't
> like my criticizing you for it, that's just tough.

So your "aha" is the test of soundness? Howya doing with
GR? Not getting it can just mean you don't know enough.

You obviously didn't follow the discussion you referred to
to the end (not that I would blame anyone for that
considering the bizzare turns the simple issue took getting
there.) It's in the sci.physics archives if you care to
look and was worked out at my request by a scientist named
Zigoteau. I framed the problem precisely for any arbitrary
signal asking for help getting a closed form solution and he
provided that solution via a truncated (at first order)
Mclaurin-Taylor series which I would never have seen myself.
The result has no prior publication that I can find. If
someone else knows of it I'll be happy to yield priority but
neither correctness nor completeness.

>>Nonetheless I saw the path to the answer and nobody else
>>involved in the noisy discussion had.
>
> Uh-huh. Do you know how that sounds to me and (at least some) others in this
> group? (If your tongue is in your cheek, it's not obvious.)

Who cares how it sounds when it the simple truth. One of
your problems is acute awareness of how things sound rather
than how correct they are.

>>Finally, back to reconstruction, you are using a frequency
>>domain argument to justify some obscure new view of the
>>time domain process.
>>In the time domain, wherein all the processes _actually_
>>[emphasis added by BS for effect] reside,
>
> Oh, ho ho ho ho ho ho! I can hear a lot of theoretical physicists (and not a
> few engineers) guffawing at _that_ belly-buster!

No part of the process is perfomed in the frequency domain.
No belly-buster there. It can be analyzed in that domain
but the actual operations are all time domain.

> One of the remarkable things about the physical universe is its apparent
> _duality_. Is light a particle or a wave? Well, it depends on what sort of
> experiment you perform. One way of looking at it is that light simply "is",
> and the way experiment varies with the form of perception.
>
> So from your point of view (see above) light is "really" a particle, and we
> can forget all that "wave" jazz? Or is it the other way 'round?

You sure know how to twist an argument into new territory.
You're reminding me of Will.

Actually, though, light really is a particle. See Feynman,
"QED: The Strange Story Of Light and Matter", for
elucidation of how particles can exhibit properties that are
wavelike without appeal to any kind of duality.

>>the reality is very simple.
>
> Reality? "I reject your reality and substitute my own!" -- Adam Savage

Exactly.

>>Redundant information
>>is removed from the continuous signal, leaving discrete
>>samples for storage and processing.
>
>
> That's an interesting way of looking at it. But it would be more correct to
> say that, in the time domain, the samples represent the minimum amount of
> information needed to unambiguously represent the signal. The samples don't
> actually remove anything.

The signal between the samples is removed and can be done so
safely because the information is redundant and still
carried in the samples if the signal was bandlimited before
the sampling. The fact that information about the signal is
in the samples is not at all the same as the signal being in
the samples. It requires calculation to restore the signal
from the information in the samples and that calculation is
performed by a reconstruction filter.

Maybe it's necessesary to define a signal here. It is a
physically measurable property that varies continuously with
time. Samples are not a signal.

>>Later, the original
>>continuous signal is restored by a filter which reconstructs
>>the original signal from the samples (exactly because the
>>information to do so is still there in the sampled form).
>
>
> Again, you just don't get. The filter doesn't restore anything -- the
> original signal is always present in the samples.

You don't get it. The information is in the samples, the
signal itself isn't. It is necessasary ultimately to
reconstruct the signal from the information that's in the
samples.

>>For obvious reasons that is called a reconstruction filter.
>>That the ideal reconstruction filter cannot be implemented
>>and requires approximation is totally beside the point
>>because that is true of so many things in audio that it is moot.
>
> Have I ever said otherwise?

I thought that at one point you had used the imperfection of
practical reconstruction to support your position (which I
hope to figure out some day.)

>>I know you don't want to continue this but I just can't
>>understand what, exactly and technically, you find wrong
>>with what I've said and what everyone in DSP would agree with.
>
> Agreement is not proof of correctness.

The consensus of experts, including a whole host of
doctorates, sure starts to approach such proof. I don't
think you are equipped to overturn what is commonly
understood by those actually working in the field.

> I've repeatedly stated that
> "reconstruction filter" is meaningful only in a limited sense, and that it
> leads to a fundamental misunderstanding of what sampling really is.

Yeah, you've stated, and stated, and stated it. It takes
more than repitition to convey a point, whatever that may be
in this case.

> Bob, how many articles have you read by so-called professionals that equate
> sampling with digitization? Does that consensus indicate truth?

Another Willism. I've never said sampling is equivalent to
digitization (or quantization as it is more conventionally
called.) I _fully_ understand the difference and the
technical consequences of the difference.


Bob
--

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

A. Einstein
Anonymous
September 11, 2005 8:42:38 AM

Archived from groups: rec.audio.pro (More info?)

>> "You people" simply refuse to even _consider_ a point of view that
>> differs in the slightest from what you already know to be "true".

> I think you are falsely accusing. The problem with your point of view is
that it
> is based on an extremely narrow use of words. In order for information to
flow,
> it requires both parties to work at trying to "get" the other one. You
seem
> to like to sit on your throne and have the world revolve around your
meanings
> and definitions.

> And by the way, your narrow view of the term "reconstruction" doesn't
> change my belief that its use is proper in the slightest. I.e., it IS
> still true that it reconstructs in the sense intended by everyone else
> in the world but you.

Thank you for being direct and honest.

By the way, there are other people who agree.
Anonymous
September 11, 2005 9:56:55 AM

Archived from groups: rec.audio.pro (More info?)

William Sommerwerck wrote:

> Because the original signal is always present.

There is no "signal" in the sense of somtehing you can audition. There
are bits, and that's it. After reconstruction, there is signal, and you
can hear it.

--
ha
Anonymous
September 11, 2005 9:56:56 AM

Archived from groups: rec.audio.pro (More info?)

>> Because the original signal is always present.

> There is no "signal" in the sense of something you can audition.
> There are bits, and that's it. After reconstruction, there is signal,
> and you can hear it.

Not quite...

It's true the signal is not "auditionable" when it's in "bit space" <grin>.
But we were talking about the time _after_ D/A conversion. The
"reconstruction" being discussed is _not_ filtering.

The original signal (re)appears during D/A conversion. The filter isn't
needed to reconstruct it. It's there, right at the output of the converter,
upstream of the filter.

Do you honestly believe that if you connected an amp at that point, you
_wouldn't_ hear the signal?
Anonymous
September 17, 2005 6:27:34 AM

Archived from groups: rec.audio.pro (More info?)

"William Sommerwerck" <gizzledgeezer@comcast.net> writes:

>>> Because the original signal is always present.
>
>> There is no "signal" in the sense of something you can audition.
>> There are bits, and that's it. After reconstruction, there is signal,
>> and you can hear it.
>
> Not quite...
>
> It's true the signal is not "auditionable" when it's in "bit space" <grin>.
> But we were talking about the time _after_ D/A conversion. The
> "reconstruction" being discussed is _not_ filtering.
>
> The original signal (re)appears during D/A conversion. The filter isn't
> needed to reconstruct it. It's there, right at the output of the converter,
> upstream of the filter.

I noticed no one has responded to you for a few days now - they're
probably tired of beating their heads against a wall. On the other
hand, my skull is thick...

It seems like you are looking at the signal just after the D/A in the
frequency domain. It is true (in theory) that the entire original
signal spectrum is present at the output of the D/A. However, there's
more. The original signal, the information at "baseband" (i.e., from
-Fs/2 to +Fs/2), is replicated infinitely. In fact a digital signal
[theoretically], even one that approaches zero at +/- infinity in the
time domain, has infinite energy.

However, it turns out that such a signal also has gaps in its
time-domain representation. In fact, the two phenomena are duals
of one another - i.e., the one causes the other, and vice-versa.

So when you apply a theoretical brick-wall filter to the output of
the D/A converter, you are both reconstructing and destroying. In
the time domain, you are reconstructing the signal at the points in
time between the sample points. In the frequency domain, you are
destroying the images above Fs/2.

> Do you honestly believe that if you connected an amp at that point, you
> _wouldn't_ hear the signal?

Actually, that is correct (i.e., you won't hear the signal), in
general. You will hear something stranger than the original signal.

You are thinking of a special case, i.e., when the sample rate is high
enough that the images from the D/A fall outside of the audible
range. If that weren't the case, e.g., if the sample rate were 4 kHz,
then you'd here a lot of garbage due to the image from 4 to 12 kHz,
and some of the one from 12 to 20 kHz.
--
% Randy Yates % "And all that I can do
%% Fuquay-Varina, NC % is say I'm sorry,
%%% 919-577-9882 % that's the way it goes..."
%%%% <yates@ieee.org> % Getting To The Point', *Balance of Power*, ELO
http://home.earthlink.net/~yatescr
Anonymous
September 17, 2005 6:35:53 AM

Archived from groups: rec.audio.pro (More info?)

Randy Yates <yates@ieee.org> writes:
> [...]
> You are thinking of a special case, i.e., when the sample rate is high
> enough that the images from the D/A fall outside of the audible
> range. If that weren't the case, e.g., if the sample rate were 4 kHz,
> then you'd here a lot of garbage due to the image from 4 to 12 kHz,
> and some of the one from 12 to 20 kHz.

Wups, a couple of errors here. The images I gave correspond to a
sample rate of 8 kHz, not 4 kHz. For a 4 kHz sample rate, the first
image goes from 2 to 6 kHz, the second from 6 to 10 kHz, etc. You
would hear a LOT of garbage in this signal.

Also, "hear" is not spelled "here." :) 
--
% Randy Yates % "How's life on earth?
%% Fuquay-Varina, NC % ... What is it worth?"
%%% 919-577-9882 % 'Mission (A World Record)',
%%%% <yates@ieee.org> % *A New World Record*, ELO
http://home.earthlink.net/~yatescr
Anonymous
September 17, 2005 8:37:03 AM

Archived from groups: rec.audio.pro (More info?)

> It seems like you are looking at the signal just after the D/A in the
> frequency domain. It is true (in theory) that the entire original
> signal spectrum is present at the output of the D/A. However, there's
> more. The original signal, the information at "baseband" (i.e., from
> -Fs/2 to +Fs/2), is replicated infinitely. In fact a digital signal
> [theoretically], even one that approaches zero at +/- infinity in the
> time domain, has infinite energy.

> However, it turns out that such a signal also has gaps in its
> time-domain representation. In fact, the two phenomena are duals
> of one another - i.e., the one causes the other, and vice-versa.

> So when you apply a theoretical brick-wall filter to the output of
> the D/A converter, you are both reconstructing and destroying. In
> the time domain, you are reconstructing the signal at the points in
> time between the sample points. In the frequency domain, you are
> destroying the images above Fs/2.

No, you are not reconstructing the signal -- unless you consider the
"signal" to be the waveform. I don't.


>> Do you honestly believe that if you connected an amp at that point,
>> you _wouldn't_ hear the signal?

> Actually, that is correct (i.e., you won't hear the signal), in
> general. You will hear something stranger than the original signal.

> You are thinking of a special case, i.e., when the sample rate is high
> enough that the images from the D/A fall outside of the audible
> range.

> If that weren't the case, e.g., if the sample rate were 4 kHz,
> then you'd here a lot of garbage due to the image from 4 to 12 kHz,
> and some of the one from 12 to 20 kHz.

The owners of hundreds of millions of CD players would not consider this a
special case.

The "special case" is when you're working with a signal that's band-limited
to less than the human ear's bandwidth -- eg, digital answering machines.

Even in this case, the filter "recnstructs" nothing whatsoever -- simply
removes something undesirable. That is not reconstruction.


What bothers me is that people use language in sloppy, poorly
thought-through ways, and never question it. It's amazing that when a
different point of view is present, people will do _anything_ to avoid
considering it. Remember Einstein's discomfort with the statistical nature
of quantum mechanics? It threw the conventional views of cause and effect
out the window. As far as I know, he didn't live long enough to see "spooky
action at a distance".

You cannot reconstruct something that hasn't been deconstructed. Sampling
does not deconstruct the original signal.
Anonymous
September 17, 2005 3:48:20 PM

Archived from groups: rec.audio.pro (More info?)

"William Sommerwerck" <gizzledgeezer@comcast.net> writes:
> [...]

Does your IQ break 100? Maybe it's my IQ I should be questioning
since I continue to reason with you. I guess I hoped you would see.
--
% Randy Yates % "The dreamer, the unwoken fool -
%% Fuquay-Varina, NC % in dreams, no pain will kiss the brow..."
%%% 919-577-9882 %
%%%% <yates@ieee.org> % 'Eldorado Overture', *Eldorado*, ELO
http://home.earthlink.net/~yatescr
!