Turing through wine country
I may have misread you. I thought you were saying that the waveforms posted were simple on-off toggling. My mistake. I'm human.
I may have misread you. I thought you were saying that the waveforms posted were simple on-off toggling. My mistake. I'm human.
fdegrove said:
First we have to switch to three-valued variables (at least), then
we have to define what toggling means for this variable.
Repeated toggling could mean circling through the values, it
could mean going forth and back between the end values passing
the middle value in both ways, it could mean switching for and
back between only two of these values (three possibilities here),
and an infinite number of other possibilities.
RE:JUST TOGGLING.
Hi Christer,
Now you're computing already...
I suggest you/we define toggling, although I feel a half decent dictionary would do, than see what it actually does in the digital world.
I'm no expert in anything digital but I don't remember bits having an intermediate status, they're either on or of, aren't they?
Don't worry though, I follow your reasoning...it just sounds like a bit (no pun intended) of reverse engineering to me.
Cheers,😉
Hi Christer,
Now you're computing already...
I suggest you/we define toggling, although I feel a half decent dictionary would do, than see what it actually does in the digital world.
I'm no expert in anything digital but I don't remember bits having an intermediate status, they're either on or of, aren't they?
Don't worry though, I follow your reasoning...it just sounds like a bit (no pun intended) of reverse engineering to me.
Cheers,😉
Frank,
I didn't mean we should define such a concept, I rather meant
that toggling is a welldefined operation on two-valued variables,
while I have never seen anybody defining toggling for variables
with other domain sizes (like {0,+,-}). One could define such
a concept, but would be a local definition for som special purpose,
within a journal article, for instance.
I didn't mean we should define such a concept, I rather meant
that toggling is a welldefined operation on two-valued variables,
while I have never seen anybody defining toggling for variables
with other domain sizes (like {0,+,-}). One could define such
a concept, but would be a local definition for som special purpose,
within a journal article, for instance.
Hi Christer,
In theory one could, yet that would be taking things into the totally abstract domain and should be stated as such in the article proper.
If your still pondering about what Fred meant, I feel we best leave it up to him to clear up the matter...before our bits get toggled off.😉
Cheers,😉
In theory one could, yet that would be taking things into the totally abstract domain and should be stated as such in the article proper.
If your still pondering about what Fred meant, I feel we best leave it up to him to clear up the matter...before our bits get toggled off.😉
Cheers,😉
fdegrove said:
If you have an application where it is useful it is not only of
theoretical interest. Whether it should be accompanied with
a "
theroy " depends entirely on in which contextit is done and who the presumed readers are.
I see no reason that publicing the Banach-Tarski theorem in
a mathematical journal need any such warnings, However, if
you provide the popular simplified explanation of it, how to
cut an orange into five pieces and put them together
again and get two oranges both identical to the one you started
with and the expected audience is not experienced
mathematicians, then it would be apropriate to say "don't try
this at home, it won't work". BTW, I do not know enough
about topology to understand the theorem, not to see what
therotical or possible practical use it could have.
[/B][/QUOTE]
We shouldn't dig more in this mess, perhaps, but I am a bit
puzzled (although not surprised) that Fred has kept silent.
Is he still trying to find a way out, or is it just so unthinkable
for him to admit an error? I used to have problems with that
myself, but nowadays I always try to admit when I am wrong.
Admitting to having been wrong is in my opinion much better
for the credibility than not doing so, but what do I bother.
It is obviously not my credibility (high or low) which suffers.
puzzled (although not surprised) that Fred has kept silent.
Is he still trying to find a way out, or is it just so unthinkable
for him to admit an error? I used to have problems with that
myself, but nowadays I always try to admit when I am wrong.
Admitting to having been wrong is in my opinion much better
for the credibility than not doing so, but what do I bother.
It is obviously not my credibility (high or low) which suffers.
Koinichiwa,
Oh no. WE SHOULD. I have repeatedly asked for someone to provide a simple and straightforward Wave File which encodes a valid audio signal (KHz sinewave) below -90.3dbfs in 16Bit/44.1KHz PCM and WITHOUT using dither, plus to provide the Wave file.
So all that was asked was an ABSOLUTELY SIMPLE, STRAIGHTFORWARD AND PRIMITIVE empirical proof of the Thesis that CD Standard Audio has a 96db dynamic range.
NO-ONE has done so yet.
Many have presented reams of math, which remind of nothing more than of the Butchers proof of the three in Lewis Carrol's "The hunting of the Snark":
"Taking Three as the subject to reason about-
A convenient number to state-
We add Seven, and Ten, and then multiply out
By One Thousand diminished by Eight. - "
"The result we proceed to divide, as you see,
By Nine Hundred and Ninety and Two:
Then subtract Seventeen, and the answer must be
Exactly and perfectly true. - "
"The method employed I would gladly explain,
While I have it so clear in my head,
If I had but the time and you had but the brain-
But much yet remains to be said. - "
So, instead of more such lessons in math, natural history and references to any number of papers which may or may not be relevent accurate and true in the context, why does not someone just go takes a 16Bit/44.1KHz PCM File (windows .wav will do fine) and show an undithered 1KHz sinewave at -91db and -96db that started out as -91db and -96db sinewaves. Please include details of the software used in creation and the settings used.
Surely there can be NOTHING MORE EASY and it should be much faster than endless discussions and arguments?
Except of course if you simply "cannot do that" (as has been previously contended and shown to be so when using Cooledit).
Sayonara
Christer said:
Oh no. WE SHOULD. I have repeatedly asked for someone to provide a simple and straightforward Wave File which encodes a valid audio signal (KHz sinewave) below -90.3dbfs in 16Bit/44.1KHz PCM and WITHOUT using dither, plus to provide the Wave file.
So all that was asked was an ABSOLUTELY SIMPLE, STRAIGHTFORWARD AND PRIMITIVE empirical proof of the Thesis that CD Standard Audio has a 96db dynamic range.
NO-ONE has done so yet.
Many have presented reams of math, which remind of nothing more than of the Butchers proof of the three in Lewis Carrol's "The hunting of the Snark":
"Taking Three as the subject to reason about-
A convenient number to state-
We add Seven, and Ten, and then multiply out
By One Thousand diminished by Eight. - "
"The result we proceed to divide, as you see,
By Nine Hundred and Ninety and Two:
Then subtract Seventeen, and the answer must be
Exactly and perfectly true. - "
"The method employed I would gladly explain,
While I have it so clear in my head,
If I had but the time and you had but the brain-
But much yet remains to be said. - "
So, instead of more such lessons in math, natural history and references to any number of papers which may or may not be relevent accurate and true in the context, why does not someone just go takes a 16Bit/44.1KHz PCM File (windows .wav will do fine) and show an undithered 1KHz sinewave at -91db and -96db that started out as -91db and -96db sinewaves. Please include details of the software used in creation and the settings used.
Surely there can be NOTHING MORE EASY and it should be much faster than endless discussions and arguments?
Except of course if you simply "cannot do that" (as has been previously contended and shown to be so when using Cooledit).
Sayonara
Kuei Yang Wang said:
No-one? I did so several days ago and i even gave the source
code for the program that generated it. So far noone has
reported finding any errors in the program.
Of course the result looks like, and is, a square wave, how could
it be otherwise at that level. Yet, you cannot claim that that
signal is not a sine but that the one Fred posted, with three
levels, is a sine. In fact, you can never get a sine wave out
of a PCM ecoding, whatever level you use, whatever number of
bits you use. It is just a question of differences in the harmonics.
Did I hear somebody whisper Fourier? No? Oh, well. His theory
was just theory -- a beutiful one, but comletely useless for any
practical purposes, or at least he claimed so himself. BTW, I
actually brought up Fouriers remark when the head of
research of a
wellknown telecom company once claimed that universities
should do applied research and not work with theories. He was,
for obvious reasons, quite disturbed by this remark.
Watashi no atarashi jidoosha akai
(Not very useful, but I am afraid that's about how far my
knowledge of japanese goes 🙂 )
Konichiwa,
Did I just miss the Wave file? I would like to load it into Cooledit and have a look. Also, is the file "valid audio" in so far that it conforms to PCM Specifications (I have something that checks this too)?
So, where is the file?
Sayonara
Christer said:
Did I just miss the Wave file? I would like to load it into Cooledit and have a look. Also, is the file "valid audio" in so far that it conforms to PCM Specifications (I have something that checks this too)?
So, where is the file?
Sayonara
Kuei,
I did not supply the whole file, since I happened to use the
10s default in my program, which makes it large. I supplied
the beginning of the file in text form. I could mail you the file,
if you wish. The file is about 1.7MB. I could regenerate a
shorter one, but then you may have to be patient, since
Ill be away next week.
As for validity, to the best of my knowledge, reading both
Microsoft docs and other sources, it is a perfectly valid PCM
encoded wave file, and so should be valid CD format, I think,
unless burning programs always do some tricks to the files,
tacitly reducing their quality.
I did not supply the whole file, since I happened to use the
10s default in my program, which makes it large. I supplied
the beginning of the file in text form. I could mail you the file,
if you wish. The file is about 1.7MB. I could regenerate a
shorter one, but then you may have to be patient, since
Ill be away next week.
As for validity, to the best of my knowledge, reading both
Microsoft docs and other sources, it is a perfectly valid PCM
encoded wave file, and so should be valid CD format, I think,
unless burning programs always do some tricks to the files,
tacitly reducing their quality.
Hm, I just loaded the file into Steinberg Wavelab and it seems
to have problems with it. However that various software
packages cannot handle it doesn't prove anything. It could be
that all programmers have fallen into the same trap as you and
made the wrong assumptions, remember how I fell for your
argument at first, before I realized the issued was much deeper
and more intricate. Also given that the average competence of
programmers is scaring and that commercial software, at least,
is written under economical and other constraints, it could also
be that handling this type of signals would necessiate awkward
special cases or even a completely different approach to the
program so the programmers simply decide to consider this
another anomalous case that isn't necessary to handle, since
it would be too costly or beyond their competence.
to have problems with it. However that various software
packages cannot handle it doesn't prove anything. It could be
that all programmers have fallen into the same trap as you and
made the wrong assumptions, remember how I fell for your
argument at first, before I realized the issued was much deeper
and more intricate. Also given that the average competence of
programmers is scaring and that commercial software, at least,
is written under economical and other constraints, it could also
be that handling this type of signals would necessiate awkward
special cases or even a completely different approach to the
program so the programmers simply decide to consider this
another anomalous case that isn't necessary to handle, since
it would be too costly or beyond their competence.
"Of course the result looks like, and is, a square wave, how could it be otherwise at that level. Yet, you cannot claim that that signal is not a sine but that the one Fred posted, with three levels, is a sine. In fact, you can never get a sine wave out
of a PCM ecoding, whatever level you use, whatever number of bits you use. It is just a question of differences in the harmonics".
I am agreement with that and the interesting thing is whether the smallest analog voltage change that the DAC can make is really a representation of the smallest signal still representative of the full scale sine wave from the same DAC. It must be least argued that the signal is non-monotonic and differs in harmonic content to the extent the even the mathematicians should squirm at calling this an attenuated version of a sign wave. Using the juxtaposition of looking at this smallest signal possible from the DAC and trying deduce the full scale waveform leaves one with some real ambiguities as to possible signals it could be. I could argue that even smaller numbers would be appropriate to represent something that would convey an actual "real" dynamic range in the sense that analog designers would be comfortable with. To do so would arouse the same reactions as setting the orphanage on fire to count the number of escaping children in an effort to see if everyone was present at bed check, and not lurking around where they shouldn't be. I can imagine the parallels to the discussions on digital dynamic range in discussing signal to noise ratio in the analog domain. Arguments over that number would lead to disputes over how far one can hear into the noise floor. As a hands on problem solving engineer working within the constraints of the real world, I shudder when I here the words "in theory" in anticipation of the battle between what's conceivable and what's achievable. Guess who the task of minimizing the distance between those goals falls to.......
of a PCM ecoding, whatever level you use, whatever number of bits you use. It is just a question of differences in the harmonics".
I am agreement with that and the interesting thing is whether the smallest analog voltage change that the DAC can make is really a representation of the smallest signal still representative of the full scale sine wave from the same DAC. It must be least argued that the signal is non-monotonic and differs in harmonic content to the extent the even the mathematicians should squirm at calling this an attenuated version of a sign wave. Using the juxtaposition of looking at this smallest signal possible from the DAC and trying deduce the full scale waveform leaves one with some real ambiguities as to possible signals it could be. I could argue that even smaller numbers would be appropriate to represent something that would convey an actual "real" dynamic range in the sense that analog designers would be comfortable with. To do so would arouse the same reactions as setting the orphanage on fire to count the number of escaping children in an effort to see if everyone was present at bed check, and not lurking around where they shouldn't be. I can imagine the parallels to the discussions on digital dynamic range in discussing signal to noise ratio in the analog domain. Arguments over that number would lead to disputes over how far one can hear into the noise floor. As a hands on problem solving engineer working within the constraints of the real world, I shudder when I here the words "in theory" in anticipation of the battle between what's conceivable and what's achievable. Guess who the task of minimizing the distance between those goals falls to.......
http://users.csionline.net/~rmini/500Hz-96dB.zip
It compresses well 😉
According to Cool Edit, it's -97.6dB RMS, referenced to 0dB FS being a sine.
Frequency analysis shows it to be a high-distortion 500 Hz wave.
It compresses well 😉
According to Cool Edit, it's -97.6dB RMS, referenced to 0dB FS being a sine.
Frequency analysis shows it to be a high-distortion 500 Hz wave.
'Also given that the average competence of programmers is scaring"
First rule of Telecom design:
Thou shalt not touch the software!
I suggested it once and feared for the future of my employment past the end of the week. I got about the same looks that one would expect after suggesting building a house of cards bird house for a woodpecker .....
First rule of Telecom design:
Thou shalt not touch the software!
I suggested it once and feared for the future of my employment past the end of the week. I got about the same looks that one would expect after suggesting building a house of cards bird house for a woodpecker .....
Attachments
Cooledit, demo version. Dither off. 16 bit source material, 440Hz 0dBFS.
Top to bottom:
.91 dB attenuation
.95dB attenuation
.96dB attenuation, starting from source
tone at 0dBFS but with 0.001% offset
not shown: 96dB att. on a pure-AC source wave, this obviously gives a flatliner.
I do wonder, though, about the numerical accuracy of CE's attenuator.
Top to bottom:
.91 dB attenuation
.95dB attenuation
.96dB attenuation, starting from source
tone at 0dBFS but with 0.001% offset
not shown: 96dB att. on a pure-AC source wave, this obviously gives a flatliner.
I do wonder, though, about the numerical accuracy of CE's attenuator.
Attachments
Konnichiwa,
Actually, let's have a look what it REALLY looks like. We are being told it is a -96dbfs 500Hz wave encoded....
Okay, what does cooledit say?
http://groups.yahoo.com/group/Thunderstone_audiophile/files/96dboff/13.jpg
Actually (look at the status bar near the bottom) - it says:
-90.3db @ 0:12.701
Now zoom in. I find this display extremely interesting, it appears to be a 25KHz sinewave, apparently being switched on/off at rate equal to 500Hz. No offence, but this does not appear to be a "valid" audio signal at all.... (Ask Werner)
Now the FFT in Cool Edit, nothing remotely approaching -96db, the 500Hz component is at -100db by the looks of it:
But we don't only have "cooledit" here, using the Wavetools Anaylser I get:
Here again we see an input level of 90.3dbfs with a 500Hz component encoded within all that noise at -100.3db.
All in all that still looks nothing like what a -96dbfs sinewave would look like....
So, we know we can generate some sort of noise below (or rather exactly AT) -90.3dbfs. If averaged long enough this noise can be shown to have a component of the inended frequency at 10db below -90.3dbfs and thus 10db below the effective noisefloor. Congratualtions are on order it seems.
Sayonara
RobM said:
Actually, let's have a look what it REALLY looks like. We are being told it is a -96dbfs 500Hz wave encoded....
Okay, what does cooledit say?
An externally hosted image should be here but it was not working when we last tested it.
http://groups.yahoo.com/group/Thunderstone_audiophile/files/96dboff/13.jpg
Actually (look at the status bar near the bottom) - it says:
-90.3db @ 0:12.701
Now zoom in. I find this display extremely interesting, it appears to be a 25KHz sinewave, apparently being switched on/off at rate equal to 500Hz. No offence, but this does not appear to be a "valid" audio signal at all.... (Ask Werner)
An externally hosted image should be here but it was not working when we last tested it.
http://groups.yahoo.com/group/Thunderstone_audiophile/files/96dboff/11.jpg 
Now the FFT in Cool Edit, nothing remotely approaching -96db, the 500Hz component is at -100db by the looks of it:
An externally hosted image should be here but it was not working when we last tested it.
http://groups.yahoo.com/group/Thunderstone_audiophile/files/96dboff/12.jpg
But we don't only have "cooledit" here, using the Wavetools Anaylser I get:
An externally hosted image should be here but it was not working when we last tested it.
http://groups.yahoo.com/group/Thunderstone_audiophile/files/96dboff/9.jpg
Here again we see an input level of 90.3dbfs with a 500Hz component encoded within all that noise at -100.3db.
All in all that still looks nothing like what a -96dbfs sinewave would look like....
So, we know we can generate some sort of noise below (or rather exactly AT) -90.3dbfs. If averaged long enough this noise can be shown to have a component of the inended frequency at 10db below -90.3dbfs and thus 10db below the effective noisefloor. Congratualtions are on order it seems.
Sayonara
That's because none of that software properly handles the DC offset. Additionally, the waveform display in Cool Edit is flawed at that zoom level; an inspection of the file will show that it's a stream of contiguous 0's, then contiguous 1's, etc.
In Cool Edit, go to the "Analyze" menu, then select "Summary..." You will then be enlightened 😉
In Cool Edit, go to the "Analyze" menu, then select "Summary..." You will then be enlightened 😉
Tried the same now with CE Pro. Its attenuator works differently.
Never mind.
Generating the source signal in a 32 bit domain, and attenuating there with 96dB together with the addition of a DC offset of the equivalent of 1/2 LSB(16bit), and then quantizing to 16 bit (without dither), yields the same as my lowest trace above: the representation of a sine wave with a signal-to-quantisation-noise ratio of almost 0dB. This is the limit level that is still being encoded, 96dB below full scale.
Never mind.
Generating the source signal in a 32 bit domain, and attenuating there with 96dB together with the addition of a DC offset of the equivalent of 1/2 LSB(16bit), and then quantizing to 16 bit (without dither), yields the same as my lowest trace above: the representation of a sine wave with a signal-to-quantisation-noise ratio of almost 0dB. This is the limit level that is still being encoded, 96dB below full scale.
- Status
- Not open for further replies.