Steve Eddy said:
OK, I see what my mistake was now. I didn't quite realize that
we could apply exactly the same definition in both cases, so
I sloppily took 1 LSB as the floor value, while it should be the
noise floor also in this case, ie 1/2 LSB.
Of course, then we can usually still have information hidden in
the noise in the analog case, but that's another story.
Steve Eddy said:
Well, since noise by definition is not information, the CD recorded with music would have the most information on it.
BUT...
With just a tiny amount of noise added to a non-dithered digital system, you can actually put MORE information on the CD than without the tiny amount of noise.
se
I said from an information-theoretic perspective, not an electronic
one. The noise is of course the actual signal we want on the
CD. Although I meant is an theoretical example, you can
imagine a test CD recorded with white noise.
So to make it more precise, CD A contains 60 min of music
recorded to make full use of the dynamic range and CD B
contains 60 min of white noise, also making use of the full
dynamic range.
Of course, if you have access to a whilte-noise CD you could
cheat by doing a simple experiment.
Christer said:
Right.
And with dithering/noise shaping, you can have that in a digital system too. And do have it seeing as virtually every digital recording system employs some form of dithering or noise shaping.
I said from an information-theoretic perspective, not an electronic
one. The noise is of course the actual signal we want on the
CD. Although I meant is an theoretical example, you can
imagine a test CD recorded with white noise.
So to make it more precise, CD A contains 60 min of music
recorded to make full use of the dynamic range and CD B
contains 60 min of white noise, also making use of the full
dynamic range.
Of course, if you have access to a whilte-noise CD you could
cheat by doing a simple experiment.
Ok, ok. You got me.
So here's one for you:
If a hen-and-half lays an egg-and-a-half in a day-and-a-half, then how many elephants does it take to mix a martini?
se
Koinichiwa,
Yet you cannot encode a signal below -90.3db. If I where measuring an analogue system where below -90.3db I hit the noisefloor (analogue behaves somewhat different than digital near the bottom of the dynamic range) the dynamic range would be qualified as 90.3db, not 96.3db.
You can wave around any amount of math, it does not change simple facts which are beyond any dispute, as they are simple facts, which can be readily observed.
Sayonara
Steve Eddy said:
Yet you cannot encode a signal below -90.3db. If I where measuring an analogue system where below -90.3db I hit the noisefloor (analogue behaves somewhat different than digital near the bottom of the dynamic range) the dynamic range would be qualified as 90.3db, not 96.3db.
You can wave around any amount of math, it does not change simple facts which are beyond any dispute, as they are simple facts, which can be readily observed.
Sayonara
But then what the hell do the guys at Burr Brown know.......
From the following PDF from Burr Brown:
DYNAMIC RANGE
The Dynamic Range is a measure of the ratio of the smallest
signals the converter can produce to the full-scale range and
is usually expressed in decibels (dB). The theoretical dynamic
range of a converter is approximately 6 x n, or about 96dB
of a 16-bit converter. The actual, or useful, dynamic range is
limited by noise and linearity errors and is therefore somewhat
less than the theoretical limit. However, this does point out
that a resolution of at least 16 bits is required to obtain a
90dB minimum dynamic range, regardless of the accuracy
of the converter. Another specification that is useful for
audio applications is Total Harmonic Distortion.
http://katalogi.iele.polsl.gliwice.pl/en/download.php?id=4083
Go back to drinking martinis with your pink elephants. If I might digress a least significant bit , how many martinis did you drink before posting the Cap-tastrophe Theory chart?
From the following PDF from Burr Brown:
DYNAMIC RANGE
The Dynamic Range is a measure of the ratio of the smallest
signals the converter can produce to the full-scale range and
is usually expressed in decibels (dB). The theoretical dynamic
range of a converter is approximately 6 x n, or about 96dB
of a 16-bit converter. The actual, or useful, dynamic range is
limited by noise and linearity errors and is therefore somewhat
less than the theoretical limit. However, this does point out
that a resolution of at least 16 bits is required to obtain a
90dB minimum dynamic range, regardless of the accuracy
of the converter. Another specification that is useful for
audio applications is Total Harmonic Distortion.
http://katalogi.iele.polsl.gliwice.pl/en/download.php?id=4083
Go back to drinking martinis with your pink elephants. If I might digress a least significant bit , how many martinis did you drink before posting the Cap-tastrophe Theory chart?
Attachments
Well I never thought of this before, but it seems to me that indeed the 16 bits must suffice to describe the peak to peak waveform. So we get 96 dB from zero to 2xpeak. And we get just 90 dB from "ground" to 1xpeak.
Christer: of course the noise has the most information content. Any non random information such as music by definition contains redundancies. Hence it may be losslessly compressed by a suitably algorithm. Pure random noise by definiton contains no redundancies, hence may not be compressed by any means.
I'm too lazy to actually compress a white noise track on mp3.
Anyway...
Christer: of course the noise has the most information content. Any non random information such as music by definition contains redundancies. Hence it may be losslessly compressed by a suitably algorithm. Pure random noise by definiton contains no redundancies, hence may not be compressed by any means.
I'm too lazy to actually compress a white noise track on mp3.
Anyway...
Koinichiwa,
I have just asked a number of Wave generation programs to create a 1KHz sinewave at -96db. I have only one program that does not flat out refuse to do that and asks me to choose a sensible level, like -90db or higher.
I have one program (DACTest - see attached Zip) that will attempt the generation. Syntrillium Cooledit Pro shows all files generated down to -90dbfs to contain data, any below that (I tried -92, - 94 and -96dbfs) show a simple "digital silence" flatline.
QED
Sayonara
SY said:
I have just asked a number of Wave generation programs to create a 1KHz sinewave at -96db. I have only one program that does not flat out refuse to do that and asks me to choose a sensible level, like -90db or higher.
I have one program (DACTest - see attached Zip) that will attempt the generation. Syntrillium Cooledit Pro shows all files generated down to -90dbfs to contain data, any below that (I tried -92, - 94 and -96dbfs) show a simple "digital silence" flatline.
QED
Sayonara
Koinichiwa,
As you seem to think of yourself as "someone who knows how to use the software properly (and proper software)" how about you take a suitable piece of generically downloadable software, generate the file and post it together with a screenshot of the settings used to generate the file. Note that dither is NOT allowed and the format must be a 16Bit/44.1KHz WAV File generated directly or by conversion from higher formt only if no dither is applied.
Sayonara
SY said:
As you seem to think of yourself as "someone who knows how to use the software properly (and proper software)" how about you take a suitable piece of generically downloadable software, generate the file and post it together with a screenshot of the settings used to generate the file. Note that dither is NOT allowed and the format must be a 16Bit/44.1KHz WAV File generated directly or by conversion from higher formt only if no dither is applied.
Sayonara
Koinichiwa,
Nor am I.
I do not claim nyquist is "wrong".
I did what you suggested and all but one piece of waveediting software I have at hand refused to generate a 1KHz sinewave with < -90dbfs level. The one that did not refuse produced a file containing digital silence. If you dislike the results form the experiement - please provide one that is repeatable and gives results you like, assuming they are achievable.
Sayonara
SY said:
Nor am I.
SY said:
I do not claim nyquist is "wrong".
SY said:
I did what you suggested and all but one piece of waveediting software I have at hand refused to generate a 1KHz sinewave with < -90dbfs level. The one that did not refuse produced a file containing digital silence. If you dislike the results form the experiement - please provide one that is repeatable and gives results you like, assuming they are achievable.
Sayonara
Koinichiwa,
I must dispute this. I have done as close an experiemnt as I can make. The software behaves as predicted by standard math for 16 Bit digital audio. It reliably encodes -90dbfs and does not encode anything below that.
I repeat, if you dislike the math and the results of practical tests the onus is on you to prove differently.
Sayonara
SY said:
I must dispute this. I have done as close an experiemnt as I can make. The software behaves as predicted by standard math for 16 Bit digital audio. It reliably encodes -90dbfs and does not encode anything below that.
I repeat, if you dislike the math and the results of practical tests the onus is on you to prove differently.
Sayonara
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.