Is the process of audio sampling a->d deterministic ? In other words if I play a vaw for example out from my dac and then acquire the analog out with a analog to digital converter more and more times, the file obtained are bit to bit identical ?
I think not, so i am curios about how different could be, even if I do not know nere and now a tools to compare the difference; perhaps the conversion can be made in wav so a proper audio tools can extract the diff.
I ammon holidays without tools to inspect ... but curiosity is high ...
I apologize for my bad english and for the silly question, I am searching a tools to inspect the difference between different digital source ...
Mauro
I think not, so i am curios about how different could be, even if I do not know nere and now a tools to compare the difference; perhaps the conversion can be made in wav so a proper audio tools can extract the diff.
I ammon holidays without tools to inspect ... but curiosity is high ...
I apologize for my bad english and for the silly question, I am searching a tools to inspect the difference between different digital source ...
Mauro
Hi Mauro, if you do a D/A conversion and then A/D then yes each time you do it you will be altering the signal from the original. If you do it multiple times the error will compound.
I've not tried it but have seen it mentioned here a number of times. but audio diffmaker seems to be what you are after. Audio DiffMaker
Tony.
I've not tried it but have seen it mentioned here a number of times. but audio diffmaker seems to be what you are after. Audio DiffMaker
Tony.
In theory, all versions of the file would be identical. In practice, they would all be different. Subsequent A to D conversions would include what ever noise was present in the analog domain. The D to A conversions would, most likely, have added dither, which is digital noise. So, each generation would have additional noise.
no analog channel is "deterministic" if you mean noise free
you may be thinking about what some call "generation loss" - put a signal thru some number of record/measure/playback/rerecord.. cycles and see how the signal changes
any purely analog scenario - say studio tape recorder, EQ, lacquer cutting, needle drop, preamp with inverse EQ back to tape... chain is going to be laughably worse than most motherboard digital audio chipsets
but even really good audio ADC/DAC aren't going to give bit perfect loopback because of gains, scale factors, offsets, filters all have tolerances orders of magnitude worse than even 16 lsb resolution
you may be thinking about what some call "generation loss" - put a signal thru some number of record/measure/playback/rerecord.. cycles and see how the signal changes
any purely analog scenario - say studio tape recorder, EQ, lacquer cutting, needle drop, preamp with inverse EQ back to tape... chain is going to be laughably worse than most motherboard digital audio chipsets
but even really good audio ADC/DAC aren't going to give bit perfect loopback because of gains, scale factors, offsets, filters all have tolerances orders of magnitude worse than even 16 lsb resolution
Last edited:
In practice you will almost certainly not get a bit perfect copy, but... it could be arranged for you to do so, especially if the ADC and DAC were of low resolution and synchronised in their timing.
Some of the earliest computer memories were based on continuous recirculation of acoustic pulses in delay lines and relied on the pulses being received at one end and re-transmitted at the other, effectively by a 1-bit ADC and DAC, so a 'bit perfect' copy was essential.
Delay line memory - Wikipedia, the free encyclopedia
Some of the earliest computer memories were based on continuous recirculation of acoustic pulses in delay lines and relied on the pulses being received at one end and re-transmitted at the other, effectively by a 1-bit ADC and DAC, so a 'bit perfect' copy was essential.
Delay line memory - Wikipedia, the free encyclopedia
Also shocking how good digital really works is this insane 60 generations D/A -> A/D conversion.
Zmix Converter Tests
If you listened and think about it you may have another view about what is really important...
Zmix Converter Tests
If you listened and think about it you may have another view about what is really important...
Well ...
Thanks to all for the reply ...
wintermute
Audio DiffMaker is the tool I am searching for ... I will experiment with this ...
PlasticIsGood
I do not know the exact English meanigs of "deterministic", I have directly translate the word that in Italian I use in math/phys way ... "a finite system with well known initial state and subsequent state transformation ..."
In my case the same wav file, the same hardware, the same (more or less) condition (temperature, supply voltage, rf noise ... etc ...); if the system is deterministic the result file must be identical bit to bit ...
Is obvious that the the condition do not be costant so there will be different result but I want to investigate HOW different can be ... non from a infomatical prospective (two file with 1 bit different are completelly different) but from an psychoacoustics prospective ...
I need to compare a lot of source chain with the "absolute ear" of a friend but I cannot test every change made with live a/b/x test with him (for logistics evident problems ...)
So I am trying to reconvert ALL the test in digital from analog an submit all the digital result to him in a simple a/b/x test ... I am curios to compare what he can "ear" with instrumental relevated difference ...
I will pass the Audio DiffMaker to an FFT analysis to compare the human hearing response with the instrumental perspective of the same difference ... if all is not the noise level of course
The unique other way is remain in analog but I have to transport to my workshop one of a very BIG high quality tape recorder that we own ... but is a complete different story
Thanks to all for the reply ...
wintermute
Audio DiffMaker is the tool I am searching for ... I will experiment with this ...
PlasticIsGood
I do not know the exact English meanigs of "deterministic", I have directly translate the word that in Italian I use in math/phys way ... "a finite system with well known initial state and subsequent state transformation ..."
In my case the same wav file, the same hardware, the same (more or less) condition (temperature, supply voltage, rf noise ... etc ...); if the system is deterministic the result file must be identical bit to bit ...
Is obvious that the the condition do not be costant so there will be different result but I want to investigate HOW different can be ... non from a infomatical prospective (two file with 1 bit different are completelly different) but from an psychoacoustics prospective ...
I need to compare a lot of source chain with the "absolute ear" of a friend but I cannot test every change made with live a/b/x test with him (for logistics evident problems ...)
So I am trying to reconvert ALL the test in digital from analog an submit all the digital result to him in a simple a/b/x test ... I am curios to compare what he can "ear" with instrumental relevated difference ...
I will pass the Audio DiffMaker to an FFT analysis to compare the human hearing response with the instrumental perspective of the same difference ... if all is not the noise level of course
The unique other way is remain in analog but I have to transport to my workshop one of a very BIG high quality tape recorder that we own ... but is a complete different story
What are we meaning by bit perfect? The thing is that a perfectly sampled copy of a waveform but with an arbitrary delay will not be bit perfect when you look at the two files i.e. the numbers in the two files will be completely different - and I don't mean simply not aligned. However, the delay could be removed in software to get you back to a bit perfect copy (or almost).
This is a great video explaining how 'bit perfect' similarity between digital recordings is not likely to be achieved, but nor is it important in determining the quality of your copy.
Xiph.Org Video Presentations: Digital Show & Tell
This is a great video explaining how 'bit perfect' similarity between digital recordings is not likely to be achieved, but nor is it important in determining the quality of your copy.
Xiph.Org Video Presentations: Digital Show & Tell
Audio DiffMaker
I agree with your statement, I have read this simple words from Audio DiffMaker and I think it can solve my problems " ... Audio DiffMaker is a freeware tool set intended to help determine the absolute difference between two audio recordings, while neglecting differences due to level difference, time synchronization, or simple linear frequency responses ..."
I know that difference in TIME domains (jitter ...) can really make the difference but I think that Audio DIffMaker do not remove them since that difference is not a time sync problems ...
I agree with your statement, I have read this simple words from Audio DiffMaker and I think it can solve my problems " ... Audio DiffMaker is a freeware tool set intended to help determine the absolute difference between two audio recordings, while neglecting differences due to level difference, time synchronization, or simple linear frequency responses ..."
I know that difference in TIME domains (jitter ...) can really make the difference but I think that Audio DIffMaker do not remove them since that difference is not a time sync problems ...
If you take a band-limited signal, sample it, convert to digital, convert back to analogue, pass through reconstruction filter then you get back an exact copy of what you started with - provided the filters are perfect, and you ignore noise.
If you do this twice, then the digital data obtained will be the same provided the sampling is synchronised - but taking into account any delays through the filters.
Real life is not so simple as you can't have:
1. perfect filters
2. perfect synchronisation
3. no noise
If you do this twice, then the digital data obtained will be the same provided the sampling is synchronised - but taking into account any delays through the filters.
Real life is not so simple as you can't have:
1. perfect filters
2. perfect synchronisation
3. no noise
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.