Does vinyl really sound better?

Phil_aka_Pip

Fretwired said:

Phil_aka_Pip said:

ICBM said:

Fretwired said:

Another interesting article .. Rega recently sold 5,000 turntables in a single month, the most since the company was founded 41 years ago.

Is that because they're one of the only few makers left?

I can't imagine overall turntable sales are at anything more than a tiny fraction of the level in the 80s.

They're not, but there are still makers of turntable gear such as Pro-Ject (Henley Designs), Michell, SME, Thorens, and probably even Linn still peddling their wares. Good luck to 'em

But there are fewer makers than in the 1970s ...

so?

Fretwired

Phil_aka_Pip said:

Fretwired said:

Phil_aka_Pip said:

ICBM said:

Fretwired said:

Another interesting article .. Rega recently sold 5,000 turntables in a single month, the most since the company was founded 41 years ago.

Is that because they're one of the only few makers left?

I can't imagine overall turntable sales are at anything more than a tiny fraction of the level in the 80s.

They're not, but there are still makers of turntable gear such as Pro-Ject (Henley Designs), Michell, SME, Thorens, and probably even Linn still peddling their wares. Good luck to 'em

But there are fewer makers than in the 1970s ...

so?

There's not as much competition in an expanding market .. Rega are one of the few volume makers of quality turntables hence the sales figures.

monquixote

Danny1969 said:

I didn't explain myself very well. What I mean is with a 16 bit CD sampled at 44.1 we have about 65 thousand levels of amplitude which are then effectively joined up by the filters. The groove in the record can contain more data then this and ironically the failings of the mechanical lathe will actually contribute to the amount of data as a mechanical component can't do exactly the same cut each time. So under a very powerful means of observation or laser measurement you could find more than 65 thousand different heights in the groove, thus the record is capable of holding more data or information as nonsense and unreadable as it will be it's there. Now the CD contains more than 65 thousand levels of amplitude but the deficit is constructed by the filters in a predictable way. There will be more randomization in the record groove if we studied that as data. 

None of which would help in the reproduction of music and none of it will help in the area of frequency response and dynamics range ...... in both areas the CD will out perform a record. 

Another interesting point in terms of what sounds better rather than what is technically better is digital mixing (called in the box when the whole thing is mixed in the digital domain of your DAW) or out the box where we feed out of the DAW's multiple outputs into the analog channels of a mixer. Now the in the box method is obviously immune from any noise or distortion but the out of the box mix with added distortion and inevitable noise if often preferred despite being more expensive. Because again what sounds better isn't whats technically better on paper 

I think what you are saying is it contains more data, but less information. 

Rocker

This will not clarify things one little bit but it is an interesting read nonetheless.  Enjoy:

http://www.tnt-audio.com/vintage/procrustes_e.html

Cacofonix

monquixote said:

Cacofonix said:

monquixote said:

Cacofonix said:

I see.  So there is no such thing as a stereogram, then?  That's the visual equivalent.

A stereogram has information encoded in it to be found.

White noise by definition does not. 

Unless it's a sonic signal.  Your earlier post said nothing can possibly be encoded in it due to [physics].  I sense a shift.

Er no 

What do you mean by a sonic signal?

Vinyl is a physical displacement. 

Just to be clear noise is what is not signal (hence signal to noise ratio) anything you show me that has data in it is not noise by definition.

I am referring to the so-called noise floor.

Specific sonic information does exist below that noise floor.  That sonic information can be isolated.  The Earth's EM signal and that of the atmosphere, can be isolated in signals below that floor.  This is fact.  A competent audio physicist can determine the date and time of any audio recording by the em signal.

Your assertion was the laws of thermodynamics prove the impossibility of my assertion. 

This is incorrect.

I am also somewhat wary of the assertion that because you can't hear it, it has no effect.  Subsonics do have an effect, and have been proved to do so, so there is no reason why ultrasonics should not also have an effect.   If this is pleasant (I don't know whether it is or not), then a random shutting off of that effect will result in a less pleasant experience.

monquixote

Cacofonix said:

monquixote said:

Cacofonix said:

monquixote said:

Cacofonix said:

I see.  So there is no such thing as a stereogram, then?  That's the visual equivalent.

A stereogram has information encoded in it to be found.

White noise by definition does not. 

Unless it's a sonic signal.  Your earlier post said nothing can possibly be encoded in it due to [physics].  I sense a shift.

Er no 

What do you mean by a sonic signal?

Vinyl is a physical displacement. 

Just to be clear noise is what is not signal (hence signal to noise ratio) anything you show me that has data in it is not noise by definition.

I am referring to the so-called noise floor.

Specific sonic information does exist below that noise floor.  That sonic information can be isolated.  The Earth's EM signal and that of the atmosphere, can be isolated in signals below that floor.  This is fact.  A competent audio physicist can determine the date and time of any audio recording by the em signal.

Your assertion was the laws of thermodynamics prove the impossibility of my assertion. 

This is incorrect.

I am also somewhat wary of the assertion that because you can't hear it, it has no effect.  Subsonics do have an effect, and have been proved to do so, so there is no reason why ultrasonics should not also have an effect.   If this is pleasant (I don't know whether it is or not), then a random shutting off of that effect will result in a less pleasant experience.

I tell you what. Write up your theory on how you can get past Shannon's channel capacity limit and after you've collected your nobel prize you can come on here and explain it to me. 

Here is a nice summary. 

http://en.wikipedia.org/wiki/Channel_capacity

Let me know which bit of it is wrong.

Also first I've heard about the date and time thing. Do you have a citation?

imalone

Danny1969 said:

imalone said:

Danny1969 said:

Drew

It's nothing to do with floating point, if you sample something 44.4k times a second that's all the information you have whever there was more information there or not. From that point you can't improve on it, same as blowing up a picture that only contained 1 mega pixels won't get you higher res, it will just pixel-ate 

Rounding up to the nearest number is irrelevant 

Imagine you looked at a record groove with a massive massive telescope, it's possible you could detect more than 44.4K changes of amplitude in a one second run .... therefore the record holds more resolution that the CD as the CD has only got it's 44.4K levels the rest is provided by the filters

You're confusing continuous with infinite bandwidth, it's not the same. With sampling it's easy to see what the limit is, with analogue systems it's not so obvious but it is there. 

I didn't explain myself very well. What I mean is with a 16 bit CD sampled at 44.1 we have about 65 thousand levels of amplitude which are then effectively joined up by the filters. The groove in the record can contain more data then this and ironically the failings of the mechanical lathe will actually contribute to the amount of data as a mechanical component can't do exactly the same cut each time. So under a very powerful means of observation or laser measurement you could find more than 65 thousand different heights in the groove, thus the record is capable of holding more data or information as nonsense and unreadable as it will be it's there. Now the CD contains more than 65 thousand levels of amplitude but the deficit is constructed by the filters in a predictable way. There will be more randomization in the record groove if we studied that as data.

Ah, got you. Thought you were arguing there was more data as a consequence. However something like a matched filter comes into play, the write and read are operating within similar limits, in terms of frequency at least. In terms of amplitude I'm not sure what the effective resolution of vinyl is. It can't be that bad as records do work.

monquixote said:

Cacofonix said:

monquixote said:

Cacofonix said:

monquixote said:

Cacofonix said:

I see.  So there is no such thing as a stereogram, then?  That's the visual equivalent.

A stereogram has information encoded in it to be found.

White noise by definition does not. 

Unless it's a sonic signal.  Your earlier post said nothing can possibly be encoded in it due to [physics].  I sense a shift.

Er no 

What do you mean by a sonic signal?

Vinyl is a physical displacement. 

Just to be clear noise is what is not signal (hence signal to noise ratio) anything you show me that has data in it is not noise by definition.

I am referring to the so-called noise floor.

Specific sonic information does exist below that noise floor.  That sonic information can be isolated.  The Earth's EM signal and that of the atmosphere, can be isolated in signals below that floor.  This is fact.  A competent audio physicist can determine the date and time of any audio recording by the em signal.

Your assertion was the laws of thermodynamics prove the impossibility of my assertion. 

This is incorrect.

I am also somewhat wary of the assertion that because you can't hear it, it has no effect.  Subsonics do have an effect, and have been proved to do so, so there is no reason why ultrasonics should not also have an effect.   If this is pleasant (I don't know whether it is or not), then a random shutting off of that effect will result in a less pleasant experience.

I tell you what. Write up your theory on how you can get past Shannon's channel capacity limit and after you've collected your nobel prize you can come on here and explain it to me. 

Here is a nice summary. 

http://en.wikipedia.org/wiki/Channel_capacity

Let me know which bit of it is wrong.

Also first I've heard about the date and time thing. Do you have a citation?

I think one of you is talking about noise and the other one about noise. At least in one context...
There has supposedly (and don't have a citation, probably overhead it on a radio programme ages ago) been a little work done on trying to identify and timestamp broadcast audio picked up in the background of recordings (like telephone calls) for forensic purposes. In theory if you can characterise that in a meaningful way and have a good enough library you can look it up. This is not noise in the signal processing sense, just in the 'turn that radio off' sense. The term noise floor doesn't really apply, it's unwanted signal.

Cacofonix

monquixote said:

Cacofonix said:

monquixote said:

Cacofonix said:

monquixote said:

Cacofonix said:

I see.  So there is no such thing as a stereogram, then?  That's the visual equivalent.

A stereogram has information encoded in it to be found.

White noise by definition does not. 

Unless it's a sonic signal.  Your earlier post said nothing can possibly be encoded in it due to [physics].  I sense a shift.

Er no 

What do you mean by a sonic signal?

Vinyl is a physical displacement. 

Just to be clear noise is what is not signal (hence signal to noise ratio) anything you show me that has data in it is not noise by definition.

I am referring to the so-called noise floor.

Specific sonic information does exist below that noise floor.  That sonic information can be isolated.  The Earth's EM signal and that of the atmosphere, can be isolated in signals below that floor.  This is fact.  A competent audio physicist can determine the date and time of any audio recording by the em signal.

Your assertion was the laws of thermodynamics prove the impossibility of my assertion. 

This is incorrect.

I am also somewhat wary of the assertion that because you can't hear it, it has no effect.  Subsonics do have an effect, and have been proved to do so, so there is no reason why ultrasonics should not also have an effect.   If this is pleasant (I don't know whether it is or not), then a random shutting off of that effect will result in a less pleasant experience.

I tell you what. Write up your theory on how you can get past Shannon's channel capacity limit and after you've collected your nobel prize you can come on here and explain it to me. 

Here is a nice summary. 

http://en.wikipedia.org/wiki/Channel_capacity

Let me know which bit of it is wrong.

Also first I've heard about the date and time thing. Do you have a citation?

I see.  So in terms of information theory and the noisy channel coding theorem, both of which you seem to be positing as undeniable truth, you want me to come up with an opposing theory.  Why do I need to do that?

By definition, both of the above are theoretical, therefore unproven.  So your assertion of their indisputable validity is false.

As to citation, no, it was in Sound on Sound a couple of years ago.  However, your implication seems to be that if I can't quote it, I may be lying about it.  That's true, I might be. My unproven assertion against yours.  Yours relies on factualising a theory or two, mine relies on pure assertion as to audionics and the existence of a job advert long since expired.

The channel capacity argument posits a theoretical limit on signal transmissibility based on measuring loss in signal as more data is passed.

That theory relies on testing apparatus (which generates its own internal noise) rather than actual signal transmission and so is flawed, as it presupposes its own conclusion, and also relies on the concept of, say, electron saturation in a wire.  However, it does not take into account the possibility of coding within electrons, or the idea that an electron can itself carry more than one state or travel  at varying speed, or somehow be enticed along a specific path within the conductor.  These are also theoretical possiblities discounted ab initio by noisy channel theory.

In practice, it has been proven that data can be transmitted at faster and faster rates through a particular channel by utilising different transmission and receiving apparatus, and coding information.  It is by no means certain that saturation is in fact a practical possibility.

monquixote

I think one of you is talking about noise and the other one about noise. At least in one context...
There has supposedly (and don't have a citation, probably overhead it on a radio programme ages ago) been a little work done on trying to identify and timestamp broadcast audio picked up in the background of recordings (like telephone calls) for forensic purposes. In theory if you can characterise that in a meaningful way and have a good enough library you can look it up. This is not noise in the signal processing sense, just in the 'turn that radio off' sense. The term noise floor doesn't really apply, it's unwanted signal.

Yeah the thing about signal to noise ratio is if it's the thing you are looking for then it's the signal not the noise. 

I'm talking about the thermal noise floor under which everything gets lost.

Cacofonix

monquixote said:

I think one of you is talking about noise and the other one about noise. At least in one context...
There has supposedly (and don't have a citation, probably overhead it on a radio programme ages ago) been a little work done on trying to identify and timestamp broadcast audio picked up in the background of recordings (like telephone calls) for forensic purposes. In theory if you can characterise that in a meaningful way and have a good enough library you can look it up. This is not noise in the signal processing sense, just in the 'turn that radio off' sense. The term noise floor doesn't really apply, it's unwanted signal.

Yeah the thing about signal to noise ratio is if it's the thing you are looking for then it's the signal not the noise. 

I'm talking about the thermal noise floor under which everything gets lost.

Oh, the Shannon-Hartley theorem.  I get it.

Edit: Even under that theory there is no assertion of total loss.  The theory is only total loss at infinity, which is a hypothetical extreme only, which is illustrated by the graph.  There is no noise floor under that theorem, merely an increasing probability of signal loss.

monquixote

Cacofonix said:

As to citation, no, it was in Sound on Sound a couple of years ago.  However, your implication seems to be that if I can't quote it, I may be lying about it.  That's true, I might be. My unproven assertion against yours.  Yours relies on factualising a theory or two, mine relies on pure assertion as to audionics and the existence of a job advert long since expired.

Nope, genuinely interested to learn something if you have a link.

The other stuff in your post didn't make a lot of sense so I'll leave it there. 

Cacofonix

monquixote said:

Cacofonix said:

As to citation, no, it was in Sound on Sound a couple of years ago.  However, your implication seems to be that if I can't quote it, I may be lying about it.  That's true, I might be. My unproven assertion against yours.  Yours relies on factualising a theory or two, mine relies on pure assertion as to audionics and the existence of a job advert long since expired.

Nope, genuinely interested to learn something if you have a link.

The other stuff in your post didn't make a lot of sense so I'll leave it there. 

Why, because it exposed your so-called proofs as theoretical only? The graph says everything.  There is no true noise floor, only a hypothetical data loss as data saturation reaches a hypothetical infinity, which in the case of a wire is the point at which it acts as a fuse, and blows, and most audio equipment won't get to that stage.

Article here: http://www.soundonsound.com/sos/jan10/articles/forensics.htm
Job was in same issue IIRC.

imalone

Cacofonix said:

That theory relies on testing apparatus (which generates its own internal noise) rather than actual signal transmission and so is flawed, as it presupposes its own conclusion, and also relies on the concept of, say, electron saturation in a wire.  However, it does not take into account the possibility of coding within electrons, or the idea that an electron can itself carry more than one state or travel  at varying speed, or somehow be enticed along a specific path within the conductor.  These are also theoretical possiblities discounted ab initio by noisy channel theory.

In practice, it has been proven that data can be transmitted at faster and faster rates through a particular channel by utilising different transmission and receiving apparatus, and coding information.  It is by no means certain that saturation is in fact a practical possibility.

Basically, this is all irrelevant, as information theory only relies on statistical arguments which are pretty much independent of the actual mechanisms. That's also why statistically based theories are incredibly strong.

monquixote

Cacofonix said:

monquixote said:

Cacofonix said:

As to citation, no, it was in Sound on Sound a couple of years ago.  However, your implication seems to be that if I can't quote it, I may be lying about it.  That's true, I might be. My unproven assertion against yours.  Yours relies on factualising a theory or two, mine relies on pure assertion as to audionics and the existence of a job advert long since expired.

Nope, genuinely interested to learn something if you have a link.

The other stuff in your post didn't make a lot of sense so I'll leave it there. 

Why, because it exposed your so-called proofs as theoretical only? The graph says everything.  There is no true noise floor, only a hypothetical data loss as data saturation reaches a hypothetical infinity, which in the case of a wire is the point at which it acts as a fuse, and blows, and most audio equipment won't get to that stage.

Article here: http://www.soundonsound.com/sos/jan10/articles/forensics.htm
Job was in same issue IIRC.

I'm at the stage where I'm not sure if you are trolling. 

You do know that gravity and evolution along with everything else in science are just theories right?

You may not believe in the second law of thermodynamics, but I don't think it's up to me to explain it to you.

Cacofonix

imalone said:

Cacofonix said:

That theory relies on testing apparatus (which generates its own internal noise) rather than actual signal transmission and so is flawed, as it presupposes its own conclusion, and also relies on the concept of, say, electron saturation in a wire.  However, it does not take into account the possibility of coding within electrons, or the idea that an electron can itself carry more than one state or travel  at varying speed, or somehow be enticed along a specific path within the conductor.  These are also theoretical possiblities discounted ab initio by noisy channel theory.

In practice, it has been proven that data can be transmitted at faster and faster rates through a particular channel by utilising different transmission and receiving apparatus, and coding information.  It is by no means certain that saturation is in fact a practical possibility.

Basically, this is all irrelevant, as information theory only relies on statistical arguments which are pretty much independent of the actual mechanisms. That's also why statistically based theories are incredibly strong.

We were talking about an assertion of total data loss on a vinyl recording  above a certain frequency due to thermodynamics.  The assertion was that no data would exist above human audible levels.  I was then pointed to proof, which turns out to be only a theory based on testing with apparatus which introduces its own noise into the equation.  Noise can be either in real terms, negating the concept by bringing into corporeality, or in hypothetical terms, by the equally valid theory that observation of a phenomenon changes the result, unless it's maths, in which case, conclusions must inexorably follow because "all other things are equal" - but in the context of a world which does not comply with that theory.

If the statistical argument is irrelevant, why was it pointed out:
(a) as irrefutable proof with a demand that I produce my own theoretical refutation; or
(b) at all?

It is also a bit disingenuous to say that your pet theory proves your point, but, if it should turn out that it doesn't because:

(a) I can understand it; and
(b) on a detailed analysis it is found not to fit the current circumstances (and even run counter to your own argument); and/or
(c) there may in fact be no "saturation point" for transmission of data, because the theory has introduced its own bottleneck by presupposing a limit on the reception apparatus;

then the only reason it doesn't is because it's hypothetical only, and theories are better than fact because [maths].

A speaker does not act in the same way as the testing apparatus, or the human auditory system, which itself is not fully understood.  So the test, and the supporting theory, are flawed.  Hence the theory does not support the conclusion.  There may be audio data in a vinyl recording which sits above the limit of human hearing. 

This is therefore one possible difference between audio and CD, which has an imposed real limit at a cut-off point.  There is a possibility, even under the theories put forward using hypothetical noise floors, that audio data above (IIRC) 22kHz can exist in a vinyl recording.  Even approaching infinity in terms of data saturation , there is a probability of data integrity (and therefore note and timbre information) remaining within the signal.

Cacofonix

monquixote said:

Cacofonix said:

monquixote said:

Cacofonix said:

As to citation, no, it was in Sound on Sound a couple of years ago.  However, your implication seems to be that if I can't quote it, I may be lying about it.  That's true, I might be. My unproven assertion against yours.  Yours relies on factualising a theory or two, mine relies on pure assertion as to audionics and the existence of a job advert long since expired.

Nope, genuinely interested to learn something if you have a link.

The other stuff in your post didn't make a lot of sense so I'll leave it there. 

Why, because it exposed your so-called proofs as theoretical only? The graph says everything.  There is no true noise floor, only a hypothetical data loss as data saturation reaches a hypothetical infinity, which in the case of a wire is the point at which it acts as a fuse, and blows, and most audio equipment won't get to that stage.

Article here: http://www.soundonsound.com/sos/jan10/articles/forensics.htm
Job was in same issue IIRC.

I'm at the stage where I'm not sure if you are trolling. 

You do know that gravity and evolution along with everything else in science are just theories right?

You may not believe in the second law of thermodynamics, but I don't think it's up to me to explain it to you.

I am aware that gravity and the scientific method are about hypotheses, yes, but I am also aware that antithesis often results in lots of heat and not much light, as seems to have been the case here.  You have not proved your assertion.

Additionally, insulting references rather than analysis of a countervailing view do not advance knowledge, it seems to me.

I am confused that you are asserting as fact matters which exist only on a probability spectrum, and using arguments and theories which actually contradict your assertions, specifically in a real world situation such as frequencies within a vinyl recording.  You are entitled to your view on which is better, but to assert a position is fine, it seems to me, if it is based on fact rather than assertions and pseudo-proofs.

You stated as a truth that there is a noise floor, based on thermodynamics, and pointed me to a theory which creates its own conclusion by imposing a hypothetical bottleneck on receipt of data when in other respects it hypothesises a perfect system.

To what extent is a theory valid when its presuppositions lead inexorably to its conclusion?  There must be a saturation limit because our testing apparatus is necessarily limited in scope.  This puts the cart before the horse. 

That theory can be rephrased as "there is no such thing as light because my eyes are closed".

That said, I was testing your assertion that there could be no audio data on a vinyl recording above the imposed limit on a CD recording.  The theories that you have pointed out tend to oppose, rather than support, that assertion.

Incidentally I had an interesting chat to a former professor of physics on the entropy principle.  If the whole is indeed greater than the sum of its parts, and those parts are expanding outwards, entropy is increasing.  However, if there are an infinite number of black holes, then there is a situation where entropy is decreasing in localised areas, and there is therefore a plausible argument that entropy will at some point be reversed as the black holes agglomerate sufficient matter (with the additional pull of their whole being greater etc) to exert a countervailing pull.  Unsure how the data on waves being able to escape black holes fits in, but I'm working on it. 

Do you think that merits a Nobel prize?  What's your theory?

imalone

monquixote said:

Cacofonix said:

monquixote said:

Cacofonix said:

As to citation, no, it was in Sound on Sound a couple of years ago.  However, your implication seems to be that if I can't quote it, I may be lying about it.  That's true, I might be. My unproven assertion against yours.  Yours relies on factualising a theory or two, mine relies on pure assertion as to audionics and the existence of a job advert long since expired.

Nope, genuinely interested to learn something if you have a link.

The other stuff in your post didn't make a lot of sense so I'll leave it there. 

Why, because it exposed your so-called proofs as theoretical only? The graph says everything.  There is no true noise floor, only a hypothetical data loss as data saturation reaches a hypothetical infinity, which in the case of a wire is the point at which it acts as a fuse, and blows, and most audio equipment won't get to that stage.

Article here: http://www.soundonsound.com/sos/jan10/articles/forensics.htm
Job was in same issue IIRC.

I'm at the stage where I'm not sure if you are trolling. 

You do know that gravity and evolution along with everything else in science are just theories right?

You may not believe in the second law of thermodynamics, but I don't think it's up to me to explain it to you.

Actually, that sound on sound article is worth a read, it's got stuff on the forensic analysis of the background noise due to the electricity grid.
It's still not about extracting a signal beneath the noise floor though, because that EM is your signal and you're trying to extract it by either modelling its contribution to the overall noise power spectrum, which requires you to have models of the other components and will only give you parameters, not a 'signal' in the normal sense, or just directly, which requires it to dominate all the other noise.
On the other hand, the bit about Richard Nixon, where some of the world's best experts were trying to get a recording back after it had been swamped with noise? Nothing.

Cacofonix said:

imalone said:

Cacofonix said:

That theory relies on testing apparatus (which generates its own internal noise) rather than actual signal transmission and so is flawed, as it presupposes its own conclusion, and also relies on the concept of, say, electron saturation in a wire.  However, it does not take into account the possibility of coding within electrons, or the idea that an electron can itself carry more than one state or travel  at varying speed, or somehow be enticed along a specific path within the conductor.  These are also theoretical possiblities discounted ab initio by noisy channel theory.

In practice, it has been proven that data can be transmitted at faster and faster rates through a particular channel by utilising different transmission and receiving apparatus, and coding information.  It is by no means certain that saturation is in fact a practical possibility.

Basically, this is all irrelevant, as information theory only relies on statistical arguments which are pretty much independent of the actual mechanisms. That's also why statistically based theories are incredibly strong.

We were talking about an assertion of total data loss on a vinyl recording  above a certain frequency due to thermodynamics.  The assertion was no data would exist above human audible levels.  I was then pointed to proof, which turns out to be only a theory based on testing with apparatus which introduces its own noise into the equation.

A speaker does not act in the same way as the testing apparatus, or the human auditory system, which itself is not fully understood.  So the test, and the supporting theory, are flawed.  Hence the theory does not support the conclusion.  There may be audio data in a vinyl recording which sits above the limit of human hearing. 

This is therefore one possible difference between audio and CD, which has an imposed real limit at a cut-off point.  There is a possibility, even under the theories put forward using hypothetical noise floors, that audio data above (IIRC) 22kHz can exist in a vinyl recording.  Even approaching infinity in terms of data saturation , there is a probability of data integrity (and therefore note and timbre information) remaining within the signal.

You're really not getting it.
In any stream, whether it's height of the vinyl track, a radio wave or the tension on a string between two tin cans you have a parameter to measure. It's pointless to try to confuse the issue by saying there could be extra information in a different mechanism, either the way you're reading it ignores it, or it interferes with your measurement, in which case great, now you've got an unwanted and unpredictable contribution. If you can't disentangle it that's exactly what noise is, if you can disentangle it (say someone has decided to cut a bizarre vinyl track where the stereo is encoded at a slope, so it leaks into your vertical) that disentangling is dependent on how well you read both components.
There is quite possibly signal a vinyl above human hearing, see the CD-4 stuff I mentioned a while ago, the technology can be pushed that far, it doesn't mean there is.
Genuine noise, thermal, electrical, mechanical, is unpredictable. This is the point, you can't extract it because you can't say what its value was at a given point and so you can't remove it and so you can't recover the original signal. What you can do is average it out, but this requires a model for the signal, say you're expecting a pure tone, you can take a long enough measurement and the noise can be averaged out (normally you'd fourier transform to see it), while the steady predictable signal remains. It has nothing to do with the speaker you're using.
(body is 94 characters too long...)

imalone

(...)
What you can hear. The ear is a pretty bizarre device, we do know a fair bit about it http://en.wikipedia.org/wiki/Cochlea frequency is spatially encoded along the cochlea: how far a particular vibration makes it along the length determines which cells get excited and in turn what frequency you hear (a bit like seeing green or red because those cells are excited). I'd put money on being able to hear a constant tone below the noise decibel level because effectively there's a degree of time averaging and because noise power is spread out over a spectrum, while the spread of a pure tone depends on your sampling duration.
Not going to spend all evening reading background on this, but this is one of the first papers I turned up http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3705086/ effectively these monkeys hear tones at 20dB below the presented noise SPL, increasing noise means it makes them take longer to respond, while the threshold they respond at also goes up accordingly.

monquixote

Cacofonix said:

You stated as a truth that there is a noise floor, based on thermodynamics, and pointed me to a theory which creates its own conclusion by imposing a hypothetical bottleneck on receipt of data when in other respects it hypothesises a perfect system.

To what extent is a theory valid when its presuppositions lead inexorably to its conclusion?  There must be a saturation limit because our testing apparatus is necessarily limited in scope.  This puts the cart before the horse. 

That theory can be rephrased as "there is no such thing as light because my eyes are closed".

That said, I was testing your assertion that there could be no audio data on a vinyl recording above the imposed limit on a CD recording.  The theories that you have pointed out tend to oppose, rather than support, that assertion.

Shannon talks about how a noisy channel behaves and asserts that without noise infinite bandwidth is available (provided you have infinite power). The second law of thermodynamics explains the inevitability and irreversibility of noise. 

Cacofonix

Actually, that sound on sound article is worth a read, it's got stuff on the forensic analysis of the background noise due to the electricity grid.
It's still not about extracting a signal beneath the noise floor though, because that EM is your signal and you're trying to extract it by either modelling its contribution to the overall noise power spectrum, which requires you to have models of the other components and will only give you parameters, not a 'signal' in the normal sense, or just directly, which requires it to dominate all the other noise.
On the other hand, the bit about Richard Nixon, where some of the world's best experts were trying to get a recording back after it had been swamped with noise? Nothing.

That's the point, it is below the noise floor.  It is within the signal even if saturated.  You have a benchmark and it therefore negates the concept of noise.

As to the Nixon recording, I think you are adding in your own interpretation. In the immediate analysis we are talking about an audio stream encoded to vinyl and decoded out. The Nixon recording was overdubbed several times.  This is a different thing.  One is ambient noise, within which a signal can exist based on the existing data, one is a situation where data has been overwritten several times.  That is not saturation of a medium, but re-recording on it.  There is still room for data to exist, but it has been altered in random patterns.

You're really not getting it.

Sorry bud, bit fick.

In any stream, whether it's height of the vinyl track, a radio wave or the tension on a string between two tin cans you have a parameter to measure. It's pointless to try to confuse the issue by saying there could be extra information in a different mechanism, either the way you're reading it ignores it, or it interferes with your measurement, in which case great, now you've got an unwanted and unpredictable contribution. If you can't disentangle it that's exactly what noise is, if you can disentangle it (say someone has decided to cut a bizarre vinyl track where the stereo is encoded at a slope, so it leaks into your vertical) that disentangling is dependent on how well you read both components.


I am not disputing that.  What I am saying is that to presuppose that no data can exist in that environment is plainly false.  I am saying that there may be data in the existing stream that is not being measured, not that it could be enhance using a different setup. On the equipment before you - the cutting lathe, the stylus, speakers etc, there may be noise introduced.  However, that does not presuppose that no audio data can exist in that area as the noise is not necessarily pervasive.  Even if it were, in the theory put forward, the data integrity may be reduced as noise approaches infinity, but it may still be there. It may be perceivable even if not technically "audible", as in subsonics, and therefore to assert without more that it doesn't exist and/or affect the listener is simply unproven.

Also, I am challenging the view that the medium that you are presupposing is the relevant one, i.e. the audio reproduction system, is false.  It is the listener who is the "receiver" in that analysis, and that person's eardrums are not their only audio-sensing apparatus.  Indeed Beethoven's ears weren't up to the job at all, so to attach massive importance to the type of system, the string between the cans in your analysis, is not the whole story.  The apparatus may be more sensitive than your analysis would suggest.

There is quite possibly signal a vinyl above human hearing, see the CD-4 stuff I mentioned a while ago, the technology can be pushed that far, it doesn't mean there is.
Genuine noise, thermal, electrical, mechanical, is unpredictable. This is the point, you can't extract it because you can't say what its value was at a given point and so you can't remove it and so you can't recover the original signal. What you can do is average it out, but this requires a model for the signal, say you're expecting a pure tone, you can take a long enough measurement and the noise can be averaged out (normally you'd fourier transform to see it), while the steady predictable signal remains. It has nothing to do with the speaker you're using.
(body is 94 characters too long...)