The Second Law of Thermodynamics

[pkane2001]

22 minutes ago, esldude said:

One has to accept the logical leap between signal and thermodynamics. Should I mention dither?

 

Leap accepted. Shannon made the connection between entropy and information entropy 70 years ago. As long as signal is information that's represented by a physical system, the second law applies to it also.

 

The best way to eliminate the unnecessary entropy in the sound path is to have a direct, short wire directly into the brain. Of course, it should be soldered for best results, as Frank often recommends  

 

[pkane2001]

47 minutes ago, esldude said:

I was aware that Shannon used entropy in his 1948 paper.  I didn't understand that it was directly the same as thermodynamic entropy. 

 

I mentioned dither because I thought I remembered an analysis showing subtractive dither can be used with all the good effects of additive dither without changing information entropy.  As well as showing that good implementations of additive dither come very, very close to not changing entropy.  

 

Would FLAC not be a case of doing something to a signal without degrading it, and without changing information entropy? There is a thermodynamic change in entropy to perform those operations, but it would not seem to be informational entropy change.  Is that incorrect?

 

I don’t think random dither ever results in information preservation. Additive or subtractive is the same, some amount of information is destroyed in the process. It is designed to not affect the important parts of the signal, and may help reduce some unwanted artifacts, but it still destroys some information, usually about 1/2 a bit.

 

There are many things in this world that temporarily decrease entropy, in an apparent contradiction to the second law. The explanation is usually that this is a temporary and/or local condition that’s balanced by an increase in entropy elsewhere or elsewhen. This might be, for example, the generator that causes an increase in entropy by burning fuel that is converted to electricity that is then used to  power the computer to encode the FLAC file.

 

Life itself is an apparent contradiction of the second law, as it results in a decrease in entropy by causing information to be ‘created’ out of organic material and water and oxygen. This is also balanced by all the ‘exhaust’ that life generates in the process that ultimately increases overall entropy. So, it goes, the second law is upheld after all

[pkane2001]

4 minutes ago, esldude said:

http://www.aes.org/e-lib/browse.cfm?elib=13033

 

I believe this is the paper on subtractive dither I remembered.  Don't see anywhere to get it other than the AES currently. 

 

You would have to transmit the applied subtractive dither to the other end in order to subtract it.  

 

Ah, so add it then subtract it. That will not destroy information in the signal itself, but it’ll certainly use extra computing resources resulting in more entropy in the end.

[pkane2001]

7 minutes ago, esldude said:

So thermodynamic entropy yes, but this wouldn't appear to alter informational entropy.  BTW, I added a link to a paper you can see on the web.  Has an MQA connection. 

 

They are the same, that’s the beauty of Shannon and the following interpretations. Everything in the universe can be treated as information and entropy itself is defined in terms of information.

 

What I think you’re saying is that the digital audio signal information is not subject to increased entropy. That’s true (assuming no losses or errors) but the second law of thermodynamics can be violated locally and/or temporarily, as long as the overall larger system entropy continues to increase.

[pkane2001]

4 minutes ago, esldude said:

Yes, my view on it was digital audio signal information is not subject to increased entropy the way JA's statement,  "....in absolute terms, doing anything to the signal can only degrade it" would imply.  I understand the thermodynamic laws (not as well versed in informational terms though I get the connection).  I was not implying the laws are broken, as obviously total system entropy increases, but in isolated ways that doesn't appear to alter digital signals.  

 

I don’t have the full context, but have to assume JA’s statement was about analog signal rather than digital.