Skip to main content
Topic: Bit-"perfect" copies and generation "loss" (Read 8558 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Bit-"perfect" copies and generation "loss"

An audiophile friend -- who does not see things scientifically -- remains unconvinced that, e.g., a WAV or FLAC file copied from orig source (e.g. HD) to back-up (e.g., another HD) is bit perfect and suffers no generation loss (all else being equal, which it usually is). Not sure I can ever convince folks like this they have nothing to worry about.
However, I'm wondering if some sort of automated experiment hasn't been conducted to prove that audio files, several generations subsequent, are still bit perfect. This experiment would have to be automated as a proper stopping point may be thousandth or millionth-generation.

A possible pseudo-code for experiment design may look like this:
File (orig) > Copy [File (orig)] > Copy[ Copy [File (orig)]] > Copy [Copy[ Copy [File (orig)]]] > …

Bit-"perfect" copies and generation "loss"

Reply #1
If you go to Windows' command line prompt via the 'cmd.exe' application, there is a file comparator called 'comp.exe'

i.e.


C:\Windows\Username\comp.exe FileA.wav FileB.wav

It does a byte-for-byte comparison of two given files and outputs if there are any differences.

If it instead comes back with "Files Compare OK" that means File A and File B are the exact same, byte-for-byte.


This command-line app exists in Windows 9x/2000/XP.  I don't have access to Vista offhand to know if it also exists there.

Bit-"perfect" copies and generation "loss"

Reply #2
If you go to Windows' command line prompt via the 'cmd.exe' application, there is a file comparator called 'comp.exe'
C:\Windows\Username\comp.exe FileA.wav FileB.wav
It does a byte-for-byte comparison of two given files and outputs if there are any differences.
Right, but my query was in the context of...

File (orig) > Copy [File (orig)] > Copy[ Copy [File (orig)]] > Copy [Copy[ Copy [File (orig)]]] > …

Where each copy is re-copied, bit-checked (and logged or tracked), copied again, and again ... to, perhaps, millions of subsequent offspring.

Bit-"perfect" copies and generation "loss"

Reply #3
Your friend needs to understand that generation loss doesn't apply like that with digital. Sure, it's a fundamental of information theory, but where it applies in digital is that from one generation to the next, the nice sharp edges that distinguish 0s from 1s get smoothed a bit. But there's still enough contrast to enable the 0s and 1s to be identified, regenerated, and if necessary retransmitted. Hence the wonders of lossless transmission and lossless copying, generation after generation, world without end.

You don't get progressive subtle degradation from copy to copy, or with increasing length of transmission; you do have to determine a maximum possible quantity of information at the beginning, so no amount of tweaking is going to increase quality, either.

Errors with digital are obvious and glitchy, not subtle degradation.

But he probably still won't believe you.

Bit-"perfect" copies and generation "loss"

Reply #4
Well, send FLAC to someone in china (or wherever far away from where you live) - it will go through a vast number of routers, switches, endless kilometers of cables, optical, coaxial, LAN, you name it, it will go through a number of media converters from optical to LAN, LAN to coax, coax to optical, coax to LAN, and guess what, file will be received as it was sent. Not only FLAC, but also jpeg, txt, doc, tiff, you name it. It is why internet is working and banks do their transactions - digital is error free as long the equipment is working as it should.
It is not analog transmission as in record to tape that the signal will get degraded. Try sending pure analog signal, for example from tape through so many wires.
If the equipment is not working, you will get either no transmission, or serious degradation of integrity of files (HDD bad sectors, for example).

Bit-"perfect" copies and generation "loss"

Reply #5
An audiophile friend -- who does not see things scientifically -- remains unconvinced that, e.g., a WAV or FLAC file copied from orig source (e.g. HD) to back-up (e.g., another HD) is bit perfect and suffers no generation loss (all else being equal, which it usually is). Not sure I can ever convince folks like this they have nothing to worry about.
However, I'm wondering if some sort of automated experiment hasn't been conducted to prove that audio files, several generations subsequent, are still bit perfect. This experiment would have to be automated as a proper stopping point may be thousandth or millionth-generation.

A possible pseudo-code for experiment design may look like this:
File (orig) > Copy [File (orig)] > Copy[ Copy [File (orig)]] > Copy [Copy[ Copy [File (orig)]]] > …



Perhaps you could mention that if digital data suffered from generation loss, then the internet would not work. Data gets copied over and over and over again, to get from one computer to another. Does he believe that his Word documents will slowly degrade from each generation of copying?

Bit-"perfect" copies and generation "loss"

Reply #6
Does he believe that his Word documents will slowly degrade from each generation of copying?


LOL!

Yes, I've heard that adverbs are the first to go.  Followed by adjectives.

The document eventually degrades into "see spot run".

Bit-"perfect" copies and generation "loss"

Reply #7
Digital files are effectively binary if you break them down enough. A big stream of 1s and 0s. No matter how many times you correctly copy a 1, it's always a 1. It cannot become a 0.9 or a 0.8 (as you get with the degredation of analogue signals) and hence you cannot get gradual degredation of digital files. It's all or nothing. The file either copies correctly, works and is an absolutely perfect copy of the original, or it breaks, becomes corrupted and doesn't work. There is no in-between.

As soon as a 1 becomes a 0 (due to a serious problem when copying the files), you have an error in the file. This isn't a slight degredation in quality, it's simply a full blown failure of the file. This type of error will prevent copied apps from working properly, result in corrupted Word files and all the rest of it when copying various types of files. Hence digital copying mechanisms are designed to withstand such problems. When copying files, CRC checks are automatically carried out to ensure that what has been copied remains the same between the source and destination. Assuming files pass these CRC checks after being copied, none of the bits in that file have become corrupted and hence what you have is an absolutely identical copy of the source file.

Bit-"perfect" copies and generation "loss"

Reply #8
Quote
C:\Windows\Username\comp.exe FileA.wav FileB.wav
This command-line app exists in Windows 9x/2000/XP. I don't have access to Vista off hand to know if it also exists there.


FWIW, this does exist/work with Vista..

Bit-"perfect" copies and generation "loss"

Reply #9
Contrary what most "scientifically-minded" people will tell you, a mass-copying test is NOT a good way to "prove" this - more like the opposite.

This is because the supposed ones and zeroes do not actually exist. There are no ones an zeroes - there are only things which are INTERPRETED as ones and zeroes (this is part of the method "coding"). On a CD for example, not all "valleys" are identical - rather, if something is to be interpreted as one or zero depends on a "threshold value" - and there usually then is a large safety buffer to make sure, that this threshold value works reliable. But this also means: What is interpreted as ones and zeroes, can degrade, just as the paper on a book from which you read letters, can degrade, thus resulting in misinterpreting the coded symbols (i.e. you misreading the letters, because they become difficult to distinguish).

Now, what this means for your masscopying test, is that the longer you run the test, the more probable it is that you WILL find errors! Problem is: The errors will not be "generation loss", but instead simply degradation or defects of the media TO which things are copied. Thus, transport of digital data is NOT guaranteed safe. Not even if a first comparision reports identical data for the source and the target. The reason for this, is that the actual health of the MEDIA gets abstracted away by the coding: What you see with a file-comparision on a computer isn't the health of the carrying media (i.e. hard drive) but simply if the code-interpretation CURRENTLY worked reliable.

To summarize: Generation loss may not exist, but media degradation as well as transport hazards affect digital data just as much as analogue data is affected by it. The difference is simply: With digital coding, you dont see the degration, until its too late - minor difference get abstracted away by the coding, so that things reliably get correctly interpreted as ones and zeroes... even if the media isn't in top shape anymore - but when the media degraded enough, you will instantly get maximum damage in your coded data. Phrased differently: Digital coding either works perfectly, or fails totally - and the path from one to the other is hidden by the coding.
I am arrogant and I can afford it because I deliver.

Bit-"perfect" copies and generation "loss"

Reply #10
As the replies increase it should be painfully obvious that this discussion does not belong in the audio section of this forum and as such I'm moving it to the off-topic forum.
Is 24-bit/192kHz good enough for your lo-fi vinyl, or do you need 32/384?

Bit-"perfect" copies and generation "loss"

Reply #11
If he still won't believe you introduce him to the concept of MD5 checksums to verify the files are original and identical.

Bit-"perfect" copies and generation "loss"

Reply #12
MD5 is ok for rock and metal, but for classical SHA-1 does a better job of verifying the integrity of the sound stage and depth between notes. 

Seriously, if they can't understand how making a binary copy of a binary file is lossless I doubt they'd understand cryptographic hash functions.

Bit-"perfect" copies and generation "loss"

Reply #13
Thx for everyone's comments thus far.
Seriously, if they can't understand how making a binary copy of a binary file is lossless I doubt they'd understand cryptographic hash functions.
You're right. Earlier, hlloyge noted the following that may be the best ("simplest") explanation for the great unwashed:
Quote
Well, send FLAC to someone in china (or wherever far away from where you live) - it will go through a vast number of routers, switches, endless kilometers of cables, optical, coaxial, LAN, you name it, it will go through a number of media converters from optical to LAN, LAN to coax, coax to optical, coax to LAN, and guess what, file will be received as it was sent. Not only FLAC, but also jpeg, txt, doc, tiff, you name it. It is why internet is working and banks do their transactions - digital is error free as long the equipment is working as it should.
I did refer my friend to this thread and I think I've convinced him. He's one of those rich audiophiles  -- one who thinks $$ "solutions" (i.e., the ad infinitum cycle of market-hyped product-/gear-buying) -- and not, scientific solutions --are the key to everything!

Bit-"perfect" copies and generation "loss"

Reply #14
Contrary what most "scientifically-minded" people will tell you, a mass-copying test is NOT a good way to "prove" this - more like the opposite.

This is because the supposed ones and zeroes do not actually exist. There are no ones an zeroes - there are only things which are INTERPRETED as ones and zeroes (this is part of the method "coding"). On a CD for example, not all "valleys" are identical - rather, if something is to be interpreted as one or zero depends on a "threshold value" - and there usually then is a large safety buffer to make sure, that this threshold value works reliable. But this also means: What is interpreted as ones and zeroes, can degrade, just as the paper on a book from which you read letters, can degrade, thus resulting in misinterpreting the coded symbols (i.e. you misreading the letters, because they become difficult to distinguish).

Now, what this means for your masscopying test, is that the longer you run the test, the more probable it is that you WILL find errors! Problem is: The errors will not be "generation loss", but instead simply degradation or defects of the media TO which things are copied. Thus, transport of digital data is NOT guaranteed safe. Not even if a first comparision reports identical data for the source and the target. The reason for this, is that the actual health of the MEDIA gets abstracted away by the coding: What you see with a file-comparision on a computer isn't the health of the carrying media (i.e. hard drive) but simply if the code-interpretation CURRENTLY worked reliable.

To summarize: Generation loss may not exist, but media degradation as well as transport hazards affect digital data just as much as analogue data is affected by it. The difference is simply: With digital coding, you dont see the degration, until its too late - minor difference get abstracted away by the coding, so that things reliably get correctly interpreted as ones and zeroes... even if the media isn't in top shape anymore - but when the media degraded enough, you will instantly get maximum damage in your coded data. Phrased differently: Digital coding either works perfectly, or fails totally - and the path from one to the other is hidden by the coding.


So you're saying that EVENTUALLY, over time, if you right click copy paste, right click copy paste, etc, etc, there will be a total failure. Approximately how many times would one have to do this before you reach total failure? - how many licks does it take...

 
SimplePortal 1.0.0 RC1 © 2008-2019