Can we hear a bit Options

[Roseval]

In a thread on a forum the author claimed he could hear if a bit was read incorrectly by a CD player. He didn’t tell which one of the sixteen but lets assume the worst case: the most significant bit is wrong.
If I understand redbook audio correctly this means signal volume is doubled or halved.
One word= 1/44100 so approximately 23 µs (thanks Menno) of music.
So here we have the digital equivalent of a scratch in vinyl but can we hear a spike or a dip of 23 µs?

This post has been edited by Roseval: Apr 30 2008, 00:24

[menno]

miscalculation: not ms, but μs

[Paulhoff]

A bit, one bit, sounds like "The Princess and the Pea" to me. It's BS.

http://childhoodreading.com/Edmund_Dulac_a...nd_the_Pea.html

Paul

[pdq]

One bit, if it is the msb, is a full-scale click. Even though it is 23 microseconds in duration, the low-pass filtering following the D/A would stretch it out, possibly to hundreds of microseconds. I would think that this would be quite audible.

[Paulhoff]

He said BIT not Byte.

Paul

[Roseval]

1 wrong bit=1 wrong byte I presume

[Paulhoff]

Well first there are two bytes per channel not one, and there is error correction, and if that doesn't work, the bits and/or bytes are averaged between the two bytes before and after.

Paul



This post has been edited by Paulhoff: Apr 30 2008, 00:51

[menno]

1 wrong bit is 1 wrong sample
And yeah you can hear that as a click if it's the msb.

Well first there are two bytes per channel not one, and there is error correction, and if that doesn't work, the bits and/or bytes are averaged between the two bytes before and after.

Hmm, ok I guess it depends here whether we talk about a bit read incorrectly from the cd, or a bit incorrect in the sample returned by the cd player. Maybe the TS should clarify this

[Paulhoff]

Once again if it is misread, it will be known and corrected as above.

Paul

[Paulhoff]

Also the data on the CD in not in sequence to start with, it is interlaced, and there is error correction. There is a lot of processing before it goes to the DA converter.

http://findarticles.com/p/articles/mi_m0HP..._v41/ai_9683351

Paul



This post has been edited by Paulhoff: Apr 30 2008, 01:13

[Roseval]

Hmm, ok I guess it depends here whether we talk about a bit read incorrectly from the cd, or a bit incorrect in the sample returned by the cd player. Maybe the TS should clarify this

I’m afraid I can’t.
Somebody claimed he could hear if the player reads a bit incorrect.

Maybe we have many different cases.
Suppose we have
0000000000000000/1000000000000000/0000000000000000
Can we hear it?
Suppose we have
1111111111111111/ 0111111111111111/1111111111111111
Can we hear it?

Suppose we have
0000000000000000/1000000000000000/0000000000000000
And the 1 detected as an error, do we get
0000000000000000/0000000000000000/0000000000000000
for output?

This post has been edited by Roseval: Apr 30 2008, 01:21

[Paulhoff]

Hmm, ok I guess it depends here whether we talk about a bit read incorrectly from the cd, or a bit incorrect in the sample returned by the cd player. Maybe the TS should clarify this

I’m afraid I can’t.
Somebody claimed he could hear if the player reads a bit incorrect.

Maybe we have many different cases.
Suppose we have
0000000000000000/1000000000000000/0000000000000000
Can we hear it?
Suppose we have
1111111111111111/ 0111111111111111/1111111111111111
Can we hear it?

Suppose we have
0000000000000000/1000000000000000/0000000000000000
And the 1 detected as an error, do we get
0000000000000000/0000000000000000/0000000000000000
As output?

Unless error correction doesn't work, we get

0000000000000000/1000000000000000/0000000000000000

Paul

[greynol]

Paulhoff has it right.

People have a tendency to think pits and lands correspond to ones and zeros; they don't!

http://en.wikipedia.org/wiki/Cross-interle...-Solomon_coding

[Slipstreem]

Aren't uncorrectable errors muted on a standalone CD player according to the Red Book standard? If so, is the muted period long enough to be audible?

Cheers, Slipstreem.

[Paulhoff]

Aren't uncorrectable errors muted on a standalone CD player according to the Red Book standard? If so, is the muted period long enough to be audible?

Cheers, Slipstreem.

That would not be one bit and/or byte, that would be a lot more than that.

Paul

Interleave
Errors found in the CD system are a combination of random and burst errors. In order to alleviate the strain on the error control code, some form of interleaving is required. The CD system employs two concatenated Reed-Solomon codes, which are interleaved cross-wise. Judicious positioning of the stereo channels as well as the audio samples on even or odd-number instants within the interleaving scheme provide the error concealment ability, and the multitude of interleave structures used on the CD makes it possible to correct and detect errors with a relatively low amount of redundancy.

Interpolation
If a major error occurs and a sample cannot be perfectly reconstructed by the error control circuitry, it is possible to "guess" the content of the sample; that is, obtain an approximation by interpolating it off the neighbouring audio samples. While this concealment will not "fix" the error, it will make it inaudible, offering a graceful degradation of audio quality as clicks and pops are avoided.

This post has been edited by Paulhoff: Apr 30 2008, 03:46

[greynol]

Sections of audio are silenced when there is too much missing to interpolate. I think this is beyond the type error that inspired the discussion.

[Slipstreem]

Yes. I've already that particular WIKI article. What I'm suggesting is that the author of the article referred to by the OP may be mistaking a muting error for a single-bit error if he's listening to the source material on a standalone CD player.

Cheers, Slipstreem.

Agreed, Greynol. This kind of error probably falls far short of triggering muting as a last line of defence, but maybe Roseval could provide us with a link to the post he's referring to so that we can all sing from the same hymn sheet.

This post has been edited by Slipstreem: Apr 30 2008, 03:59

[Jebus]

Data isn't stored as 1s and 0s on CDs. It is stored using different lengths of valleys, which are then interpreted in such a way as to provide a bitstream. A single pit misread will not equal a single misread bit... the error correction occurs before it gets turned into binary. does that make any sense?

edit: I'm wrong. I thought the 8-14 coding mentioned below had to do with pit lengths, not binary. My bad.

This post has been edited by Jebus: May 1 2008, 05:22

[Woodinville]

Data isn't stored as 1s and 0s on CDs. It is stored using different lengths of valleys, which are then interpreted in such a way as to provide a bitstream. A single pit misread will not equal a single misread bit... the error correction occurs before it gets turned into binary. does that make any sense?

No.

There are two forms of error correction on a CD, both are digital.

The lowest level is "EFM" or eight-to-fourteen modulation". It takes each byte, and turns it into 14 bits in a fashion wherein it is possible to both correct many misreads as well as tell when a misread is uncorrectable.

Then there is an interleaving, to allow for scratches, holes, pits, etc in the CD data layer.

After that a reed-solomon code is used to error correct and error check again.

After that, if the RSC can't get it right, the player either interpolates (for a block or two) or mutes.

On most CD's there is almost nothing getting past the reed-solomon code. Usually in fact nothing at all.

Sections of audio are silenced when there is too much missing to interpolate. I think this is beyond the type error that inspired the discussion.

A single-bit error in the DAC input is almost impossible to cause.

Usually, either everything is corrected, or a large quantity of data is missing. Such is the way of digital forward error correction schemes.

[cabbagerat]

The lowest level is "EFM" or eight-to-fourteen modulation". It takes each byte, and turns it into 14 bits in a fashion wherein it is possible to both correct many misreads as well as tell when a misread is uncorrectable.

I was under the impression that EFM was used in CDs primarily because it limits the number of consecutive 1s or 0s, which can prevent clock recovery. Also, it's a DC free code, which can be important in the analogue transport of digital information. I suppose it does offer some redundency, and could be thought of as error correction.

I just did some testing in MATLAB and found that it's pretty easy to hear single bit flips down to about the 9th MSB in my (noisy) office. These errors are unlikely to occur - but do occur - inside computers an across interfaces which don't offer error correction. This isn't audiophile "slightly veiled highs and constrained bass" stuff though - it's an audible, clear click or thump.

[greynol]

Certain pit and land combinations/geometries are excluded, but this is not the same thing as redundancy coding.

There is no question that a single bit error can be heard, but as Woodinville is saying single bit errors aren't very likely to happen inside a CD player.

[Roseval]

Roseval could provide us with a link to the post he's referring to so that we can all sing from the same hymn sheet

I know this is a violation of TOS #15 (Thou shall not read Audiophile threads) but here it is:
How much an incorrect bit affects the sound depends on the particular bit and the particular music. In some cases a single bit in error will be obvious, in other cases inaudible
http://www.audioasylum.com/cgi/vt.mpl?f=pcaudio&m=31127

[Slipstreem]

Thanks, Roseval. I promise to keep my eyes shut whilst reading it.

Cheers, Slipstreem.

[cabbagerat]

http://www.audioasylum.com/cgi/vt.mpl?f=pcaudio&m=31127

One of the posters in that thread seems to be claiming that he/she can hear the difference between different .wav file players. His ears must literally have been machined from solid gold

And, linked from that thread, this page with a quote which actually makes me feel ill:

Playing the file from a deeply embedded folder instead of from top of directory: Audible.
Seems to have same effect when file path name is maximum length, regardless of directory structure. Probable reason -- buffering issue, since the music playback software takes considerably longer to load the data prior to playback in these cases.

This post has been edited by cabbagerat: Apr 30 2008, 11:48

[pdq]

Playing the file from a deeply embedded folder instead of from top of directory: Audible.
Seems to have same effect when file path name is maximum length, regardless of directory structure. Probable reason -- buffering issue, since the music playback software takes considerably longer to load the data prior to playback in these cases.