What is bit perfect?

[Rescue Toaster]

fmak, I'm not sure why this thread went this direction exactly...

Whenever someone talks about bit-perfect in PC audio, they are generally talking less about the error recovery of the CD-ROM ripping process (though that's important) and more about everything AFTER that.

The key is never doing mathematical operations on the data that will cause a loss of dynamic range, or ideally, any change in the absolute values of each sample. No EQs, no windows mixer (easier said than done), no volume controls at all, no nothing. Ideally you should be able to rip a DTS-encoded CD, play it out via SPDIF to a DTS-capable receiver, and have it work.

Successfully accomplish all that and you have a "bit-perfect" digital signal chain.

[janneman]

Originally posted by yulquen



so if I understand you correctly, it is not possible to create a
copy protection scheme for regular CD-players that
screws up the audio data at the SPDIF connector while
retaining flawless playback at the analogue outputs?
(not taking copy bit into account since it is easy to get around)

Jorn,

I'm not sure I understand the question. For one thing, the analog output is produced by running the dig output through a DAC, so the analog can only be good if the digital is good.

I am no expert on the copy protection stuff, but I understand that it is handled separately by the playback logic. In fact, as I see it, the copy protection bit is just 'one' bit in the total data stream and as such will also be corrected if necessary. It then depends on the player logic what it does with it.

So it seems that in theory that player logic can take the correct dig signal, route it through the DAC for 'perfect' analog, but then manipulate the SPDIF output data to prevent a dig copy. Anything can be done in that respect. But I don't know if that is how it is done. Maybe domeone else knows it more in detail?

Jan Didden

[janneman]

Originally posted by gmarsh

This is assuming that your scratches aren't too wide, or are spaced in such a way that the de-CIRC process corrects for it. But put a third deep scratch on your CD so that the CIRC algorthm fails, and anything could happen.

Most CD players will interpolate, producing a "non-bit-perfect" output. Some will snap/crackle/pop. Your CD player might even lose tracking and skip if it's a bad scratch.

I wouldn't mind seeing a professional CD player that will flash a LED whenever a CIRC failure occurs...

Yes, sure. As soon as the error correction 'gives up' because the damage is too large, it will generally sound horrible, sometimes whole parts of tracks are repeated or muted, as I said above, you'll probably throw out that CD.

I don't know about the CIRC failure statistics, but I once used a small add-on developed by Elektor that connected to the player logic and displayed the number of error corrections while playing a CD. Thousands, literally. I had perfectly (at least that's how it sounded) playing CDs, with 4000, 5000 error corrections during one play! So, error correction is not some kind of last-ditch rescue for bad data, it is part of life of every CD. Think about it, a scratchy CD must have many thousands of damaged frames on that CD. Still it plays correct, thanks to the error correction.

Jan Didden

[guido]

Hi,

Wasn't that between 7210 to 7220: the decoder telling the filter to do interpolation? (so NO perfect read of the disc).
I don't think you'll hear it if there is one or a few samples interpolated at one moment. 5000 'events' would be a lot, think the article stated only a few per disk.

Think i still have that article, i'll look it up.

[janneman]

Originally posted by guido
Hi,

Wasn't that between 7210 to 7220: the decoder telling the filter to do interpolation? (so NO perfect read of the disc).
I don't think you'll hear it if there is one or a few samples interpolated at one moment. 5000 'events' would be a lot, think the article stated only a few per disk.

Think i still have that article, i'll look it up.

Sigh. The 4000-5000 was error correction, NOT interpolation or another form of error concealment. Error correction is transparant. The majority of CDs have zero error concealment, the error correction taking care of the, well, errror correction. Error concealment by nature is more than 'a few samples interpolated', because if only a few samples are damaged, chances are that the error correction can fix it, transparently. Error concealment is gross and generally quite audible, like 'yikes!'....

Jan Didden

[scott wurcer]

My understanding of "bit perfect" as it is often used in the recording field has to do with sample rate conversion. Some gear ALWAYS puts a sample rate converter on the recording input path even if you are recording at the same bits/frequency as you are inputing. You then are not guaranteed to get the same bits.

BTW I've also done all those record/rip re-record experiments and with out obvious gross failure due to disk defects I always got nada, a perfect flatline when subtracting the waveforms. At an intuitive level how could software distributed on CD be so robust and music be so weak? Take one random instruction bit and invert it in your OS's kernel and see what happens (yup even on a MAC).

[erko]

Here's a link which describes a cd and errors correction system..

http://www.ee.washington.edu/consele...audio/95x6.htm

There's some more info on the net exactly how the reading of pits is done (thats where the laserbeam thats refelected is not as intense because of the pits in the structure), but I couldn't find that anymore.
But remember that the cd is already quite 'old', it's a bit over 25 years back.

Eric

[janneman]

Originally posted by erko
Here's a link which describes a cd and errors correction system..

http://www.ee.washington.edu/consele...audio/95x6.htm

There's some more info on the net exactly how the reading of pits is done (thats where the laserbeam thats refelected is not as intense because of the pits in the structure), but I couldn't find that anymore.
But remember that the cd is already quite 'old', it's a bit over 25 years back.

Eric

Eric, nice link. Like I said, these Philips & SONY engineers had their act together.

Two comments:
- realise that all that processing to read the disk is analog, RF and servo-circuitry, until at the very end the data is converted to square waves. Who said CD is a digital medium??
- I am still amazed that they can build & sell these complex engines for just a few 10s of euros. Incredible!

Jan Didden

[carlosfm]

Making a bit-perfect copy of a CD to hard disk (and then to another CD) is to copy the CD exactly as it is, including the errors.

That is a bit-perfect copy.
Correcting the errors results in a non bit-perfect copy.

Some copy-protection schemes have deliberate errors on the disc.
Some test CDs have gaps on the disc, to test CD transports.
Only a truely bi-perfect copy can make a real copy of the original disc.

EDIT: a .wav file is never a bit-perfect copy, in my understanding.
Only an image file of the complete CD, made with a special software.

[guido]

Originally posted by janneman



Sigh. The 4000-5000 was error correction, NOT interpolation or another form of error concealment. Error correction is transparant. The majority of CDs have zero error concealment, the error correction taking care of the, well, errror correction. Error concealment by nature is more than 'a few samples interpolated', because if only a few samples are damaged, chances are that the error correction can fix it, transparently. Error concealment is gross and generally quite audible, like 'yikes!'....

Jan Didden

I get what you mean, but i'm just trying to remember the elektorarticle and the stuff i did back then (with my cd650, so the article must have been based on that chipset).

From what i remember the counter in the article was used to count the number of times that the decoder told the filter to interpolate. So not 4000/5000 times but a few times per disk (depending on the quality).

The 7210 decoder does not have an output that indicates that it received a bad piece of data from the disk afaik (why would it?) Just the output to the filter to interpolate when circ fails.

According to the manual you need to physically 'brake' a disk with the fingers or use ff or reverse to generate pulses on this efab line between filter/decoder. Just playing the on purpose DAMAGED testdisk does not work: the circ can recalculate the correct samples from the damaged data.

Just wondering what the signal is that gives 4000/5000 pulses.
Best guess from me now: track loss. But that signal warns the microprocessor that track loss is about to happen. It can adjust the arm then (talking swingarm cdm's here). But track loss does not mean that the disk is read incorrectly (bad data)
Just that this is about to happen (and it might have resulted in bad data, or maybe not)

Too bad i cant find the article anymore... I know for shure i had it (i did some experiments on the 650). Maybe they talked about both options: counting track loss and interpolation.