In a discussion of error protection, Guy G. Sotomayor, Jr ggs@shiresoft.com explained how playing back a CD-ROM at 1x requires reading the disc at more than 4 megabits per second. If you are concerned about matters such as making copies accurate to the bit level or if you want to understand some of the reasons behind the imperfections of CD-R, this section is worth the time it takes. Before going into his text, let me note that a 'symbol' is a character or a value - basically, just an 8-bit byte. Each sample of a redbook waveform consists of sixteen bits for each of two channels: 32 bits, 4 bytes or 4 symbols.
The first thing to remember is that the data on the CD is stored as 14 bits/symbol rather than just 8. There are several reasons for this, but it makes reading the bits easier. Here are the criteria for selecting which patterns can actually be used:
1. 11 is not allowed
2. 1001 is as close as two 1s are allowed to get
3. 100000000001 is as distant as two 1s are allowed to get (ten 0s).Each 14-bit symbol is separated from its neighbor by 3 bits, called merging bits, coupling bits, connecting bits or packing bits. They allow the 3 rules above to be applied continuously. Remember at this point we just have a string of bits -- there is still nothing indicating where bytes start and stop. To recap, 8 bits is represented by 17 bits - a little over 2x.
Now, we look at how data are actually stored on the CD. Everything depends upon a 588-bit frame. The 588 bits are organized as follows:
Description Bits/each Total bits Sync Word 24 bits + 3 padding bits=27 27 Control Word 17 17 Audio Samples 6 samples, 2 symbols ea 17 x 12=204 Error Correction Q 17 17 x 4=68 Audio Samples 6 samples, 2 symbols ea 17 x 12=204 Error Correction P 17 17 x 4=68 Total 588.
To figure out how many 588-bit frames per second we're dealing with we go back to what we see on the output of the CD (namely 176400 bytes per second). A stereo "frame" is 24 bytes, so if we take 176400 / 24 we get 7350 frames per second. Now take 7350 x 588 and you get 4321800 bits per second. Q.E.D.
BTW, the P, Q, G, etc channels are derived from "stacking" the control words (98 of them to be exact) to get the encoded data. As you will no doubt notice, audio data has ECC applied to it. This is the CIRC that I spoke of previously. This is always there. Data because of its more exacting nature has an additional layer of ECC applied to it to further reduce the chance of an error. Hope this clears it up!
Footnote from Mike:
There are additional issues including the nature of that 'additional layer of
ECC' and the numeric values for error rates in audio and in data. Since ECC
discussed by Guy is always imposed, I usually do not consider it in the
discussions in this primer. The only specifications I have found for accuracy
on CD-R suggest a bit error rate of one in a million for audio and one in a
billion for data. Both appear to be worse than we experience in practice,
particularly the one for data; that extra layer of ECC should cut the rate by a
lot more than a factor of a thousand. Error rates for pressed discs are
much lower than those for CD-R.
E-mail me at cdrecording@mrichter.com
Return to Mike's home page