Handling CD Read Errors

Handling errors when reading a disc was a big part of the original Red Book CD standard. CDs use parity and interleaving techniques called cross-interleave Reed-Solomon code (CIRC) to minimize the effects of errors on the disk. This works at the frame level. When being stored, the 24 data bytes in each frame are first run through a Reed-Solomon encoder to produce a 4-byte parity code called "Q" parity, which then is added to the 24 data bytes. The resulting 28 bytes are then run though another encoder that uses a different scheme to produce an additional 4-byte parity value called "P" parity. These are added to the 28 bytes from the previous encoding, resulting in 32 bytes (24 of the original data plus the Q and P parity bytes). An additional byte of subcode (tracking) information is then added, resulting in 33 bytes total for each frame. Note that the P and Q parity bytes are not related to the P and Q subcodes mentioned earlier.

To minimize the effects of a scratch or physical defect that would damage adjacent frames, several interleaves are added before the frames are actually written. Parts of 109 frames are cross-interleaved (stored in different frames and sectors) using delay lines. This scrambling decreases the likelihood of a scratch or defect affecting adjacent data because the data is actually written out of sequence.

With audio CDs and CD-ROMs, the CIRC scheme can correct errors up to 3,874 bits long (which would be 2.6mm in track length). In addition, for audio CDs, only the CIRC can also conceal (through interpolation) errors up to 13,282 bits long (8.9mm in track length). Interpolation is the process in which the data is estimated or averaged to restore what is missing. That would of course be unacceptable on a CD-ROM data disc, so this applies only to audio discs. The Red Book CD standard defines the block error rate (BLER) as the number of frames (98 per sector) per second that have any bad bits (averaged over 10 seconds) and requires that this be less than 220. This allows a maximum of up to about 3% of the frames to have errors, and yet the disc will still be functional.

An additional layer of error detection and correction circuitry is the key difference between audio CD players and CD-ROM drives. Audio CDs convert the digital information stored on the disc into analog signals for a stereo amplifier to process. In this scheme, some imprecision is acceptable because it would be virtually impossible to hear in the music. CD-ROMs, however, can't tolerate any imprecision. Each bit of data must be read accurately. For this reason, CD-ROM discs have a great deal of additional ECC information written to the disc along with the actual stored information. The ECC can detect and correct most minor errors, improving the reliability and precision to levels that are acceptable for data storage.

In the case of an audio CD, missing data can be interpolatedthat is, the information follows a predictable pattern that enables the drive to guess the missing values. For example, if three values are stored on an audio disc, say 10, 13, and 20 appearing in a series, and the middle value is missingbecause of damage or dirt on the CD's surfaceyou could interpolate a middle value of 15, which is midway between 10 and 20. Although this might not be exactly correct, in the case of audio recording, it probably won't be noticeable to the listener. If those same three values appear on a CD-ROM in an executable program, there is no way to guess at the correct value for the middle sample. Interpolation can't work because executable program instructions or data must be exact; otherwise, the program will crash or improperly read data needed for a calculation. Using the previous example with a CD-ROM running an executable program, guessing 15 is not merely slightly offit is completely wrong.

In a CD-ROM on which data is stored instead of audio information, additional information is added to each sector to detect and correct errors as well as to identify the location of data sectors more accurately. To accomplish this, 304 bytes are taken from the 2,352 that originally were used for audio data and are instead used for sync (synchronizing bits), ID (identification bits), ECC, and EDC information. This leaves 2,048 bytes for actual user data in each sector. Just as when reading an audio CD, on a 1x (standard speed) CD-ROM, sectors are read at a constant speed of 75 per second. This results in a standard CD-ROM transfer rate of 2,048 x 75 = 153,600 bytes per second, which is expressed as either 153.6KBps or 150KiBps.