Error correction implementation and performance in a CD-ROM drive - technical

Hewlett-Packard Journal, Dec, 1990 by John C. Meyer

[FIGURES HAVE BEEN OMITTED]

The HP Series 6100 Model 6001A implements the error protection algorithm defined by the CD-ROM yellow book standard. This extra level of protection means that the error rate is improved from one error in 1012 bits to one in 1016.

IN AN IDEAL DIGITAL transmission channel the information that is input at the transmitter should be faithfully reproduced at the receiver. Unfortunately, because of things like noise, power-line fluctuations, and imperfections in the media, data can be corrupted before it reaches the receiver. This is why error detection coding and error correction coding (EDC/ECC) techniques are incorporated in the digital transmission system. The data on the media is encoded at the transmitting end with the EDC/ECC bits and decoded at the receiving end to correct for errors and to recover the original data. If there are errors, the EDC/ECC algorithms are designed to detect and correct a certain number of errors or to cause retransmission of the data. The box on page 46 is a primer on EDC/ECC techniques.

Because the CD-ROM disk has a very high bit density, it has an inherent error rate of [10.sup.-5] to [10.sup.-6] errors per bit. The red book standard, which has become International Electrotechnical Commission (IEC) standard 908, specifies the CD audio media format and provides a parity and error correction scheme that reduces the error rate to 10-11 to [10.sup.-12] errors per bit. All CD manufacturers provide this level of error protection.2.3 The yellow book standard,4 which is currently undergoing standardization as European Computer Manufacturers Association standard 130, or ECMA-130, specifies the CD-ROM format. This standard expands on the CD audio standard to include track definitions for digital data, mixed media disks, and digital data representation within a frame. For example, the minimum addressable section of a disk is defined by the red book as 1/75 of a second of audio data or 1176 16-bit audio samples, and by the yellow book it is defined as 2352 bytes of digital data or a disk sector. The CD-ROM yellow book specification provides an additional level of error protection that reduces the error rate to [10.sup.-15] to [10.sup.-16] errors per bit. Fig. 1 summarizes the results of error correction provided by the red book and yellow book standards.

The yellow book further defines three modes for sectors within data tracks. These modes all have a 16-byte header composed of 12 bytes of synchronization data (00, FF,..., FF, 00), three bytes of address data minute, second, frame), and a mode byte (00, 01, or 02). Mode 0 sector headers are followed by 2336 bytes of zeros. These sectors are used for lead-in at the beginning of disks, lead-out at the end of disks, and transition regions between audio and digital data tracks. The table of contents for a disk is contained in a portion of the lead-in area. Mode 1 sector headers are followed by 2048 bytes of user data and 288 bytes of EDC/ ECC data. The EDC/ECC bytes add the extra data protection over that which is available at the native interface as specified by the red book standard. Mode 2 sector headers are followed by 2336 bytes of user data. These sectors can be used for bit maps or other data that may not require the data protection provided by mode 1. Fig. 2 shows these mode formats.

Red Book Error Protection

The CD audio standard (red book) specifies two levels of parity and error correction which are implemented within the drive. The algorithm used for encoding the parity bytes is called a cross interleaved Reed-Solomon code (CIRC). As shown in Fig. 3, there are two CIRC encoders, CIRC1 and CIRC2, which provide the two levels of error protection. Sectors are divided into 98 24-byte F1-frames, which are processed through the CIRC encoder, which consists of three delay sections and two encode sections. The first delay section interleaves the F1-frames into two 12byte groups. One group is delayed for two F1-frame times (24 byte times). The interleaved F1-frames then pass through the CIRC2 encoder section, which generates four Q-parity bytes using a (28,24)* Reed-Solomon code. The second delay section delays the CIRC2 bytes from ON to 27N F1-frame times (N = 4). Time-delayed packets of 28 bytes then pass through the CIRC1 encoder, which generates four P-parity bytes using a (32, 28) Reed-Solomon code. Finally, the third delay section delays every other byte from the CIRC1 encoder one F1-frame time. At the output of the CIRC encoder, the 32-byte packets are called F2frames.

As a result of all this delaying and interleaving the original F1-frame bytes are delayed from three to 108 F1-frame times, and they wind up being redistributed over 106 F2frames. Thus, one frame's data is spread over three physical sectors on the media, thereby reducing the impact of burst errors. An additional control byte is added to the F2-frames to form 33-byte F3-frames.

F3-frame bytes are 8-to-14-bit (EFM) encoded and linked together with a 24-bit synchronization header. Three merging bits are used between bytes to maintain run length limits between transitions on the media. The result is that each 192-bit F1-frame is represented by a total of 588 channel bits on the media (see page 39 for an explanation of 8-to-14-bit encoding). Most of this redundancy can be used for error recovery, either for detection of errors and creation of erasure flags or for correction of errors using the P-parity and Q-parity bytes. An erasure flag is error location information that is obtained outside the decoding process, usually from hardware. If erasure flags are available, the number of errors that can be corrected increases because