Adaptive Coding for Self-Clocking Recording

Self-clocking of binary information involves a coding constraint whereby the maximum length of zero strings is limited to a predetermined number. For reasons of operational stability it is desirable to keep this number low. The price of such constraint is a degradation in code density. The class of serial recording which includes the well-known frequency modulation and phase encoding represents one extreme where every other symbol is mandatory, making the code density one bit for every two symbols. This paper describes a coding method which improves code efficiency while maintaining the constraint on the length of zero strings. It will be shown that adaptive coding, which is an extension of the block code concept known from information theory, offers a significant improvement. The concept is illustrated by describing a practical application in which the code density is improved from two symbols per bit to three symbols per bit pair, while keeping the maximum length of zero strings to one. It is shown that while this technique represents an improvement by one third over phase encoding, it approaches the theoretical limit for this class of codes to within 4 percent.