| The personal computer market is pushing the magnetic disk drive industry toward more compact hard disk drives operating at faster data rates and lower power. At the same time that mechanical form factors are shrinking, the demand for greater data storage continues. These conflicting objectives can only be accommodated by increasing the areal bit density through more tracks/inch or more bits/track. To date, areal data density has been increased through improvements in the head/disk interface. However, improved detectors in the playback channel can provide higher density per track. Sophisticated signal processing techniques that have been widely employed in telephony applications are now being used in the playback channel. One promising technique developed by Jay Moon at Carnegie Mellon University is Fixed-Delay Tree Search with Decision Feedback (FDTS/DF). FDTS/DF incorporates a fixed-depth Maximum Likelihood (ML) decoder implemented as a tree-search detector and a decision feedback filter to cancel intersymbol interference not accounted for by the ML decoder. This thesis examines FDTS/DF in its original form and shows that the objectives of faster data rates and lower power are in direct competition with the objective of increased storage density. By analyzing the detector using linear discriminants, the system is greatly simplified. In fact on a channel using a Run-Length Limited code with a minimum run-length constraint of 1, a tree of depth 2 can be reduced to multi-level decision feedback equalization. |
