Soft-output detector for partial-response channels using vector quantization

Turbo equalization has been shown to have excellent performance for magnetic recording channels, and permits operation at low signal-to-noise ratios. Soft-output algorithms, such as the BCJR or SOVA algorithms, are used as the detector in proposals for turbo equalization. However, these algorithms are difficult to implement effectively at the bit rates and power consumption levels required for hard drives. The paper considered an implementation of the BCJR algorithm where the state metrics are vector quantized, and the operations are performed using table lookups. Table lookup operations are efficiently implemented in both hardware and software. The proposed algorithm uses off-line computation to generate vector quantizers and lookup tables, and the only run-time operations are table lookups. Using vector quantization, the size of the lookup tables can be of reasonable size. Further, the compressed representation reduces the required state metric storage.