Biased logarithmic arithmetic in FIR filters

Logarithmic arithmetic can provide advantages especially in high-speed, moderate precision computations. A drawback of the logarithmic number system is the cumbersome addition and subtraction operations, involving lookup table accesses. A biasing scheme is proposed for substantially reducing the complexity of the lookup table. The approach is effective when repeated accumulation of products is needed, as in finite impulse response (FIR) filter implementations.