Multiplierless analog LMS adaptive FIR filters

Two adaptive finite-impulse-response (FIR) filter configurations are proposed which implement the least mean square (LMS) algorithm with no analog delay elements or multipliers. The first configuration uses a linear feedback shift register to generate pseudo-random binary sequences for applications where white noise has been traditionally used. The binary signals are delayed by a shift register and multiplied by on-and-off switching so that the resulting structure is free of the analog delay line and multipliers. For applications where colored inputs must be used, the second adaptive filter configuration uses a sigma-delta modulator as a front-end for converting the filter´s reference input into a binary sequence. It is shown that the filter tends to adapt itself into a low-pass shape to remove the high-frequency noise generated by the modulator. Computer simulation results show very good performance for such hardware efficient configurations