Modified fixed-point delayed LMS adaptive filter for achieving low adaptation delay with area-power efficiency

Least mean square (LMS) algorithms are an important class of adaptive filters which are used in various real world applications because of its simplicity and efficient convergence performance. However the LMS filters have limitations in the designing like inner product computation which do not support pipelining type of applications. The adjustment of the critical path length is also a major drawback which is overcome by using a novel partial product generator and optimized balanced pipelining in the combinational blocks. The delayed LMS filter is introduced which enables the critical path length adjustment thus enabling the optimal filter output. L/2 binary adder tree is employed in the error computation and weight update block previously performs the accumulative inner product operation but it has the major disadvantage of computation complexity causing difficulties in inner product computation. Hence in this paper, the carry look-ahead adder and the ripple carry adder has been employed instead of L/2 adder tree. Carry look-ahead adder provider quicker computation by reducing the time required to determine the carry bits as the carry bits do not wait for the previous bits to complete the computation. Ripple carry adder is a slow adder as the carry bits are processed one by one but it is simpler architecture helps in reducing the area which also reduces the power consumption. Thus the adaptive delay can be reduced in the proposed scenario with efficient area and power consumption. The simulation results also prove that the proposed modified fixed-point LMS adaptive filter performs better than the existing methodologies.