Approximate Computing (AC) waives error free computation to improve circuits performances. Adaptive Least-Mean-Squares (LMS) filters can benefit from AC, being both power hungry and inherently approximate. In this paper an approximate LMS filter is proposed, which is able to change, at runtime, the precision level by acting on an external quality knob. An auxiliary circuit enables the approximation mode, in which the update of some of the filter coefficients is frozen. The proposed filter achieves a power improvement in the range 5–32%, as function of the tolerable quality degradation.
Design of low-power approximate LMS filters with precision-scalability
Napoli E.
2019-01-01
Abstract
Approximate Computing (AC) waives error free computation to improve circuits performances. Adaptive Least-Mean-Squares (LMS) filters can benefit from AC, being both power hungry and inherently approximate. In this paper an approximate LMS filter is proposed, which is able to change, at runtime, the precision level by acting on an external quality knob. An auxiliary circuit enables the approximation mode, in which the update of some of the filter coefficients is frozen. The proposed filter achieves a power improvement in the range 5–32%, as function of the tolerable quality degradation.File in questo prodotto:
Non ci sono file associati a questo prodotto.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
