Reduced precision redundancy for low-power digital filtering

We propose a low-power digital filtering technique based on voltage overscaling (VOS) and a novel algorithmic noise-tolerant (ANT) technique referred to as reduced precision redundancy (RPR). VOS implies scaling of the supply voltage beyond the critical voltage required for correct operation. RPR involves having a reduced precision replica whose output can be employed as the final output in case the original filter computes erroneously. In addition, an LSB estimator is also employed to compensate for information loss in the LSBs. For frequency selective filtering, it is shown that the proposed technique provides 30% energy savings over an optimally scaled (i.e., the supply voltage equals the critical voltage) present day system. Energy savings of up to 65% can be achieved if a SNR loss of 1.5 dB is tolerated.