A High Accuracy Approximate Multiplier with Error Correction

A novel approximate multiplier with tunable accuracy modes which leverages an inaccurate 4 × 4 multiplier is proposed. The inaccurate multipliers achieve a power reduction of 17.98%–34.15% over corresponding accurate designs with a maximal absolute relative error of 22.2%. A Gaussian filter is implemented for image sharpening as a sample assessment and the results show that the proposed architecture can achieve 5–22 dB increment on Signal Noise Ratio (SNR) with less power and area compared to other designs. In addition, the proposed multiplier is enhanced to correct approximate results by using an extra error detection and correction (EDC) circuit for non-error resilient applications. The proposed design has the least area compared to other existing error correcting multipliers.