An Energy-Efficient StEFCal VLSI Design with Approximate Squarer and Divider Units

Approximate computing attempts to maximize area and energy savings for a trade-off between quality and efficiency. This paper investigates the approximate squarer and divider units in the calibration procedure for radio astronomy called StEFCal (Statistically Efficient and Fast Calibration). The StEFCal circuit uses two squarer units from the literature, i.e., radix-4 (AxRSU) and SquASH, and iterative-based Newton-Raphson (NR) and Goldschmid (GLD) dividers. We demonstrate the efficiency of StEFCal using the approximate arithmetic operators from the Pareto-optimal that show the area- and power-quality trade-off. The results show that using the AxRSU combined with the NR divider improves the Mean Square Error (MSE) by 4.24%, with 58x more energy saving than the state-of-the-art.