Level-crossing downsampling for quantization error reduction in sine wave estimation

Abstract

This work introduces a digital postprocessing algorithm—level-crossing downsampling (LC-DS)—for estimating sine wave parameters from sequences of quantized values acquired by standard ADCs. LC-DS emulates level-crossing sampling by retaining only transition points, reducing correlated quantization error, and accelerating least-squares regression (LSR). Its performance is benchmarked against uniform LSR and calibrated sinefit to highlight accuracy and computational tradeoffs. Across a wide dynamic range, LC-DS consistently outperforms uniform sampling and approaches the accuracy of calibrated sinefit for low-level signals, while remaining up to two orders of magnitude faster for large datasets. Unlike conventional methods, LC-DS scales efficiently with data size, enabling real-time estimation without hardware modifications. Practical and simulated experiments, including electrochemical impedance spectroscopy, confirm robustness under conditions such as signal saturation. These results position LC-DS as a compelling alternative for applications requiring both high precision and computational efficiency.