Optimization and Performance Characterization of FIR Butterworth Filters on Resource-Limited Embedded Platforms

Resumen

This work presents a comprehensive benchmarking study of Finite Impulse Response (FIR) Butterworth filter implementations on the ESP8266 (ESP-12E) embedded platform, with a detailed comparative evaluation against the widely used ATmega328P microcontroller. Identical filter configurations, with orders ranging from 13 to 51, are employed to systematically assess performance across key metrics, including CPU utilization, dynamic memory usage, processing latency, and signal fidelity. Experimental results reveal that the ESP8266 exhibits highly stable CPU utilization, remaining within 34–36% across all filter orders, in contrast to the ATmega328P, which demonstrates a pronounced increase from 28.6% to 68% as filter complexity rises. Memory consumption on the ESP8266 scales linearly from 11% to 33%, whereas the ATmega328P approaches a critical utilization level of 76% at the highest filter order. Owing to its 80 MHz operating frequency, 32-bit architecture, and 80 KB SRAM, the ESP8266 achieves a 1.89× improvement in processing speed and a 111.5× increase in available memory relative to the ATmega328P. Statistical analysis confirms that these performance differences are highly significant (p < 0.001). Signal fidelity evaluation shows close agreement with theoretical Butterworth responses, with passband ripple below 0.06 dB and stopband deviation limited to 1.23 dB. Long-duration stability testing over 24 hours further indicates zero timing violations and negligible performance drift. Collectively, these results demonstrate that contemporary IoT-oriented microcontrollers substantially extend the practical limits of embedded digital signal processing, enabling more complex filter implementations while preserving robust performance and system stability.