An Improved Accuracy Harmonic Extraction Method for Microcontrollers

Harmonic extraction is frequently used in condition monitoring and control of power systems, power electronics, and machine drive applications. The challenge is to implement fast and accurate harmonic extraction on consumer-grade microcontrollers that have limited computational resources. When the sampled length of the input signal is not multiples of the fundamental frequency, the extracted amplitude and phase values have large errors. We proposed a new method named double-interpolation FFT (DIFFT), comprising both frequency domain and time domain interpolations and fully utilizing the manufacturers’ optimized binary libraries of FFT. The method is verified with both GNU Octave and microcontrollers, including ARM-based STM32 and Texas Instruments (TI's) TMS320. Variable input frequencies of several reference signals are tested, and up to the 41st order of harmonics are extracted and compared to reference values. The new method outperforms state-of-the-art harmonic implementation on microcontrollers in both execution speed and accuracy. Memory usage and computation time are measured, underscoring the applicability and superior performance of the proposed method.