Home / Blog / Comparison Between Improved FFT Methods and Modern Spectral Estimation for Closely Spaced Interharmonics

Comparison Between Improved FFT Methods and Modern Spectral Estimation for Closely Spaced Interharmonics

Views: 0     Author: Site Editor     Publish Time: 2026-10-27      Origin: Site

Inquire

facebook sharing button
twitter sharing button
line sharing button
wechat sharing button
linkedin sharing button
pinterest sharing button
whatsapp sharing button
kakao sharing button
snapchat sharing button
telegram sharing button
sharethis sharing button

The full-phase Fourier method involves cyclically arranging and summing multiple segments of continuously sampled signals, resulting in a faster decay rate for the leakage spectrum; however, because the frequencies of adjacent interharmonics are close to those of the harmonics, it cannot effectively suppress spectral leakage caused by adjacent interharmonics. In summary, methods based on an improved Fast Fourier Transform can effectively reduce spectral leakage interference, but they often fail when adjacent interharmonics are present. This is particularly true when the system frequency deviates from the rated value (50 Hz), as the spectral signals of adjacent interharmonics interfere with the leakage spectrum of the harmonics, making it difficult to distinguish between harmonics and adjacent interharmonics based on spectral shape. Consequently, it becomes impossible to construct a correct computational model, leading to significant errors. Another class of modern spectral analysis algorithms includes the Prony method, rotation-invariant techniques, and the MatrixPencil method. The Prony method establishes a signal model with a damping factor, constructs a differential equation, and solves it to determine the frequency parameters.  Fast Fourier Transform (FFT) represents the prevailing technique for frequency-domain signal analysis, yet it suffers from spectral leakage. Representative improvement schemes cover windowing, interpolated Fourier transform, iterative Fourier transform and all-phase Fourier transform. The windowing approach mitigates spectral leakage arising from non-integer-period truncation by multiplying sampled signals with diverse window functions to attenuate signal magnitudes at the two edges of the sampling window, while its suppression capability remains relatively restricted in practice. The interpolated Fourier algorithm usually adopts three or five spectral bins around the main harmonic spectral line to interpolate and fit the genuine spectral component for leakage suppression. Even so, it cannot remove the main lobes of nearby interharmonics and is incapable of restraining the leakage interference induced by such interharmonic components. Built upon interpolated Fourier transform, the iterative Fourier method conducts sequential iterative computation for individual harmonics to pursue better estimation precision, yet it is still severely disturbed by adjacent interharmonics.

 

 

 

The all-phase Fourier transform rearranges and sums multiple consecutive sampled segments in a cyclic manner to achieve a more rapid attenuation of leakage spectra. Nevertheless, owing to the small frequency separation between harmonics and neighboring interharmonics, this approach cannot well suppress leakage triggered by adjacent interharmonics. To sum up, these improved FFT variants are capable of alleviating spectral leakage to some extent, whereas their performance degrades drastically under the coexistence of closely spaced interharmonics. This issue becomes more prominent when the system frequency drifts away from its rated value of 50 Hz. Under such circumstances, spectral components of adjacent interharmonics overlap with the leakage spectra of harmonics, making it hard to differentiate harmonics from interharmonics merely by spectral profiles. As a result, an accurate calculation model cannot be established, which eventually introduces substantial estimation errors. Another category belongs to modern spectral analysis algorithms, including the Prony method, rotation invariance techniques and the Matrix Pencil method. The Prony method constructs a damped signal model and sets up the corresponding difference equation, from which frequency parameters can be solved.

 

 

 

 

The ESPRIT algorithm builds the autocorrelation and cross-correlation matrices for sampled signals. It leverages the orthogonality between signal components of distinct frequencies to separate frequency constituents, then retrieves frequency parameters via generalized eigenvalue decomposition. As for the Matrix Pencil approach, it forms the Hankel matrix from sampled signals and achieves frequency parameter estimation using the same generalized eigenvalue decomposition technique. Operating purely within the time domain, these modern spectral analysis approaches are free from spectral leakage and fence effect. Nevertheless, they generally demand a pre-defined proper model order, exhibit strong sensitivity to noise, and yield low-precision estimations for amplitude and phase.

To tackle the challenge of precise parameter estimation for frequency components in scenarios with closely spaced interharmonics, existing research adopts zero-padding to isolate harmonic spectra from neighboring interharmonic spectra. This technique shares the same working mechanism as the chirp Z-transform; both rely on interpolation to reconstruct the envelope of the discrete-time Fourier transform. Fundamentally, zero-padding fails to enhance the intrinsic frequency resolution of the discrete Fourier transform, which greatly restricts its practical performance.


Focus on Intelligent Power - Build a Custom Electrical System

Products

Quick Links

Support

About Us

Contact us
  +86-15258336333
  +86-13058657657
  4th floor, No.2building, No.25 Yuxiu Road,Zhuangshi Street, ZhenHai Strict, Ningbo City,Zhejiang province
Copyright © DINGNUO ELECTRIC CO.,Ltd. All Rights Reserved.   Sitemap