Basic FAQ for FFT Analyzer

The input time signal is often a continuous signal that extends infinitely, and the FFT processing involves cutting it out with a time window length T (number of sampling points N) for analysis. The FFT operation is performed by taking a finite time length T from the real signal, treating it as a repeating time waveform, so when T is an integer multiple of the input signal period, the repeating waveform becomes continuous and a correct spectrum is obtained. However, if this is not the case, the waveform becomes discontinuous around the repeating points, resulting in distortion in the waveform, and a spread occurs in the spectrum centered around that frequency. As a result, the magnitude of the spectral peak decreases compared to the true value, and the corresponding amount of power (equivalent to the magnitude) leaks out on both sides of the peak.

Cutting a finite duration T from the actual signal in this way is called applying a time window, and the spread of this spectrum is referred to as leakage error. Such errors can hide small peaks that are nearby due to large power frequency components, which becomes a significant practical issue, so efforts are made to minimize this as much as possible.