by VirtinsTech » Tue Jun 24, 2025 3:04 am
Thanks for the additional explanation. These screenshots are now understood.
The method of measuring magnitude frequency response using a linear frequency sweep in the signal generator along with peak hold in the spectrum analyzer has significant limitations. To achieve higher frequency resolution at the low-frequency end, the FFT size must be sufficiently large. As a result, more peak-held frames are required to form a smooth frequency response curve. In the most optimistic case—assuming each frame produces a peak in one different FFT bin—the minimum number of peak-held frames equals half the FFT size. However, in practice, some FFT bins may be skipped and not captured by the peak hold, even if the sweep duration is very long. This can cause artificial undulations in the measured frequency response curve.
The recommended approach is the "FRswlog" method, which uses a logarithmic sweep combined with 1/96-octave analysis.