matlab-analyze-spectrum

Use pwelch for long signals where segment averaging reduces variance, or periodogram for short signals that cannot be segmented. For a quick look with automatic defaults, use pspectrum(x,Fs). Always pass the sample rate Fs explicitly and use pow2db for dB display.

PSD measures power per Hz and is resolution-independent, making it suitable for noise characterization and comparing signals. Power spectrum measures power per bin so peak height equals true tone power, making it better for reading tone amplitudes. Convert between them using the resolution bandwidth RBW = enbw(win)*Fs/segLen.

Use pspectrum for quick exploration with good defaults and a Leakage parameter. Use pwelch for low-variance PSD estimates on long signals. Use periodogram for short signals or when you need exact Fs/N resolution and explicit window control. Use pmtm for short data needing low variance without segmentation.

Either the tones are closer than the frequency resolution Fs/N, or sidelobes from the stronger tone mask the weaker one. Fix resolution by acquiring more data or using rootmusic for super-resolution; fix masking by switching to a lower-sidelobe window such as blackmanharris or kaiser with high beta.

Yes. For nearly uniform data with small timing jitter, pspectrum(x,t) resamples internally. For truly non-uniform or gapped data, use plomb(x,t), which computes the Lomb-Scargle periodogram and supports detection-probability thresholds via the Pd option.

Do not use it for filter design, time-frequency or spectrogram analysis such as STFT and CWT, audio-specific features like MFCC or pitch, or control system frequency response such as Bode plots. Those tasks belong to dedicated toolboxes and workflows.