@embr what happens mathematically is the window is multiplied in the time domain („no window“ just mesns rect window, unless you just actually take the discrete fourier transform of the entire signal) and mult in time domain becomes a convolution in frequency domain. So what happens is you get as the result your signals fft convolved with the windows fft. Since the rect windows fft has a ton of very strong side lobes it spreads a lot the energy from each bin very far to all other bins (unless your signal components happen to all perfectly fall on fft bin frequencies). Hann or honestly any commonly used window has a wider main lobe but far smaller side lobes in frequency domain, so instead of generally smearing components all over the spectrum, it deliberately smears them but in a very localized way.

which one to use for any given task can be fairly involved but Hann is a safe default choice and other than „not rect“ it doesn’t matter *that* much until you get far deeper into the weeds