spectrum updates - Githubissues

cboulay commented 3 months ago

Add nfft argument, similar to numpy / scipy's fft
Add norm argument, similar to numpy / scipy's fft
Add do_fftshift argument, only used when requested output is FULL
Use faster rfft over fft when input is not complex and requested output is POSITIVE.
Fix frequency slicing when using POSITIVE or NEGATIVE. Previously we were missing the last frequency when nfft / win_len was even (nyquist for POSITIVE, and f = 0 Hz for NEGATIVE).

The last change listed above has the potential to affect outputs because in some scenarios the spectrum is longer by 1 frequency bin than previously. I had to update some unit tests to accommodate this.

cboulay commented 2 months ago

@griffinmilsap , can you please check that this doesn't break anything important for you? To save time, you may want to use the wavelets branch from PR #19 which includes this, filterbank, and a CWT. However, this is the only PR that touches old code and might affect something and thus has the higher burden for checking.

cboulay commented 2 months ago

This fixes bugs and I need it for other new features so I'm going to merge. If this causes any problems then let me know and I'll help you update your code to work with this fixed version.

ezmsg-org / ezmsg-sigproc

spectrum updates #16