search

EoRImaging / eppsilon

eppsilon - error propagated power spectrum with interleaved observed noise
BSD 2-Clause "Simplified" License
5 stars 4 forks source link

add spectral window shift #108

Closed nicholebarry closed 4 years ago

nicholebarry commented 5 years ago

Addresses #97

This needs to be checked in the math by someone who understands it better.

bhazelton commented 5 years ago

This doesn't make sense to me. The concept of the center bin in k-space doesn't apply in frequency space, where this window is constructed and used.

nicholebarry commented 5 years ago

I'm a little confused as to how we deal with an asymmetric window function (i.e. even number of frequency channels) if the proposed shifting isn't right.

Would it be right to have a symmetric window function that doesn't peak at 1 and has two maximums about the center in the even number case?

bhazelton commented 5 years ago

We can either use a symmetric or an asymmetric window in the case of an even number of frequency channels. Currently, it's hard coded to an asymmetric window (the keyword periodic is set), but I'm open to changing this or making that option accessible from the wrapper. It'd be good to do a more thorough investigation of which one is more correct. Looking at it now my guess is that the symmetric one might be more correct...

Wikipedia has a nice section on this: https://en.wikipedia.org/wiki/Window_function#Symmetry

bhazelton commented 4 years ago

I'd like to close this in favor of the approach I suggest in #124 and implement in #127

bhazelton commented 4 years ago

after talking to @nicholebarry, we're closing in favor of #127