Closed mkolopanis closed 3 years ago
This line is suggestive:
# this builds a 3 x 64 x 64 matrix, need to transpose axes to [2, 1, 0] to get correct
# 64 x 64 x 3 shape
lm_matrix = numpy.asarray([i * lm_step - 1.0, j * lm_step - 1.0, numpy.zeros_like(j)])
lm_vector = lm_matrix.transpose([2, 1, 0]).reshape((self.skymodes, 3))
It may be that changing the transpose to 1, 2, 0
is all that is needed,
this very well could be, I think I added this when trying to make a loop a little "smarter" wouldn't be surprised if i got it all backwards.
I'll give it a try.
It's a flip and a numpy.fft.fftshift
in order to use the regular SaveOp
.
This was closed via #18
noted in #16 the DFT images appear at first glance to be different than the FFT ones. @jaycedowell noted there may be a rotation? Also it returns a 1-D flattened array of the pixels, we should have both return a comparable product. This will help in the inclusion of DFTing some antennas and adding them to the FFT images.