dft with true z

[nicholebarry]

I think maybe the true dft isn't quite right (but I could totally be wrong).

Right now it's z_exp_ztrue = exp(1.*complex(0,1)*matrix_multiply(comov_dist_los-min(comov_dist_los), kz_mpc_orig_trim, /btranspose)) I'm suggesting z_exp_ztrue = exp(-1.*complex(0,1)*matrix_multiply(reverse(comov_dist_los-min(comov_dist_los)), kz_mpc_orig_trim, /btranspose))

They don't give the same transform, I checked.

[nicholebarry]

Here is the real part of the FT matrix for a regularized z, the original true z, and the proposed true z. Notice how the regularized z is more similar to the proposed true z (look for a left-right "shift"). The images are similar for the imaginary part.

Here is the PS difference between reg z and the original true z, for reference.

Here is the PS difference between reg z and the proposed true z. Notice how the red and blue flipped in the window? That seems like a logical result to flipping the dft direction. However, the majority of the difference seems to mimic the one above, indicating that the overall de-regularization of the dft obscures this direction effect.

Here is the PS difference between the original true z and the proposed true z.

The difference PS didn't really help me. But the matrix comparison (first image) gives me confidence that this is right.

[bhazelton]

Very interesting. I need to think about this a little more.

[bhazelton]

closed by #107