(paging rhl, just FYI)

For all images, the physical scale of the image is roughly 375 microns (25 pixels on the detector).

FFT_effects_without_apodization.png shows how different pupil create different PSFs. I show the different component arising from the 3 types of the obscuration that are present in the system (detector, struts, slit+slit holder). As we suspected the crossing patter is due to the struts. 

FFT_effects_witg_apodization.png shows the results after I applied a little bit of apodization, by convolving the pupil with the Gaussian with the sigma of 0.75 pixels. We see that the reflection is diminished.

I played with this a lot before already.... I feel a bit uneasy to introduce a custom apodization, i.e., to just pick a number with which to convolve the pupil but perhaps there is no better options. If the sigma is small, then the cost of the convolving is not too high and this could be introduced in the current algorithm. 

Resulting images look ``fine'' when analyzed on cluster in early November 2019