hassan Is it ok to hijack this ticket for my function which provides 2D PSF? (following the discussion and comment by price: `Regarding the 2D PSF, I suggest that Neven provides me a function that generates his PSF, and I'll wrap it up in the API. How that gets ticketed doesn't bother me.`) Note that I would probably not provide PSF as function of wavelength as it says in the description, but x and y (which is what I believe 2D PSF class will also take).

Yes, no problem using this ticket for your function ncaplar.

@price Paul, for the lack fo clear location, I have placed the first version of the code at https://github.com/nevencaplar/PFS_Work_In_Progress/blob/master/PIPE2D-347/Provide_PSF_2D.py

As you might see, it depends on the files in Tiger, which are placed at /tigress/ncaplar/PIPE2D-450.

Very happy to help with clarification or code! Excited to see this first version implemented so that we can start working on improvements!

I've got a basic implementation going; work is on PIPE2D-343.

ncaplar, can you regenerate your images with the PSF centered on the middle pixel? Having it centered between two pixels creates an offset when binning by 5.

I do not fully understand the request/problem. You basically would want to an odd number of pixels in the provided PSF image? Can you problem be solved by shaving off last few columns/rows of the provided image? Would kind of offset is being created when binning by 5?

It doesn't matter how many pixels there are — I can deal with the edges.

I want the centroid of the image to be centered on a pixel after binning by 5. This cannot be done by changing the origin when binning, because 5 is an odd number. Currently, the centroid is halfway between two pixels. No matter where I place a 5x5 pixel box around that point, the centroid cannot be in the center of that box.

Actually, maybe it would be best if I measure the centroid and interpolate so that it's centered? That way, I can remove any coordinate differences between your code and the pipeline.

Let's not re-estimate the centroid – therein lie biases and offsets. Neven, can you generate the PSF centred in a pixel (which is 0.0, 0.0 in the coordinates we use?)

I still do not understand the problem, and I think some more explaining will be needed on Monday. Having said that, I do not expect any problems here, I can generate either 10x oversampled PSFs or somehow shifted PSF (once when I understand what the problem is).