search

githubdoe / DFTFringe

DFTFringe Telescope Mirror interferometry analysis Program.
GNU General Public License v3.0
165 stars 59 forks source link

when averaging wavefronts, ignore regions should be dealt with better #17

Closed gr5 closed 1 year ago

gr5 commented 1 year ago

example data and result: https://groups.io/g/Interferometry/topic/97668823#33021

I'd like to work on this issue. It probably won't be for a few weeks but I'd like to look at the code for this and see if I can come up with a solution.

gr5 commented 1 year ago

Thinking about this more, it's quite complicated! The wavefronts displayed are not the wavefronts stored. The stored wavefronts have HUGE amounts of tilt and defocus and each of the wavefronts to be averaged may have very different values for (for example) tilt. So when you include all the data in 2 wavefronts and average them together the new wavefront will have a different tilt from either of the other two.

If you skip a few points because they are ignored, they won't get the proper tilt and neighboring pixels, one pixel a proper average will give a very different value from one pixel over which only uses one of the wavefronts.

So for the "ignore" pixels we still need to average in something.
1) Maybe the nearest non-ignored pixel? If it's a large ignore region that could be pretty far away.
2) So maybe the unchecked zernicke's should be used to calculate the best fit replacement pixel. No I think it should use piston,tilt,defocus every time. Because 2 different igrams are expected to have different values for these 4 terms even in a perfect world. whereas coma, etc shouldn't change from one igram to the next.

I think option #2 is best. The "ignore" areas get the pixel value for a simulated wavefront with the same piston,tilt,defocus.

githubdoe commented 1 year ago

I'm surprised that I don't average the nulled wave fronts instead of the raw (.Data) member. Perhaps there is a reason that I no longer remember. But I would have you try using the .nulled wavefront member instead in the average routine in surfaceManager.

githubdoe commented 1 year ago

The problem with using the .nulled member is that then the null goes away in the final result. One could add the null back into the result.

githubdoe commented 1 year ago

The other option would be to null with only defocus and tilt. The nulling routine can be passed what zernike values should be nullded so this could be done to each wavefront just before add it to the average.

gr5 commented 1 year ago

Fixed in PR 57