search

spacetelescope / poppy

Physical Optics Propagation in Python
https://poppy-optics.readthedocs.io
BSD 3-Clause "New" or "Revised" License
216 stars 72 forks source link

StatisticalPSDWFE divide by zero #452

Open philhinzatucsc opened 3 years ago

philhinzatucsc commented 3 years ago

I have a newbie question/ bug. I have just started using poppy to estimate wavefront error effects on the PSF for Keck/SCALES. To do so, I have a created a FresnelOptical System model and then use StatisticalPSDWFE to create an error for one of the optics. This results in a warning:

. . ./poppy/wfe.py:349: RuntimeWarning: divide by zero encountered in power psd = np.power(rho, -self.index) # generate power-law PSD

I see why it's complaining. r^-3 is infinite when r=0 (center of the array). So I went to wfe.py and added a really crude hack: rho[rho == 0] = 0.00001 #get rid of infinity.

Now this works fine and I can happily continue modeling. Is this a real bug, or am I missing something about how to properly use StatisticalPSDWFE?

mperrin commented 3 years ago

@jlumbres could you please take a look at this issue and provide your take on this divide by zero issue with StatisticalPSDWFE? Thanks much..

jlumbres commented 3 years ago

Oh, hi @philhinzatucsc!

This issue happened with me when I wrote PowerSpectrumWFE. I did a similar solution, too. I overwrote the one pixel where spatial frequency k=0 to be 1*dk (spatial frequency spacing), did my math, then I overwrote the power at that same pixel location to be 0 after the math was done. I set the power to 0 because it was an arbitrary choice on my end. You can see my code here.

philhinzatucsc commented 3 years ago

Ok, thanks for the confirmation. I am new enough to github and poppy that I am not sure what the next step is. For my own work, my internal hack works fine, but presumably I or you should be fixing this for others? I am not sure how to submit a bug fix.