search

spacetelescope / webbpsf

James Webb Space Telescope PSF simulation tool
https://webbpsf.readthedocs.io
BSD 3-Clause "New" or "Revised" License
119 stars 63 forks source link

Small (~0.5 deg) rotation of NIRCam coronagraphic PSFs relative to on-sky 335R observations #673

Open kdlawson opened 1 year ago

kdlawson commented 1 year ago

PSFs generated by WebbPSF appear to be rotated relative to on-sky PSFs. This is especially visible at large separations:

(Note: data are F300M + 335R coronagraphy; WebbPSF models include distortion and data are not distortion-corrected)

Talking with @mperrin, it sounds likely that this is caused by pupil shear: pupil shear for the on-sky data is causing a small rotation of the PSFs which isn't included in the WebbPSF model. Setting nrc.options['pupil_rotation'] = -0.5 gets much closer to the on-sky PSFs' orientation:

However, since the translation components of shear (nrc.options['pupil_shift_x'] and nrc.options['pupil_shift_y']) are not coupled to the rotation effect in WebbPSF, simply adopting empirical X/Y shear measurements won't correct the rotation.

Assuming some nominal shear effects by default may be beyond the scope of this project. However, I think that at least including a default pupil_rotation value to approximately correct this would improve functionality for many users who might otherwise miss this nuance.

Beyond that: in cases where pupil_rotation is not set while pupil_shift_x and pupil_shift_y are, automatically populating a corresponding pupil_rotation would be a really helpful functionality.

Thanks much! Please let me know if there is additional info I can provide.

mperrin commented 1 year ago

I have re-confirmed that for regular NIRCam imaging, there is no need to adjust the rotation parameter.

Here's an example of a NRC ALONG PSF from yesterday (the LW parallel taken to the routine WFS), and the right-out-of-the-box simulation (no adjustments applied other than loading yesterday's WFS OPD, and aligning in position and flux).

This confirms that the overall NIRCam pupil-to-detector rotation is simulated correctly, and the discrepancy in the simulation has to be from the Lyot wheel rotation. I.e. it's correct to adjust via the nrc.options['pupil_rotation'] as you're doing, and NOT by changing the overall nrc._rotation parameter.

im

To triple check, I did repeat the same simulation with adjusting nrc._rotation, and it's unambiguously not a good match for the data.

im_rot

mperrin commented 12 months ago

The models relevant for this will be improved by PR #675