Simulation of a cold optics misalignment in NIKA2, and the way to recover from it with the external optics
On December 8 we have checked the alignment of the cold pupil inside the NIKA2 cryostat, using an ecosorb slab, sliding it from the edge of the window toward the center and marking the position when it had an effect on the detectors response. Doing it from 4 sides allowed to determine that the cold pupil was misaligned on the window by 2mm horizontally and 10mm vertically.
We don't have enough margins on the cryostat air-compressed feet to compensate this misalignment, we have to do it with the external optics. Removing the M3 mirror and using the telescope laser on the elevation axis, we tilted the M6 mirror so that the laser spot on the window rose by 10mm, in order to match the telescope pupil with the cold aperture stop footprint on the window, hence aligning the external optics with the internal optics.
The most probable explanation for the cold pupil (aperture stop) misalignment is an error of adjustment of the shims which are used to fix the internal optics and various temperature screens concentrically within each other. In other words, it's not the sole cold pupil which is misaligned, but the whole cold optics inside the cryostat.
The images below show a simulation of this problem and the solution we chose to compensate this misalignment of the optics with the M6 mirror.
Image 1 = cryostat face view ; image 2 = cryostat side view ; image 3 = footprint on cryostat window (the small colour crosses represent the contour of the primary of each field that can be seen on the array, we can see the window is not exactly at a pupil surface of the system because these contours are not quite at the same place); image 4 = footprint on array (image) ; image 5 = footprint on primary mirror with vignetting of the rays from the aperture stop (the cold pupil).
1) Nominal optics: everything well aligned.
2) Tilt of the internal optics: 10mm too high at window, 0.35 degree tilt downward so that plate near array in transmission through polariser stay at same height. The red circle on the window represent roughly the contour of the internal pupil we found using the ecosorb, while the small colours crosses represent the contour of the primary for each field).
3) Extra tilt of M6 to compensate for the internal optics tilt and thus align external optics with internal optics.
Some additional explanations:
With the test using the eccosorb we found that the internal optics in the cryostat (its optical axis) crosses the window 1cm higher than its center. The thing we did with M6 is changing its angle so that it points 1cm higher than the center of the window, hence matching the axis ox the external optics with the internal optics. We used the laser of the elevation axis in order to visualize this tilt on the cryostat window. Another way to see the problem is that yesterday we found that the window was 1cm lower than the real optical axis of the cold optics. In the previous run we aligned the external optics to the center of the window, while now we aligned it with the real internal optics axis.
So if we had an offset between the radio axis (where the center of the instrument array points in the sky) and the optical axis (where the telescope points in the sky) then what we did yesterday had no impact on this. It is possible we have such offset between the optical axis and the radio axis, but this is degenerated with the pointing model. The only way we could find to align these two correctly would be to use a source independent of the pointing model. There's only two possibilities: put a point source at the center of the telescope image plane which is between M5 and M6 (ex using the sky simulator with fake planet in front of the cryostat as it was done in the lab in 2015), or use the axis of rotation of the wobbler which is at the center of the telescope optical axis.In the latte case we can't wobble looking at a source because we re-introduce the degeneracy with the pointing model, but we could wobble on a blank sky and use the change of background power due to spill over of the ground radiation on the edge of the primary. Though it would be super if we could do that, in both cases I don't see how we could implement such test soon. So we have to live with this uncertainty on the optical vs radio axis possible misalignment.
The good news is that as long as we are not too far off, and that the field distortion is not bad on its edges (which is the case in the way I designed the optics), then a small mismatch has practically no impact on the data quality.