New update on one-meter secondary mirror.
We have been working to calibrate the test setup with the N-position method, described earlier. We have a calibration, and work should resume within a few days at most.
We decided to collect two sets of data - one at five steps (360/5 = an interferogram every 60º), and one at seven steps (360/7 = one every 51.43º). This was necessary due to the presence of a dimple (from an internal fracture or delamination) on the mirror used for the calibration. Below are the set of 12 images. You can see the dimple (and the astigmatism) rotating from one image to the next; the mirror was physically rotated to make each image.
The rotated images, when averaged, should remove the aberrations and artifacts of the subject mirror, and leave only the aberrations and artifacts of the test setup.
To make a long story longer, we could not use the mirror being polished for the test because rotation preserves rotationally symmetric errors (aka toolmarks) - they cannot be average out, as they show up in all images. We had to use a similar mirror with heavy astigmatism and the aforementioned dimple.
Averaging the 12 images resulted in a series of dips where the dimples were - the recess was just too large to average out with 12 images (that is, one twelfth of the depth of the dimple was still too large for purposes of calibration). The dimple artifacts can be seen in five obvious locations in the image below, about halfway between center and edge. There are some more subtle artifacts there as well.
So a mask was created for each of the 12 images, zeroing out the dimple area, and the resulting average was much cleaner. Here is one of the masks that was used to filter out the dimple from a single interferogram.
Averaging the 12 masked images left some very light circular high-frequency toolmarks. Here is the average of all the masked images showing that the echoes of the dimple are gone. At this point we were nearly done; only some circular (and very shallow) toolmarks remain.
No sign of the dimple artifacts.
The above image, therefor, represents a mix of two things: an interferogram of the test setup showing its aberrations, and the toolmarks from the mirror used to make the calibration. The next step is to use a low-pass filter to ‘ emove' the toolmarks. (The toolmarks are high-frequency, so filtering out high-frequency artifacts will remove the toolmarks and leave MOST of the other data intact.)
which were cleaned up with a low-pass filter. Here are images we used to evaluate how much filtering to use. From left to right: Unfiltered, 5-pixel filtering, 10-pixel, 20-pixel.