At the beginning of AA implementation, we checked balance (between each segment of QPD), demodulation phase (this is equavilent demodulation phase set in DGS, which can rotate the input I/Q signals by a matrix. In our case, we rotate signals to I phase.) and rotation angle (this is the quavilent rotation of QPD). All of these were done without the lock of AA loop. And we found some change from day by day.

Since now we have AA locked. I checked them again and optimized accordingly. The old work gave already quite good gain/phase/angle values. For example, in the first two attached figures, we can see the balance situation of QPD1 and 2. The balance is already quite good. Since we could get exact values from figrue 1 and 2. I calculated the required gain to balance them. After applying these gain, the balance situation is shown in the attached figure 3 and 4. We can see that the balance becomes much better. But if this will change again from day by day should be examined later.

(Old gain is 1 for each segments.) New gain is shown in the following table

After that, I also optimized the demodulation phase. The result is shown in the attached figure 5 and 6.We could see that only negligible signals go to Q phase. This should also be checked again later.

New phase is shown in the following table

Then, the rotation angle was also optimized. The optimization result is shown in the attached figure 7 and 8. However, we could see that pitch to yaw coupling has become already around 15% in WFS1/2. This should also be checked later if they will change or not.

Old rotation angles are all zero. New angles are shown in the following table

In the end, a new matrix was developed to close AA loop. The loop filters/gains are the same with the old case. A comparison of diagonalized signals are shown in the attached figure 9. It can be seen that they don't have large difference.