METADATA:
Cosmics with both reverse full field and full field.
Statistics:
RF: About 3,800,000 tracks
FF: About 180,000 tracks

Yuri suggested that we try to reverse the alignment in terms of which part of the sector we move with respect to another. Before we used to rely on the outer sector survey parameters and align inner ones. Now we are attempting to move the outer sector with respect to inner ones.

I ran a few passes and here are the results for the Run-21  reverse full field cosmic data:

The above alignment procedure was done starting with the existing parameter set.
Below is the same OUTER SECTOR alignment process (3 passes, 0 indicates the initial displacement) starting from the zeroed out parameters.

I continued the alignment procedure and also repeated it with the Full-Field cosmic data. Below are the results:

The above plots show that RF was brought to a below 100 um level and using that as a starting point for FF shows that deltaY and deltaZ are in the right place and much improved compared to the starting point (0 in pass coordinate). deltaX for the FF does not converge but I believe that is due to a pretty small data sample.

Bellow is the comparison for the delta pT as a function of pT for 1) existing alignment (final with nominal alignment procedure); 2) zero alignment; 3) final alignment with reversed alignment procedure.

1)  2)  3) 

Below is the sigma of the above resolutions as a function of pt for all these cases

The prePass, which stands for the existing alignment parameters before this study started, is slightly worse than both non-aligned and aligned versions.

UPDATE 2023.04.28

Looking at the SuperSector positions (X, Y, Z) for the prePass [left] and Pass016 (with no SuperSector correction applied) [right] we see that we need SuperSector correction after Pass016:

Update: June 1, 2023
I ran SuperSector alignmetn procedure a few times but it looks like it diverges in the Y direction, even though X and Z seem to be going the right way.

After discussions during the meeting it was suggested to increase the statistics for the procedure.
SUPERSECTOR ALIGNMENT COTNINUED
I increased number of events on which I run SS alignment to around 300,000 this is enough for all the fits to work with very well.
The result is that the procedure does not quire perform as it should - it diverges from the zero point.

Good thing is that the very initial position (which is same between all ideal - inner, outer, ss and my-outer and ideal inner ideal SS) stage is not as bad under 200 um.