Leginon

Is sq mag in LM and hl mag in M mode ?

Yes

We suspect the problem may be from a bad modelled stage calibration though I am surprised since the stage has not been touched in a very long time. I switched the square exposure movement to stage and it seems to be better than before. i will repeat a full modelled stage alignment on Monday. If you have any alterate explanations please let us know and we will try them out. Thanks.
Bob

If sq mag is in LM mode and hl in M, there will be image shift hysteresis in sq. Preset cycling and lens normalization in general can only remove so much hysteresis. What I've observed is that the lower the mag sq is the worse, probably because we then deal with hysteresis both in objective lens as well as projection system. A full square is something I have never seen though. More typical is 1-3 um. just enough to make you wonder which hole is the one you targeted at the high mag.

In general, the longer you leave the lens in LM mode, the longer it takes for it to recover to the stable normalization image shift as well. One thing I do during set up is to go through my work flow of acquiring square/hole/exposure at least once before I align my presets. It normally uncorrect whatever I did before going through it after atlas collection, indicating that the first alignment so soon after staying at
very low mag is not a stable value once your workflow brings it to a better place.

Regarding stage calibration, when you do stage calibration, please observe how reproducible the movement is. You can check by displaying the correlation peak cursor during the measurement. It may tell you if you have other problems. In Navigator node, there is a position movement test tool at the far right. You can tell it to return to the current position 10 times from different direction from a given distance. You should get within 1 um, I think.

If problem is not stage, check if you've the same image stabilization after the "rehearsal" as i did. You may also try to move to a little higher mag on sq. Also, I think Polara might already have a lens normalization option, you can try to use it, even though it takes more time.

Anchi

Anchi,
I performed the stage reproducibility test a few times and that appears to have been the problem. when I moved it 100 microns the error was in the 3-20 micron error range at the sq preset but when I repeated the test after recalibration and modeling the stage at sq preset the largest error I got was 0.1 microns at 0deg, 45 degrees and 90 degrees move directions when moving 100 microns away. The tecnai lens normalization is already set to normalize proj and obj between LM to HM.
Thank you.
Bob

The problem was definitely the modeled stage calibration. It was made worse because the hole targeting was done in queueing mode making the centering of the square image used for targeting move further and further from center eventually to the point where the grid bar was in the center and the holes were not visible. Thanks.
Bob

Glad to know it is solvable.