Leginon

The fundamental tracking algorithm has not changed much in Leginon. We always run with fixed compustage model that predicts well the xy position in the low-tilt range now a day. That makes low tilt behaving almost always at good as high tilt. 3.3 does make how many low tilt xy are predicted with the original model at low tilt and adjustable. It may help with the problem you describe.

Problem of low tilt not behave usually come from backlash correction requirement of the computstage, it turns out. I have learned from different scopes that not all of them behave the same, even if the same model. We have one Krios that needs a pre-position always in the negative direction, one does not need it, probably because of the relaxation required to settle. Tecnai and Talos compustage wants it towards the untilted direction because of the weight of the cryo-stage. The first two tilts in a particular direction are especially sensitive to it. 3.3 has fei.cfg that allows a choice of the behavior rather than the one-size fit all approach that we used to have.

If you are referring taken a lower mag image when the tracking becomes bad, that is more relevant to high tilt as far as I know, and no, we don't have that, still.

You are correct. Here is the problem when I use symmetrical tilt scheme, as Wim Hagen's method. Even with smallest angle steps (1 degree), there is significant backlash (>500nm) when changing from positive to negative at beginning, then the image is shifted about 30-50 % at 3A pixel size, and will be even larger at higher mag. I tried to model the shift manually with separate directions, but the calculated shift seems never kicked in. Again, my version is v20000, a bit old.

Another observation is that when we use continuous tilt from one direction to another, we only can tilt from negative to positive directions (say -35 to 55), but not the other way around (say from 35 to -55). Our compustage actually perform better in negative angle range for high tilt, so we would like to use 35 to -55 instead.

Regards,
Jian

Then I recommend that you try out the alpha tilt correction as implemented in 3.3 pyscope/fei.py (I moved tecnai.py to fei.py). You may have to change InstrumentData entry name field for the scope to Krios and instruments.cfg to the same. Yours is one of the first Krios I worked on. I can not remember if yours is using this new name, yet.

For testing, you might be able to just replace your current pyscope on the scope side as a hack. We have done a lot of addition but not much removing.