Motion corr preaverage frames

Motion Correction
1

Dear all,
is there a way to prealign a certain amount of frames beforehand? I might have saved too many fractions for motion corr to work properly (too less dose in each fraction).
Best
Dario

2

Are you referring to an EER dataset and the EER Number of Fractions parameter of an Import Movies job?

3

IME it works even with very low dose (e.g. 0.15 e-/A2/fm for 20 frames, 3e- total, it’s fine).

If it’s really not working you can try reducing the number of patches/knots for Patch Motion, or instead use MotionCor3 and use -Group to group frames before alignment.

5

OK, looks good! Those parameters are not low dose or finely fractionated IMO, in particular > 1e-/Å2/fm is actually a high per-frame dose.

Shawn (MotionCor2+ author) recommends throwing the first frame if the per-frame dose is > 1 e-/Å2, and grouping 2 frames if it is ~0.3 e-/Å2 or less. BTW 0.3e-/Å2/fm or less is required in order to get Thon rings from water. In SPA I usually shoot for ~80-100 frames ~0.3-0.6 e-/Å2/fm. For ET it’s more like 0.15e-/Å2/fm for 20 frames, because CTF estimation can be difficult at higher tilts.

With respect to dose rate for the K3 it should be ~15 e-/px/s for non-CDS and ~7e-/px/s for CDS. On FalconX it should be substantially lower, on DE Apollo it could be ~30 eps. (Based on published data, I have not used them).

4 Likes
6

Falcon 4i is good to about 12 e-/pix/s, vanilla 4 is OK up to about 10 e-/pix/s. I usually aim for 6-8 depending on mag and other conditions.

1 Like
7

It’s a falcon 3.
Do you also drop the first frame like Daniel suggested?

8

Not initially. If it jumps excessively, then I will.

9

@Dario the numbers I gave are just rules of thumb Shawn determined by testing various parameters with small ~100 image apoF datasets. @rbs_sci is right, in general you should see how your own data behaves.

1 Like