Empty frames in movies leads to silent problems

I noticed that if a movie contains a fraction of frames that are composed of mostly zeros (probably caused by accidental beam loss), the full frame motion correction job will finish without any noticable symptoms, but the patch motion job will generate a blank micrograph.

When the movies are directly motion-corrected using the patch motion correction job, the resulting blank micrographs will lead to obviously wrong CTF parameters. (If manual curation job is run on one such movie only, it will crash. Probably due to NaN.) Of course, no particle will be picked from all-zero micrographs. Therefore the problem will not propagate to later analysis.

However, if one such movie is first full-frame motion corrected, it will generate a normal looking micrograph and normally behaving particles. The problem comes when the user does patch motion correction later and re-extracts particles based on previous results. Now there will be particles extracted from the all-zero micrographs. If such all-zero particles are directly sent to a 3D refinement job, the refinement will behave strangely (which is how i noticed this problem). 2D classification would filter these particles out though.

This might be a very rarely occuring scenario. I am posting this as a note, in case it happens to someone.

Thank you for reporting this observation.