Björn Forsberg

Thanks for the detailed report! A couple of follow-up Qs; 1. Do both runs reach the same resolution? 2. Do both runs reach the same sampling density? 3. What happens...

Looks identical from a sampling point-of-view. There's something strange about the sampling in the CPU-run though, judging from the GIFs. The sampling points at iterations 2-6 (or thereabouts) are all...

Your issue resulted in a very interesting discussion, and I believe we can offer some insight. Relion uses the initial --healpix_order to determine the most basic [tessellation](https://en.wikipedia.org/wiki/Tessellation) of the Tilt/Rot-space,...

I should perhaps also clarify and emphasize that this will only ever be an effect when you use symmetries, and more of an effect he higher the symmetry. Higher-symmetry usually...

Convergence doing its thing! Nice result. Any noticeable differences in the densities?

That confirms or at least strongly indicates that the increased density at certain points along the symmetry unit edges is due to the same excluded-area effect. I'll try to get...

These are results of a C1-run on something with high symmetry, right? P.S. It much easier to refer to your observations if you make it a numbered list rather than...

The "drift" you mention (observation 2) is intended measure which RELION does actively to avoid issues arising from the finite sampling grid and lingering bias. It's a perturbation which you...

I'm not sure there is a "problem", at least not a major one. It might be a problem for what you are specifically trying to do, if you are trying...

Unless I'm mistaken, you can simply change which axis the C-symmetry is defined around by changing line 613 in [symmetries.cpp](https://github.com/3dem/relion/blob/master/src/symmetries.cpp) from ``` fileContent.push_back("rot_axis " + integerToString(pgOrder) + " 0 0...