SegRefine3DGrid: ValueError: slice step cannot be zero

[Juha Huiskonen]

Hi Carsten,

SegRefine3DGrid crashes with the following error:

File "/apps/spring/0.84.1470/lib/python2.7/site-packages/emspring-0.84.1470-py2.7.egg/spring/segment3d/refine/sr3d_diagnostics.py", line 1581, in prepare_upper_part_of_figure

    mean_out_of_plane, each_reference, pixelinfo)

  File "/apps/spring/0.84.1470/lib/python2.7/site-packages/emspring-0.84.1470-py2.7.egg/spring/segment3d/refine/sr3d_diagnostics.py", line 1049, in generate_simulated_power_from_latest_reconstruction

    projection_parameters = projection_parameters[::ten_evenly]

ValueError: slice step cannot be zero

I think I used parameters from a previous SegmentRefine3D, plus defaults from SegRefine3DGrid (attached). Any idea why this is happening? Many thanks.

Best wishes,

Juha

[Carsten Sachse]

Dear Juha,

 The error arises from a very small number of projections to be matched:

Number of projections azimuthal/out-of-plane angle = (1, 1)

 This will not give you a sensible result. To be honest, the entire parameter file does not look suitable for a proper alignment search. This will not give you meaningful results.

Number of iterations                     = 2

HR - azimuthal and out-of-plane search restraint in degrees = (0.0, 0.0)

HR - X and Y translation range in Angstrom = (0, 0)

MaxR - Maximum resolution aim            = True

MaxR - azimuthal and out-of-plane search restraint in degrees = (0.0, 0.0)

MaxR - X and Y translation range in Angstrom = (0, 0)

Absolute X and Y translation limit in Angstrom = (1, 1)

 I suspect  that you want to limit the computation time due to the expensive grid search. I would still recommend to use the default parameters for total of five cycles but limit the refinement to low and medium resolution. Five cycles are required to somewhat stabilize the refined structure. It is normally sufficient to match the layer lines. For example, if your highest resolution layer line from your sum of the power spectrum ( i.e. experimental) is located at 9 Å,  there is no need to refine at unbinned data. With the pixel size of 1.35 Angstrom, you could bin by a factor of 3 still sufficient to capture that layer line easily. Just examine the comparison between experimental and simulated spectra through your various resolution refinements and you would know which level of detail is sufficient.

 In any case, the reported error is obscure and will be fixed in the future.

Best wishes,

Carsten

[Juha Huiskonen]

Hi Carsten,

Thanks for your reply. Indeed these parameters were for testing and to speed up things. The highest resolution layer line is at 1/6.0 A so I thought I should try to include that. I will try the parameters you suggested and binning by factor of 2.

Best wishes,

Juha