Downsampling job error - could not broadcast input array from shape (120,120) into shape (800,800)

Troubleshooting
1

Hi,

I am trying to downsample my particles from a 3d homo-refine job that were extracted at box crop 1260px / Fourier crop 560 px to 800/400 so I can get around NU-refine memory issues. I am running into this error

could not broadcast input array from shape (120,120) into shape (800,800)

suggesting it cant downsample the dimensions correctly. What am I missing?

2

Can you post a screenshot of the inputs for the downsample job?

3

Here is the input screen. The particles are after 2d/3d cleaning from homogenous refinement job from an extract job @1260/560 px binning.

image
image1457×1033 178 KB

4

I’m not sure exactly why it is dying, but I’m also not sure if these are exactly the parameters you want.

“Crop to box size” does not downsample - it reduces the box size, without adjusting the binning. So as you have it, it should first trim the particles down to 800px (from 1260px) and then downsample those 800px particles to 200px. Is that what you want?

I am not sure how Downsample handles a stack which was extracted with both an unbinned and downsampled particle blob (e.g. 1260/560) - maybe that is where problems are arising?

5

I figured cutting the box size down would be good to reduce the amount of stuff/voxels NU-refine has to process. So yes, I was aiming to trim both box and fourier crop further since the resolution may get higher as well.

I set box size to 1260 (so downsample to 1260/200) and similar error. If I don’t specify a boz size it runs…so it looks like shrinking the box size is a problem.

6

I realize since I’m hitting the resolution limit at 1260/560, so downsampling without box size reduction will just cut the resolution.

7

I figured cutting the box size down would be good to reduce the amount of stuff/voxels NU-refine has to process. So yes, I was aiming to trim both box and fourier crop further since the resolution may get higher as well.

But if you fourier crop further, you will limit the resolution - Fourier cropping to 200 from 800 will increase your pixel size 4x - so if your original pixel size was 1Å, it will now be 4 Å (limiting your acheivable resolution to 8Å). Is this what you want?

8

Based on your other thread, I would initially try just downsampling to say 1000px, and see how it goes - that will still give you a Nyquist limit of ~2.3 Å