Hi,
Resampling a mask using volume tools gives volumes that do not have a strictly bounded 0,1 range. This can cause issues for other software (and I imagine also may lead to issues within cryosparc?
Masks that are made “fresh”, i.e. thresholded & soft-edged in vol tools, rather than just resampled, do not show this behavior:
Cheers
Oli
Hey @olibclarke – this is indeed a side effect (read: bug :)) of how we do the mask interpolation in Fourier space within Volume Tools, which can cause ringing in real-space. We’ve noted this now and will change to a real-space trilinear interpolation in the future.
Thanks for reporting!
Valentin
EDIT: I spoke too soon (apologies!). We actually DO perform real-space interpolation for masks within Volume Tools (and in all refinement jobs) as of CryoSPARC 3.3+. I was, however, able to reproduce a small negative minimum value after resampling a mask in Volume Tools. Still investigating, but we think this small negative value might an artefact of 32 bit floating point arithmetic within the linear interpolation call.
Is the same thing happening with mask/volume interpolation in refinement jobs? If so maybe checking that too! 
Hey @olibclarke – please see my edited comment above! We’ll keep you posted on the results of this investigation.
Quick update: we’ve identified the cause. These negative values do indeed manifest as a result of floating point arithmetic, but not where we originally suspected (interpolation). They are a result of an ifft(fft(mask)) operation that occurs within the Volume Tools job prior to resampling. We apply this redundant transform in order to support other (optional) filtering operations in the same code path.
Thanks for flagging this – we’ll be sure to figure out a proper fix for this in an upcoming release.
