MRSCloud is based on FID-A scripts but has been substantially modified to speed up the simulations. It also doesn’t use makebase.in but has a custom function to directly write out the .basis file from the .raw files that FID-A simulations create.
No, you shouldn’t have to. I believe that if you’re using LCModel it can internally re-interpolate the basis set to the desired resolution (not sure about FSL)
I reached out to the MRICloud folks, response pending
Just to clarify, are you sure your dataset does actually have 2080 data points, i.e., can you please share the exam card that was used to generate this data? This feels like the pre-echo zero-padding that a lot of Siemens raw data have, and I suspect you might need to cut the first 32 points to get a well-phased dataset during pre-processing.
unrelated, but I have a theory for the error that FSL threw in your initial post - did you download that basis set from the LCModel website? These are encrypted in such a way that renders them unusable outside of LCModel (but no one should use them anyway as it is unclear what sequence/scanner was used to generate them)