That’s a good point, we didn’t really think of that in our pipeline (we wanted to keep the data as untouched as possible) and were looking for ways to make it faster.
I get why you are suggesting 256 Hz, as it is a power of two - that being said, it might be even faster to go for 250 Hz (as we can keep 1 out of 4 time points). Is there any advantage going for 256 rather than 250 in that case?
Coming back to the problem itself, do you think it might be worth investigating? We have used the same pipeline for previous pre-processing, and I am wondering if what might have been a silent integer overflow could translate to erroneous data (typically, positive numbers becoming negative, which in the case of matrices would overwrite some of the trailing data).
Thank you !