I'm running into the same issue. For some reason, Picoscope seems to be truncating part of the last matrix; it absolutely isn't an issue with matlab opening the file. I can see this even by checking the filesize -- Tstart, Tinterval and length take 35, 38, 31 bytes respectively. Channels A and B given 2000020 samples each should take (20 + 2 + 2000020 * 4) * 2 bytes; all told the file should be 16000308 bytes; picoscope is consistently giving a file with 16000244 bytes.
For this one, the difference is a few bytes; but I have another file I need binary output from where it's off by 70ish MBs; all from that last channel.
(I don't actually think this has to do with the last channel at all -- this actually seems a lot like something bad is happening to the file handle object before it's finished flushing it's contents)
(http://www.mathworks.com/help/pdf_doc/m ... format.pdf
for where I got that math).
I'm going to e-mail links of the smaller example to the support e-mail. Running 126.96.36.199 on Win7 x64.
Also; it'd be really nice to have mat bin files of just the integer output from the ADC itself. As it is; storing a 32 bit float which is pulled from an 8 bit sample is annoying.
Have a good one,