 
            Thanks, Sylvain. That makes sense. I probably wouldn't have spent so much time on it if the lib gave some indication that it wasn't setting those rates.
Well, we don't really know exactly what rates are supported by the hardware or not since there isn't exactly good documentation on it ... and the lib will just pass that to the hardware as long as the range fits into the "registers". This way users can experiment ...
But if you want reliable operation you should stick to "known" working samples rates like 250k 1M 1.024M 2M 2.048M 2.4M
Cheers
Sylvain