If you're interrupting on every conversion in fast-as-possible mode, you will (my prediction, but do the arithmetic) spend all your time in the ISR, since the samples are arriving faster than you can pick them up. If you're only interrupting during the pulse, and all your work during the pulse is done by the ISR, that should be workable -- save the data during the pulse, then process it in main between pulses. In this case, the amount of work you do in the ISR directly affects how many samples you can collect, so you want to minimize that. Did you implement the GPIO wiggle in the ISR I suggested? If you have a scope, that should tell you directly how long your ISR is taking.
↧