[gerthworm]

One other consideration we had this year:


1) CAN enables you to do a lot more (we especially like the current draw measurement of the SRX's), but with a daisy-chained architecture could also be a liability - if your CAN cable comes unplugged from the RIO, your whole robot is dead.

Our general philosophy of recent year(s) has been PWM-style controllers for drivetrain (where you need lots of raw super reliable power) and SRX/CAN controllers for things like shooter motors or arms which need finer control. Just a rule of thumb, though...

Updates after talking to our electrical lead:

1) I was mistaken, Spark controllers were not known to have an issue. SD540's did have a low-voltage interesting behavior, but that was fixed in the B revision.
2) PWM controllers are also much lower cost. In reality, that's what would drive us to use them unless there is a sensor associated with that particular mechanism.