Our team is building an x-drive, and we are debating between integrated and shaft encoders. Space and ports available are not issues, just which is most accurate. We ask this, because our current robot’s integrated encoders do not go the same distances when told to go say 500 encoder counts.
Thanks!
We just transitioned one of our robots from IMEs to Shaft Encoders because we were getting too much variance from the left side to the right side of our tank drive (like 300-400 diff). This was our first year trying IMEs. We had decent luck using a shaft encoder to control our flywheel last season, so I wanted to try IMEs for drive wheels to keep inputs free this year.
I don’t know if we got a bad IME or what, but after thinking about it, measuring closer to the end of the gearbox versus the beginning seems it would be more accurate. After switching to shaft encoders tonight, we tested it out and were always within 10 ticks from left to right on our tank drive.
Quadrature encoders for sure. I have found that IMEs are far less accurate and can occasionally cause disconnects from static electricity discharges. Quadrature encoders are probably the most accurate sensor Vex makes after the potentiometer.
Actually, button and limit switches are way more accurate than quad encoders.
The reason that difference is there is actually not because shaft encoders are much more accurate (even though they are) but it’s because the ime’s have a lot more ticks per rotation than a shaft encoder.
To agree with everyone else, shaft encoders are quite a bit more accurate and they are MUCH less problem prone than the ime’s. The ime’s seem to almost always have problems 
So a IME is measuring the number of rotations of the gear. This gear goes under a torque ratio (I’m pretty sure) before the shaft in the motor is turned. This means that the white and black gear turns several times for each time that the shaft turns. Using this logic,the ime probably is more accurate unless the gear is spinning too fast to get an accurate reading. Another thing is an ime’s tic might be less degrees of rotation in the shaft then an encoder’s tic so being 100 off with an ime might actually be better than 10 with an encoder.
I guess it comes down to reliability. I loved the concept of having 8 IMEs attached to the I2S port on the cortex, but we had too many problems with getting 1 of 2 IMEs working accurately, let alone more. We might try them again for next year’s game, depending on what that is. But between Quad Encoders for our 2 tank drive and a potentiometer for our lift we are set for code knowing the status of the robot without them.