Our shooting mechanism rotates up and down on a hex shaft driven by a hex gear on the shaft. Picture here: http://imgur.com/S4qvkzd. While having a good amount of slop between shaft and gear is alright for gearboxes, it makes our design difficult, as the arm can rotate by a few degrees without moving the gear. This presents a problem when using a PID loop to control the angle, as getting to the position quickly or driving while maintaining the angle causes the arm to bounce up and down fairly rapidly. We’ve tuned the PID loop such that the effect isn’t terrible, the there are still ranges in the motion of the arm that bounce a lot. Does anyone have an idea as to what we could do to the gear or the shaft that would (robustly) remove the slop between the two? Tell me if any more specific information would help.
I would use a shim between the gear and the shaft. Probably won’t need more than a couple thou to close up the gaps.
If it’s wedged in there pretty solidly, it’ll be difficult to get it out (or, for that matter, to get the gear off the shaft).
If possible, you should bolt the gear directly to the arm while putting your encoder on the last stage pinion. This will minimize the effect of the slop in the hex bores, while also getting the most accurate arm position.
There are ways to mechanically reduce the play, like pinning the connection, shimming, and the like. However, one thing that works rather well is preloading the connection with a counterbalance spring. We do this quite often by using surgical tubing to partially support the weight of various arms. It has the added benefit of often reducing the size of the actuator needed to provide the motion.
But, if you really want to do it mechanically, use a couple #4-40 screws and attach a hex shaft collar to the face of your sprocket. Position the sprocket with the collar loose, then tighten it up. (Disclaimer: never actually done this, but seems like it ought to work.)
I like Todd’s idea better that this, but you could bearing retaining compound. The downside is you will not dissemble your mechanism without heat and maybe a press.
Controls wise: maybe a deadband and keep the D term small or zero.