That's pretty interesting. So it has a bore for an input shaft and the clutch locks the pulley to it? With a second option of locking both the shaft and pulley to the main assembly? I did a lot of looking around when I first started this project and thought I was gonna run dual motor for turn/mill, but eventually decided to avoid that whole mess and just throw a huge servo at it lol. But at this point it's looking like I probably will add a brake.
So anyway, I finished the mill turn. got it all wired up and started testing. Things are never quite as simple as you hope they will be. Couple issues. First of all, the gt3 8m profile has more backlash than I expected. At a radius of 3 inches, it's about 10 thou. Not so good. Not your typical backlash though. Up to a certain amount of torque the belt has enough grip to allow no backlash at all, but when applying enough load to the spindle, the belt "slips" and the teeth bottom out in other direction, so this would be very hard to compensate for. After more digging, indeed the 8m profile is not as good for positioning accuracy and 5m is much better for this. Second issue, this is my first time dealing with servos so I'm learning a lot about how they operate. With average gain settings, the servo felt very spongy when grabbing onto the 5 inch pulley attached to it and applying load by hand. I could push it off position a couple degrees before it would correct and pull back to center. Running the auto tune cycle still came up with numbers that were too low to get a rigid feel. After increasing main gain and I gain up to about 75% of max, it was starting to feel pretty rigid, but can still push the pulley off position a small amount before it corrects. Increasing I gain even further helps but eventually it gets into a bad oscillation. A mill turn spindle isn't exactly a common thing for a servo to deal with. It has to hold a rock steady position while loads are being applied. Usually servos are introduced to load as they are commanded to move which is much easier for a pid loop to deal with. This is the opposite. When holding a steady position and applying load to it, the pid loop needs to see error and correct for it. It's also hard because of the lack of gearing. In my case, a single encoder count is about a half a thou of movement at a 5 inch radius. With a servo on a ballscrew, a single encoder count is 50 to 100 times less linear movement. Much easier to hold a position in that case.
So long story short, I think I'll most likely be dropping down to a 5mgt profile belt to reduce backlash and add the brake. Had to see for myself if this could be done without a brake, but it's starting to look like most likely not if I want best surface finish possible. Luckily the existing design will work with the new pulleys and it has room for the brake. They will also use the same 2012 taper lock bushings. Might even be able to return the pulleys I have.
All that being said, still very impressive to see what this mill turn can do. Wrote a program testing out my spindle swap macro. First had it do some indexing as a 4th (at insane speed lol) then ran the swap macro, had it spin up to 3k as a spindle, stop, swap back to 4th axis, do some more indexing. Very cool
Oh and I found something else really cool about this servo. It has an absolute zero position input. You can wire it up to give it an input signal that tells it to go to encoder zero. So after running as a spindle, i can call that signal in my spindle reset swap to automatically re home the 4th axis to an accurate position. No need for homing sensor