- Joined
- Feb 8, 2014
- Messages
- 11,176
''I'm having a bit of a problem with my machine: When cutting a circle, my Y axis dimension is always 0.002 smaller than the X axis dimension and this problem is getting worse, it used to be 0.001'' difference. When cutting a rectangle, the X and Y dimensions are always correct to +/- 0.0001''. I can't find any backlash in the system, and the math that calculates the circle seems to be correct in the program. I suspect a servo tuning problem, but not sure. I need to take a deep dive into this to figure out. For the moment, I can bore the critical dimensions to the correct size in a second operation, I left them undersize to accommodate the known error.'' From https://www.hobby-machinist.com/threads/press-feeder.82699/#post-725777
Well it turns out the above statement is not quite true. I rechecked the backlash and found about 0.004'' in the X axis, but only in the middle of the travel. At about 80% travel I can only measure about 0.0006'' backlash (guess where most of the work is done). The maximum backlash I can find in the ballscrew thrust bearing is about 0.0005'' which contributes to the 0.0006'' I measured at the 80% travel. The Y axis only seems to have about 0.001'' backlash, and this seems to be in the ball screw and not the thrust bearing. The way that I have been getting away with this is because with linear scales on the table, the system automatically compensates for backlash to the degree that it can. But there is no substitute for a tight machine. On linear moves the system has time to compensate for the error, but when cutting a circle, there isn't time to completely compensate as the direction changes when cutting around the corner. Slowing down the feed rate does help and is what I may do for the very short term. I did confirm that the program is correctly calculating the trig for generating the circle, it seems to be dead on.
What is the best option to fix this problem?
The ball screw is obviously worn in the center of travel, but not so much toward the ends, so reballing with oversize balls in its current condition is not really a good option. I would wind up with way too much preload on the ends.
I can probably retune the servos to tighten up the reaction time, but at best this is a temporary band aid fix. This needs to be done anyway, so I'll do it. This will solve another small problem I'm having with surface finish.
Have the ballscrew rebuilt. ~$600, and in that case I would send in both the X and Y screws. So something in the $1000 range for both. The problem here is that the machine will be down for about 10 days, shipping + 5 days repair time. And I can't really afford to have the machine down for that long right now. Maybe finish the current project then repair the machine. This is probably the best option, and will give me some time to do some other maintenance on the machine.
Buy new ballscrews from Thomson or Hiwin. About the same price as a rebuild, but I would have to turn the ends fit the bearings, and ballscrews are hard, not an easy job. The advantage is that the machine would only be down for a day and I would buy double nuts.
Or I could add a backlash compensator, an air cylinder to preload and bias the table in one direction. Would most likely cost about $300 or so for parts. It would work, but not really a good option.
Decisions.... Decisions
Well it turns out the above statement is not quite true. I rechecked the backlash and found about 0.004'' in the X axis, but only in the middle of the travel. At about 80% travel I can only measure about 0.0006'' backlash (guess where most of the work is done). The maximum backlash I can find in the ballscrew thrust bearing is about 0.0005'' which contributes to the 0.0006'' I measured at the 80% travel. The Y axis only seems to have about 0.001'' backlash, and this seems to be in the ball screw and not the thrust bearing. The way that I have been getting away with this is because with linear scales on the table, the system automatically compensates for backlash to the degree that it can. But there is no substitute for a tight machine. On linear moves the system has time to compensate for the error, but when cutting a circle, there isn't time to completely compensate as the direction changes when cutting around the corner. Slowing down the feed rate does help and is what I may do for the very short term. I did confirm that the program is correctly calculating the trig for generating the circle, it seems to be dead on.
What is the best option to fix this problem?
The ball screw is obviously worn in the center of travel, but not so much toward the ends, so reballing with oversize balls in its current condition is not really a good option. I would wind up with way too much preload on the ends.
I can probably retune the servos to tighten up the reaction time, but at best this is a temporary band aid fix. This needs to be done anyway, so I'll do it. This will solve another small problem I'm having with surface finish.
Have the ballscrew rebuilt. ~$600, and in that case I would send in both the X and Y screws. So something in the $1000 range for both. The problem here is that the machine will be down for about 10 days, shipping + 5 days repair time. And I can't really afford to have the machine down for that long right now. Maybe finish the current project then repair the machine. This is probably the best option, and will give me some time to do some other maintenance on the machine.
Buy new ballscrews from Thomson or Hiwin. About the same price as a rebuild, but I would have to turn the ends fit the bearings, and ballscrews are hard, not an easy job. The advantage is that the machine would only be down for a day and I would buy double nuts.
Or I could add a backlash compensator, an air cylinder to preload and bias the table in one direction. Would most likely cost about $300 or so for parts. It would work, but not really a good option.
Decisions.... Decisions
