Backlash Compensation with Polar Coordinates
- spumco
- Offline
- Platinum Member
Less
More
- Posts: 1829
- Thank you received: 739
10 Apr 2023 23:08 #268798
by spumco
Replied by spumco on topic Backlash Compensation with Polar Coordinates
Lots of questions...
A Big Ol' Stepper has a ton of holding torque at zero speed, yes. Often more so than an equivalent sized servo does at zero speed. And steppers don't (generally) dither.
But most steppers aren't as accurate as servos for rotational positioning - including closed-loop steppers. Aside from some specialty motors, the ones you're likely to get are typically 1000-line (4000 count post-quad). Whereas modern servos are 17-bit and up resolution.
Internet Commando time: I've read, but don't have direct knowledge (other than observation of my own stuff) that beyond 8x or maybe 10x microstepping that steppers are more 'ish' when it comes to positional accuracy. Like you command a move to 11 microsteps between two full steps, you're going to get 'about' 11 microsteps. Maybe 8 to 13. And if the encoder resolution is only 4000ppr, you aren't going to get anything better than that.
4000/200 steps = 20x microsteps max if you set it to alarm at 1 encoder count (unrealistic). Half that is pushing it. So you set it to 10x, and use a 50:1 or 100:1 reducer so you don't care about 3-4 counts at the motor - that's maybe a 'tenth or two at a 4" radius.
I like steppers. I have them on all my linear axes on my mill (and my lathe). But I went with a servo for my 4th axis specifically because I wanted better rotational repeatability and homing to the servo index.
Next...
Back-driving a worm-gear - not likely. There may be compliance, but if there's a live stepper or servo attached you aren't going to back-drive a 40:1 or 80:1 worm gear. But if there's slop between the worm and wheel you'll certainly back-drive through the backlash without a brake on the wheel.
The above does not include roller-cam (globoidal) drives. I think those have low enough friction that they can be back-driven even at high ratios and pre-loaded cam followers. Same-same for harmonic, cycloidal, and planetary reducers.
As to the backlash vs stepper... even if the stepper is holding like a rock, any backlash and compliance in the undefined 'gearbox' you mentioned will permit the part to move under machining forces.
So... if you don't need continuous machining, how about a manual 4th and 5th indexer? You can fabricate a couple rotating tables that lock with a through-bolt (and maybe have Belleville springs to pop it open for indexing). And the indexing could be done with an array of pockets & ball bearings pressed in the pockets. Loosen the draw bolt, rotate to the desired angle, tighten it up...Profit.
I was thinking about this some today. Maybe a conical/tapered boss in a matching pocket, with small holes drilled across the mating line. Glue or press precision dowels on the face of the bottom female cone and matching grooves in the upper half lock them rotationally when they're pulled together. The cones act as a taper and locate the movable part so you don't need a sliding surface (like with a cylindrical boss) for concentricity.
Like a poor-mans Hirth joint. Except with a built-in taper to align everything nicely (but now that I think about it, Hirth joints are tapered and should be self-aligning...duh). You could fit a large number of small dowels around the circle and be able to index to a pretty fine resolution - not just every 10 degrees or so. And if you're lucky any machining errors will be averaged out (or reduced) by the large number of pins & grooves.
I'm tempted to build something like that to bolt to my 4th...
Or... you could buy one of those ebay or aliexpress "Harmonic Gearbox 5th Axis" things. About $500, and they have actual harmonic drives in them (not fake with belts or planetaries). The trunnion version looks to be about $800. Already comes with steppers.
Give it a shot, and re-make some parts out of steel if it's too flexible. What's your time worth? You could be a 5-axis Ninja in the time it takes for shipping instead of going in to a mechanical design decision-hell for months.
A Big Ol' Stepper has a ton of holding torque at zero speed, yes. Often more so than an equivalent sized servo does at zero speed. And steppers don't (generally) dither.
But most steppers aren't as accurate as servos for rotational positioning - including closed-loop steppers. Aside from some specialty motors, the ones you're likely to get are typically 1000-line (4000 count post-quad). Whereas modern servos are 17-bit and up resolution.
Internet Commando time: I've read, but don't have direct knowledge (other than observation of my own stuff) that beyond 8x or maybe 10x microstepping that steppers are more 'ish' when it comes to positional accuracy. Like you command a move to 11 microsteps between two full steps, you're going to get 'about' 11 microsteps. Maybe 8 to 13. And if the encoder resolution is only 4000ppr, you aren't going to get anything better than that.
4000/200 steps = 20x microsteps max if you set it to alarm at 1 encoder count (unrealistic). Half that is pushing it. So you set it to 10x, and use a 50:1 or 100:1 reducer so you don't care about 3-4 counts at the motor - that's maybe a 'tenth or two at a 4" radius.
I like steppers. I have them on all my linear axes on my mill (and my lathe). But I went with a servo for my 4th axis specifically because I wanted better rotational repeatability and homing to the servo index.
Next...
Back-driving a worm-gear - not likely. There may be compliance, but if there's a live stepper or servo attached you aren't going to back-drive a 40:1 or 80:1 worm gear. But if there's slop between the worm and wheel you'll certainly back-drive through the backlash without a brake on the wheel.
The above does not include roller-cam (globoidal) drives. I think those have low enough friction that they can be back-driven even at high ratios and pre-loaded cam followers. Same-same for harmonic, cycloidal, and planetary reducers.
As to the backlash vs stepper... even if the stepper is holding like a rock, any backlash and compliance in the undefined 'gearbox' you mentioned will permit the part to move under machining forces.
So... if you don't need continuous machining, how about a manual 4th and 5th indexer? You can fabricate a couple rotating tables that lock with a through-bolt (and maybe have Belleville springs to pop it open for indexing). And the indexing could be done with an array of pockets & ball bearings pressed in the pockets. Loosen the draw bolt, rotate to the desired angle, tighten it up...Profit.
I was thinking about this some today. Maybe a conical/tapered boss in a matching pocket, with small holes drilled across the mating line. Glue or press precision dowels on the face of the bottom female cone and matching grooves in the upper half lock them rotationally when they're pulled together. The cones act as a taper and locate the movable part so you don't need a sliding surface (like with a cylindrical boss) for concentricity.
Like a poor-mans Hirth joint. Except with a built-in taper to align everything nicely (but now that I think about it, Hirth joints are tapered and should be self-aligning...duh). You could fit a large number of small dowels around the circle and be able to index to a pretty fine resolution - not just every 10 degrees or so. And if you're lucky any machining errors will be averaged out (or reduced) by the large number of pins & grooves.
I'm tempted to build something like that to bolt to my 4th...
Or... you could buy one of those ebay or aliexpress "Harmonic Gearbox 5th Axis" things. About $500, and they have actual harmonic drives in them (not fake with belts or planetaries). The trunnion version looks to be about $800. Already comes with steppers.
Give it a shot, and re-make some parts out of steel if it's too flexible. What's your time worth? You could be a 5-axis Ninja in the time it takes for shipping instead of going in to a mechanical design decision-hell for months.
The following user(s) said Thank You: CtodLinuxcnc
Please Log in or Create an account to join the conversation.
- spumco
- Offline
- Platinum Member
Less
More
- Posts: 1829
- Thank you received: 739
10 Apr 2023 23:23 #268799
by spumco
Replied by spumco on topic Backlash Compensation with Polar Coordinates
Holy (*&%*&^!@ - I just checked Aliexpress and you can get a brand-new 25-series through-hole harmonic reducer for $250 shipped.
It seems prices have come down since HDS' patent ran out. No more ebay shopping for me...
It seems prices have come down since HDS' patent ran out. No more ebay shopping for me...
The following user(s) said Thank You: tommylight
Please Log in or Create an account to join the conversation.
- CtodLinuxcnc
- Offline
- New Member
Less
More
- Posts: 17
- Thank you received: 2
11 Apr 2023 23:57 - 11 Apr 2023 23:58 #268856
by CtodLinuxcnc
Replied by CtodLinuxcnc on topic Backlash Compensation with Polar Coordinates
Thanks for the input. Ive decided since this is my first build i will keep it simple and focus on a robust three axis. If i do a five axis i want to do it right and it appears it’s impossible without a pricey harmonic drive. There are cheaper drives available but I’m finding their holding torque to be below what I’m looking for.
I’m writing towards launching a side hustle machining aluminum parts for people so if that goes well i will reinvest in myself.
Much appreciated,
Christian
For now i am going with closed steppers as that’s what my budget permits.
I’m writing towards launching a side hustle machining aluminum parts for people so if that goes well i will reinvest in myself.
Much appreciated,
Christian
For now i am going with closed steppers as that’s what my budget permits.
Last edit: 11 Apr 2023 23:58 by CtodLinuxcnc. Reason: Typo
Please Log in or Create an account to join the conversation.
- CtodLinuxcnc
- Offline
- New Member
Less
More
- Posts: 17
- Thank you received: 2
12 Apr 2023 00:00 #268857
by CtodLinuxcnc
Replied by CtodLinuxcnc on topic Backlash Compensation with Polar Coordinates
If you go through with it let me know how it works out
Please Log in or Create an account to join the conversation.
Time to create page: 0.057 seconds