Most of the Delts'a out there have universal joints in them. The tolerance / wear / wobble issues with them are well documented. Going from the fancy bearings in a LISA back to the universal joints would not improve the accuracy.
None of the Delta printers print perfect parts right out of the box. The same is true of cartesian printers. My i3 did not print perfect cubes right out of the box. If you are picky enough, it still does not print perfect cubes. Calibration is everything on any of these printers. Doing calibration takes a lot of time. There's no magic shortcuts where you spend 10 minutes and it's done. The alternative is to womp everything up on a *big* CNC so all the locations are within zero tolerance of where they should be. If you want it to *stay* right, make it all out of something that's not humidity / temperature sensitive.
The only thing that nut rotation on the shaft does is change the math. I don't see how it impacts accuracy. Coding up math is a one time thing.
On all these printers (Delta or what ever) The angles between each axis matter. They rarely get tweaked (exempt for bed level). Distances traveled per step matter, people spend a *lot* of time on that. Twists during travel matter on all of these printers, again rarely tweaked. With a cartesian printer *some* of that is a more 1:1 map of error to source.
One example on the I3:
There are two smooth rods that hold the carriage for the extruder. If they are not parallel, The carriage probably rotates as it travels. That gives you an curve error in X,Y, and Z.
On any Delta:
You need to get the hub and shoulder offsets in correctly. You need to have the right arm lengths. You need the correct lead screw distances. The names of the numbers are different on other Delta's, but the need to get them right is still there. Get any of them wrong and you get a curved error in X,Y, and Z. Zero any of the three drive shafts wrong and you have a curved error in X,Y, and Z.
Software correction:
If you *know* exactly what any of the errors are, you could put them into the motion software and correct for them. Things like twist in an i3 or bow in a smooth rod could be exciting to describe. There's enough math horsepower on a PC to deal with (and pre correct) for any of these errors. The limitations are:
1) Somebody has to do the math
2) Somebody has to write the code
3) Somebody else has to collect accurate data
4) Everything has to stay the same (no drift).
5) It's all got to be debugged and verified (QA matters ...)
Based on what I've seen, 3-6 are the real limits. If you pick a small CPU, then 2 can hit a wall. On a big printer, there's not a lot of need for a tiny CPU card. There's *lots* of CPU's out there.
Accuracy:
The same magic math that gives you speed changes as you move up and down the shaft on a Delta gives you a change in shaft steps vs head motion. If you are getting a 2:1 speed boost, you are taking a 2:1 step hit. With my 1/2" screws and 16 micro steps on a 200 step motor, the shafts give about 4 microns per step. Of course the steppers are 5% on each of the 200 steps, so that's just resolution. Accuracy wise there's about a 2 micron error just from the steppers. Even if I multiply those by two, they aren't real big.
Do I believe the rest of my plywood monster is good to 10 microns? No. If the wood moves .1 to .2 % in humidity, No real movement in temp, but a lot with humidity.
[www.performancepanels.com]
I'm not going to hit those sort of numbers. The posts will move 700 microns. Can I paint it? Yes. Will that keep out humidity - nope. It just slows things down. I've spend a lot of time testing that sort of stuff. It's a good idea, just not a complete solution. Might I make a gizmo to simplify manually measuring the distance to say 0.1 mm? Indeed I might.
Temperature is another variable. Tell me what temp you calibrated at, the temperature now, and the materials. I can make a pretty good guess in software to compensate for the first order stuff. Would I do that? - probably not. Am I going to tightly constrain my cast aluminum build plate in X and Y - no. I'd rather not encourage it to warp.
None of the Delta printers print perfect parts right out of the box. The same is true of cartesian printers. My i3 did not print perfect cubes right out of the box. If you are picky enough, it still does not print perfect cubes. Calibration is everything on any of these printers. Doing calibration takes a lot of time. There's no magic shortcuts where you spend 10 minutes and it's done. The alternative is to womp everything up on a *big* CNC so all the locations are within zero tolerance of where they should be. If you want it to *stay* right, make it all out of something that's not humidity / temperature sensitive.
The only thing that nut rotation on the shaft does is change the math. I don't see how it impacts accuracy. Coding up math is a one time thing.
On all these printers (Delta or what ever) The angles between each axis matter. They rarely get tweaked (exempt for bed level). Distances traveled per step matter, people spend a *lot* of time on that. Twists during travel matter on all of these printers, again rarely tweaked. With a cartesian printer *some* of that is a more 1:1 map of error to source.
One example on the I3:
There are two smooth rods that hold the carriage for the extruder. If they are not parallel, The carriage probably rotates as it travels. That gives you an curve error in X,Y, and Z.
On any Delta:
You need to get the hub and shoulder offsets in correctly. You need to have the right arm lengths. You need the correct lead screw distances. The names of the numbers are different on other Delta's, but the need to get them right is still there. Get any of them wrong and you get a curved error in X,Y, and Z. Zero any of the three drive shafts wrong and you have a curved error in X,Y, and Z.
Software correction:
If you *know* exactly what any of the errors are, you could put them into the motion software and correct for them. Things like twist in an i3 or bow in a smooth rod could be exciting to describe. There's enough math horsepower on a PC to deal with (and pre correct) for any of these errors. The limitations are:
1) Somebody has to do the math
2) Somebody has to write the code
3) Somebody else has to collect accurate data
4) Everything has to stay the same (no drift).
5) It's all got to be debugged and verified (QA matters ...)
Based on what I've seen, 3-6 are the real limits. If you pick a small CPU, then 2 can hit a wall. On a big printer, there's not a lot of need for a tiny CPU card. There's *lots* of CPU's out there.
Accuracy:
The same magic math that gives you speed changes as you move up and down the shaft on a Delta gives you a change in shaft steps vs head motion. If you are getting a 2:1 speed boost, you are taking a 2:1 step hit. With my 1/2" screws and 16 micro steps on a 200 step motor, the shafts give about 4 microns per step. Of course the steppers are 5% on each of the 200 steps, so that's just resolution. Accuracy wise there's about a 2 micron error just from the steppers. Even if I multiply those by two, they aren't real big.
Do I believe the rest of my plywood monster is good to 10 microns? No. If the wood moves .1 to .2 % in humidity, No real movement in temp, but a lot with humidity.
[www.performancepanels.com]
I'm not going to hit those sort of numbers. The posts will move 700 microns. Can I paint it? Yes. Will that keep out humidity - nope. It just slows things down. I've spend a lot of time testing that sort of stuff. It's a good idea, just not a complete solution. Might I make a gizmo to simplify manually measuring the distance to say 0.1 mm? Indeed I might.
Temperature is another variable. Tell me what temp you calibrated at, the temperature now, and the materials. I can make a pretty good guess in software to compensate for the first order stuff. Would I do that? - probably not. Am I going to tightly constrain my cast aluminum build plate in X and Y - no. I'd rather not encourage it to warp.