YACTLQ (Yet Anoter Closing the Loop Question ;) )

I use my cnc mainly to drill pcbs, its a moving gantry mill with steppers,g540 and a mesa 5125.

On very long programs (2000+ drills ) it tends to deviate a little from its point, been working for months to fix it but can not find anything yet that leeds to this error

It was on my plans to purchase optical encoders and magnetic linear scales to close the loop. Since those are very expensive in here (Argentina) I can currently buy one at a time(optical firts, linear scales latter this year). I have read a post from PCW that linuxcnc can be configured to improve accuracy with optical encoders alone, instead of just stopping on ferror like everybody else says.

My question is if I understood this correctly and if all of this is feasible.

Also, does anybody knows if HEDSS optical encoders are any good?

thanks to all

Before going to that extreme I would try some things to eliminate the possibility of lost steps.

First thing I would try is doubling all the step times (steplen,stepspace, dirsetup, dir hold)

It is known that the minimum g540 timing values do not work reliably

I dont have the machine turn on rigth now but I have the setttings you recommend on another post, I guess they where like four times the minimum

Is having optical encoders to improve accuracy an "extreme" ? (honestly dont see why)

Because it may well be fixing the wrong problem.
I would try to find out why and where you are losing/gaining steps or perhaps having mechanical issue before an expensive change that may not fix your actual problem.

PCW wrote:

Because it may well be fixing the wrong problem.
I would try to find out why and where you are losing/gaining steps or perhaps having mechanical issue before an expensive change that may not fix your actual problem.

Besides this, Did I undestood correctly that linuxcnc can use an optical encoder to improve accuracy? wich are the drawbacks?

I think it is possible to close a PID loop with a high resolution scale and a stepgen in velocity mode driving a high enough uStep ratio step drive and get static accuracy that is limited by the scale accuracy. This works because the PID loop can pull the step motor back and forth up the step motors displacement vs torque curve by taking uSteps.

One disadvantage of this control method compared to a normal torque mode servo is that the step motor torque curve vs displacement is not very linear.

Is having optical encoders to improve accuracy an "extreme" ? (honestly dont see why)

Cheaper is attach an standard encoder to the axis or stepper shaft.

One disadvantage of this control method compared to a normal torque mode servo is that the step motor torque curve vs displacement is not very linear.

In low RPM there is low torque for a stepper, just when it needs a higher than normal torque to compensate the deviation.

Saludos desde Argentina.
VNR

PCW wrote:

I think it is possible to close a PID loop with a high resolution scale and a stepgen in velocity mode driving a high enough uStep ratio step drive and get static accuracy that is limited by the scale accuracy. This works because the PID loop can pull the step motor back and forth up the step motors displacement vs torque curve by taking uSteps.

Are you saying that i should go for the scale first instead of the optical encoder?

PCW wrote:

One disadvantage of this control method compared to a normal torque mode servo is that the step motor torque curve vs displacement is not very linear.

I didnt understand this, maybe a practical example?

Txs a lot!

PCW wrote:

driving a high enough uStep ratio

10 ustep on the g540 are enough?