[sundernlightning]

Hi! First post here on these forums.

I'm controlling an XY gantry put together in a H shape, with the X axes below the higher Y axis. I'm using an Arduino Uno with a CNC shield and TMC2208 controllers.
I'm using 16x microstepping on nema 17 motors with 200steps/rev. The steppers are mounted to lead screw guided carriages. The motion is smooth and free of noise and apparent skipping of steps.

The repeatability test that I ran on my gantry involved moving a total of 90mm in 10mm incremental steps (#1...#9), repeated over 5 runs. The position of the moving gantry was measured using a laser triangulation sensor (https://www.bannerengineering.com/us...rt.803940.html)
I am much more interested in the repeatability of my system, than the accuracy. If 1mm is 1.01mm that is completely fine.

I found it interesting to see in the data that, within each run, the distance traveled (per commanded 10mm move) varied for each of the 9 moves. (data highlighted red)
The baffling thing about the data is that, between all 5 runs, the distance traveled by each numbered segment (#1..#9) was almost identical. (data highlighted green)

What can I attribute this behavior to???

See the data attached in jpg:

Thanks,
Erik

[peteeng]

Hi Eric - Steppers move from "magnet" to magnet and have a mechanical accuracy. Typical commercial motors are +/-5% per step non accumulating. You can get 2% motors. So for a 200step motor ie 1.8degs per step you have a 1.8*0.05= +/-0.09deg accuracy or "zone" that the motor will stop at, at each full step. So you can now figure out the "zone" the motor should stop in mechanically if you know the lead of the screw (then there's also backlash to consider and hysteresis which means you will get a different measurement going one way vs the other). Then you have the electronic accuracy. Microstepping is a way to sort of balance the motor between mechanical steps. This has an accuracy as well.

So there is an averaging going on as the motor moves it stretches and compresses slightly around each technically correct spot (mechanically and electronically). Over many moves it averages to correct. Cheers Peter S

[PCW_MESA]

That looks like ballscrew error to me, assuming the runs were all done over the same absolute positions...

[routalot]

May we have the chance to view the lead screws?I'm guessing that with Nema 17's involved they aren't large and may not be ballscrews.I would expect a stepper to complete one revolution when the code requires it,with or without microstepping.

[ger21]

I would do the same test with microstepping set to half or full steps.
This should eliminate errors from microstepping.

I also agree that it could be pitch error with the ballscrew.

And I'd also adjust your steps per unit, as your average travel is slightly over 10mm.

[peteeng]

Hi Erik - what is the lead of the screw and the steps /unit you have the controller set at? Peter

[sundernlightning]

Hi, thank you all for your insights. Some more information about my setup.
@peteeng, how does the nonaccumulating error manifest itself during microstepping. It is just +/-%5 or +/-2% of my my microstepped amounts?
@routalot, It is in fact not ball screws, but lead screws. Here is an image: https://photos.app.goo.gl/EPKYVGLDafqCXYg8A.
@get21, I performed the same test with microstepping changed from 16x to 4x and saw the same behavior.
@peteeng, the lead of the screw is 14mm/rev and i'm using 228.571 steps/mm. Here is a link to the inexpensive drive i'm using: https://www.robotshop.com/en/panowin...iABEgIMCfD_BwE

[109jb]

Rather than get regurgitated information, just go to one of the most repsected sources when it comes to steppers

https://www.geckodrive.com/support/s...esolution.html

https://www.geckodrive.com/support/s...or-basics.html

[peteeng]

Hi Erik - I won't try to explain your machines behaviour, I will explain what should happen. Did you do this test in only one direction. You need to do it in two directions to remove hysteresis or to understand how much hysteresis is in your system. Plus if possible measure backlash to account for that as well.

So from the beginning, which I've been told was only 13.8 billion years ago... The motors have a mechanical tolerance spec but not an electronic tolerance on the usteps. This is because depending on speed, torque and all the other variables we cannot determine the exact location of where the ustep physically is. What we do know is that when the motor gets to a full step it centralises to the motor tolerance which for general spec motors is 5% of a step. So looking at the calcs attached.

1) a 14mm lead means that 200steps = 14mm and you have 228 (at 14 usteps per step 14/200*14=200) which is a big delta. ie the lead screw is poorly made, possible. Did the lead come with a tolerance or grade? Or you have a big backlash. So calibrate your screw over a large run and a small run and see what happens
2) One step is 0.070mm and has a tolerance of +/-0.0035mm
3) one ustep, if we use 14 usteps to the step is 0.005mm fuzzy and there are 14x200=2800 usteps/rev
4) 10mm is 142.857 steps so every step the error is averaged out mechanically and it's the last 0.857 steps that are electronic. If system is perfect then this is exactly 10mm all good. Max error is 1 ustep so in fact we will land somewhere between 9.995mm and 10.005mm which is what happens on your run at the first 10mm, past that it gets a little bit off. You need to work on your 228 and figure this a bit better I think and if possible include backlash compensation

5) BUT when you where at "zero" we do not know if the motor is at a full step or a ustep. If its a ustep then it takes off from an unknown point (starting error) could be +/-0.005mm then corrects at first step then gets to the remainder usteps and could be out 0.005 again so total position error could be 0.01mm in the perfect system.

So all of this is sort of academic if you are aiming at an accuracy or repeatability of 0.1mm as these figures will stand up to that as long as your mechanics are good. If you are aiming at 0.01mm then its borderline and you need to use much better mechanics and forget about trying for 0.001mm. Hope this helps & check my maths...Peter

I just noticed in the attached that I write the tolerance as +/-5degs should be +/-5%

[109jb]

1) a 14mm lead means that 200steps = 14mm and you have 228 (at 14 usteps per step 14/200*14=200) which is a big delta. ie the lead screw is poorly made, possible.

This is wrong. He has 16x microstepping which is 3200 microsteps per revolution. so, 16*200/14 = 3200/14mm = 228.5714. Exactly what he said he has it set for. I have never seen a stepper driver with a setting for 14x microstepping. So the screw is fine.

[peteeng]

Thanks for the check JB - see corrected attached. Not sure why the 14 popped into my head. Commentary same intent. I'm not sure how controllers decide on the remainder steps. In this case does it truncate to 13 or round up to 14 have to ask the programmer. Maybe its smart calculates the closest which would be 14 steps.... Peter

[routalot]

From the images the 14mm pitch looks like quite a large multiple of the diameter.With such a coarse pitch in relation to the diameter it only needs a tiny discrepancy in the amount of rotation for a positioning error to occur and the previous posts have listed or linked to explanations of what those may be.

[109jb]

Thanks for the check JB - see corrected attached. Not sure why the 14 popped into my head. Commentary same intent. I'm not sure how controllers decide on the remainder steps. In this case does it truncate to 13 or round up to 14 have to ask the programmer. Maybe its smart calculates the closest which would be 14 steps.... Peter

Not sure I fully understand your question, but this article qives a really good explanation of microstepping:

https://www.zaber.com/technical-arti...pping-tutorial

[peteeng]

Hi JB - Good article. Zaber explain things quite well. The error I discuss above is called the quantisation error. The stepper can only work in discrete steps or microsteps. So when Erik wants to move 10mm the calculation shows a decision has to be made at 142 full steps plus 13 or 14 micro steps. What Zaber say is they aim at a resolution of 0.5 usteps. So my guess is that they calculate in half usteps and truncate to last ustep so they would end up at (142*16)+13 usteps moved. Or they may round up, I'll read further to see if they mention this. Peter

[sundernlightning]

Thanks for the link explaining the quantisation error. I'll have to read up on this.

[109jb]

The quantization error is quite simply the difference between the idealized programmed position and reality. Nothing more. Nothing less. Every microsteps has an ideal position. For example on a 1.8 deg stepper at 10x microstepping, each microsteps would ideally be 0.18 deg. So at say the 4th microsteps the ideal position would be 0.72 deg, and Zaber is claiming there stuff will get it there within 1/2 if a microsteps. So 0.72 +/-0.09 deg. Basically for 10x microstepping the motor position would be the commanded position +/-0.09 deg.

A stepper driver does nothing to decide if it should move 10 microsteps or 11, or 764. It simply receives a pulse and moves by whatever microsteps increment is selected. If it receives 10 pulses it moves the motor 10 microsteps.

The CNC controller like mach3, LinuxCNC, uccnc, etc is what has to decide how many step pulses to send to the driver. For example if each microsteps is a theoretical 0.001" and the user programs a move of 0.0005", it is the controller that decides if a pulse should or should not be sent. It is the controller that decides to round up or round down, or whatever.

- - - Updated - - -

The quantization error is quite simply the difference between the idealized programmed position and reality. Nothing more. Nothing less. Every microsteps has an ideal position. For example on a 1.8 deg stepper at 10x microstepping, each microsteps would ideally be 0.18 deg. So at say the 4th microsteps the ideal position would be 0.72 deg, and Zaber is claiming there stuff will get it there within 1/2 if a microsteps. So 0.72 +/-0.09 deg. Basically for 10x microstepping the motor position would be the commanded position +/-0.09 deg.

A stepper driver does nothing to decide if it should move 10 microsteps or 11, or 764. It simply receives a pulse and moves by whatever microsteps increment is selected. If it receives 10 pulses it moves the motor 10 microsteps.

The CNC controller like mach3, LinuxCNC, uccnc, etc is what has to decide how many step pulses to send to the driver. For example if each microsteps is a theoretical 0.001" and the user programs a move of 0.0005", it is the controller that decides if a pulse should or should not be sent. It is the controller that decides to round up or round down, or whatever.

[hallenErik]

Thanks for the link explaining the quantisation error. I'll have to read up on this.
--------------------------------------------------------------------------------------------------------------------


My stepper motor website:skysmotor.com- Machinery Design • - • Stepper Moter Development

"Necessity is the mother of invention." - Author unknown. as you type.