Calibration deviation…0.072, 0.062, why?



  • I apologize in advance for the length of this, I'm not sure if my expectations are realistic it, if I'm not giving the motor enough "juice" or if something else is the cause to my lack of repeatability and accuracy right now.

    I'm using 40mm 1.8 steppers and my currents are set at 1200 @ with a 12v PC 600w power supply. I have a KosselMini and I'm using 1.8° Nema 17 with 16 tooth pulleys and GT2 belt @ 16x microstep should get me 100 steps per mm, correct me if I'm wrong or miss something. I am calibrating manually (with no probe but autocalibrate) and use a one inch offset where I use my precision 1.0000" spacer to touch off of. It seems as though moving in the negative Z direction moves a different amount than when moved in the positive Z direction (if this is true with Z, it would not surprise me If all movements are slightly off).

    Stacking two precision spacers on top of each other and touching off at 1.9998-2.0000" (taking into account tolerance stacking as much as I can without being at the machine shop) produces near zero error (very similar calibration) which makes me believe the movements are correct and the motor may not have enough torque.

    Are my expectations unrealistic? 1.8° Nema 17 using 16 tooth pulley with GT2 belt, precision linear rails, optical end stops, machined horizontal rails making a pretty darn precise triangle for the top and bottom (assembled on granite because granite is machined and lapped) and my diagonal rods are carbon fiber and are within a few thousandths of an inch of each other. I measured the distance between centers by by using gage pins (precision pins to verify that a hole is a specific diameter, very precise) and measuring with a certified 12" Fowler dial caliper. I know my rod lengths are within a few thousandths of an inch of being correct if I don't have the exact number, I'm extremely close. My inclination is to keep my diagonal rod length fixed.

    I switched motors because the set/grub screw on the pulleys on my other XYZ motors kept loosening then sliding on the motor spindle until the belts wore against the bed and ruined the print. Put pulleys on these motors and tightened the set screws as hard as I could.

    Should the auto calibrate deviation get to 0.000 consistently given what I believe to be a pretty accurate machine? I've gotten the deviation down below 0.060 but I am having a hard time keeping it there now since I've changed the motors (which are slightly different motors). I'm being picky about this because I am trying to print parts for another printer and I wanted to print at a later height of 0.06mm hoping for high enough part accuracy where I don't really have to post process the part.

    Is this realistic?? I think I might be pushing what these particular motors can do but don't want to just crank up the current without getting a second opinion. I know I'm probably pushing it at the later height I am trying to shoot for but theoretically, I should be able to hit it and it should be repeatable if my math is correct and I do not have any overtravel issues.

    I really want some opinions, it feels like I'm so close to this and I am thinking I should just dial up the current and try it, but I wasn't to know if I'm missing something here. Drivers are rated up to 2.0 amps without really worrying about cooling the chips right? Do I just need more current? 1.5A? :]

    Thank you in advance, it's frustrating because I feel like I should be able to do what I want to do and I am not entirely sure where my problem with this machine is.



  • I can't see anything about your bed? Is it granite? Dimensions?

    What is your frame made from and what are it's dimensions?

    You my have precisely assembled a frame but if it's components are not rigid enough this might account for the calibration.

    Set motor current to 75% of rated maximum and forget about it



  • Run an auto-calibration with the # of factors set to S-1 (instead of S6 or whatever). That will tell you your probe heights at each point. Do that a few times, see if they're steady, or shift around. If they're steady, you should note some will be further out of plane than others, this is what pushes your deviation up. That may give you a better picture of what's going on. These gremlins can be tough to chase down, but just looking at the deviation alone is only looking at a small amount of the data.



  • @DjDemonD:

    I can't see anything about your bed? Is it granite? Dimensions?

    What is your frame made from and what are it's dimensions?

    You my have precisely assembled a frame but if it's components are not rigid enough this might account for the calibration.

    Set motor current to 75% of rated maximum and forget about it

    I have robotdigg 2040 corners and a bunch of 2020 shimmed and fixed together. My bed is a 220 mm dia pcb heater with a 2 mm (I'm guessing) aluminum heat spreader screwed directly to the base triangle extrusions which is flush with the the corners. I have a difficult time believing rigidity is much of an issue for me.

    https://postimg.org/image/s0g1ri6hn/

    https://postimg.org/image/5v0jeuiq3/

    https://postimg.org/image/yxetht2gr/

    Ignore the zip ties ? it's a work in progress ?

    Idk the motor number or the max rated current hahaha that's why I was asking for a recommended current, I don't want to take the bed off to check ? okay I heard 1.5A hahaha



  • @kraegar:

    Run an auto-calibration with the # of factors set to S-1 (instead of S6 or whatever). That will tell you your probe heights at each point. Do that a few times, see if they're steady, or shift around. If they're steady, you should note some will be further out of plane than others, this is what pushes your deviation up. That may give you a better picture of what's going on. These gremlins can be tough to chase down, but just looking at the deviation alone is only looking at a small amount of the data.

    Okay, I'll do just that. I almost suspect a bit of effector tilt but I haven't reprinted DJdemon's mount yet for my Kossel with the robotdigg metal effector. I'm thinking if I can get a good enough calibration to make a few parts then I'll be set.



  • I'd be suspicious of the bed not being very flat my bed is 6mm cast aluminium machined tooling plate and a 2040 frame. I get 0.004 deviation using smart effector.

    Rolled aluminium especially 2mm will deform when heated.



  • @DjDemonD:

    I'd be suspicious of the bed not being very flat my bed is 6mm cast aluminium machined tooling plate and a 2040 frame. I get 0.004 deviation using smart effector.

    Rolled aluminium especially 2mm will deform when heated.

    Oh I know the bed isn't particularly flat, got a new bed on order but hoping bed mesh compensation could do a good enough job. I think effector tilt may or may not be responsible for some of the error. I'm thinking of machining my own bed and using a heated chamber…I'll do it on my CoreXY instead though I don't feel like making an enclosure for this Kossel.



  • I don't know if I've ever seen a belt driven actuator that specifies a better repeatability range than 0.05 mm. I think this is a limitation of belt drives of any type. I think your machine is operating as well as it could be expected to. I think you'll be able to print small layers just fine. A little inaccuracy doesn't preclude high resolution 😛



  • @NoSkillzEngineer:

    @DjDemonD:

    I'd be suspicious of the bed not being very flat my bed is 6mm cast aluminium machined tooling plate and a 2040 frame. I get 0.004 deviation using smart effector.

    Rolled aluminium especially 2mm will deform when heated.

    Oh I know the bed isn't particularly flat, got a new bed on order but hoping bed mesh compensation could do a good enough job. I think effector tilt may or may not be responsible for some of the error. I'm thinking of machining my own bed and using a heated chamber…I'll do it on my CoreXY instead though I don't feel like making an enclosure for this Kossel.

    Thing is, the calibration routine presumes the bed is flat, although it can be tilted if using 8 factor - S8. So to achieve a really low calibration deviation you need to be calibrating onto a flat bed. If you aren't then any level of precision in your build will not give you single micron deviation values. Grid/mesh compensation can reliably enable full bed printing.



  • @DjDemonD:

    @NoSkillzEngineer:

    @DjDemonD:

    I'd be suspicious of the bed not being very flat my bed is 6mm cast aluminium machined tooling plate and a 2040 frame. I get 0.004 deviation using smart effector.

    Rolled aluminium especially 2mm will deform when heated.

    Oh I know the bed isn't particularly flat, got a new bed on order but hoping bed mesh compensation could do a good enough job. I think effector tilt may or may not be responsible for some of the error. I'm thinking of machining my own bed and using a heated chamber…I'll do it on my CoreXY instead though I don't feel like making an enclosure for this Kossel.

    Thing is, the calibration routine presumes the bed is flat, although it can be tilted if using 8 factor - S8. So to achieve a really low calibration deviation you need to be calibrating onto a flat bed. If you aren't then any level of precision in your build will not give you single micron deviation values. Grid/mesh compensation can reliably enable full bed printing.

    Good news! I got the new bed in which appears to be much flatter and much thicker Al heat spreader! I still have to bore the holes out for 5mm screws but it should product a better result later tonight.

    It is worth noting that I have been cheap and have not changed my Buildtak sheet which is sorta…kinda...super marred haha so....yeah. I have made some foolish mistakes.

    Another good thing that happened is that my friend gave me a very precise mini protractor that I can use to measure effector tilt!
    quick question...how exactly is each axis defined? Does the axis start from the tower and directly outward so that they meet in the middle creating the (0,0)? This might seem like a dumb question, the reason I ask is because the digital protractor only measures one direction at a time. For better understanding, on a Cartesian machine if I was leveling the bed I could only do X + and X - at a time.

    This is the tool I was given: https://www.mscdirect.com/product/details/73896714

    It was free so I am hoping this is precise enough to do what I need it to do.

    So my question is: Can I use this to measure each axis individually for effector tilt well enough? It is worth noting that I would have to remove the hot end and place it directly on the effector and stack my 1" spacers on each other to use as a probe surface against the underside of the effector, taking into account the minor tolerance stacking of my spacers. I would calibrate it, measure the length of the hot end, then install the tool head and reset the zero verifying by measuring the offset with calipers.

    I was thinking eliminating the tool head might make measuring the effector tilt easier since each spacer is 1" tall and is about 1.651" in diameter. The Dia doesn't matter but it is wide enough to tell whether the effector is flat against it or not.


 

Looks like your connection to Duet3D was lost, please wait while we try to reconnect.