1. Got a question or need help troubleshooting? Post to the troubleshooting forum or Search the forums!

Z Axis - precision and accuracy

Discussion in 'Mods and Upgrades' started by Ziggy, Mar 25, 2014.

  1. SteveC

    SteveC Well-Known Member

    Joined:
    Jun 12, 2013
    Messages:
    894
    Likes Received:
    316
    Microstepping is a product of the stepper driver. Any problems with microstepping are going to just sit on top of that typical +/-5% step angle accuracy.

    Remember that the step angle inaccuracy is non-cumulative from step to step.

    I guess that a microstep's holding torque will also depend on the rotors magnetic field strength. Problems with holding torque will look like a quantized and fully cumulative position error. It would accumulate over all layers of the print and not be cyclic.
     
  2. nickster

    nickster Member

    Joined:
    Oct 7, 2013
    Messages:
    82
    Likes Received:
    52
    @SteveC - Yup. +/-5% Step accuracy roughly equates to +/- 1 microstep. Do you mean additive accumulate?

    On my latest hardware, I measured 0.005mm Z repeatability and accuracy, but that is like an order of magnitude greater than a microstep at 0.000441mm . Before I started mods, I could get over 0.25mm displacement by jiggling the Z screws and displacing the Znuts. That is over 500x greater than a micro step. Anybody else have measured Z numbers for mods?

    Two other sources of ribbing that are often overlooked are hobbed bolt concentricity and filament diameter variation. It must have been a heck of party when my original bolt was hobbed. The same goes for some of the eBay and Amazon replacements. The machinist was evidently thoroughly lubed and we are not talking cutting oil. I ended up hobbing my own. When the bolt teeth are not concentric or do not take a uniform bite in the filament, you can get a cyclic pattern in flow rate for every rotation of the bolt. This beats against the application rate of the print. Similar issues occur with light spring tension letting filament slip, except on the flip side, too much spring tension causes excessive bite and deformation of the filament.

    Say you have some low/typ quality filament that is running 1.75mm +/- 0.05mm diameter with uniform circular cross section. This translates to a +/-6% variation in flow rate. Say you are printing with 0.2mm LH 0.5mm wide.
    Since the Layer Height is fixed (..right) the printed layer width can vary between 0.47mm to 0.53mm. 0.06mm doesn't sound like much except consider when you print 3 perimeters with inner perimeter first. This creates a compounding effect of up to 2.5x or 0.15mm total error in perimeter thickness due to worst case filament variation.

    Back to the original example. A decrease of 0.005mm at a LH of 0.2mm means the filament width expands from 0.5mm to .5128mm or by 0.0128mm in width.
     
  3. SteveC

    SteveC Well-Known Member

    Joined:
    Jun 12, 2013
    Messages:
    894
    Likes Received:
    316
    My understanding is that that +/-5% is in the manufacturing tolerances for the rotor. It should have nothing to do with microstepping which is a result of the driver (16 microsteps/step in our case). Any correlation is a coincidence.

    Yes, additive. (An accumulator is a register fed by and adder which feeds back to the adder;) )
     
  4. nickster

    nickster Member

    Joined:
    Oct 7, 2013
    Messages:
    82
    Likes Received:
    52
    Just an observation that a source of error is the +/-5% mfg tolerance which is about the same as +/-1 microstep. People are getting hung up on microstep accuracy and resolution affecting their print quality. There are a lot of other things as talked about on this thread that have to be fixed on the machine before a microstep becomes and issue. Most Robo's are off more than an order of magnitude. Probably talking serious money for equipment just to measure a microstep of Z displacement.

    Not so sure it is cumulative. If you walk thru the tool flow, all the way to the number of steps sent to the steppers, position info is absolute until the final conversion to steps. So yes, the final Z height of the part will be limited by the precision of the metric to screw conversion factor, but that is in the noise and not what we are talking about. Marlin gets a commanded absolute Z height and it knows its current exact Z height in steps. Haven't looked at the code in a while, but Marlin calculates the number of steps required to get as close as possible to the commanded Z height. It sill knows its new exact Z height as measured in steps, so I don't think there is a cumulative effect.
     
  5. SteveC

    SteveC Well-Known Member

    Joined:
    Jun 12, 2013
    Messages:
    894
    Likes Received:
    316
    Nickster, that's all good. The only thing I was trying to point out is that any error in the stepper & driver executing a microstep command will be cumulative. If the software requests 10 steps but the stepper ends up doing 9 because of torque inadequacy or something then that missed step is forever and will add to the next missed step. I did not mean the software generation of the microstep commands which I totally agree would not be cumulative like you say unless there is a bias in that final conversion step.
     
  6. nickster

    nickster Member

    Joined:
    Oct 7, 2013
    Messages:
    82
    Likes Received:
    52
    hehehe...yes and no. You are talking about a stepper not exactly tracking the exact microstep position because of torque/friction and not about missing a step/dir at the controller.

    Let me see if I remember this. A single full step is a complete phase reversal in one of the windings. (.7*W0+ .7*W1+) (.7W0- .7W1+) (.7W0- .7W1-) (.7W0+ .7W1-) for 4 full steps. Half steps would be (W0+) (.7*W0+ .7*W1+) (W1+) (.7W0- .7W1+) (W0-) (.7W0- .7W1-) (W1-) ...

    Nice Figures and Tables that show the relationship at the end of the spec:
    http://www.pololu.com/file/0J450/a4988_DMOS_microstepping_driver_with_translator.pdf

    There is a state counter in the stepper driver that advances up and down with step/dir pulses. Micro stepping is just adjusting the current ratio between adjacent windings to give the illusion of smooth motion with sinusoidal current. You can manually deflect a stepper many micro steps worth of rotation, and it will bounce back to the correct microstep position. You could theoretically deflect 31 micro steps had have the stepper bounce back to the correct position. Without missing (two) full steps (32 microsteps), you can have up to a 31 micro step lag from friction/torque, and not have a cumulative error. With 32 microsteps of lag, the stepper is happy enough either going forward or backwards.
     
    2 people like this.
  7. John Rygg

    John Rygg Active Member

    Joined:
    Feb 21, 2013
    Messages:
    111
    Likes Received:
    31
    Ziggy thank you for your work. n
    I think we need dedicated sticky thread for x ,y, & z axis tuning !

    can you tell me where you sourced your metric threaded rod and what was the part number ?
    mine look like they are not very straight and want to replace them when I do my y axis upgrade kit

    thanks
    John
     
    2 people like this.
  8. SteveC

    SteveC Well-Known Member

    Joined:
    Jun 12, 2013
    Messages:
    894
    Likes Received:
    316
    Nickster, You are totally right. I forgot how the microstep positions are held. I probably took a course that included brushless motors but it would have been close to 30 years ago and obviously I have not used it since. Thanks!
     
  9. Lew

    Lew Member

    Joined:
    Dec 15, 2013
    Messages:
    41
    Likes Received:
    9
    Ziggy,
    I am attempting to build a copy of your AutoLeveling add-on. However, the instructions are a bit sketchy in a few respects. Could you please clarify a few points:
    1) More detail of the collar/set screw on the rod. It is not obvious how this is constructed and attached. Is this fastened to the servo lever? How?
    2) Length of the vertical rod.

    Thanks, otherwise nice work!
     
    Hermes_aile likes this.
  10. Mike Kelly

    Mike Kelly Volunteer

    Joined:
    Mar 11, 2013
    Messages:
    6,967
    Likes Received:
    2,275
  11. Lew

    Lew Member

    Joined:
    Dec 15, 2013
    Messages:
    41
    Likes Received:
    9
    Thanks, Mike, but that thread does not give any detail on the mechanical aspects of the probe rod. What I need is a really good close up photo and description on how the servo is connected to the collar and probe rod.
     
  12. Mike Kelly

    Mike Kelly Volunteer

    Joined:
    Mar 11, 2013
    Messages:
    6,967
    Likes Received:
    2,275
    It's more to keep relevant information as well contained as possible. Useful information tends to get shuffled around and lost in the mix.
     
  13. Sasabs

    Sasabs New Member

    Joined:
    Feb 12, 2015
    Messages:
    39
    Likes Received:
    4
    What are stabilizere? I would appreciate a link where I might obtain them
     
  14. Mike Kelly

    Mike Kelly Volunteer

    Joined:
    Mar 11, 2013
    Messages:
    6,967
    Likes Received:
    2,275

Share This Page