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

R1+ Auto-Level not really accurate

Discussion in 'General Questions' started by Schlomo, Feb 2, 2016.

  1. Schlomo

    Schlomo Member

    Joined:
    Dec 14, 2015
    Messages:
    36
    Likes Received:
    12
    I've been successfully printing for a while now with my R1+, but have been having issues with my first layer adhesion in some cases due to the Auto-level not actually being accurate. I can auto-level my bed, but when printing a claibration square along the outside of my build platform the first layer deposit is raised and lowered. For my bed, the left side is much closer ~0.2mm while the right side is closer to 0.35-0.4mm. Yes, it was actually printing in the air on the right side at one point. By manually shimming the right side, i got the numbers closer to 0.2 on the left and 0.25-0.3 on the right, but it is still not perfect.

    Is there something in Marlin that could be affecting this? My understanding was that the Auto-level should be able to take care of this problem. Is there a limitation to the auto-level calibration after a G29 is run? Is a manual bed level even necessary?

    If pictures help i can print something and show.
     
    jim3Dbot likes this.
  2. mark tomlinson

    mark tomlinson ༼ つ ◕_ ◕ ༽つ
    Staff Member

    Joined:
    Feb 21, 2013
    Messages:
    23,912
    Likes Received:
    7,338
    The way Marlin does this it is not terrific at leveling Side-to-Side (X axis) in the case where it is initially off.
    It rather assumes this is pretty much level already.
    It does seem to take into account variations in the bed if the X is leveled, but it doesn't do a good job of leveling the X if it is very far out initially.

    Totally how Marlin does this IMHO.
     
    jim3Dbot likes this.
  3. Schlomo

    Schlomo Member

    Joined:
    Dec 14, 2015
    Messages:
    36
    Likes Received:
    12
    Is there something you would suggest that could help? Custom G code? Modification to my Firmware?

    The shim i put in was about .3mm on the right side under both magnets.

    My prints come out ok, but i end up needing to lower my first layer height to ensure i get good adhesion to my bed. But this can lead to excess presssure in my hotend and i get a very over-extruded first layer...
     
  4. mark tomlinson

    mark tomlinson ༼ つ ◕_ ◕ ༽つ
    Staff Member

    Joined:
    Feb 21, 2013
    Messages:
    23,912
    Likes Received:
    7,338
    If you are manually leveling it, then shims under the bed rail mounts is probably easier and more precise.
    There is nothing I can suggest in terms of GCode other than the Z offset.
    Usually that is enough to tweak it in (since that only applies to the first layer).\ once you have the X fairly level side-to-side.

    The M565 command in the startup GCode...
     
    jim3Dbot likes this.
  5. jim3Dbot

    jim3Dbot Active Member

    Joined:
    Jun 1, 2015
    Messages:
    246
    Likes Received:
    124
    My oH My oH My..........I just stumbled across your thread..............I have been fighting with auto-level for two 10hr days.....I have printed mostly small objects.....all okay......I have printed 2" x 8" pieces, long side parallel to the Y axis...Okay, but had to move bed center to a new location because of heater not under far edge. Thus heating the bed caused warping, glass is an excellent heat Non-Conductor....A handheld IR thermometer revealed immediate temperature drop-off on glass where the heater was not.....

    Now, I finally have a requirement, for a customer, to print ten front panels. They are about 7" x 3.5" x 0.125" thick. I decided to print with the 7" side parallel to the x axis. Two plus days later.....I have my first print. Absolutely! The auto-level software is the problem.....I ran estimated 50 to 75 prints, only about 3 or 4 loops.....checking first layer & tweaking as necessary. I am printing on low resolution, 0.3mm height.....25mm/sec speed.

    I leveled the x axis to .1mm, sheet of paper, obtained very close level......tweaked rods. After auto-level, left side would cause extruder to click, too low...right side too high....lots of extrusion, no stick'y.........tweaked again...no good.........paused print, removed rubbish, released motors, checked level, bad.......don't want to get in all the test attempts I tried, but what finally worked was:

    Placed two #6 steel flat washers under left side magnets......placing a semi-colon to the left of the G29 Start code, 'no auto-level', leveled X axis with lead screw twists, set offset .......I am now printing........I am done with auto-level...period..nada...no mas............my IR sensor is destined for homing z axis zero. Having said all that, auto-level will work on some beds but not mine, I am positive of that................Appears the best solution may be an aluminum bed, trammed correctly, great conductor of heat.........no ABL
     
    mark tomlinson likes this.
  6. jim3Dbot

    jim3Dbot Active Member

    Joined:
    Jun 1, 2015
    Messages:
    246
    Likes Received:
    124
    Mark, you fricken nailed it.............
     
    mark tomlinson likes this.
  7. mark tomlinson

    mark tomlinson ༼ つ ◕_ ◕ ༽つ
    Staff Member

    Joined:
    Feb 21, 2013
    Messages:
    23,912
    Likes Received:
    7,338
    While I added auto leveling to the two beta units we have (5 bucks plus a little time) and I use it, there is no substitution for just manually leveling the build plate.

    I get that it is easier to learn and that it can account for a non-perfectly-flat-build plate, but it is deceiving for new users who assume the printer 'will just take care of it'.

    Shipping is harsh and sometimes these thing arrive all shook up.
     
  8. WheresWaldo

    WheresWaldo Volunteer ( ͠° ͟ʖ ͡°)
    Staff Member

    Joined:
    Feb 18, 2015
    Messages:
    5,905
    Likes Received:
    3,593
    Also it is a lot easier to level Y axis first then the X axis. It isn't so much the Marlin Auto-leveling as a combination of the poorly designed Mendel style of printer plus the fact that both ends of the X axis weigh different amounts, plus the flex in the bed itself. If you can get then almost level, first front to back then side to side, marlin has less work to do to overcome all the deficiencies in this printers design and execution. If you need a better auto-level then look at Gantry style printers, such as Ultimaker and BigBox. If you trust the BigBox (= E3D) guys they depend on Marlin and it's auto-leveling system for their automated level, so you simply can't place all the blame on Marlin.
     
    jim3Dbot likes this.
  9. Schlomo

    Schlomo Member

    Joined:
    Dec 14, 2015
    Messages:
    36
    Likes Received:
    12
    Well, im glad im not the only one. Good to know there are inherent issues with the bed leveling routine, as I put too much trust in the G29, i guess.

    I will work to manually level the bed and see if i can get it very consistent. At this point, i have 0.5mm shimmed on the right back and right front sides. I'll get it all level and go from there. Just hope i dont have any slippage in my z axis...
     
    jim3Dbot likes this.
  10. WheresWaldo

    WheresWaldo Volunteer ( ͠° ͟ʖ ͡°)
    Staff Member

    Joined:
    Feb 18, 2015
    Messages:
    5,905
    Likes Received:
    3,593
    One more thing, the reason I know the printer design itself does not lend itself to accurate leveling is that I have two different manufacturers of microswitches on my Robo. One has a distinct click and the other does not. While executing every single G29 command you can hear the switch click when the head is at one end and no clicks whatsoever when the extruder is at the other end. The whole X Axis pivots on the nozzle. That is a poor design for a printer. Mendel design dates back to 2009, even the now much cheaper Prusa i3 (3rd iteration) is a much more rigid printer than the Mendel. It is not all Marlin's fault!

    Unfortunately, none of this really helps us, so the best we can do is to level Y then level X mechanically. Make it as level as you can without pressing on the build plate. The paper method works well for this and shimming the magnets will get you as level as the thickness of your shims. Then Marlin does a fairly good, although not perfect, job of maintaining that level.
     
    #10 WheresWaldo, Feb 3, 2016
    Last edited: Feb 4, 2016
    jim3Dbot likes this.
  11. jim3Dbot

    jim3Dbot Active Member

    Joined:
    Jun 1, 2015
    Messages:
    246
    Likes Received:
    124
    Thanks Waldo & Schlomo,

    Y axis is good, X considerable low left side, Yep, I guess I may have been quick on ABL, but I remember reading, over a year ago, when I purchased the robo, 'you can print at extreme angles & ABL would take care of it'. I sorta took that as gospel. No regrets with Robo or Marlin firmware, with simple changes as new lead-screws & E3D v6, the printer functions admirably. I will attempt to remove the semi-colon in the start G-code with the added shims installed, but first I gotta get those front panels printed. As always, nice hearing from a master robomeister.............
     
  12. KTMDirtFace

    KTMDirtFace Well-Known Member

    Joined:
    May 18, 2015
    Messages:
    1,247
    Likes Received:
    461
    Noticed this too, when the print head is on the left side probing.. only the left switch clicks.. the whole assembly pivots on the print head, because the right side is too heavy for it to simultaneously lift the motor on that side. When it probes the middle and the right side both switches click for me.
     
  13. mark tomlinson

    mark tomlinson ༼ つ ◕_ ◕ ༽つ
    Staff Member

    Joined:
    Feb 21, 2013
    Messages:
    23,912
    Likes Received:
    7,338
    Yep. Downside to the extruder for the probe.
    Others have come up with probes that are not the extruder (one was a manual probe board and the other uses IR--@jim3Dbot).
    That is one advantage to those.
     
  14. jim3Dbot

    jim3Dbot Active Member

    Joined:
    Jun 1, 2015
    Messages:
    246
    Likes Received:
    124
    This prevents bow'in the glass via the snozzle.........but you still need to be level......... DSC08586.JPG DSC08587.JPG DSC08588.JPG
     
    mark tomlinson and Mike Kelly like this.
  15. Schlomo

    Schlomo Member

    Joined:
    Dec 14, 2015
    Messages:
    36
    Likes Received:
    12
    So. I got mine dialed in.

    Turns out, the front 2 magnets were placed from the factory too far apart. So, even if i shimmed something up, by pressing on the opposite side it would lift up the other sides magnets so they werent touching and the side i was pinching would touch. This explains my insanity while leveling the bed. It would be too high on one side one day, and too low on that same side the next.

    I was able to use an exacto blade to open up the outside of the y carriages magnet holes to accept both magnets simultaneously.

    Second, i took a distance from the nozzle to the bed seperated by a 0.03mm feeler gauge so that the distance from each point of the bed was an actual number. I based everything off of the center of the glass being 0. So, the rear left was -0.9mm, the rear right was +0.5mm, front left was -0.5mm and front right was +0.3mm in relation to the center being 0.

    Quickly found the abs value difference between the left and right sides, found some washers that were within 0.1mm of those numbers, and put them in place. Now my bed is as good as its going to get. Each side is within 0.05mm as measured by feeler gauges, although i have a slight upward warp in the x direction as i approach the center of the glass.

    TL;DR - My magnets weren't both touching on the front side. Enlarged holes and shimmed now it works like a charm.
     
    jim3Dbot and mark tomlinson like this.
  16. mark tomlinson

    mark tomlinson ༼ つ ◕_ ◕ ༽つ
    Staff Member

    Joined:
    Feb 21, 2013
    Messages:
    23,912
    Likes Received:
    7,338
    My replacement/upgrade bed had that issue with one magnet. I totally forgot that.
     
  17. jim3Dbot

    jim3Dbot Active Member

    Joined:
    Jun 1, 2015
    Messages:
    246
    Likes Received:
    124
    Great you have it dialed in, it is quite maddening until you locate the source of the problem,,,,Good going.......One of these daze, I will print a x axis mount for a dial gauge, to do precisely what you did......I have purchased shim stock, back in the day & may revive that practice.......Take Care................jimmy
     
  18. WheresWaldo

    WheresWaldo Volunteer ( ͠° ͟ʖ ͡°)
    Staff Member

    Joined:
    Feb 18, 2015
    Messages:
    5,905
    Likes Received:
    3,593
    Marlin is quite capable of printing on a sloped bed, that is really not the issue here, it is the fact that the Z axis pivots and the bed flexes. Attempts to fix it by replacing the bed with something more rigid will only bring mixed results. The biggest contributor is the X axis design and the fact that it is not secure. Of course that is also what makes the Robo's use of the extruder kind of innovative. No offset needed from auto-level probe to nozzle, it makes for a compact setup. To be honest, I have looked at trying to fix this issue with a redesign of how X is treated. @jim3Dbot is almost 100% there, if he were to fix the location of the Z axis brass nuts to the X axis idler arms, that in conjunction with replacing those same brass nuts with anti-backlash nuts and his IR proximity sensor, he could eliminate all the rocking and one side lower than the other first layers, other than the deviations caused by bed flex.

    Just thinking out loud.
     
    jim3Dbot likes this.
  19. jim3Dbot

    jim3Dbot Active Member

    Joined:
    Jun 1, 2015
    Messages:
    246
    Likes Received:
    124
    @WheresWaldo Thanks for your input, much appreciated.......Just a little confused on a few statements?

    Fixin' the brass nuts to the X carriage, if I'm understanding correctly, what happens 'when' something goes wrong, and the nozzle buries itself deep into the glass, no x lift.

    I'm thinking the lash would come into play only during z motor reversals, which is during z homing & between print retractions. The printed adapters fit snug into the x axis idler, really no backlash there.
     
  20. jim3Dbot

    jim3Dbot Active Member

    Joined:
    Jun 1, 2015
    Messages:
    246
    Likes Received:
    124
    Short of expensive ball screws & split nut anti-backlash nuts, I have seen backlash addressed with software that momentarily sped up the motor during direction reversals to eliminate mechanical backlash...just remembered that....does not really apply here.......Hey thanks Schlomo & Waldo for interesting topic....................Take Care
     

Share This Page