bed leveling issues

Discussion in 'Calibration, Help, and Troubleshooting' started by metajack, Mar 31, 2019.

  1. metajack

    metajack Member

    Joined:
    Apr 16, 2016
    Messages:
    8
    Likes Received:
    10
    So ever since I first build the bigbox, I've struggled with bed leveling, but I've managed to deal with it somehow. Lately that has been to use bilinear bed leveling with a 7x7 grid, which is working ok. However, it seems no one else is having these kinds of problems, so I must be doing something wrong.

    I've been trying to find the root cause of this now that I'm trying to use the printer more.

    Here's the output from the mesh bed leveling probe. Note I have it doing 3 probes and averaging them at each point (MULTIPLE_PROBING = 3).

    Recv: 0 1 2 3 4 5 6
    Recv: 0 -0.239 -0.123 +0.059 +0.112 -0.091 -0.146 -0.339
    Recv: 1 -0.275 -0.135 -0.269 +0.184 -0.134 -0.221 -0.327
    Recv: 2 -0.304 -0.132 -0.154 -0.012 -0.155 -0.228 -0.344
    Recv: 3 -0.294 -0.195 -0.254 -0.108 -0.263 -0.257 -0.401
    Recv: 4 -0.416 -0.302 -0.343 -0.277 -0.358 -0.401 -0.527
    Recv: 5 -0.454 -0.421 -0.468 -0.425 -0.512 -0.553 -0.619
    Recv: 6 -0.577 -0.665 -0.690 -0.608 -0.707 -0.724 -0.864

    An an image (though this image is from a second run, not the same one as the above data):
    newplot.png


    Things I've done:
    • I've made sure the left and right edges of the z-bed are the same height from the floor. I measure from the floor near the lead screws to the bottom piece of acrylic on the z-bed.
    • I've recommissioned the XY axes. I loosened everything, then went all the way through commissioning.
    Is it possible that my piece of glass is warped? Is there something else I'm not doing that I should?
     
  2. Old_Tafr

    Old_Tafr Well-Known Member

    Joined:
    Mar 7, 2016
    Messages:
    721
    Likes Received:
    92
    Is the new "bilinear bed levelling" consistent?
    Does the built in to Rumba (9 point?) work? But is or is not consistent
    Assumed that both levelling systems work independently of each other?
    Do you have a coating on the glass and do the levelling before or after coating, e.g. Uhu glue is quite a ridged layer (even with my patented tooth brush application method ! ) and isn't xxx with sliding a piece of paper under the nozzle (Rumba 9 point method) Substitute your own word for "xxx"

    You don't mention what the original problems are/were with bed levelling, and you need to mention what spec it is, e.g. Rumba etc.

    If the mechanical build is sort of out of kilter but nothing is loose then I would expect some consistency.

    If there is a loose or misaligned part (e.g coupling at the Z motors) then I would expect some inconsistency which was not necessarily repeatable

    If you turn the glass 180o does the contour picture turn 180o too? You could also flip it to use the underside as the top surface.

    Order another piece of glass. always a good investment.

    Another way to check how flat it is would be some engineers blue and a piece of plate glass, (usually flat within less than a thou) see your local engineer on how to check a surface.
     
  3. Ephemeris

    Ephemeris Well-Known Member

    Joined:
    Nov 25, 2015
    Messages:
    388
    Likes Received:
    241
    Old-Tafr's suggestions are all good.

    Remember what you are testing with the probe is the distance from the X-guide rods to the top of the bed. Because 8mm rods are a tad undersized for our XY system, rod sag will tend to warp the height map which will show the middle of the bed higher (closer to the carriage) than the edges of the bed. If your bed was perfectly flat, the height map would basically show nothing but rod sag.

    You should manually level based on that instead of the distance to the lower support (which for various reasons is likely to me less flat than the X-guide rods are straight.)
     
  4. metajack

    metajack Member

    Joined:
    Apr 16, 2016
    Messages:
    8
    Likes Received:
    10
    This is a stock BigBox Pro except that I've put a BuildTak flex plate system on the glass and updated the firmware to Marlin 1.1.9.

    The leveling seems relatively consistent. Always higher in the front than the back, and the left higher than the right. There's a hump going along the yaxis in the center of the plate.

    When I first built and commissioned the BigBox, the main issue was the nozzle getting too close to the glass. I tried adding some paper under the 0,0 spot to give the IR sensor something more consistent to read. I don't think I got any consistent prints at that time.

    At some point I tried upgrading to some firmware that was posted here that had mesh bed leveling enabled (what you call the rumba 9x9), and would do the leveling semi-manually with the piece of paper between the glass and the nozzle. That worked alright and I was able to complete some prints finally on painters tape affixed to the bed. The issue now was that every 2-3 prints it would just misread the Z height or something and drag the nozzle through the tape destroying it. Then I'd have to relevel the bed, retape it, and try again. So while it worked, it was frustrating and I basically stopped trying to print anything.

    Now that the BuildTak flex plate is on the glass, I can use the IR sensor for auto bed leveling, and I've not had any issues with it randomizing my z-height at the start of prints. I've managed to print a number of things now pretty well, but as I'm getting more ambitious about how much of the build plate is used, I'm running into problems. I've been trying to solve them with finer and finer leveling meshes, but this seems a bit silly.

    I turned the BuildTak removable piece 180 degrees and saw some of the features swap sides. I then turned the glass 180 degrees as suggested. This resulted in the overall profile not changing. It's till high in front and lowest in the rear right. Here's the data and image from rotating the glass plate and the buildtak surface both 180:

    Recv: Bilinear Leveling Grid:
    Recv: 0 1 2 3 4 5 6
    Recv: 0 -0.035 +0.023 +0.023 +0.090 -0.098 -0.153 -0.238
    Recv: 1 -0.087 +0.065 +0.068 -0.049 -0.030 -0.112 -0.292
    Recv: 2 -0.099 +0.032 -0.012 -0.017 -0.078 -0.141 -0.292
    Recv: 3 -0.052 -0.012 -0.106 -0.023 -0.059 -0.186 -0.353
    Recv: 4 -0.174 -0.063 -0.133 -0.354 -0.225 -0.271 -0.428
    Recv: 5 -0.218 -0.241 -0.229 -0.293 -0.242 -0.170 -0.572
    Recv: 6 -0.371 -0.350 -0.429 -0.167 -0.357 -0.646 -0.731
    Recv:
    Recv: X:286.00 Y:153.00 Z:10.00 E:0.00 Count X:45760 Y:27540 Z:15150



    Screenshot 2019-03-31 12.39.01.png


    And here is the glass plate flipped over, and covered in painters tape to prevent weird IR sensor readings:

    Recv: Bilinear Leveling Grid:
    Recv: 0 1 2 3 4 5 6
    Recv: 0 -0.082 +0.137 +0.145 +0.223 +0.185 +0.070 -0.030
    Recv: 1 -0.003 +0.036 +0.051 +0.085 +0.142 -0.030 -0.146
    Recv: 2 +0.060 +0.083 +0.106 -0.010 +0.080 +0.005 -0.143
    Recv: 3 -0.114 +0.018 -0.065 -0.055 +0.014 -0.076 -0.167
    Recv: 4 -0.198 -0.056 -0.132 -0.187 -0.156 -0.197 -0.370
    Recv: 5 -0.223 -0.216 -0.297 -0.380 -0.308 -0.440 -0.579
    Recv: 6 -0.497 -0.408 -0.396 -0.467 -0.581 -0.615 -0.696
    Recv:
    Recv: X:286.00 Y:153.00 Z:10.00 E:0.00 Count X:45760 Y:27540 Z:15167

    newplot (1).png

    I can measure the left and right sides from the x-axis bars instead of the floor and adjust the lead screws to fix the tilt. I have no idea how to address the forward to rear tilt.
     
  5. Old_Tafr

    Old_Tafr Well-Known Member

    Joined:
    Mar 7, 2016
    Messages:
    721
    Likes Received:
    92
    eeeeeK lots of stuff

    I didn't realise that this BuidTak plate sat on top of the glass plate, itself on top of the heated bed.

    First, Although E3D don't do this, not on the old print farm BigBox I bought from them, I have a support under the middle of my heated bed.

    I think that the more items you build up, heated bed, glass, BuildTak the more likely you are to have flexing problems with one thing not sitting flat on another.

    As always the simplest solution is the best. I would go for just the glass on the heated bed and get another piece of glass so that you can make a comparison, at least for testing. It is fairly easy to get a flat piece of glass, not sure what the BuildTak is made from but I suspect this is harder to get flat.

    I have always had an issue about getting the bed level wrt the X and Y framework. (i.e. not wrt to the bottom of the frame) In itself the mechanical bits resulting in the bed not being "level" wrt the frame would be compensated by the levelling process (regardless of the maker of the levelling system) BUT it is simpler in the end to ensure that the frame is level 90o angles are 90o etc. Included in this that the bed is level with the X and Y which means measuring from the X rods down to the bed. Having built the BB several years ago and not exactly documenting my extra steps I'm a bit rusty on this, but what I did has make gauges of a fixed length to measure down from the rods to the bed, exact length is immaterial just that the distance all round is the same.

    I would make life simpler...

    Ensure that the frame is square and that the bed is a fixed distance to the X rods.
    Just glass on heated bed.
    Ensure that it is exactly flat on the heated bed, nothing to lift the glass at any point.
    Run the levelling and see what you get.
     
  6. gonzalo

    gonzalo Well-Known Member

    Joined:
    Mar 17, 2016
    Messages:
    255
    Likes Received:
    146
    I have the buildtak flex plate system on a FLAT cast aluminum bed and I still have some issues, I believe that the buildtak base with the embedded magnets is not perfectly flat but. That been said... that’s why we have MBL right? If the surface was truly flat then we would not have any kind of leveling concerns. What I do is make sure that all 4 corners are level to the nozzle by manually adjusting and then do a manual MBL of 16 points (4x4) and store it. My problem is that I can’t get consistent results from print to print. Either the ir probe is bad or there is a firmware or code or board issue where the mesh is not stored or retrieved correctly. The first 1 or 2 prints are spot on but the day after the nozzle will be digging into the bed or 1mm away, totally random.
     

Share This Page