Mini Differential IR sensor on BuildTak issue

  • Following my issues with the 5mm Glass I installed a BuildTak black surface and managed to get the sensor to trig with an acceptable ~2mm distance

    Things went well and I was able to print a few Benchy and so on.
    Next on my tasklist is printing the panel due case with e3d edge, which as far as I understand is some kind of PET which shall stick like crazy.
    Well it doesn't really in the few test I did and if it did slightly it was only in very few areas of the contour… never managed to do much because after 1 turn my nozzle has attracted so much material that I have to pause and clean.

    So I wanted to try make sure my bed was flat and noticed the middle point was a bit stange. I then decided to triple each point with 5 or 10mm shift for the 2 repetitions
    For the middle point I actually did 4 extra points north south east west.
    I looked at the buildtak very closely and there is absolutely no damage or dirt in the center spot and I am pretty sure I could feel or see a 0.5mm dip or hill quite easily

    Bed probe heights:
    -0.079 -0.068 -0.066
    -0.163 -0.208 -0.090
    -0.073 -0.365 0.029
    -0.041 -0.200 -0.092
    -0.052 -0.097 -0.090
    -0.030 -0.210 -0.072
    -0.060 -0.133 -0.064
    -0.086 0.369 -0.001
    -0.099 -0.089 0.131
    -0.558 -0.238 -0.123 -0.060 -0.149

    again just for fun:
    0.032 0.037 0.037
    -0.030 -0.070 0.062
    0.063 -0.198 0.176
    0.043 -0.072 0.002
    0.062 -0.001 0.017
    0.069 -0.100 0.032
    0.039 -0.008 0.039
    0.093 0.438 0.067
    0.011 0.000 0.228
    -0.493 -0.141 -0.001 0.039 -0.059

    I noticed that the material is slightly transparent to the eye. So is it possible that it is semi transparent to IR and (occasionally) getting a second reflection from the underlying glass? Unfortunately this doesnt explain the other strange point which is positive. meeeehhh
    Anyway I think that with 32 points these 2-3 stange points will not have too much weight in the calculation. I'd still be interested to know if other people did measurement on Buildtak and got similar results. I guess my recomendation would be if you see that you have 2-3 outliers in the 10 point calibration you might want to make use of the new feature of the last firmware (thx for the 64 points!).

    Last question, because maybe I am doint it wrong. So I know babystepping is on the todo list. But in the meanwhile what I do is adjusting the
    G31 X0 Y35 Z0.2 P500
    in my config.g and obviously I have to rehome and autocalibrate. Is there a faster way?

  • administrators

    Those figures do look odd. What probing speed are you using? It's the F parameter in the M558 command.

    You can fine adjust the Z=0 height using G92. For example, if the nozzle is at 0.2mm and you want to lower tne Z=0 plane by 0.05mm, send G92 Z0.25.

  • I am not sure because my M558 command doesn't specify it:
    M558 P1 X0 Y0 Z0 H2
    so it must be the default for firmware 1.15c
    It is quite slow I would say 1-2mm/second looking at it and counting 3seconds plunging time for about 4-5mm depth.
    thank you for the G92.

    update: added F60 which slowed down the calibration a bit still acceptable if you don't have to run it 10 times…
    Anyway I was able to print the panel due case overnight (not perfect but still ok after 2-3 false starts and messing with Z height)
    and now when I

    -0.047 0.009 -0.043
    0.004 0.011 -0.026
    0.178 0.196 0.268
    0.032 0.111 0.060
    -0.009 -0.064 -0.041
    -0.137 0.041 0.014
    -0.327 -0.311 -0.020
    -0.038 -0.028 0.261
    -0.053 -0.132 -0.011
    0.080 -0.268 -4.601 -1.808 -1.718

    I removed the bed and it is certainly rotated since last time but the problem has gone a bit worse due to the rotation it is hard to say if it has spread actually.

    my "new theory" (yes I can have many per days) is that the buildtak top surface gets smoother with use ( only a few prints so far) and that changes the IR reflection.
    I think I shall have taken the white one because black certainly has less pigment (you can see through) and the top surface would reflect less light.
    But those number are too much off for that…

    update 2: removed the 3 culprit points => bed is now perfectly flat and printing.

  • I wonder if having a "probing target" with a known offset that is placed on the bed for probing would solve any sort of variation in beds – sure, it won't be good for the "autocal before every print" type, but I don't think delta printers need to calibrate anything but the homed height before a print. The full calibration could be done with the target on the bed, and then before every print the printer just gets the homed height from the bed (if the bed will provide even one, reliable probe point).

  • the probing target would have to be moved on the bed manually I guess but that could work.
    The problem with such a solution is that you lose the advantage that the autocalibration provides for positional drift (tilt) of the bed with time. And for a delta that could be the bed, a column or anything.
    I am pretty sure now that the white buildtak would actually work much better so I am going to order some and report here.

  • Update on IR mini differential. I just started to print after running a few calibrations on WHITE BUILDTAK

    second calibration results
     -0.040 -0.026 -0.037
    -0.081 -0.067 -0.028
    0.031 0.062 0
    0.021 -0.047 -0.053
    0.018 -0.01 0.037
    -0.02 -0.087 -0.081
    -0.009 0.008 -0.04
    -0.072 -0.038 -0.061
    0.022 -0.029 0.022
    -0.008 -0.042 -0.027 -0.032 0.01

    bed check:
    0.027 -0.021 0
    0.042 -0.009 -0.021
    0.069 -0.021 -0.011
    0.003 0.04 -0.007
    0.013 -0.018 -0.012
    0.021 -0.016 -0.008
    0.018 0.041 0.004
    -0.036 -0.039 0.009
    0.058 0.001 0.006
    0.039 -0.013 0.016 -0.011 0.04

    it's night and day… essentially my first layer is just perfect now the deviation is under 30microns... and as for my impression about the filament I tried the ngen (copolyester) vs edge (pet+) and the first is so much easier to work with perfect layers with default setting and no fideling.

    so the answer is WHITE BUILDTAK

    Also a friend lent me a seek thermal camera so I took a few shots and a thermal movie of the bed and duet in action.

  • administrators

    Really useful info - I like the look of the white buildtak as a surface as well, it will be interesting to hear how it holds up over time.

  • I will report on that. It definetly doesn't like nozzle crashing into it.
    I upped the temperature with edge filament to 240°C and I am doing a print tonight first layer adhesion is actually better ( I used too low temperatures due to some wrong indication on the e3d sticker spool box and also because I was lacking experience)
    I still get a lot of goo around the nozzle which I need to babysit clean every 5-10 layer or so… I wonder what is wrong in my slicer settings. The ngen (colorfab) didn't do that.
    problem with the goo is that it detaches and hardens always at the worst place and then you get a nozzle crash...

Log in to reply