Smart Effector Consistency Query



  • Hi all, been having some trouble getting my smart effector to read consistently. For some background, the SE is on a Tevo LM with duet wifi 1.03. Previously this was using the BLTouch. The BLTouch was set up with:

    M558 P9 H15 F1200 T2000 A10 R0.4 S0.015
    without issue.

    Firmware is 2.02RC4 with associated web control and wifi server 1.21.

    First up my config is attached so as not to exceed character limit.
    0_1543372058876_config.g

    Now the bed.g:
    0_1543372069515_bed.g

    The issue I am encountering is inconsistent readings. I created a repeatability test when I had the BLTouch which I used to calibrate for minimal error using the Snn parameter in M558. The macro is as below:
    0_1543372082338_macro.g

    All connections are made with 18AWG silicone wire stripped from a 9-conductor Belden cable. Crimps are tight on the conductors with some jacket deformation since the molding is rather thick. All molex XT connectors have the little J-hook "tab" massaged outwards to give more positive contact on the pin. I followed the wiki on setting up and comissioning the effector. The only step I did not perform was setting an h compensation value at every point of the bed since I did not previously need it on the same machine.

    Effector assembly procedure was as below:

    • heat heatsink away from assembly and tighten cold heatblock and throat coated in IC DIamond to produce thermal fit

    • place heatsink on bottom of effector 2 hours after cooling, place red plastic spacer on top, add red loctite and tighten top nut until no tactile play is available, then tighten 1/8 turn more.

    Heatsink fan is moved away from electronics by DJDemonD's 10mm spaced duct printed in PETG. E3d-supplied 24V fan with cabling going out the side, then angled down away from electronics going up to the connector.

    Magball arms are 400MM by Haydn oriented N-> S on all neighboring points.

    All measurements were made when the hotend is cool (no fan) and no parts fans. No heater current. All thermistors read about 20C. MCU temperature reads average 40C.

    Endstops are height set with trigger screws protruding 5mm as measured with depth pin on caliper.

    Belts before and after conversion were tightened using the thingiverse tool with the rotating knob. Tension was always set to "3" on the device.

    All tests are after several subsequent G32 runs until the calibrations converged as so and an M500 applied:

    Calibrated 8 factors using 16 points, deviation before 0.042 after 0.031

    My first clue that something was wrong was when I ran a G29 and got this result:
    0_1543371209218_TLMBed.JPG

    This is at odds with the result I got from the BLTouch (which gave me a good first layer)
    0_1543371249582_tlmheight.JPG

    Following this rabbit hole, my repeatability test returned this result over three different days of testing:
    [spoiler]
    9:25:20 PM
    Stopped at height -0.060 mm
    9:25:18 PM
    Stopped at height -0.104 mm
    9:25:16 PM
    Stopped at height -0.066 mm
    9:25:15 PM
    Stopped at height -0.079 mm
    9:25:13 PM
    Stopped at height -0.110 mm
    Stopped at height -0.129 mm
    Stopped at height -0.066 mm
    Stopped at height -0.116 mm
    9:25:11 PM
    Stopped at height -0.110 mm
    Stopped at height -0.129 mm
    Stopped at height -0.066 mm
    9:25:10 PM
    Stopped at height -0.110 mm
    Stopped at height -0.129 mm
    9:25:08 PM
    Stopped at height -0.110 mm
    9:25:06 PM
    Stopped at height -0.060 mm
    9:25:04 PM
    M140 S60
    Stopped at height -0.041 mm

    9:40:51 PM
    Stopped at height -0.248 mm
    9:40:50 PM
    Stopped at height -0.311 mm
    9:40:49 PM
    Stopped at height -0.298 mm
    9:40:48 PM
    Stopped at height -0.342 mm
    9:40:47 PM
    Stopped at height -0.342 mm
    9:40:45 PM
    Stopped at height -0.298 mm
    9:40:44 PM
    Stopped at height -0.198 mm
    9:40:43 PM
    Stopped at height -0.311 mm
    9:40:42 PM
    Stopped at height -0.198 mm
    9:40:40 PM
    M98 P"0:/macros/BLT Repeatability Test"
    Stopped at height -0.136 mm

    5:23:29 AM
    Stopped at height -0.223 mm
    5:23:28 AM
    Stopped at height -0.205 mm
    5:23:27 AM
    Stopped at height -0.211 mm
    5:23:26 AM
    Stopped at height -0.198 mm
    5:23:25 AM
    Stopped at height -0.192 mm
    5:23:23 AM
    Stopped at height -0.186 mm
    5:23:22 AM
    Stopped at height -0.217 mm
    5:23:21 AM
    Stopped at height -0.186 mm
    5:23:20 AM
    Stopped at height -0.205 mm
    5:23:19 AM
    M98 P"0:/macros/BLT Repeatability Test"
    Stopped at height -0.161 mm
    [/spoiler]

    The bed leveling attempts all exhibit that very jagged pattern across the bed (which is glass without any sticker). Nozzle is a clean Micro-Swiss .4mm in all tests.

    Is this amount of variability normal for the smart effector? None of the traces are damaged, fan wires away from the PCB, and all the LEDs illuminate and I do not see any visibly damaged or "crusty" SMD components. Sensitivity is at default and it is very difficult to complete a G29 without bumping Snn in the M558 to 0.05. THe bed calibration also does not appear to apply correctly since low layered objects come out squished with effector clicking on some sections of a circle and then too high to adhere to the glass in other sections of the same circle.

    I think I'm missing something and hoping you can point me in the right direction. Thanks!



  • I have the same setup as yourself, there is a few differences in my config.g file,
    M574 X2 Y2 Z2 S1 ; Set active high endstops
    M558 P5 R0.4 H5 F1200 T10000 ; Set Z probe type to effector and the dive height + speeds
    but this looks more like a mechanical issue to me, have you lubricated the Magball arms? as Hayden recommends?


  • administrators

    The first height map in your post shows the classic "crinkle crisp" pattern that indicates backlash in your motion system. The height map has ridges and valleys running in the X direction, but what is a ridge at minimum X turns into a valley at maximum edge. The reason for this pattern is that is probes the mesh in alternating X directions. Because of backlash, the height that the probe registers depends on whether you approach the point from the low-X or the high-X direction, except at X=0.

    Here's a macro file that I use to check for backlash (I've added a .g extension so that I can upload it):

    0_1543397271800_TestBacklash.g

    It does 10 probes at X=-100 Y=0, approaching the point from more negative X each time. Then it does 10 probes at the same point but approaching from the opposite direction. Here is a typical result:

    09:26:48
    G32 bed probe heights: -0.114 -0.109 -0.109 -0.107 -0.109 -0.107 -0.107 -0.104 -0.104 -0.107, mean -0.108, deviation from mean 0.003
    09:26:23
    M98 P"0:/macros/TestBacklash"
    G32 bed probe heights: -0.124 -0.119 -0.119 -0.119 -0.117 -0.114 -0.114 -0.114 -0.114 -0.114, mean -0.117, deviation from mean 0.003

    There is a difference in the mean of 0.009mm depending on which direction the point is approached from. This is the amount of backlash. Coupled with the observation that my delta has more motion noise than it used to, it tells me that I need to lubricate something.

    A similar test can be run for backlash in the Y direction, by exchanging all X and Y coordinates in the script.



  • Wizards, one and all. I knew I had to be missing something. I put a little pat of SuperLube on each balljoint and ran the test again.

    6:27:06 AM
    Stopped at height -0.172 mm
    6:27:05 AM
    Stopped at height -0.185 mm
    6:27:03 AM
    Stopped at height -0.166 mm
    6:27:02 AM
    Stopped at height -0.172 mm
    6:27:01 AM
    Stopped at height -0.166 mm
    6:27:00 AM
    Stopped at height -0.166 mm
    6:26:58 AM
    Stopped at height -0.166 mm
    6:26:57 AM
    Stopped at height -0.172 mm
    Stopped at height -0.178 mm
    Stopped at height -0.166 mm
    6:26:56 AM
    Stopped at height -0.172 mm
    Stopped at height -0.178 mm
    6:26:55 AM
    M98 P"0:/macros/BLT Repeatability Test"
    Stopped at height -0.172 mm

    I still have a little bit of tweaking but that's very, very close to the mean error of .009mm

    Thank you!


 

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