Bed leveling issues.

hi i have a delta printer with a bltouch. If I start it with the eeprom without the transformation matrix, and level it by hand, it is perfectly flat in its 300 mm diameter. But if I run the command G32, after finishing the leveling process without errors, and retesting with a sheet of papper, the bed is as if it was very tilted, that is, the point X150, Y150 is much higher than the point x -150x and-150, which is not true.

The z probe data is:

15:47:43.351 : X:0.00 Y:0.00 Z:487.638 E:0.0000
15:47:52.715 : Z-probe:10.36 X:-70.00 Y:-70.00
15:47:57.491 : Z-probe:11.05 X:-70.00 Y:0.00
15:48:02.350 : Z-probe:11.27 X:-70.00 Y:70.00
15:48:08.060 : Z-probe:11.19 X:0.00 Y:-70.00
15:48:13.013 : Z-probe:11.56 X:0.00 Y:0.00
15:48:18.006 : Z-probe:11.88 X:0.00 Y:70.00
15:48:23.746 : Z-probe:10.78 X:70.00 Y:-70.00
15:48:28.620 : Z-probe:11.64 X:70.00 Y:0.00
15:48:33.651 : Z-probe:12.38 X:70.00 Y:70.00

According to these data, there is an inclination, but instead of correcting it, he accentuated it, making the low side even lower and the high side even higher.

This is on a Delta with 0.9.2.9.

This is the configuration.h file
https://drive.google.com/file/d/1BcltVdDvll5PSo5jo05wOuEeB3V0RiTf/view?usp=sharing

Comments

  • the same with z correction, after running G33 or G32 the problem is accentuated instead of improving.
  • I would 1st update to ver 1.0.4dev i believe the older version had some bugs in the bed levelling.
  • MartinH said:
    I would 1st update to ver 1.0.4dev i believe the older version had some bugs in the bed levelling.
    there are some many other bugs en 1.0.4dev, (al least on delta printers) I tried 1.0.2, 1.0.3, and 1.0.4 dev, and until now 0.9.2.9 is the most stable. The issue with auto level is the unique issue I have. I need to solve this in 0.9.2.9
  • I just tried the same in 1.0.4 dev and exactly the same thing happens, the side that was higher, after running G33, and then G32 is even higher. and the side was a little lower, now it is 2mm from the nozzle.
  • When you repeatedly run
    G30 P0
    on one point does the height returned change significantly or does it stay swinging around an average number?
  • Repetier said:
    When you repeatedly run
    G30 P0
    on one point does the height returned change significantly or does it stay swinging around an average number?
    In 10 attempts it only had a margin of variation of 0.01. between 489.40 and 489.41
  • So repeatability is good. What is the plane equation G32 returns along with measured values. Question is matches the equation the measured rotation.
    And another one - why do we measure a rotation if you say without autoleveling the paper method shows it is flat. That is a contradiction except if at that point G32/G33 was active adding to the correction.
  • Repetier said:
    So repeatability is good. What is the plane equation G32 returns along with measured values. Question is matches the equation the measured rotation.
    And another one - why do we measure a rotation if you say without autoleveling the paper method shows it is flat. That is a contradiction except if at that point G32/G33 was active adding to the correction.
    I don't need the flat correction (G33), I just activated it to test if anything changed. I'm only interested in being able to level the bed. Now I'm printing, when I'm done I try the G32 and tell you what the result is.
  • Repetier said:
    So repeatability is good. What is the plane equation G32 returns along with measured values. Question is matches the equation the measured rotation.
    And another one - why do we measure a rotation if you say without autoleveling the paper method shows it is flat. That is a contradiction except if at that point G32/G33 was active adding to the correction.

    18:25:40.417 : X:0.00 Y:0.00 Z:488.000 E:0.0000
    18:25:46.963 : Z-probe:16.30 X:-70.00 Y:-70.00
    18:25:50.547 : Z-probe:17.74 X:-70.00 Y:0.00
    18:25:54.312 : Z-probe:18.85 X:-70.00 Y:70.00
    18:25:58.789 : Z-probe:17.04 X:0.00 Y:-70.00
    18:26:02.460 : Z-probe:18.02 X:0.00 Y:0.00
    18:26:06.167 : Z-probe:19.26 X:0.00 Y:70.00
    18:26:10.657 : Z-probe:16.62 X:70.00 Y:-70.00
    18:26:14.258 : Z-probe:18.31 X:70.00 Y:0.00
    18:26:18.047 : Z-probe:19.98 X:70.00 Y:70.00
    18:26:18.960 : plane: a = 0.0048 b = 0.0194 c = 18.0146
    18:26:18.969 : Transformation matrix: 0.999989 0.000000 -0.004791 -0.000093 0.999813 -0.019356 0.004790 0.019356 0.999801

  • Is the same with 3 points method.
  • Equation for z is z = a * x + y * b + c
    So with a and b positive this means bed goes down with increasing x and y.
    Looking into measured values you see that this is in deed what you measured, so computation of this correction matrix seems correct.
    For example x 70, y 70 you get z = 0.0048 * 70 + 0.0194 * 70 + 18.0146 = 19.71 compared with 19.98 measured. So at that edge an error of 0.2mm is to be expected.
    At 0,0 it is easy 18.0146 compared to 18.02 measured.

    These differences will be there and come from imperfect geometry of course. What you get is a regression plane that minimizes the mean error over all measured points. With 3 points you get an equation where the 3 points measured would fit exactly, but all other points you would manually measure will be different.

    What I see is also that the difference is nearly 3.7mm which is a lot. Did you adjust the end stop offsets to have same distance to bed after homing? That should be the second stage in delta calibration right after checking bed is also as perpendicular to the z axis as you can. This should reduce the starting error a good part.

    How many extruder does the printer have? If you have extruders with offset different then 0/0 you might need to go lower after homing then the 4mm I saw. Problem is that activating the extruder could trigger an endstop and then position is already wrong at the top since xy motion is limited at top to only a few mm with 4mm space left to end stops.


  • I still can't get this to work. I think there is some problem in the results returned by the G32. If I compare them with the measurements of the G33 (5x5) I have a variation of + - 0.2mm with G33, while the G32 (3x3, although I also tried 5x5) always returns a difference of 2mm measuring in the same area. With G32 there is always a difference of 2mm or more between the lower left and upper right point, no matter how the bed level is. Even when I manually level and make a pefect print across the entire radius of the bed, running the G32 command I always have a 2mm difference. I already discard that the BLTouch have any problem because the results of the G33 are more correct

    a curious fact is that, the lower the area that claims to be the highest, the more difference there is between it and the lowest.

    G33

    23:18:42.070 : Z-probe:16.48 X:-63.88 Y:78.12
    23:18:49.575 : Z-probe:16.68 X:-28.87 Y:78.12
    23:18:57.022 : Z-probe:16.72 X:6.13 Y:78.12
    23:19:04.474 : Z-probe:16.71 X:41.13 Y:78.12
    23:19:11.921 : Z-probe:16.66 X:76.13 Y:78.12
    23:19:20.700 : Z-probe:16.41 X:-63.87 Y:43.12
    23:19:28.102 : Z-probe:16.50 X:-28.87 Y:43.12
    23:19:35.521 : Z-probe:16.59 X:6.12 Y:43.12
    23:19:42.951 : Z-probe:16.59 X:41.12 Y:43.12
    23:19:50.378 : Z-probe:16.61 X:76.13 Y:43.12
    23:19:59.177 : Z-probe:16.37 X:-63.87 Y:8.11
    23:20:06.572 : Z-probe:16.49 X:-28.87 Y:8.11
    23:20:13.990 : Z-probe:16.60 X:6.13 Y:8.12
    23:20:21.417 : Z-probe:16.56 X:41.13 Y:8.12
    23:20:28.955 : Z-probe:16.40 X:76.13 Y:8.11
    23:20:37.643 : Z-probe:16.53 X:-63.87 Y:-26.89
    23:20:45.074 : Z-probe:16.62 X:-28.87 Y:-26.89
    23:20:52.521 : Z-probe:16.65 X:6.13 Y:-26.89
    23:20:59.997 : Z-probe:16.57 X:41.13 Y:-26.89
    23:21:07.408 : Z-probe:16.42 X:76.13 Y:-26.89
    23:21:16.109 : Z-probe:16.46 X:-63.87 Y:-61.88
    23:21:23.589 : Z-probe:16.77 X:-28.87 Y:-61.87     <<<<<<<<
    23:21:31.053 : Z-probe:16.79 X:6.13 Y:-61.87
    23:21:38.574 : Z-probe:16.69 X:41.13 Y:-61.87
    23:21:45.984 : Z-probe:16.33 X:76.13 Y:-61.88      <<<<<<<<
    23:21:48.676 : Info:Distortion correction matrix:
    23:21:48.676 : 42, 26, 22, 23, 27
    23:21:48.680 : 47, 40, 33, 33, 31
    23:21:48.680 : 50, 41, 32, 35, 48
    23:21:48.680 : 38, 31, 28, 34, 47
    23:21:48.680 : 43, 19, 17, 25, 54

    G32
    23:24:29.268 : Z-probe:15.33 X:-65.00 Y:-65.00   <<<<<<<<
    23:24:36.881 : Z-probe:15.77 X:-65.00 Y:0.00
    23:24:44.594 : Z-probe:16.47 X:-65.00 Y:65.00
    23:24:53.212 : Z-probe:15.92 X:0.00 Y:-65.00
    23:25:00.930 : Z-probe:16.31 X:0.00 Y:0.00
    23:25:08.825 : Z-probe:17.00 X:0.00 Y:65.00
    23:25:17.595 : Z-probe:16.06 X:65.00 Y:-65.00
    23:25:25.354 : Z-probe:16.55 X:65.00 Y:0.00
    23:25:33.239 : Z-probe:17.33 X:65.00 Y:65.00    <<<<<<<<<




  • You can not compare G32 with G33. You always do G32 S2 first which starts software correction for the rotation part. Then you run G33 to fix problems with bed bending - at that point the rotation is already corrected so the difference should be less - the 2mm are corrected by rotation. 

    If you look into your marked part
    23:21:16.109 : Z-probe:16.46 X:-63.87 Y:-61.88
    23:21:23.589 : Z-probe:16.77 X:-28.87 Y:-61.87     <<<<<<<<
    23:21:31.053 : Z-probe:16.79 X:6.13 Y:-61.87
    23:21:38.574 : Z-probe:16.69 X:41.13 Y:-61.87
    23:21:45.984 : Z-probe:16.33 X:76.13 Y:-61.88      <<<<<<<<

    you see z goes first up then down. That is not from rotation and no flat bed would do that. So this is not correctable with G32 rotation. This comes assuming bad is flat in reality - from geometry errors in your delta. And that is also why you shoudl first calibrate geometry parameter using e.g. escher delta calibration tool for 6 parameter and then start correcting the remaining errors using G33 so the regular error is at minimum possible.

    Since that error will exist until the top you might also want to correct it for complete Z and not only until Z=1 which is the default.
  • Repetier said:
    You can not compare G32 with G33. You always do G32 S2 first which starts software correction for the rotation part. Then you run G33 to fix problems with bed bending - at that point the rotation is already corrected so the difference should be less - the 2mm are corrected by rotation. 

    If you look into your marked part
    23:21:16.109 : Z-probe:16.46 X:-63.87 Y:-61.88
    23:21:23.589 : Z-probe:16.77 X:-28.87 Y:-61.87     <<<<<<<<
    23:21:31.053 : Z-probe:16.79 X:6.13 Y:-61.87
    23:21:38.574 : Z-probe:16.69 X:41.13 Y:-61.87
    23:21:45.984 : Z-probe:16.33 X:76.13 Y:-61.88      <<<<<<<<

    you see z goes first up then down. That is not from rotation and no flat bed would do that. So this is not correctable with G32 rotation. This comes assuming bad is flat in reality - from geometry errors in your delta. And that is also why you shoudl first calibrate geometry parameter using e.g. escher delta calibration tool for 6 parameter and then start correcting the remaining errors using G33 so the regular error is at minimum possible.

    Since that error will exist until the top you might also want to correct it for complete Z and not only until Z=1 which is the default.
    The bed is a glass, I have several glasses, they all give the same values, so if the bed is flat, but the measurements are not, there must be an error in the measurements. But I can't figure out where that error is. Measurements are repeatable with a variation of 0.02mm. The geometry of the printer is fine because without any kind of corrections, only with manual leveling, it makes a perfect first layer, and in large pieces I have a difference in dimensions between 0.02 and 0.08 depending on the material.

    I can't think what else to try to get this working
  • So with your manual leveled delta when you just measure a bed height map the errors are lower then +/-0.2mm?

    From my experience these values look like errors from delta geometry. This does not mean xy size will differ that much and the error is not very big. Changing an tower angle by 0.5° might change it more.

    What bothers me more is that the values do not swing around 0 for G33 which they should if run after G32 S2.
  • Repetier said:
    So with your manual leveled delta when you just measure a bed height map the errors are lower then +/-0.2mm?

    No, always about 2.0mm between the higher and lower points. 0.2m is the variation of each point when I repeat measurements

    From my experience these values look like errors from delta geometry. This does not mean xy size will differ that much and the error is not very big. Changing an tower angle by 0.5° might change it more.

    What bothers me more is that the values do not swing around 0 for G33 which they should if run after G32 S2.

  • You mean a second G32 reduces from 2 to 0.2mm error without changing anything? That would be wrong - should always get the same correction. But you showed G32 S2 which should be the first test btw, and afterwards you run G33 and there error are of course less since G32 is already actively reducing the 2mm error from rotation. That is what you showed. If it is different your example with values is misleading.
Sign In or Register to comment.