Yet another "XYZ calibration failed. Please consult the manual."
Hello all. First post. I have scoured the forum for other's with the same XYZ calibration failure, no solutions for me unfortunately.
First, my printer is basically an MK2 upgraded to an MK2.5s then upgraded to almost an MK3s. I upgraded the MK2.5s with an Einsy Rambo v1.1b bought from Ultimachine.
I am also running FW 3.9.0 RC1 but have also tried 3.8.1 with same results.
My first calibration after I made the upgrade worked just fine. But it was strange because I thought the Y axis was backwards. Sure enough, I found that in the real MK3s, the Y motor had been turned around from the way it is in mine. No problem, I just changed the firmware to invert the Y motion.
Well, with the Y axis inverted, the XYZ calibration fails. It stops essentially at the end of point #4. It appears that it really did find the point, at least from all the motions it seems like it did. I also captured output on the serial port from a working calibration (Y axis backwards) and from the failing calibration (Y axis corrected). I cannot tell from the output why it gives up.
I have measured how much space is between the Y at end stop and the calibration point at both ends of the bed, and they are virtually identical. So why it would like the points in reverse order but not correct order makes no sense to me.
I found in one forum where someone says the Pinda probe should land on the first calibration point after an auto-home. Mine lands close, but not exactly on top. It looks like Y is about 1mm toward the outside of the point, and X is about 3mm on the inside of the point. After an auto-home, it leaves my extruder at X=12 and Y=6. If I knew how to adjust the position, I am sure I can get the probe to land directly on top of the point. Does anyone know if this is necessary, and if so which variables in the firmware source control this?
By the way, as an MK2, the calibration reported that all axis were perpendicular. With my Y axis backwards, it reports the same thing after calibration. I don't think things are skewed enough to cause this.
Does anyone know how to decipher the outputs I captured on the serial port? If so I can upload them.
I am really stuck. If anyone has some good ideas, please let me know. Thanks in advance.
Bill
RE: Yet another "XYZ calibration failed. Please consult the manual."
Just a bit more of interesting information.
I captured log files from the serial port with the Y axis non-inverted in FW, but backwards from the way it is supposed to travel, and Y axis inverted. In the backwards case, the calibration passes. The 4 points found were: (1046,755), (21813, 765), (21839, 19955) and (994, 199962). Note that the two points sharing the same Y have very similar values, and points sharing the same X have very similar values.
Now, without touching the hardware at all, just inverting the Y direction in FW, the calibration yields the following 4 points: (1043, 370), (21839, 841), (21812, 19565) and (1046, 19953). For these points, those that share X value are very similar. But the ones that share a Y value (or should) differ by approx. 400. Example 841 vs 370, and 19953 vs 19565. Mind you these are the same points, just the direction from which the points are approached is different.
I can understand that given these numbers that the FW thinks things are really messed up and would generate an error. The bed would have to be a parallelogram and no longer a rectangle.
How can the algorithm produce good numbers from one direction and not the other?
RE: Yet another "XYZ calibration failed. Please consult the manual."
It gets curiouser and curiouser.
I moved the frame about 3mm forward to make the distance from the end stop to the calibration point be the same no matter which direction the y axis moves. It was enough that now I can calibrate. But, since it measures slightly different y values, it calculates skew. With the y axis backwards, the calculated skew is 0.01. With the y axis normal, and no other changes, it calculates skew of 0.10, or 10 times the value from the other direction. Now I am concerned that if the skew is calculated incorrectly, that it will make a square printer print out of square.
RE: Yet another "XYZ calibration failed. Please consult the manual."
Today I did some more experimenting. When I calibrate XYZ, the numbers come out fairly consistent from run to run. They are not exact, but pretty close. One thing I did was extend my Y axis a bit. I thought it might have been hitting the end for the last 2 calibration points. Now there is plenty of room.
While at it, I squared the frame up. I used machinist squares and tweaked and tweaked. I think it is about as square as I can get it.
Here are some typical XY values I get from monitoring the serial port:
1101, 756
21894, 450
21883, 19953
1119, 19612
You will notice that the X values are about the same for the 2 pairs of points. The Y values differ by some 300 or so, yet I can move the extruder from one side to the other and everything lines up exactly. There is no way there is 3mm of difference. And you will notice if you plot these points on a graphs, they don't represent a rectangle, they represent a trapezoid.
How can this be? I have adjusted the PINDA up and down. No difference.
The only thing that gets a rectangle is to reverse the Y axis direction. The calibration then comes at the points from the other direction, but it finds them in a rectangle. What is it about the direction of the sweep that would result in a better or worse estimate of the location of the calibration pad?
Does anyone have any ideas for me to try?
RE: Yet another "XYZ calibration failed. Please consult the manual."
All,
I believe I found the reason for this. One difference between the MK2.5s and the MK3s is the fact that the Y motor is reversed. When I replaced my controller with an Einsy Rambo, I simply reversed the direction for the Y motor in the firmware, I did not physically reverse the motor. For normal movements, this worked just fine. However, for the calibration, the search direction is not reversed by the firmware change. I had previously noted that calibrating backwards worked.
Well, I rotated the connector for my Y motor, so I can undo the firmware direction change. Now, the calibration sweeps the points from the opposite direction, same as with a backward Y motor. Now the calibration is spot on.
One of the weirdest issues I have found. Now all is good.