Tec4 problems with levelling

Post Reply
Beta
Posts: 19
Joined: Sat Mar 03, 2018 6:23 am

Re: Tec4 problems with levelling

Post by Beta »

Finally, after all of the above, I changed the level and then let the machine go through its levelling routing. Afterwards, I let the machine move to various points with the distortion correction turned on, and measured the distance between the bed and the tip of the nozzle. Results are heights I had to tell the machine to go to in order to just touch the reference block at each point:

Bed temp: 55C
Nozzle temp: 120C
Reference height: Aluminum block with a height of 9.99mm after sitting on the heated bed for 30 minutes

X: 10.0 mm
Y: 10.0, Z: 9.86 mm
Y: 102.5, Z: 9.82 mm
Y: 204.0, Z: 9.76 mm

X: 127.5 mm
Y: 10.0, Z: 9.88 mm
Y: 102.5, Z: 9.88 mm
Y: 204.0, Z: 9.82 mm

X: 230.0 mm
Y: 10.0, Z: 9.80 mm
Y: 102.5, Z: 9.82 mm
Y: 204.0, Z: 9.70 mm

Again, inconsistent results where I expected the correction to deal with any deviations but instead had to tell it to move the build plate upwards in order to lightly touch the tip of the nozzle. Centerline seemed consistent within 0.06mm, whereas the left side of the plate seemed lower than the right, and the front being lower than the rear

Beta
Posts: 19
Joined: Sat Mar 03, 2018 6:23 am

Post by Beta »

And any analysis wouldn't be complete without a heat map. Attached is the heat map of Standard Deviation of the Z-axis distortion corrections as measured by the printer. I've oriented the data so the top left of the table conforms to the back left side of the print bed. All measurements are in mm.

Plate levelling was run 5 times, so each XY point has 5 Z-correction values for the data set.

As you can see, as you traverse from back-left corner to front-right corner of the build plate, the deviations become significantly greater.
Attachments
Heatmap of Z distortion standard deviations.PNG

Beta
Posts: 19
Joined: Sat Mar 03, 2018 6:23 am

Post by Beta »

The data seems to point that the z-probe (which is also the Z-min end stop) is the problem. It seems neither highly accurate nor greatly repeatable in its measurements.

Conclusions:
As the Z-sensor is also the z-min end-stop, this makes homing Z a variable - The Z-position of the build plate at home is solely dependent upon the accuracy and repeatability of the Z-probe. If the z-probe is malfunctioning or inaccurate, Z-home will never be the same. Even if it is functioning to spec, Z-home will be dependent upon the tolerance of the Z-probe. This will lead to inconsistent printing results unless the tolerances of the Z-probe is very small versus the layer height.

This contrasts to the X and Y axis which have optical break-beam end stops which are highly accurate and repeatable. While I understand the firmware supports multiple end-stops for each axis, I haven't yet looked at the controller board to see whether it does as well. Improved accuracy would likely result if the Z-axis had a fixed end-stop independent of the levelling probe - at the very least, position 0 on the Z-axis would no longer be a variable.

Beta
Posts: 19
Joined: Sat Mar 03, 2018 6:23 am

Post by Beta »

Last night I got to thinking about the heat map I posted above. Yes, the data supported the fact that the distances measured by the z-probe were rather weird. But the data fell into a systemic pattern instead of a random one - if the z-probe was malfunctioning you would normally expect some degree of randomness. But the heat map clearly shows a pattern that isn't random. So I went back to what would be the causes.

The answer was subtle: Turns out the z-axis nut that rides along the z-axis screw was not completely positioned inside the print bed carrier arm. It is very difficult to see or even detect (even after going over the machine 3 times). The nut was turned in just such a way where it seemed as if it was solidly positioned, but in fact was only supporting the carrier arm by its corners. This introduced a slight wobble in the z-axis.

After finding that, I reassembled the machine, leveled and calibrated the z-probe then ran through the same test sequences above. The standard deviation of the results was less by a factor of 10, and what remained was random. Printing now works everywhere on the print bed, and even the XY calibration routine works perfectly.

So, here's me humbling myself and apologizing. You were correct, it was ultimately mechanical.

User avatar
gfeliksdal
Site Admin
Posts: 405
Joined: Sat Feb 25, 2012 10:40 pm
Location: Netherlands
Contact:

Post by gfeliksdal »

Dear Beta,

Good to hear you found the issue after the extensive search. Guess the devil is in the details...

Post Reply

Return to “Troubleshooting”