#Inconsistent Z-offset

35 messages · Page 1 of 1 (latest)

sudden fox
#

I am debugging why my 2.4 is having inconsistent GQL. I tried a PROBE_ACCURACY samples=100 today, and I noticed that the Z is drifting. I have already replaced the Z motor bracket, realigned the belt, and tightened the Z belt. However, the problem continues. Does anyone have any idea what might be causing this?

I am using an inductive probe and the tests were done with the machine cold

image.png
sudden fox
#

I also ran the test on each corner:

image.png
dry crown
#

Does it keep drifting even further, or does it at some point stop?

#

It's not a huge difference (0.02mm), and at least to me it seems a bit too consistent in all four corners to be something mechanical.

#

i haven't used an inductive probe in a couple years and it was all cheap knockoff ones, but they were pretty temperature sensitive (and would develop some kind of offset with temperature). Even if you're not running any heaters, there's still electronics and stuff (like extruder driver/motor) and the probe I guess could heat itself up by a bit. Was this after a "cold start"? If you leave the printer on for a couple hours and then run the same probing, does it do the same thing?

sudden fox
#

Isn't 0.02mm enough for the QGL macro to abort? I keep getting aborts after 5 attempts or early aborts because it detects that the error is increasing instead of decreasing

#

The machine was cold for 12 hours

dry crown
#

depends on how close you set the tolerance in the qgl settings. I have it set to a bit higher, since that close tolerance in z height doesn't seem to matter at least for what I'm printing.

dry crown
sudden fox
#

I just ran another test. My hypothesis is that if this is a z-drive issue, the drift would also be noticeable when testing on the z-offset endpoint. So I did the following:

•    Installed the klipper_z_calibration plugin and ran PROBE_Z_ACCURACY with 250 samples.
•    Then, I moved the head a bit forward and ran PROBE_ACCURACY with 250 samples as well.

These are the results:

image.png
#

Some conclusions:

  • The first 2-3 touches on the switch are trash, but then it becomes very reliable
  • The probe slowly drifts up but then it becomes stable after around 100 retries
  • Both the switch and the probe show a slow drift downwards, that could be either mechanical or temperature, but I don't think that's the source of my issue
#

It looks like the probe is the source of my issue.

tawdry wren
#

Thanks for testing. I'm having a similar drift (in amplitude) with my Tap.
As already suggested, have you tried completely heat soaking your machine?

#

Your drift is 0.005mm over >200 samples.
I realized that for me, the drift is reset when moving the gantry, which hints at a mechanical issue. Can you confirm?

sudden fox
# tawdry wren Your drift is 0.005mm over >200 samples. I realized that for me, the drift is re...

The mechanical Z endstop switch has an initial drift of 0.025 mm over the first 2-3 activations. I am not sure what causes it, but I guess it might be due to some internal pressure build-up inside the switch. After the first three activations, it shows a slow drift of 0.005 mm over 200 samples, as you mentioned.

The probe shows a drift of around 0.010 mm over the first 100 samples but then switches to a slow drift similar to the mechanical probe.

The mechanical switch is an Omron, but the probe is a €5 model from Aliexpress.

My machine is not enclosed, so I was running all tests at room temperature. Hopefully, I can finish the panels today and then rerun the tests.

sudden fox
#

I finally got around to replacing the probe. I chose Unklicky because it appears to be more reliable than the Omron mechanical switch. After running a series of tests, here are the results compared to the inductive probe:

image.png
#

Unklicky seems much better; it doesn’t exhibit the drift seen in the first 100 tests.

#

I then conducted another test, this time with 2,500 samples (which took several hours), comparing Unklicky with the Omron mechanical Z-offset switch.

image.png
coral pawn
#

What is your probe speed?

sudden fox
#
  • Both switches exhibited drift, but Unklicky drifted at a slower rate. I believe this is primarily a mechanical issue with the Z-axis. The slower drift with Unklicky is likely due to the faster test script rather than the switch itself.
  • Both switches showed some inconsistency in the initial samples, but Unklicky took longer to stabilize. This might be because the switch is brand new and still settling in.

I’m hopeful that Unklicky will resolve my issues. The Z-offset variation has been so problematic that I’ve already damaged a PEI sheet.

coral pawn
#

Let me see if I can give you an example

#

probe accuracy results: maximum 6.527320, minimum 6.524820, range 0.002500, average 6.527070, median 6.527320, standard deviation 0.000750

#

That's just 10 samples, but it stabilizes very quickly. When I was probing at the default of, from memory, 20mm/s, I didn't get near as much consistency.

#

It's slow, yes, but I almost never have retries.

#

With Klicky, I also use Auto-Z, but just from memory the z offset is pretty consistent from run to run.

sudden fox
coral pawn
#

I guess my point is, with mechanical switches, slow down.

#

I'm sorry, I meant retries during a probe of a point. Like when the returns vary too much and it starts over.

#

This is from a cold start Retries: 1/5 Probed points range: 0.005000 tolerance: 0.007000

sudden fox
#

Do you get those Probe samples exceed tolerance. Retrying... ?

coral pawn
#

I could also hear my Klicky swith body slamming into the print bed.

#

Just try slowing down and see if things improve 🙂