Auto bed level unstable

Is there a reason I have to keep changing the Z-offset for auto bed leveling?

Having given up totally on my Delta printer, I thought that since I had the sensor, I might as well install it on my Prusa. After chasing down a bunch of end stop problems that didn't exist before, and turning off "safe" z homing (which should really be titled "ensure crashing the axes homing") I got it to where it will probe in 4 places and run the z axis up and down during the print.

But between every single print, I have to change the Z-offset.

So it costs more, is a pain in the pants to set up, takes more time to start a print, and still must be adjusted for every single print. Were there any advantages?

Also, I've got one of DC42's sensors for sale.

Your endstops are 'moving'.
Losing steps on the motors.
Loose screw on one or more pulleys.
For the ir sensor, bed surface not reflecting ir consistently.
Tilting of the effector.

If your firmware supports it try building with Z_MIN_PROBE_REPEATABILITY_TEST enabled. Load it and then do a G28 followed by an M48 to run the repeatability test.
I'm using RC8BugFix, a BLTouch probe.and built with AUTO_BED_LEVELING_BILINEAR which works fine. A quick run of M48 on my Folger 2020 results in the following:

Send: M48 P4 X100 Y100 V4
Recv: M48 Z-Probe Repeatability Test
Recv: Positioning the probe...
Recv: Bed X: 100.000 Y: 100.000 Z: 3.190
Recv: 1 of 4: z: 3.192 mean: 3.1925 sigma: 0.000000 min: 3.192 max: 3.192 range: 0.000
Recv: 2 of 4: z: 3.155 mean: 3.1737 sigma: 0.018750 min: 3.155 max: 3.192 range: 0.037
Recv: 3 of 4: z: 3.185 mean: 3.1775 sigma: 0.016202 min: 3.155 max: 3.192 range: 0.037
Recv: 4 of 4: z: 3.180 mean: 3.1781 sigma: 0.014073 min: 3.155 max: 3.192 range: 0.037
Recv: Finished!
Recv: Mean: 3.178125 Min: 3.155 Max: 3.192 Range: 0.037
Recv: Standard Deviation: 0.014073

I also have safe homing defined and it works well to position the BLTouch at the bed center before lowering for the G28.

Ive always had issues with auto level and offset with every print.
I just ran the probe repeat and got this:

SENDING:M48 P4 X100 Y100 V4
M48 Z-Probe Repeatability Test
Positioning the probe...
Bed X: 100.000 Y: 100.000 Z: 5.692
1 of 4: z: 5.693 mean: 5.6926 sigma: 0.000000 min: 5.693 max: 5.693 range: 0.000
2 of 4: z: 5.693 mean: 5.6926 sigma: 0.000000 min: 5.693 max: 5.693 range: 0.000
3 of 4: z: 5.693 mean: 5.6926 sigma: 0.000000 min: 5.693 max: 5.693 range: 0.000
4 of 4: z: 5.694 mean: 5.6930 sigma: 0.000531 min: 5.693 max: 5.694 range: 0.001
Finished!
Mean: 5.692954 Min: 5.693 Max: 5.694 Range: 0.001
Standard Deviation: 0.000531

This is a Makerfarm i3 with home brew microswitch probe on servo.

Edited 1 time(s). Last edit at 12/19/2016 10:37PM by Floyd.

Quote
stephenrc
Your endstops are 'moving'.
Losing steps on the motors.
Loose screw on one or more pulleys.
For the ir sensor, bed surface not reflecting ir consistently.
Tilting of the effector.

End stops moving -- I don't think this is a problem. The IR sensor was functioning as the only Z stop, and it was hard mounted to the extruder motor. The X end stop is butted against the X motor mount, nowhere for it to go. The Y end stop could be better, in fact I'll CAD something up tomorrow. But, here's the thing: it did a G28, a G29, and then started printing. It didn't strike the Y end stop twice, so the whole thing is working from the same Y reference.

losing steps on motors -- I've been printing for 12 hours straight with no steps lost. I'm confident that this isn't the issue.

Loose screw on pulleys -- See above. Printing for 12 hours straight with no slippage. Though I will make it a point to check them.

Bed not reflecting IR signal consistently -- Most likely culprit. I switched from a mirror to transparent glass with no changes to my red PCB heater.

tilting effector -- My effector hasn't tilted in weeks. It's over there in my parts bin with the rest of my Kossel. This issue cropped up when I installed the sensor on my Prusa. One thing about a cartesian printer is they don't tend to suffer extruder tilt.

I'm not going to try to fix it. The sensor has already been removed. I've come to the conclusion that auto leveling solves fewer problems than it causes. I'll admit I am a tremendously impatient man. I could take the time to learn how to do that repeatability test and then figure out what numbers in that report matter, whether bigger or smaller numbers are good, and what to do about it, figure out whether I need to use a different sensor, if so CAD up a new mount, rewire my printer AGAIN, determine whether or not it's the nature of my print surface, try five or six things to fix that, and maybe eventually getting it to sort of work...or I can keep twiddling the corner screws like I've been doing for months now and get reliably good parts.

That said, let's leave this thread here and continue to compile advice and solutions. I was unable to find any good search terms on Google for auto leveling problems, so anyone who's willing to put up with the process of getting it to work should have somewhere to look.

I've seen results similar to what you are describing with the inductive and capacitive sensors. Some people have good results with those types of probes. But not everybody has good results. I prefer a micro switch on a Z-Probe leg. That works very reliably.

But the other thing that is very helpful is to set your Z-Probe-Offset such that the nozzle is about .3mm or .4mm above the bed. And then just dial in the height using the Z-BabyStepping while the skirt is being printed.

And I know I'm biased on this, but I've added several ways to get into the Z-BabyStepping very quickly in the Marlin UBL code base. The first way is you can over ride the 'Kill' switch on your LCD Panel to do Z-BabyStepping. If that is not acceptable, you can just double click the encoder wheel while in the Status Screen and it will go to Z-BabyStepping.

The reason this is valuable is because if you need to adjust the nozzle, it can be done very quickly and easily. And at that point, having your endstops adjusted exactly correct becomes much less important. In fact, it becomes almost the opposite. In order to protect the nozzle and the printer's mechanics, you deliberately adjust the nozzle a little bit higher than it should be to leave some room for error. And just dial in the correct height when you start a print.

Edited 2 time(s). Last edit at 12/25/2016 02:43PM by Roxy.

This is a really old thread, but I decided to post here. Running Marlin on Prusa i3 metal frame. I too have encountered inconsistency using the bed leveling feature of Marlin. And I am using a standard inductive sensor obtained from Aliexpress and an aluminum hot bed. So, here's what I have discovered. The sensor itself is very stable, the M48 command reports something around 0.0002mm deviation. However! It is very sensitive to how clean the bed surface is! I usually use hairspray before printing. So, the thinnest layer of hairspray glue on the surface changes how the sensor reacts to metal underneath it. If you got a lot of glue residue from several uses, it gets even worse. And, although it's all non-magnetic and the sensor is supposed to "shoot" thru it, in reality it doesn''t. So whenever a particular spot of the bed is not clean enough, the sensor thinks it's closer.

Another thing... I am a long time user, I started 3d printing around 12 years ago. Now, after several messed up prints I decided to do some tuning to the printer, and I discovered that some of the Pololus are set to a pretty low current. Could have been the potentiometers moved due to vibration. Check your machines. Mine were set to around 1 volts for Y and Z, but 0.4v for X and extruder, which is pretty low. (check the pot slider with voltmeter relative to Gnd).

Also, steppers often skip if the rails are (again!) not clean and not oiled. You have to do it from time to time, at least once per year.

Hope this helps someone.

Edited 3 time(s). Last edit at 05/23/2021 12:19PM by AngelOfGrief.

Quote
AngelOfGrief
This is a really old thread, but I decided to post here. Running Marlin on Prusa i3 metal frame. I too have encountered inconsistency using the bed leveling feature of Marlin. And I am using a standard inductive sensor obtained from Aliexpress and an aluminum hot bed. So, here's what I have discovered. The sensor itself is very stable, the M48 command reports something around 0.0002mm deviation. However! It is very sensitive to how clean the bed surface is! I usually use hairspray before printing. So, the thinnest layer of hairspray glue on the surface changes how the sensor reacts to metal underneath it. If you got a lot of glue residue from several uses, it gets even worse. And, although it's all non-magnetic and the sensor is supposed to "shoot" thru it, in reality it doesn''t. So whenever a particular spot of the bed is not clean enough, the sensor thinks it's closer.

Another thing... I am a long time user, I started 3d printing around 12 years ago. Now, after several messed up prints I decided to do some tuning to the printer, and I discovered that some of the Pololus are set to a pretty low current. Could have been the potentiometers moved due to vibration. Check your machines. Mine were set to around 1 volts for Y and Z, but 0.4v for X and extruder, which is pretty low. (check the pot slider with voltmeter relative to Gnd).

Also, steppers often skip if the rails are (again!) not clean and not oiled. You have to do it from time to time, at least once per year.

Hope this helps someone.

I agree. Bed cleanliness is critical. I have a home built under-bed piezo sensor on my delta, which determines nozzle contact with the borosilicate glass bed. M48 consistently reports 0.000000 standard deviation under Marlin. However, Marlin UBL often fails if the bed and nozzle aren't scrupulously clean. I'm talking motes of dust clean, and no chance that old hardened plastic at the nozzle tip might get flattened a few tenths of a micron during the leveling process. I rarely do a bed leveling (maybe a few times a year, or after a nozzle change), but when I do, I manually move the effector on my delta down to a sheet of 1200 grit wet/dry paper on the bed, and polish the nozzle tip beforehand.

And here is another consideration which just occurred to me. I use an inductive sensor which is mounted to the left of the nozzle, I also use the Prusa mechanics with two Z-stage motors. Which actually means, that whenever the motors are slightly out of sync with each other, the X stage is not strictly horizontal. Although bed leveling is supposed to take care of this, tilt of the print head actually changes the Z offset of the sensor relative to the nozzle! Think rotating it. Nozzle gets closer to the bed, sensor gets farther. To get around this, we must mount the sensor in front of or behind the nozzle, and not by its side! I'll remount the sensor and see how it goes.

p.s. Piezo sensor looks like a cool idea, I wonder how it works when the machine generates so much vibration, how do you filter out the "clean" signal?

Edited 2 time(s). Last edit at 05/28/2021 07:03AM by AngelOfGrief.