How much variation can auto bed leveling compensate for?

Question

I just upgraded an original CR-10 with the CRTouch bed leveling sensor and upgraded the firmware from stock to Marlin. In the CURA slicer, I added a preamble to every print to perform bed leveling. The bed leveling is running a 5 by 5 (25 point) leveling sequence. The sequence is not started until the bed and hot end are at temperature.

When using the above configuration with a bed level test (5 squares, 1 in each corner and at the center of the bed) the squares on the right top and right bottom printed so thin I could see through them!

(In this image the top is the left side of the bed and the bottom is the right.)

I then used OctoPrint to report on the bed height and found that the left side of the bed was about -0.5 mm and the right about +0.5 mm.

Is it incorrect for me to assume that bed leveling could have compensated for a 1.0 mm difference between the lowest and highest points on a 3D printer's bed??

To verify that all the software and hardware were actually performing (some) bed leveling, I observed the Z-axis stepping motor while printing the skirt for the above bed-level test pattern. I did see where the Z-axis stepping motor turned about 1/4 of a turn in the positive (up) Z-axis direction as the hot end traveled from the (low) left to the (high) right side of the bed. So, bed leveling appears to be working. But not enough to make a good first-layer print on the right side of the bed.

---

Added later...

I needed to add to the question to address manual bed leveling. As asked by a 3dprinting.stackexchange.com contributor.

• This is not a question regarding manual bed leveling. This is a question asking how much can auto bed leveling compensate for. Specifically, can a 25 point scan (a 5 by 5 point grid) before every print compensate for 1 mm linear (not a dip but a straight flat surface) difference from the left side to the right side of the glass build surface.

• Before posting this question, in a effort to fully understand the situation, I did verify the CR-10 printer could pass the bed leveling test. I did this by reverting to manually leveling the bed before the second bed leveling test. I did this (manual bed leveling) just as if no CRTouch was installed. Just like I would have over the previous 6 years using this printer and glass bed. The results were great. Please note what I am inferring here ... in order to pass the bed level test, I had to manually bed level the CR-10. The CRTouch did not make any difference. The first test failed despite using a CRTouch and 25 point alignment. And the second test would have likely passed with out using the CRTouch after the bed was manually leveled.

Answer 1

From the image and your observations:

"the squares on the right top and right bottom printed so thin I could see through them"

Hints to a different problem than your posed question. Depending on what type of levelling mesh and the amount of probing points you use, the bed surface will be replicated. The accuracy of that mesh, and the type (e.g. a plane versus a bed surface scan), is controllable by you, but remember that the resolution of the probing points must be higher than the defect you want to counter measure. See Automatic Bed Leveling (ABL) with a sensor (BLTouch, inductive, capacitive), how does it work? for a description how bed levelling works.

The problem with the economic printers like the CR-10 is that they employ a single drive Z-axis together with a questionable solution using a roller system over the aluminium profiles, it is a cheap solution often causing many problems when the printer ages (see many questions here). Note that I'm not bashing the printer, but just pointing out the weaknesses.

The graph and your observations that the right side of the printer prints "more thin" square's, hints to problems with the misalignment of the X-Z gantry. You most probably need to realign the roller carriages, re-tighten the eccentric roller or replace rollers. This has shown to work for many others with similar symptoms.

Hence, the question "How much variation can auto bed leveling compensate for?" shouldn't be the question for this problem, I've seen compensations for beds that where at least 10 mm higher on one side and still the head followed the bed surface. Or examples where a coin was placed on the bed, all imperfections are perfectly accounted for.

Answer 2

From the comments, it appears the general consensus is that automatic bed leveling can only compensate for about 0.5 mm of Z-axis difference from the lowest to highest points on a bed. And that 1.0 mm is too much variation for automatic bed leveling to compensate for.

I wanted to add an answer to my question as it would have saved me hours of time wasted on running tests on my printer while thinking 1.0 mm compensation was well within the capability of any automatic bed leveling scheme.

I wanted to add text from the Marlin website regarding the exact problem I am having above. The Marlin comments align with Oscar's answer which I have already accepted.

One Side High, One Side Low

• Problem: Bed probe done, leveling compensation enabled, and the first layer comes out out too squished on the right side, or too high on the left side, or vice-versa.
• Solution: Make sure the probe is always perpendicular to the bed as the carriage moves. If the wires or Bowden tube are pulling on the carriage, or if the X gantry is even slightly twisted, the probe will not remain perpendicular and its readings will be garbaggio. There is a feature called MECHANICAL_GANTRY_CALIBRATION that can be used to compensate, but it is always best to tackle any mechanical issues first.

(If this is not the case I will gladly modify my answer based on any additional comments. Thanks for all the feedback!)