Is there a better way to calibrate?

A quick note. This is not about the auto leveling. That seems largely ok, or at least within margin of error.

This is a question on using the card on the final point of the autoleveling process. I find I cannot for the life of me pick the correct point. I lower the nozzle to the card, feel some tension, lower some more, the card creases. Then just to be sure, I raise and lower the nozzle by the same amount again, card doesn’t crease.

When I got to a point I was happy with, card creasing and decent tension, I ran a print, and found it was over extruding (seeing squishing), So I’d increase the z-offset.

Next time I ran the calibration again, took the nozzle down to what felt like the same amount of tension. Taking the z offset from the previous time in to account, I raised it by the same amount, set and go. Result: under extruding, poor adhesion, and poor fill.

At the moment I have to essentially babysit the first layer, adjusting the z-offset until the layer seems smooth with decent fill, then let it finish. What I’d really like to do is have a way to apply that z-offset to the saved calibration, but I’m not sure how.

Side note: I am currently doing a print that has a lot of travel across the layer which is leaving track marks. It appears there is no Z-retraction on travel commands, or it is not enough, or I’m still over extruding? OR is it related to the calibration?

Hi ! This description is backlash compensation. This feature doesn’t work yet on SM2. I don’t know why they are not fixing it. They are taking a lot of time with Firmware update on not important feature but they need to focuse on Bed leveling and Backlash !

You might also be running into the effects of the aluminum frame heating up. It might be fine for the first run and then after the aluminum frame heats up, it gets closer because the frame heating up moves it upwards. You might try heating your bed up for 30 min. and then level.

Thank you, Isaac, for that answer. I’ve ran into this exact situation with my A250 as well. It was still present as of Firmware version 1.8 when I last calibrated (haven’t tried with later firmware). At least now I know I’m not going crazy!

I know this post is a bit old, just wanted to add that I do have backlash compensation up and running.

The latest firmware has backlash compensation with M425 enabled, the only caveat is it’s not able to be stored in firmware. That’s no big deal though as I created a custom “Home” macro in Luban that contains the M425 command, as well as adding it to my slicers so all gcode files contain the appropriate command before it homes.

The syntax is here:

I measured Z backlash using the calibration card and sending + and - Z movements in custom increments of 0.01mm. I measured X and Y using the following model:

I found each axis on my machine needs 0.03mm of backlash compensation (very good, in my opinion!) so my files contain the following:
M425 F1 S0 X0.03 Y0.03 Z0.03

Your mileage may very! Good luck!


@brent113 That’s awsome! Do you know if the G29 Auto calibration for Z works also? MEASURE_BACKLASH_WHEN_PROBING

It does not - and worse, if you do a G29 it will almost certainly crash the tool into the bed. The only way to trigger an autocalibration via gcode is detailed here: How to do a larger than 3 x 3 matrix level

MEASURE_BACKLASH_WHEN_PROBING is not defined in this firmware, and running G29 will not do anything productive on this machine, unfortunately.

However, it’s fairly easy to measure Z backlash manually, just by feel with the card, and by setting Luban Z jogs to a custom value of 0.01mm.


Thanks, I won’t try it then :grinning:

@brent113, how can you get a good level bed when the backlash isn’t compensated for? Seems like a chicken/egg scenario. I do an auto 5x5 and only the very middle of the bed has good first layer. Everywhere else the bad is too low. I switch to manual and the right top and bottom get closer to a good first layer but the the middle is too close. The left top and bottom still are too far away. The more manual levels I do the closer it seems to get but I don’t know how much backlash is causing.

I have a macro I run when I start the machine:

This turns on backlash compensation, and you can run auto bed levelling from the touchscreen.

That being said, backlash doesn’t matter if you always measure in the same direction, which the bed probing does - it always triggers as it moves towards the bed. That means all the measurements will be accurate relative to each other, regardless of backlash.

The only thing that could change is the z height at the very end - if you set the height as it’s coming down or going up, you’ll get different numbers. If you turn backlash compensation on first, then that’s not an issue.

From what I’ve seen with the bed’s level - mine is too far out to even try to fix. You could shim it to bring it up to level, but I think it’s beyond software compensation. I think you would need at least a 10x10 grid to accurately map the bed surface, it’s like the moon.

My solution is to just buy a glass build plate and attach that. I got fine prints using the center 150mm of print bed, but anything larger than that I had your issues - corners way too low.

Other than that, use a raft - that will make a level surface for you, at the cost of a lot of wasted filament :confused:


Thanks a ton, @brent113. I think i have a grasp now… So basically I need to:

  1. Get a glass build plate.
  2. Manually calibrate.
  3. Print backlash test print.
  4. Set M425 macro with my specific compensation settings.

Wouldn’t it be great to have a magnetic glass build plate? :joy:

I agree, that’s why I have one! The factory provided removable build plate sits on it just the same.


Sold! You can still use the auto-level i would assume?

Yes, the inductive sensor references the spring steel core of the removable sheet, not the PCB underneath.

It’s worth repeating the obvious here: glass is a thermal insulator. This will drastically affect bed heat up times and maximum temp. Other printers, like the Ender, it’s typical to put insulation under the heated bed to help.

I have other plans with a silicone 750W AC heater, spare cables are on order and when they show up I’ll assemble it all.

Recommendations for glass?

I bought this glass:

It was a compromise between:
+ Pretty close size fit, a bit longer even.
- Losing some width on the build surface
- Having screw holes, but the magnetic sheet and removable build plate offer decent cover over the holes
+ Not having to buy an oversized plate and cut it down myself
- 3mm thick, wish I could’ve gotten 4
+ Good price

If this doesn’t work out for me in the long haul I’m going to get a 4mm borosilicate plate and a diamond cutoff wheel and just cut my own custom plate.

Here it is set on top of the build plate for size reference (currently set up for laser with the machine):


Nice. Why 4mm? Strength?

Brent, I would love to see your current setup. My bed inconsistencies are preventing me from pulling off larger builds and glass seems to be the way to go.

Are you using the heated bed, then EZFlex magnetic plate, then the Tornado glass? And then the Snapmaker Build plate on top, or is that out of the equation at this point?

You have some great posts that have saved me a lot of time. I very much appreciate that!

1 Like