Auto Bed Levelling Query

I'm not exactly sure how the Marlin auto bed levelling works. I think it probes the bed and adjusts the z height in each location to compensate?

(Apologies in advance for any incorrect technical terms!)

I was thinking about an independant auto bed levelling system and have a rough plan of how to mechanise the current M90 manual system. Essentially I'd have a worm gear and micro continuous rotation servo in each corner of the bed to adjust the height. There would also be a min end stop to limit the downward travel of the bed. I would also have a limit switch mounted to the X carriage somewhere close to the hot end. This would need to be manually flipped down so it is a few mm below the hot end tip when calibrating bed height. I would need some push switches to control the servos which are limited by the end stop switches.

The procedure I had in mind is this:-

1. Flip down the z X carriage limit switch to it's position a few mm lower than the hot end
2. Lower all 4 corners of the bed to their low limit switches
3. Move the hot end to say X90 Y90 Z5 (new Z limit switch would be at approx Z1)
4. Switch the servo in the same corner to raise up the bed until it touches the new Z limit switch and stops
5. Repeat the process for each corner
6. I assume that the Z height would need tweaking unless the new Z limit switch could be accurately re-positioned each time

I have no idea if this would work and if it could be controlled by the existing board somehow. I think I could make it work as a standalone system with basic wiring and cut-out limit switches but it's not my forte, just thought I'd throw it out and see if there's any merit to it or if I'm already miles behind the times!!

Any comments welcome of course

That is a very complex solution.

You can level the Menderl90 bed very accurately using a cheap dial gauge (I get mine to within a few microns between corners).

There is one distinct disadvantage to auto-levelling over the manual method and that is you end up with a print which exhibits sloping layers (due to the compensation).

I prefer my prints to have layers that run parallel to the bed surface and manual bed levelling gives you that for what is really a little effort.

Regards,
Neil Darlow

---

I try to write with consideration for all nationalities. Please let me know if something is unclear.
Printing with Mendel90 from fedora 25 using Cura, FreeCAD, MeshLab, OpenSCAD, Skeinforge and Slic3r tools.

Hi Neil, yes it is complex compared to the manual method. I guess I'm looking more at the development of the printers to make them simpler to use (overall!) and bed levelling, although simple, is not something I'd expect an end user to have to do. Similar to aligning a laser printer I suppose. While software isn't my area, I'd imagine it would be reasonably straightforward to implement the procedure above into an automated one.

I assumed that the existing auto bed levelling system in Marlin would have a funny effect if it wasn't level in the first place as the Z will always be vertical regardless of the build plate orientation.

I have no real issues with manual bed levelling, it's just one of my small list of things I think 3D printers need to make them more accessible and generally easier to calibrate. A few of the others are:-

1. Bed glass clamps which are not office bulldog clips
2. Or a build surface which is suitable for several materials (I've ordered a GeckoTek build plate which may turn out to be good... hopefully!)
3. Auto filament size detecting
4. Auto pause / stop on filament out or significant snag up. I've had my X carriage pulled halfway up the machine a couple of times due to the roll getting a little tangled!
5. Auto Z height calibration

Just a few thoughts that go through my head while I'm using the printer!

Ross

If you don't like bulldog clips but you can use 3mm picture clips instead. They allow the full build area.

---

[www.hydraraptor.blogspot.com]

Hi Chris, I'll have a go at an ABS cam type clip over the next few days but I'll look at the picture clips too so thanks for the tip. With regards the bulldog clips, I was thinking something printed as it makes sense to me that all the other parts are bespoke for the machine seeing as it has the ability to make them. Assuming they don't go all soft anyway!

I might be missing something (as usual ) but I don't see how 3ddue's idea would cause any sloping in the print layers?
If I understand it correctly this solution would mechanically level the bed so it was perfectly flat, presumably before each print? No changes in Z height would be necessary within a layer. It looks to me as if it's intended to simply automate on a per print basis what would otherwise be a manual job every week/month/3 months, whatever.
I could of course have this completely wrong.......
D

Quote
maddog7
I might be missing something (as usual ) but I don't see how 3ddue's idea would cause any sloping in the print layers?

I believe the algorithm for compensating for an non-level bed only does so on a layer by layer basis. If it's only slightly out of level, you probably wouldn't notice or the effects would be minimal. But if take it to an extreme, imagine a bed that's 45 degrees out of level and you were printing a 10cm cube. Instead of having sides that were perpendicular to one another, your box would end up being more like a rhombus with each successive layer being shifted making for sloped sides.

The idea behind the auto leveling is to get it close, but it doesn't have to be perfect. While getting it perfect in the first place is probably best and reduces the amount of number crunching the controller needs to perform, if the printer gets moved around, prints in different environments that would require recalibration, or the design just isn't conducive to dialing it in and having it stay calibrated, auto bed leveling is another option.

The levelling will rotate the model to match the plane of the bed, so a cube will still be a cube. If you did it at an extreme angle the only thing that would be wrong is the nozzle tip is not inclined, so the top surface would be stepped instead of flat.

---

[www.hydraraptor.blogspot.com]

I guess I'll stand corrected.

If it's any consolation, what I was proposing was purely to automate the existing manual bed levelling to make it simpler for an end user. I don't know too much about the software based auto levelling feature but I'm impressed if it corrects the vertical angle like Nophead is saying. I'd imagine that would make slicing for overhangs etc a bit more complex!

I also found a 3d printer which appears to have covered some of my suggested general 3D printer improvements earlier in this thread - [robox.cel-uk.com]

Auto bed levelling, auto z-height etc. Not open source and reviews aren't great regarding actual print quality but for me it's a step towards mainstream users.

The Robox produces some amazing quality prints out of the box. The print quality of the beta machines was a bit hit and miss (I think the needle valves didn't work as intended and the firmware wasn't making use of the feedback loop) but it incorporates a lot of nice technology that if nothing else, makes 3D printing something than anyone could do without first having to take a degree in thermoplastics.

@3ddude,
The three point levelling in Marlin does the same as I pioneered here: [hydraraptor.blogspot.ca]. The grid version fits the best plane through many points using least squares and then uses that plane to rotate the model in the same way.

---

[www.hydraraptor.blogspot.com]

@ Quackingplums - I have only read a couple of reviews and the quality was an issue so glad to hear it has been improved. There are times when I'm calibrating that I would like to have all the auto setups. But I guess that now I know a lot more about my M90 and printing in general though!

@ Nophead - Nice work! Sounds like a great solution