4
$\begingroup$

I've never used a printer with auto bed leveling, but my understanding is that most or all of them don't actually level anything, but rather compensate for the unlevel bed in firmware by transforming the coordinates. It seems to me this would necessarily introduce aliasing artifacts in all your prints by making it so that line widths are no longer a whole number of microsteps - I immediately observed such an effect on top/bottom skin when I tried to use steps-per-mm tweaking in firmware to compensate for a dimensional error rather than fixing the mechanical source of the error, and concluded that it was a dead-end approach.

If the auto-leveling firmware only makes adjustments with the Z motor as it moves, rather than transforming all three coordinates, it seems like that would be mostly or entirely mitigated, but with slight errors in dimensional accuracy dependent on how tilted the bed actually is.

Do any of the printers with (or add-on kits for) auto bed leveling actually level the bed mechanically, with servos attached to the adjustment knobs? Why isn't an approach like that more common?

$\endgroup$
  • $\begingroup$ I'm not sure if there's any commercial software that does this, but a well-designed raft can be a level surface, even on a unlevel bed $\endgroup$ – Daniel M. Jun 28 '19 at 12:16
6
$\begingroup$

Automatic bed levelling is not magic; it still requires you to level the bed properly (as level as possible). The upside of automatic bed levelling is that it compensates for small deviations like a slightly slanted surface or a (somewhat large) dent in the surface (as long it is probed and can be digitized by the firmware). It will keep the nozzle at a distance to the bed that it maintains proper distance to the bed for the filament to adhere properly (first layer adherence is key for successful prints). The slight imperfections are smeared out over about 10 mm (set in the firmware), this way you do not need transformations for the whole print (so if you deliberately make the bed very skew, the print will follow the Z axis, not the direction perpendicular to the bed).

While systems to level or align the bed exist, it is not very practical and expensive as it requires more parts, that is why it is not commonly used. Apart from the suggested printer in this answer, printers with e.g. 4 ball screw Z movement lead screws exist (mostly printers for companies, not for use at home); ball screws are way more expensive, but also way more accurate than trapezoidal lead screws. A low accuracy is preferable as such systems generally have no guiding linear rods (as that would mean that you fix the plane/alignment of the build platform!).

| improve this answer | |
$\endgroup$
  • $\begingroup$ So does it just adjust the Z motor as the print head moves, without making any transformations to X/Y? $\endgroup$ – R.. GitHub STOP HELPING ICE Jun 27 '19 at 22:26
  • $\begingroup$ Regarding cost, it seems like servos for three adjustment points, using a the touch sensor mounted by the nozzle as feedback for control, wouldn't be terribly expensive. But maybe it's just unnecessary and unnecessarily complex it the bed is manually leveled close enough that the adjustment in firmware works with only local Z correction. $\endgroup$ – R.. GitHub STOP HELPING ICE Jun 27 '19 at 22:29
  • $\begingroup$ @R.. Yes, only transformations in Z. Also the more parts the more complex. The K.I.S S. principle is not that bad. :) $\endgroup$ – 0scar Jun 27 '19 at 22:32
  • $\begingroup$ I'm also a bit confused with how it would help with a dent (or worse, warping) in the bed itself - even if it were corrected, the bottom surface of your prints would be correspondingly messed up, and/or the warpage would propagate all the way up the layers. $\endgroup$ – R.. GitHub STOP HELPING ICE Jun 27 '19 at 22:32
  • $\begingroup$ There’s only so much you can do for a bed that isn’t flat. You have to either compensate for a dent or bend in some way or get a flatter bed. $\endgroup$ – T. M. Jun 27 '19 at 22:35
3
$\begingroup$

A delta machine does not natively have cartesian planes in its coordinate space. "The bed" is a mathematical abstraction that must match the physical bed plane. Small mechanical errors or offsets distort the expected planar abstraction to introduce wave, cup, and bowl artifacts. "Leveling the bed" for a delta consists of calibrating out those distortions.

A delta machine is constantly adding small positional quantification artifacts to the print. This could be beneficial, as there is a small constant noise introduced into the position of the print, rather than jumps which are a simple function of position.

A delta machine could adjust the z-axis to be perpendicular to the observed bed with no additional print artifacts.

| improve this answer | |
$\endgroup$
  • $\begingroup$ Thanks for sharing your experience with, and insight in Delta robots! These machines seem to get less attention. $\endgroup$ – 0scar Jun 28 '19 at 15:03
  • $\begingroup$ I have worked with "delta" machines, off and on, for about 20 years. They are interestingly different. Although not new 20 years ago, the availability of faster computation resources has eliminated the difficulty of controlling the kinematics. Better optimization algorithms improve the ease of calibration. $\endgroup$ – cmm Jun 28 '19 at 15:07
  • $\begingroup$ Yes, this is a nice addition to the answers! I've actually been wondering about how Deltas get this acceptably right; I would think you'd need much finer-grained stepping. I've been vaguely thinking of trying to make a fully-3d-printed Delta machine. $\endgroup$ – R.. GitHub STOP HELPING ICE Jun 28 '19 at 15:16
  • $\begingroup$ Very interesting! Please keep sharing, I think more people like to read about this. To explain, I meant as hobby machines, in industry, pick and place, measuring, etc. devices are frequently Delta style robots. $\endgroup$ – 0scar Jun 28 '19 at 15:18
2
$\begingroup$

The rail core II does this by having a 3 separate stepper motor + lead screws attached to the bed. Along with a z probe this allows the bed to leveled automatically. Although this will not make the bed any flatter meaning it will only align the bed in the correct XY plane with respect to the gantry.

https://www.kraegar.com/railcoreii

| improve this answer | |
$\endgroup$
1
$\begingroup$

The only 'consumer' approach to this would be providing adjustment feedback to the user, after the bed has been probed to determine an optimal position so that most of the bed is in the same plane.

This seems to me like a 'free' enhancement, but users don't seem to be overly worried by the distortions introduced (or perform the calibrations manually already).

| improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.