Have anybody tried the autolevel with normal Cartesian machine(not delta) ? have you guys found any difference ? I see there should be slant in base it wont be right angled.
Tried it wit adjusting the Z for the angle of the bed. Didn’t like the fact that the Z is moving all the time because of it.
Also tried by using the nozzle&metal-bed as “endstop” which didn’t work reliably.
The “auto-leveling” (which is a horribly inaccurate description of what it does) algorithms fully rotate the coordinates in 3d space, so the printed object will have the correct angles, but will be reoriented to the orientation of the platform. Note that the algorithm does not correct for a cartesian machine where the axes are not square (at right angles to one another), only got a platform that is not parallel to the X/Y plane.
It’s a nice idea, but if you actually look at how the axis movement is discretized, you’ll realize that the process inherently introduces artifacts that reduce print quality (of course, whether these artifacts are actually visible depends on whether there are any bigger print quality issues to hide it). It also makes Z backlash a huge concern, where it is usually not something we need to worry about because in normal operation, the Z axis will always hit a layer from the same direction.
I am concerned about the angle between walls and base (say x-z or y-z axis). also with high pitch leadscrew there will be additional step included. my other concern is if the bed is near flat and if still z compensation is performed will it give worse results? I havent found 1 comparison photos of with and without z compensation.
The algorithm does handle the transformation properly to avoid issues with the x/z and y/z angles, but as I said, it only compensates for the platform being out of tram, not for axes that are out of square.
The math says that the artifacting should be worse for a near-tram platform than for one that isn’t even close, though at some point nozzle geometry becomes a factor.
so you mean if there is 5 degree tilt then object is built 5 degrees tilted? means a straight tower wont have layers on top of another?
Correct.
What does the autolevel algorithm do if the platform is not flat BUT it is well leveled?
I mean, for example, in the four corners the platform is at same distance from the nozzle, but in the center it’s not.
Im running autobed on my mendel90 and it think it works great!
@Samuel_Ureta the standard algorithm probes three points and assumes that they create a plane. There’s a another version that i believe probes 9 points, but it still assumes that the platform is flat. I’ve only seen that one on deltabots though, so it might be adjusting the cartesian -> delta coordinate transformation, which could result in layers following the curvature of a bowed platform.
I’ve seen other algorithms for probing a grid over the entire work area to detect a non-planar surface, but in order to compensate for that, you need slicer-level integration so that you can build up a planar raft on top of the non-planar surface.
@Whosa_whatsis oh, I believe that the 9 points one was for that.
I’m thinking about an approach that maybe doesn’t need slicer-level integration: The firmware does the probing and it creates a matrix of heights, and in the first layer it moves z-axis following the surface. So does in second layer but the z-moves are multiplied by a factor <1 so it simulates a surface a little flatter. Maybe doing this in 3-4 layers (customizable) can lead to a near flat surface and the rest of the gcode can be interpreted as usual, and the error would be only on the bottom layers and it would be very small, assuming non-flat printing platforms have very small differences in height (but sufficient for adhesion problems).
Think it’s a good idea?
@Samuel_Ureta flow will also need to be adjusted and with dynamic layer height for first 3-4 layers effects can be interesting. Also if first layers are not removable raft then again print will be bit skewed. From whosa explanation i dont like the idea of layers not being on too of another
update on auto level. i tried it today. took more time then i though due to the dimension change in machine and as i use max endstop for x and y there were marlin issue. fixed by getting latest marlin.4 probes are not enough because my psu is bit weak and due to 2 fans and led and servo the servo twitches.
@Whosa_whatsis there are some vertical artifices but the prints are more smooth to feel then before.