Uneven calibration on Kossel XL

Hi everybody - I assembled a Kossel XL printer with the aid of the beautiful DC42's guide he published on his blog, as well as through reading many of the builds I found on this forum.

I had to assemble it and disassemble it two or three times before being able to sort out most of the issues (defective parts didn't help, either).

I got to the point where (https://duet3d.com/wiki/Calibrating_a_delta_printer):

-towers are parallel to eachother: linear level says that all of the towers are perfectly straight on both sides, so, although the couples might not be equally spaced, the statement should still be valid.
-rods are the same length: I'm using the magnetic arms and balls which came with Duet3D Smart Effector, so I'm assuming those are the same length, as stated on the rods itselves.
-spacing between the ball studs on carriages and effector are the same, within 0.05mm, checked with digital calipers.
-carriages are not warped or rotated around z axis: I'm using the PCB carriages supplied with the Smart Effector, and linear rails/sliders, so this shouldn't be the case.
-there is no play in the joints.

-the upper and lower triangles are not perpendicular to the towers: I had to tilt them for tensioning the belts (as suggested on ultibots manuals) and for achieving the perfectly straight towers (Robotdigg metal corners were not perfect, sadly). That shouldn't be an issue as long as the towers are straight. Of course since the lower triangle is slightly tilted, the print plate which I mounted on it was tilted as well.

And here we get straight into the problem. I added some washers for compensating the lower triangle tilt and making it perpendicular to the towers. Checked it with a linear level and it says the bed is flat in all the directions. I then run autocalibration, followed by a mesh grid compensation, and no matter how I adjust the pcb carriage adapters, the heightmap I get is always the same (I even printed a round 0.15mm thick layer as wide as the plate, and it indeed resulted in the heightmap layer pattern)



By going through some random tries, consisting in adding or removing 0.5mm washers from the 3 spacers the bed is mounted on, I was able to achieve this



But the first layer still looks very uneven.

One thing I want to point out, is that although the linear level (it's a 20cm level) says the bed is flat in all the directions, the small bullseye level which came with the smart effector shows different tilts as I move it around the bed.



Any idea of what I'm doing wrong or what I could try? Could it be the aluminium plate not being fine machined? I thought I could try getting a round glass I could put atop of it, but I decided to hear some opinions before investing more money in this.

Thank you for your patience,
Alu


[Edit: I posted this in the wrong section - I was trying to post in the Delta Machines section. Could someone please move the thread? Thanks]

Edited 2 time(s). Last edit at 02/16/2018 06:33PM by Alu.

I don't use mesh bed leveling on my Deltas anymore, although the XL 35cm bed isn't milled flat either. Your heightmap looks good, did you try to print without MBL? What's your 1st. layer height?
I calibrate the printer every day with the probe points in bed.g and thats good enough.
Adding a second instance of leveling is only worth it, when the first instance is spot on.

Hi o_lampe, and thank you.

Indeed I tried with both MBL enabled and disabled, and best results were obtained with it being disabled. I'm therefore only using the heightmap as a visual reference for seeing what's going on with autocalibration.

I tried to print at 0.15mm first layer height until now since my aim is to print in the range of 0.075/0.05 mm - currently trying a dual process slicing in Simplify3D where I set it to print the first couple of solid layers at 0.2mm and then switch to a higher resolution.

I am now using 7 factors calibration with 17 points, and this is the result:



15:53:25 Calibrated 7 factors using 17 points, deviation before 0.032 after 0.019

16:11:52 M665 Diagonal 363.509, delta radius 180.321, homed height 471.535, bed radius 140.0, X 0.037°, Y -0.056°, Z 0.000°

16:11:46M666 Endstop adjustments X0.02 Y-0.11 Z0.10, tilt X0.00% Y0.00%


Which don't look that bad to me, but the first layer still looks rather uneven. Currently printing another 0.15mm circle to see how it looks with the latest calibration results. Uploading some photos soon.

Quote
o_lampe
I don't use mesh bed leveling on my Deltas anymore, although the XL 35cm bed isn't milled flat either. Your heightmap looks good, did you try to print without MBL? What's your 1st. layer height?
I calibrate the printer every day with the probe points in bed.g and thats good enough.
Adding a second instance of leveling is only worth it, when the first instance is spot on.

Here are some photos of the prints I made.

#1 try:



#2 try:





Edited 2 time(s). Last edit at 02/18/2018 07:53PM by Alu.

All I can think off is adding height correction to the probe points on the right ( H-parameter in the bed.g file ) and hope it will influence the bed mesh data, too.
You could also edit the heightmap file itself, but do on your own risk...

Quote
o_lampe

All I can think off is adding height correction to the probe points on the right ( H-parameter in the bed.g file ) and hope it will influence the bed mesh data, too.
You could also edit the heightmap file itself, but do on your own risk...

That is indeed something I haven't tried after the last build of the printer.

I didn't do that because I noticed that when I did that previously, I get inconsistent probing trigger heights. They can vary up to 40/50 um. Not sure why that's happening. So I just left the default 0.1 trigger height for now.

I will try editing the height map manually, definitely sounds like fun. Hopefully I won't crash the nozzle too badly


Also, just noticed Duet 3d wiki has been renewed, I love it!

I know this might sound dumb, but, mesh grid compensation has to be run after autocalibration , am I correct?

I've always done so, as stated on the Duet 3d wiki - but I'm unsure because it's even stated that you should cancel every bed transformation before starting the procedure.

The mesh grids i obtain before and after calibration are, obviously, very different.

This is what I get if I run the heightmap calculation right after manual endstop and delta radius calibration (Tower X,Y,Z and center are at the same height when at z=0 height).



I guess this clearly shows that two of the towers are misplaced. Autocalibration should compensate for this, so, It should be correct to run the mesh grid calculation after running autocalibration (?).