Still Not Printing Level, Could Rod Length or Delta Radius be wrong?
-
Alright, printing at the same bed.g file posted except I changed it to S7.
Calibration Below:
; This is a system-generated file - do not edit
; Delta parameters
M665 L312.301 R145.735 H367.910 B140.0 X-0.127 Y0.218 Z0.000
M666 X-0.361 Y-0.131 Z0.492 A0.00 B0.00The print of the outer circle improves in that the variance now between layer heights around the circle only looks to be about 0.05mm or so, which is within the FSR margin of error perhaps.
So I guess the question is why are the arm measurement wrong?
The ball cup joints have a diameter of roughly 6.50mm, so the only way the arms could be ~312mm is if somehow the pivot point for the joints is not at the center?
-
So if I measure with a universal FSR height of -0.1mm and do 12 points at 140mm and 12 points at 70mm and a center point, I get the following calibration at S7:
; This is a system-generated file - do not edit
; Delta parameters
M665 L310.877 R145.827 H367.975 B140.0 X-0.337 Y0.108 Z0.000
M666 X-0.311 Y-0.162 Z0.472 A0.00 B0.00Then a 0.3mm cylinder height with grid infill with 130% first layer width looks like the following below (the ridges and valleys are non-existant):
-
Adding a link to the post below from the SeeMeCNC forums, it looks very helpful as the issue appears to be similar (in modifying the rod lengths, the dimensions might change on printed items, which I have not checked on yet):
-
Yes, and see my contribution to that thread at http://forum.seemecnc.com/viewtopic.php?t=11441#p102296. If calibration with S7 gives you straight lines in that test, you can correct any XY scaling error using https://duet3d.com/wiki/G-code#M579:_Scale_Cartesian_axes.
-
Thanks, it does look like this is the right fix.
A 30mm test cube came out as 29.28 and 29.31 per side with the height being correct, so I started a print I've been trying for a while now and scaled it up in X and Y in my slicer by 102.4% to account for that. I'll compare it to my previous prints and see how if the variations look better. It has several little square cutouts around 20mm across a 200mm+ Y length so any differences will show up quickly.
I suppose I'll try the M579 in config.g as follows to replace scaling in the slicer?
[c]M579 X1.024 Y1.024[/c]
On the rod lengths, if I can probe at 150mm or so on the midpoints between the towers, those are the probe points needed to get as close to the calculated rod length as possible?
If I add in a bunch of other probe points at a radius of 75, will those hinder the correct calculations for the rod lengths?
I think what I may do is do a 50mm radius, 100mm radius and 150mm radius and 9 total probe points at the midpoints around the bed. I'll run a receipt paper height test again and use those numbers for a final S7 calibration and see how it does.
It's nice to get this solved, for a while I thought it was something I was doing wrong in the printer build.
Also, on the steps per mm in the config file, is it ok to use fractional or decimal steps per mm? Say 80.15 if I wanted to calibrate to that?
-
For the auto calibration to determine the rod lengths with any accuracy, you should probe between the towers as far from the centre as possible, as long as your FSRs are still responding well (check the trigger height) and the joints are still moving freely. If those conditions are satisfied, probing at a smaller radius too won't make things worse.
Be careful: I remember Michael Hackney posting that if you probe a long way out from centre and between and outside a pair of FSRs, probing can make the bed tilt and give inaccurate results for that and succeeding measurements.
-
I'm adding the following forum post as well as a link to this post for future reference in discussing the rod length and steps per mm relationship on the "ridges and valleys" problem.
Of course, it reinforces the theory that DC42 is on every possible 3D printer forum posting about delta calibration.
-
I also got this part of the problem sorted out, please see my thread on "Easy Checks on steps/mm"
Thank you a lot for the references to the other discussions…
-
In our Spire delta printers we use a Fisher like bed leveling where the hotend pushes the bed. The solution is beautiful in it's simplicity but imposing several obvious problems like different push strength in different bed areas.
There is also a tempered glass bed always a tiny bit curvy.
I had all sorts of problems with correct print dimensions and print sticking at the beginning.
The problems went away with cutting calibration factors down to four.
I'm not sure why, perhaps the added errors of bed probing and some unevenness in the frame added up, but 7 point calibration were putting errors in rod lengths which I was dead certain about.
The prints have been longer in one tower direction and different from one calibration to another.
With 4 factor calibration these problems are gone. -
The calibration algorithm assumes that the bed is flat and the geometry is accurate except for those things that auto calibration adjusts. If your build has other types of geometric error or a non-flat bed, it will give incorrect results.
Calibrating diagonal rod length automatically is not advised unless you are able to probe points a long way outside the triangle formed by the towers.
