In my last post I had decided to stick with 3 point levelling. Although that seemed to be working for smaller prints, I was having spots with really poor adhesion on the outside edges of the bed. Presumably due either to flexing up/down or irregularities in my glass.
As a consequence I’ve switched back (at least for now) to full grid levelling. It adds about 3 minutes to the start of the print, which isn’t unreasonable.
First thing I noticed on my first larger print (a mount for the Sbase, TL Smoothers and a Mosfet breakout board spanning about 280mm) was just how much the z was moving on long travels. I watched the level through the terminal:
G28; Home, not strictly neccessary G1 Z0 Y0 Z10; move to first probe point. Note, my printer doesn't home to 0,0 G32; You'll get outputs like DEBUG: X0 Y0 Z1.32424242 or similar
and found that I had almost a 2mm of runout in X direction, and nearly 3 in Y. That being said, the machine did do a fantastic job of compensating for it – look at the first layer:
However, it’s clearly suboptimal, and appeared to be slowing down long moves so Z (which has a much lower max feedrate) could keep up. Fortunately I still have the bed on screw ‘towers’ rested on an M3 Nyloc, so it was super easy to use the probe to get a more or less level bed. Open up your favourite terminal (I use an old Nexus 10 with a dud batterty on the wall in my printer cave/garage with the octoprint touch interface on) and enter the following Gcode (no semicolons or after, they’re just my comments)
M561; Clears the current compensation settings (I won't admit how many times I did this before I realised I still had the compensation on!) G28; Home all axis G1 X80 Y80 Z10; Move to a sensible first probe point G30 Z0; Probe the bed and set that to Z=0 G1 X240 Y80 Z10; Move to 2nd probe point G30; Watch the reading here, keep sending G30 and twiddling the knob nearest till you get about 10 G1 X160 Y240 Z10; 3rd Probe G30; Watch & twiddle again G1 X80 Y80 Z10; G30; See how far off 10 this is. If it's significant start from G30 Z0 again
This is a bit of a faff I’ll admit, but considering how far out of level I knew my bed was (and not having springs) I didn’t really want to just jog it at z=0. If you’re confident about knowing where the high spot is you could do that instead. Do yourself a favour as well and print some thumbwheels for the nylocs, it’ll save youself all types of effort.
I managed to get mine to about .3mm difference across the probe points, which gets me to +- 0.4 across the whole bed. A number I’m comfortable with considering the size (the centre 200*200 is about 0.1). I don’t think my set enough is either innately flat enough or stiff enough to get reliably closer to 0.