Next time I get back to my printer,

Next time I get back to my printer, I will get to again go through the pleasure of bed leveling. My printer started life with a 6mm plywood print bed, quickly adding to that a heater PCB and I think 5mm glass clipped directly to the bed. I was having such issues with getting the bed level I decided that the plywood base must not have been stiff enough and was bending, so changed that to 12mm. I was still having issues so I mounted my heater PCB up off the bed and clip it and the glass together.

My problem remains that no matter which way I level the bed (this is a Sells Mendel with printed bed springs for adjustment), I always end up with high and low spots, as if the glass is dished. It’s always enough to screw up the first layer if I print on about 30% of the surface. Any thoughts on likely causes? Glass came warped or became warped with heating (standard replacement glass from a hardware store)? Smooth rods are bent or are flexing causing the X and Y carriages to slightly change height throughout their travel? Are there any slicing tricks to try to deal with this, like adding rafts in certain areas at -Z levels?

Do you have any idea how much warp there is in your bed? I’ve been using a 3mm brushed aluminum plate, which has some warp, with ABS juice but I run my Z really tight to the surface. 0 is just a hair below 0 (0.2mm layers).

Before I start a layer I move the Z up 10mm and then run the gcode and xyz meet at the start point so the print head isn’t dragging without extrusion. The first layer comes in really squished and tight across the bed which requires a little bit of trimming on the face surface when done but no problems with lifting or adhesion.

Plywood is a poor choice for the bed base, as it tends to warp when exposed to heat. I’m using 6mm thick FR4, which is very stiff and temperature stable, yet still fairly cheap. For a regular-sized bed (mine is twice as wide) 4mm should do, which cost less than 10 bucks. You can work it just like wood or metal - the same tools cut it nicely and you can even tap threads into it.

Does the bed material matter too much when the PCB and glass are on standoffs away from it? I made sure there was enough slop in the PCB mounting holes so I don’t think it’s being pulled by the base warp.

Due to the way it’s made, float glass (aka totally regular glass) will be almost perfectly flat. In my experience i’ve never noticed the glass being warped or getting warped after some heat cycles.
If your base is warped, it will always somewhat transfer to your glass plate. You can minimize this by using only three clips or kapton strips (in the right edges, namely those that appear to be the closest to the nozzle) to hold the glass down.
Slic3r allows you to set a separate layer thickness for the first layer - i’d recommend setting this to your nozzle diameter, since this is the largest layer height Slic3r will accept.

Ok, so got back to my printer. I really opened up the mounting holes on the heater PCB and left the top nuts off so that it would be decoupled from any carriage warp. The PCB now stands off the carriage and is clipped to the glass. Sells Mendel bed screws kind of interfered with the PCB, so I cut the non-conductive edges of the PCB to make room for a screw driver to make life easier.

Now that I’ve gone through leveling again, I find that the center, front right and back left corner (one diagonal of the print surface) are pretty well leveled with each other (with my current z opto setting, between .175mm and .200mm), however the other two corners are both lower, about .325mm and .340mm. This makes me think that the Y carriage is rolling slightly from one side to the other as it moves from front to back.

Initial guess is that the mounting pieces for the Y smooth rods are slightly different (since they each required some cleanup before assembly) and are holding the rods at slightly different heights from one end of the printer to the other. With a little more thought, I’m guessing now that my triangles are not perfectly equilateral. I will test the hypothesis another time.

Well, indeed moving my vertices along the structural rod allowed me to undo the roll along the Y axis. The four corners and center are now within 0.03mm instead of 0.15mm as before. Very exciting!

Next up: Whosa whatsis’s article on correcting Z artifacts.