1
$\begingroup$

I've been fighting on and off with issues with my build plate seemingly not being perfectly flat. I've ordered a machined straight-edge and some new feeler gauges to diagnose the problem (and test any potential replacement) better, but I'd like to get an idea of what options are available for obtaining a flatter build surface. I'm not really interested in approaches like BLTouch since I want accurate flat bottoms on my prints, not just consistent adhesion.

My printer is an Ender 3 and I believe my problem is a mix of wear to the Buildtak clone surface and warping in its backing plate. Mine is the newer model with removable buildplate and clips, and the backing is reportedly a "glass fiber" (FR-4) material that's hard to find genuine replacements for.

Glass seems like an obvious option - I might even put a buildtak clone on one side of it and use that most of the time - but I'm somewhat concerned about weight and whether having glass moved by the Y axis is going to limit acceleration. I've had (seemingly unrelated) problems on and off with layer shifts, which seem to be mostly Marlin's fault not mechanical, but I'm scared to introduce another factor that might encourage them.

Are there other non-glass options I should consider that would provide a rigid, flat backing? I'm very happy with the buildtak clones and their adhesion properties, so my leaning it to look for something that makes a good backing for them rather than a material that's intended to be built directly upon.

Measurement results: Using a machined straightedge and feeler gauges, the surface seems to be at least 0.10 mm but less than 0.15 mm lower within a couple inches of the bed center. This is plausibly just wear on the buildtak-clone surface, so I think I'll try to see if I can get by with just replacing that for now. But I'm still interested in the general topic of the question.

| improve this question | | | | |
$\endgroup$
  • $\begingroup$ Why should you put Buildtak on a glass sheet? Glass is a perfect bed material in itself... I even can get POM to stick to it. Also, the additional weight has never been a problem for my i3 type printers, even with an enlarged build surface of 200 x 300 mm with 3 mm aluminium heated bed plates. $\endgroup$ – 0scar Feb 12 at 7:25
  • $\begingroup$ @0scar: Even with low or no heat? I agree glass has some nice properties especially for special materials, but they seem like tradeoffs. Even with the same kind of surface on top I wonder about things like thermal mass. $\endgroup$ – R.. GitHub STOP HELPING ICE Feb 12 at 13:18
  • 1
    $\begingroup$ I would just stick the BuildTak onto the aluminium or a sheet of steel. $\endgroup$ – Trish Feb 12 at 17:36
  • $\begingroup$ @Trish: I was actually wondering if a sheet of steel would work well. I'm hesitant to put it direct on the aluminum plate there because it's a consumable part and I'd probably warp/scratch/otherwise damage the aluminum plate next time I have to replace it. What thickness of steel would be suitably rigid? $\endgroup$ – R.. GitHub STOP HELPING ICE Feb 12 at 17:46
  • $\begingroup$ uh... depends. 0.25mm is super light and can be easily cut with shears, it doesn't add too much thermal mass. 0.5 mm is less floppy and more durable. Springsteel is more durable than mild steel, but more expensive. If you dent it, you might be able to fix those two thicknesses even. 1mm is kinda heavy duty and doesn't budge easily, but you need someone to cut it for you. and you likely can't bend-pop it. $\endgroup$ – Trish Feb 12 at 18:51
4
$\begingroup$

It all depends on what you mean by "flat".

Is the problem that the build plate isn't flat (perfectly planar), or is the problem that the distance to the build plate varies based on X and Y coordinates? They are very different problems.

"Bed Leveling" is the process of allowing the firmware to know the Z position of the build plate for every (X,Y) location. Some printers support measuring a mesh of points and interpolating the mesh. This compensates quite well both for warped build plate, and for bad positioning mechanisms which add some position dependent Z-axis offset.

If you really need a flat, planar, build surface, perhaps because you are printing mirrors or precision parts, glass should work well. It is still, and will not brook any inelastic bending. It will shatter before it takes on a curve. If possible, you could consider a Pyrex® bed, since it has a lower coefficient of thermal expansion and won't warp as much with a temperature gradient through the glass.

I tried to find a spec on how flat your glass might be. I'm assuming it is "float glass", which is made by floating molten glass on a pool of molten tin. I didn't find a spec, but I found this answer to a similar question:

I doubt you will find such a spec- float glass isn't manufactured for that purpose and it isn't really in the manufacturer's interests to maintain such a spec. that said, the stuff is surprisingly flat, just as a side effect of how it's made. you don't say what your needs are other than near optical tolerances . might just be good enough for you. consider though that granite surface plates of certified flatness can be had fairly inexpensively these days.

So, it is clearly flat. I can't tell you how flat. And the flatness depends on not applying stress to the glass that causes it to bed. Thicker is stiffer (probably at the third power of the thickness ratio).

| improve this answer | | | | |
$\endgroup$
  • $\begingroup$ While I mainly want consistent distance from nozzle to bed for proper adhesion independent of position, I do want decent flatness/dimensional accuracy too. If there are problems with the bed large enough in magnitude to mess up the former without position-dependent Z, then with position-dependent Z they're going to noticably mess up dimensional accuracy, flatness of surfaces, extrusion amounts in the first few layers, etc. I really don't want to be fighting with interaction of all these things. As stated in the question I'm going to investigate with a machined straight-edge to determine... $\endgroup$ – R.. GitHub STOP HELPING ICE Feb 12 at 17:18
  • $\begingroup$ ...how bad the bed actually is, and whether there are other plausible mechanical causes (worn rollers? play in gantry? etc.) that might be behind the apparent irregularities, but so far there don't seem to be. $\endgroup$ – R.. GitHub STOP HELPING ICE Feb 12 at 17:20
  • 1
    $\begingroup$ BTW I would try to get borosilicate glass if I do, but it seems hard to find ones that aren't fraudulently mislabeled. (Unrelated but interesting: even Pyrex brand glassware is fraudulently mislabeled - it's not pyrex anymore.) $\endgroup$ – R.. GitHub STOP HELPING ICE Feb 12 at 17:30
1
$\begingroup$

You are almost certainly attempting to optimize a component rather than optimize the system. The quality of the final produced prints depends on the combination of error sources. In brief, this means there's no point in getting a bed whose flatness is, say less than 50% of the variation in Z-axis positioning, in extrusion flow rate, and so on.

It is true that it's nice to have a material which won't warp over time so recalibration isn't necessary. Variation with temperature is unlikely to matter unless you anticipate making prints with a large difference in the temperature setpoint (maybe for radically different materials), and even then you should be able to calibrate for a couple different setpoints.

| improve this answer | | | | |
$\endgroup$
  • $\begingroup$ I don't see how this answers the question. Error in flatness of bed is orders of magnitude larger than error in filament diameter, Z positioning, etc. $\endgroup$ – R.. GitHub STOP HELPING ICE Feb 12 at 17:09
1
$\begingroup$

Even with glass, some warping may still remain.

One option is to use this procedure:

Basically you first measure the flatness, then you add material below the bed to compensate.

| improve this answer | | | | |
$\endgroup$
  • $\begingroup$ Nice! That's basically exactly what I did with the existing non-rigid buildplate (see my self-answer) so I'm glad to know it's a thing ppl do. $\endgroup$ – R.. GitHub STOP HELPING ICE Feb 17 at 17:19
  • $\begingroup$ It's not an answer, but if the resulting non-evenness is limited to less than a layer thickness, I would use mesh leveling with progressive smoothing, so that after a few layersthe hotend goes flat. $\endgroup$ – FarO Feb 17 at 17:42
  • $\begingroup$ I wouldn't, and I specifically addressed why in the question - I want dimensional accuracy and flatness of bottom of print. Mesh leveling with progressive smoothing is going to preclude that. The height of a cube will vary by x,y position. $\endgroup$ – R.. GitHub STOP HELPING ICE Feb 17 at 18:51
0
$\begingroup$

For evaluating the extent of a surface flatness problem and success of a solution, a real machined straightedge and decent feeler gauges are a huge help.

Using them to measure where the problem was, I found that a not-perfectly-flat but non-rigid build surface like the one I have can be corrected with shims, at least as long as the underlying bed is not warped. I measured the error in the center of the build surface as very close to 0.1 mm, and placed a 24x24 mm square of 0.1 mm thickness blue tape on the aluminum bed under the removable build plate, then clipped it back on. Measuring again with the straightedge showed no significant inconsistencies.

As a bonus, the feeler gauges are wonderful for height adjustment. After confirming flatness, I needed to re-level the build, and got it perfect on the first try with no trial-and-error feedback loop by homing the nozzle, then moving it to Z=0.1, and then at each adjustment point, adjusting until:

  1. the 0.1 mm feeler gauge could slip under the nozzle with some pressure, but without compressing the spring, and
  2. the 0.15 mm feeler gauge could not slip under the nozzle without visibly lowering the bed and compressing the spring

I'm still interested in other options that might be lower-maintenance.

| improve this answer | | | | |
$\endgroup$
  • $\begingroup$ You could use a BL touch to draw a map of the non-evenness to know where and how much to correct it. The BL Touch (and clones) should be accurate enough. Surely faster than feeler gauges, which anyway wouldn't allow you to visually distinguish between no-compression of the spring and a compression of 20 microns. $\endgroup$ – FarO Feb 18 at 9:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.