I'm tired of trying to tram a warped bed.

FWIW, about heated beds. https://plus.google.com/u/0/+PrestonBannister/posts/5h2n5Nr2hSW

Note the comment from Kyle offering to make a properly machined and assembled bed.

Thanks! Your point in there about steel and glass having lower thermal conductivity than aluminum and therefore having more even heat distribution makes sense to me… More fuel for my bimodal printing idea.

@mcdanlj Though you can make steel quite thin, but glass not so much - especially for a removable bed surface.

In truth, wondering if the cost trade-offs are seriously bimodal. The ideal solution would be a thick aluminum plate with a milled-flat top, then a removable thin spring-steel build plate (held by magnets). That would be the ideal solution. Dead-flat top surface, with fast heat-up.

Problem is that 400mm milled plates (for the other printer I am building) seem to be about $300 (or more). That is a lot.

Aiming at smaller cost, bought a 400mm square of sheet aluminum for ~$50 (delivered). The plate is flat (near as I can tell) in one dimension, a bow of <1mm in the other. Pressing the sheet against a flat surface, fairly small force is needed to bring the plate to flat. Attaching the plate to a bit of extrusion might be enough. Maybe.

Maybe the lowest-cost way to get a flat bed is to use glass, with an aluminum sheet glued to the bottom. Not ideal in term of heat times, but cheap and flat. Could still use a spring steel removable build plate.

Or is there some other approach?

Whether you can make it flat when it’s cold is a different question from whether it’s still flat when heated.

For me, I have no question whether I’m going to buy milled cast plate. Either ATP-5 (probably from Midwest Steel Supply) or MIC-6 (which I can buy on my way home from work) depending on total cost; they are both good enough and definitely better than rolled plate. I’m almost certainly going to use AC heat pad, an SSR, and a TCO to protect against the common SSR failure mode.

While I really like the idea of the kinematic mount, I’ll save that for another day—after I have, or have access to, at least a CNC router if not a mill. There’s enough slop in the mounting holes on the X5S bed supports that I can bury M3 in the plate with loctite, not showing any exposed screw head on the printing side of the plate, and expansion won’t hurt.

300mm x 300mm Keenovo heater: https://www.amazon.com/approx-KEENOVO-Universal-Flexible-Silicone/dp/B00V81ZI70 — same price from Amazon as the Keenovo aliexpress store. Yeah, I’ll have to wait a little while for the edges to heat up if I’m printing something big.

Bought mine through Aliexpress. (This is not for the X5S.)

I’m confused by you saying it would be $300 or more for milled plates, by the way. I see 400x400, 450x450, and 500x500 variously in your comments, but none of those sizes closely approach $300 AFAICT. 450x450mm 1/4"is $65 for ATP-5 from Midwest Steel. 3/8" 20"x20" is still under $100, and probably about another $25 or so to ship.

@mcdanlj Will admit some confusion, on this topic, in terms of process, alloy, and cost. Went through the choices months ago, and made my best guess. In the above picture I have a 450mm square aluminum bed with a 400mm heater. (The print surface is meant to be 400mm.)

Pricing the cast ATP-5 plate from MidWest, the same size square should be a bit over $100 shipped. The zinger is “flat within .015 inch” or 0.38mm … which is a bit large compared to layer height. Could get lucky. Or could need to flatten the plate, somehow. If I need to flatten the cast plate, is that still an advantage over less expensive sheet?

Punted. Ordered the sheet, to see if that could be sorted.

The “right” way to do this would be to get a proper milled top surface, but that seems to get us into the ~$300 price range.

I had the same worry, until I learned that the “flat within .015 inch” is a max spec across the full width of the uncut piece and not typical; also any variation across the full width is not representative of the variation on a cut, which I’ve typically seen described as “mirror-flat”. It’s already a “proper milled top surface”, apparently done with a huge facemill. See http://www.practicalmachinist.com/vb/general-archive/material-question-what-mic-6-a-97627-post299545/#post299545 and http://www.howardprecision.com/aluminum/aluminum-cast-tool-jig-plate-stock-list/alca-5 (Alca 5 and ATP-5 are the same for our purposes; same alloy and similar manufacturing processes, as I understand it.)

Milling rolled sheet wouldn’t help you much because the internal grain orientation and stresses lead to uneven expansion when it is heated. Cast plate doesn’t have those internal stresses. Different alloys give you different properties; MIC-6 is 7000-series (purer aluminum IIRC) and is softer and more difficult to mill; ATP-5 and Alca 5 are both 5083 aluminum-magnesium alloy and are harder and mill more easily; they have less thermal expansion and I think around 1% lower thermal conductivity. All three have consistently good results reported for 3d printing.

On a related note, did a bit of work with the TronXY X5S print bed. Added 1/8" cork to insulate the bottom. put three lengths of aluminum extrusion underneath, between the screws.

Before the print bed sagged between the rails. Adjusted to ensure the bed is parallel to the gantry. Now the aluminum bed is dead-flat, near as I can tell.

Clipped on the plastic (plexiglass?) print bed. Started dialing in the nozzle height … and found there is now a slight crown(!) to the print surface. (From the clips pressing on the thin plastic?)

Right.

Maybe I should just print on the aluminum. :confused:

A picture is worth a thousand words sometimes. I have referenced the bed supports to the frame and the bed to the supports, but haven’t yet tried to level the bed to the nozzle on average and I intentionally set Z a bit high for now. I wanted to get a sense what things look like now. This is with glass clipped with the supplied bulldog clips to the bed, with bed and nozzle preheated for about 10 minutes prior to measuring.

I entered 25 data points from a bed leveling session into a TSV file:

-0.91 -1.08 -1.29 -1.3 -1.28
-0.62 -0.78 -0.93 -0.97 -0.92
-0.53 -0.64 -0.77 -0.81 -0.76
-0.51 -0.5 -0.65 -0.73 -0.69
-0.41 -0.56 -0.7 -0.74 -0.76

gnuplot> splot ‘grid.tsv’ matrix using 1:2:3 with lines

It looks vaguely hyperbolic. There’s only so much I can do with this.
missing/deleted image from Google+

Measuring bare aluminum, the overall pattern looks similar.
missing/deleted image from Google+

gnuplot is autoscaling the axes because I didn’t provide a scale. This is useful for seeing the overall shape of the data. Here’s what it looks like after many rounds of leveling using the bed level Marlin feature to help calculate how much to turn the leveling screws (0.5mm pitch M3 screws) to adjust; trying to have minimal tension on the springs; still .4mm variance between lowest and highest points is the best I seem to be able to do with the shipped bed at 60⁰ and nozzle at 200⁰
missing/deleted image from Google+

One more round got me to .33mm total variance and .1mm in the middle of the board (middle ~150mm).

Makes me think that if I clipped mirror to the sides that do not have the “leveling” screws, I might reduce that variance further and get acceptable prints until I get a better bed. But that’s a project for another day.

I don’t have mirror available right now, but clipping glass on at the sides gives me worse variance; .61mm total and .36mm in the middle (as previously defined). This surprises me since I’ve clipped it along the relatively flat axis so as not to bend it over the “saddle” that shows in the data for the aluminum. I’m wondering whether the glass actually has a stronger tendency to warp than aluminum, because of a lower thermal conductivity relative to its coefficient of expansion, thus a larger temperature gradient from bottom to top of the glass plate. I’m curious enough to buy mirror to test, but this seems to confirm the idea of using PEI on cast aluminum.

It’s easy to find 300mm x 300mm (or 12" x 12") PEI sheet, but 330mm x 330mm (or 13" x 13") seems to be scarce.

[EDIT]

Actually, all the variance is front-to-back; it’s pretty flat, just not level. So probably @N_Y_NakiYozora 's point about exact position of the clips mattering is exactly what I’m seeing here, and playing with it I could probably improve this a lot. I’ll have to explore that when I next have a chance.

Clipping only at the back reduced overall variance with the glass, but it’s still showing more variance than the plain aluminum.

The metal store I visited this morning had no MIC-6 and hadn’t heard of ATP-5, and wondered whether 5083 was steel. :slight_smile:

One approach…

For a (conservative) 12" (304mm) square bed (small for rails):

$ 21 - BuildTak Printing Surface.
$ 35 - BuildTak FlexPlate.
$ 53 - 025" ATP-5 Aluminum plate from MidWest Steel (shipped).
Subtotal ~$109.

Plus…
$ ?? - (magnets to bore/bond into the plate & hold the bed)

For a (daring) 16"(406mm) square bed.
$ 21 - BuildTak Printing Surface.
$ 50 - BuildTak FlexPlate.
$ 74 - 025" ATP-5 Aluminum plate from MidWest Steel (shipped).
Subtotal ~$145.

https://www.midweststeelsupply.com/store/castaluminumplateatp5
https://www.buildtak.com/shop/https://www.buildtak.com/shop/

BTW, if you buy borosilicate glass, then it will not change dimension when heated. More expensive than ordinary glass, and still a poor thermal conductor, but better than plain glass.

@Stephanie_A I also bought ATP-5 from Midwest. I took your advice and asked them to wrap it in extra cardboard, but that was clearly ignored. I bought two plates, one for the X5S and one for my i3 clone. Both plates have some scratches in the plastic covering that look like the surface may be scratched. Planning to mill it (maybe this weekend) for a kinematic mount before putting the Keenovo pad on it.

I have a different order from Howard Precision’s “random rack” http://www.howardprecision.com/random-rack including some Alca-5 (which seems to be functionally identical to ATP-5) and I’ll see how that’s packaged when it arrives.

Promised update, a little late: Howard Precision’s pieces were a bit better packaged, with less swarf included, and with with layers of cardboard between pieces. Also, the pieces had a much cleaner edge and were cut more square than the pieces from Midwest.

They don’t have as easy an online order experience. You end up actually talking to a person after putting together a quote online. The quote was inclusive of shipping. They don’t take credit cards online; they take them by fax or phone; they are clearly oriented towards taking POs from businesses. But they did a good job. While I am generally satisfied with Midwest, I think I’ll check with Howard Precision first.