"Tree-branching" when printing multiple small parts

[ChrisERAU]

I was printing some small propeller blades on .1mm layer height, and I got a bunch of weird tree branch looking things on each propeller. I saw that they were forming as the extruder moved to each one. I saw this same phenomenon on Makerbots earlier. Any ideas? What settings do I need to adjust?

[ednisley]

a bunch of weird tree branch looking things on each propeller

Is the solid model manifold? Those branches look nearly the same on each blade, with more plastic than I'd expect from insufficient retraction, so examine the G-Code using a previewer (gcode.ws, for example) to verify that they're not built right into the program. A non-manifold model will cause all manner of trouble in the slicer and you can't fix the damage by tweaking the slicer.

If the model is not manifold, then you must repair your model. The techniques for that depend on which modeling program you're using, but ideally you will fix the model rather than submit it to the uncertain mercies of an auto-repair program like Netfabb.

If the model is manifold and the G-Code previews properly, then you need much lower speeds (perhaps a minimum layer time of 15 seconds or so), much more cooling, and probably more retraction.

The skirt threads around the pillars look very, very narrow. Have you worked through the extruder calibration & Z-axis calibration guides (see the sticky posts) to perfect those settings?

[ChrisERAU]

I should've put a ruler in the pics for scale, but these propellers are about an inch tall. I made them in CATIA, which doesn't allow for non manifold geometry. The only way something might've happened would be in the conversion to stl. The preview looked perfect in S3D. I'll check the gcode asap.

There is a slightly related issue with the bed. It's raised in the middle by a very small amount. I printed a 2-filament wide grid over the entire need, and the middle area was adequately squished (no gap between lines), but the lines were thinner and had gaps in between towards the edges of the bed. I'll edit this post later with pics. I levelled the bed and set the z height before this. Because of all this, I printed the props in the middle of the bed. Bed adhesion was very good.

EDIT: added pics

[ednisley]

It's raised in the middle by a very small amount

Other folks have observed that making the X-axis belt hella tight can bow the middle of the gantry downward by enough to make the bed seem bowed upward. I haven't observed that on my M2, but if you did a really, really good job of snugging that belt down, maybe backing off a bit would help... [grin]

something might've happened ... in the conversion to stl

That's possible, because the STL format tesselates the surface into triangles, but the triangles meeting "at the same point" all use distinct XYZ coordinates for that point. They should all be identical, but it's entirely possible for numeric rounding to create different values that will break the STL mesh. Slicers can fix up small errors on the fly and should (be able to) handle that situation, but that's not a guarantee and I've seen cases where the results went distinctly pear-shaped.

A tiny model may also have non-adjacent triangles with close-together vertexes that should be distinct, but numeric rounding mashes them together. That breaks the model, because now you have surfaces sliding inside each other in ways you didn't expect.

One way around all that is to use lower resolution for the shapes; I'd expect CATIA would have a control for the output resolution. In round numbers, resolution below about 0.01 mm doesn't buy you anything and (some) slicers allow you to crunch the points to a specific resolution, but doing it with the modeling software lets it pin the vertexes to the nearest grid point before creating the STL.

The slicer should tell you if the model is manifold, so pay attention to that report!

[Rara]

I had this same problem when I was trying out PET-G on a turbine impeller. I'm trying really hard to remember what fixed it; I think I had my retraction set too low. Unfortunately there were a number of things found wrong with my settings so many variables were changed at once via group consensus, but I think retraction was the big one.

[3dPrintingMD]

I'm assuming this is 6 of the same model. What happens if you just print 1 copy ?

[ChrisERAU]

I always set my tessellation quality to the highest that CATIA allows. It can result in some giant STL file sizes lol.

I'm assuming this is 6 of the same model. What happens if you just print 1 copy ?

Yup, 6 identical models. Printing just one eliminates the branching, but then the part ends up overheating and the upper layers get a little melty. This is at outline speeds of 7 mm/s and 100% fan. My fan has a duct added on for additional cooling efficiency. Clearly it's not enough I print the black Makergear PLA at 215C with a PEI covered bed at 70C. Maybe dropping it to 210 might help?

I had this same problem when I was trying out PET-G on a turbine impeller. I'm trying really hard to remember what fixed it; I think I had my retraction set too low. Unfortunately there were a number of things found wrong with my settings so many variables were changed at once via group consensus, but I think retraction was the big one.

My retraction settings are below. Some people have a retraction vertical lift of a couple mm - mine is 0. Maybe that's something worth tweaking?

Other folks have observed that making the X-axis belt hella tight can bow the middle of the gantry downward by enough to make the bed seem bowed upward. I haven't observed that on my M2, but if you did a really, really good job of snugging that belt down, maybe backing off a bit would help...

I got my printer last week fully assembled. Haven't touched the belts, but they seem snug enough to eliminate play. Not tight enough (I think) to bend all that steel underneath.

[ChrisERAU]

Also, slightly related issue: I have holes in the top layers of my prints (engine nacelles in this case) where the top infill areas are under a square centimeter. I think it's due to low infill underneath - 15% fast hexagons. It's weird because I've done this same print with 15% hex infill on Makerbots and it turned out fine. If a Makerbot can do it, I think the M2 should, right? The first pic shows the holes, the second shows the bottom surfaces. There's an attachment peg in each one which I printed at the same time as the two engine nacelles. The first layers are pretty squished together...is there such a thing as too much ?

[ednisley]

set my tessellation quality to the highest that CATIA allows.

That is probably a mistake, simply because a 3D printer cannot possibly reproduce that much detail. Back it off to 0.01 mm and you won't notice any difference in the results, other than a dramatic reduction in slicing time and probably an improvement in print quality.

Attempting to print extremely small segments causes the G-Code interpreter to pause the nozzle motion, because the USB link simply can't transfer that much G-Code as fast as required. If the linear segments are shorter than about 0.1 mm, you're kidding yourself about the quality: remember, you're creating an object by squishing layers of hot goo atop each other.

Worse, none of the retraction settings apply to slowdowns due to buffer clogging; the nozzle simply sits there and drools until the interpreter catches up and motion resumes.

Not tight enough (I think) to bend all that steel underneath.

You'd be surprised; less than 100 microns vertically doesn't require all that much bendiness...

However, if the apparent surface is flatter than about 0.10 mm, you're good to go. Split the difference so that the central part is 0.05 mm too short and the outer sections are 0.05 mm too high; you'll have no problem with adhesion or overall print quality.

If the difference is more than 0.10 mm, then trading off adhesion with squashing may be messier. But that definitely isn't causing the stringing problems you're seeing!

Run off a set of five thinwall hollow boxes, measure their heights & positions, and report back! [grin]

http://softsolder.com/2015/09/04/thinwa ... er-images/

My retraction settings are below.

The retraction speed looks low: bump it to 60 mm/s or higher. If it's too high, you'll hear the extruder motor clunk as it skips some steps, which is a Very Bad Thing for print quality. If it's too low, you'll see blobs where the nozzle drools as the hot end pressure gradually drops.

You'll need Z hop (a.k.a. retraction lift) to avoid having the nozzle bump into previously printed objects during travel moves between them. That's more of a problem with thin angled objects than vertical slats, but it'll be obvious when it happens.

the part ends up overheating and the upper layers get a little melty

If you can't print a single blade without having it get melty, then your minimum layer time is too low and maybe a few other things need tweaking. I'm using 10 seconds with PETG and this absurdity came out surprisingly well:
viewtopic.php?f=7&t=2985#p19083



A few posts later, Jules trumped everyone:



Bottom line: get one blade working before you try six...

[Jules]

If you're going to use only 15% infill, you want to use at least 4 or 5 top layers. (Those holes sometimes form over skimpy infill, you need enough top layers to make sure it gets filled in.)

Bottom layer looks fine - might be a bit over-extruded, but that's better than not enough.