Consistent misprint indicating a firmware bug?

[Jackson He]

I've printed this part three times, each with exactly the same problem: after the "dome" is completed, the extruder sends out much more filament than needed (probably by 2x), causing a big mess at the end. The part file is here: https://dl.dropboxusercontent.com/u/19394555/MIC-Logo%20ball%204.stl. I was using slic3r with ajf's config.

[ajf]

Actually, I think that's an issue with part itself.  I tried printing on one of my Kossels and had a similar soft-serve ice cream effect.  

What's happening is that, because the spire is so small and the hotend is staying on it for so long, the plastic doesn't have enough time to cool between layers.  The solution would be to print multiple copies at the same time, which you can do with the 'more' button in Slic3r.  In fact, I'd suggest that you do separate STLs for the two halves and then print each half in sets of 4-6.  That way as the hotend moves from part to part the layers have time to cool.

[Jackson He]

Thanks. Will try to do that.

[ajf]

Another thing to consider, especially for small parts, is how you model them.  Ideally, you want to make X/Y dimensions that are multiples of your nozzle size (0.4mm in the case of MOD-t) and Z that is multiple of layer height. That will make it easier for the slicer and you'll get better part quality.

[Jackson He]

So indeed ajf is right--it's not the firmware that's at fault. 

In the above picture I have the results of three attempts. 

• The first is made with a separate "pillar" of the same diameter built next to the part I wanted, just so that there is a doubling of the extrusion volume in the upper portion. The outcome is not correct, but is much improved from the "soft-service ice cream cones" I got previously.
  
• The second is made without any additional part--since I only needed to make one of these in the end, I decided to experiment with slic3r parameters. For this one I changed the slowdown_below_layer_time value from 20 to 1. This has about the effect as the previous run. Still not right. But not worse.
• The third is printed through the NewMatter Store. The problem here (not shown) is that the store build is not customizable. I had to spend much time trimming the support material, which I'd rather avoid. But at least it gave me the confirmation that the top portion could be printed correctly with this printer.

Still tinkering. Still learning.

JH

[ajf]

I usually go with multiple copies of small parts because it's easy and I don't need to make delay part.  Think that if you made a wall, say 1.6mm x 8mm x height of your actual part, you'll get the desired quality without doing multiple copies or using support. Also, spacing them farther apart will add more delay.

[Jackson He]

I now feel it's really some bug in slic3r or the firmware, because watching the printer make the part, I am pretty sure the machine is extruding too much material for the top layers. The above is where I paused the printer (and subsequently I stopped it by turning off power). All this extra material has no where to go but to make a mess outside of the build volume.

JH

[ajf]

Here is the part printed alongside a cooling/delay part using my current Slic3r settings as posted to GitHub.  I do think the part itself needs to be modeled to account for the nozzle diameter and intended layer height to optimize the print quality.  But, don't think there's bug in the firmware that's causing over extrusion, and there certainly isn't one in Slic3r.  There are settings in Slic3r that could be better tuned to your particular spool of filament (actual average diameter, temp, extrusion multiplier.)

Attached STL.

[Jackson He]

aj:

With due respect, I'm still not convinced. I mean, if Cura can make the part without a cooling tower, why should slic3r need it? Is it because Cura is running the printer at a lower temperature?

JH

[ajf]

Well, I don't know what Cura settings you're using.  But if you when you printed from the cloud interface, it may be that the support that was printed gave it enough time to cool. Couldn't really tell you without looking at the gcode or the Cura config that generated it...

Have you tried printing the STL I attached?

[Jackson He]

aj:

The Cura is from their cloud service. Yes, it does generate support material, but not up above the "dome". So the support does not help cool the part. And because it's from their cloud service, I don't know how to get to the gcode or the config file.

No, I have not tried your stl file. I trust that it'll work, especially since you've printed it successfully already, but I'm looking the simplicity of a tool that gives me the output I wanted without having to tweak the part, or adding other parts. 

Actually, this part below was printed successfully without a cooling tower, nor any support, and the head section is much smaller than the cylinder in the part we were talking about. The stl and gcode are here: https://dl.dropboxusercontent.com/u/19394555/Turkey%20Chopstick%20Rest.stl, https://dl.dropboxusercontent.com/u/19394555/Turkey%20Chopstick%20Rest.gcode. The part is so small that I printed it using 100 um layers.

Again, I'm not disputing the effectiveness of your method at all. I'm simply trying to get a better understanding of how to successfully use the printer/slicer combo with the least per-part tuning.

Best regards,

Jackson

[ajf]

Hey, no worries. It's really not my method.  Actually, I think the very first part I printed had a cooling wall like that, it was of a printed extruder for the original Rostock delta printer.

Again, I'm no expert on 3D printing, I mostly print parts for building printers or calibration. 

Would be interested in what you get when printing my STL...

[Joe Jaszczak]

Just curious. Did you know you can turn off the supports when printing in the store? When you select the model to print And lick send to store click the wrench icon and you can turn off supports. I wonder if it would still print as well.

[Jackson He]

@Joe jaszczak: No, I didn't know that. Thanks for pointing that out. I tried that, and the print is on the right in the image below.

@ajk: I tried your STL with the wall, and the result is on the left in the image. I think it's about the same quality as what you got.

The store print is printed using their fine setting (100 um layers). But still the quality isn't that good. The part has been partly fixed before I took this picture. Most importantly, I cut off a couple of extra loops of filament on the outside of the arches. What's especially bad is the pillar portion under the dome, with lots of extra bits sticking out.

The ajf print came out as the best so far. The worst part is in the undersides of the arches, but that's true of all the parts printed without supports, and it's not worse on this part than others.