Surface of Bridges from below

[JRE 0]

Hi!

 

I am trying to design a sealing channel from the below side of a flat structure. Unfortunately, the surface roughness is too high and the channel depth is not accurate.

 

Therefore, I performed some tests with a test geometry and with different settings, but cannot really see an improvement. Attached picture is with and without the epxerimental bridge setting (brdige skin density set to 100%). I attached the test geometry .stl file as well as the g-code.

 

The test plate should have 1.4 mm deep channels, they turn out to be only approx. 1.3 mm deep (ony layer offset?), the roughness is very high as can be seen on the photos. From Cura I cannot see why the lines are so far apart (compare Cura screenshot and photo).

 

Filament: Winkle PLA HD, official Cura profile, no brim

Layer size: 0.1 mm

Ultimaker S5

Reminder: The images are taken fom the side facing the buildplate. The channels are facing downwards and rotating the geometry is not an option.

 

I am quite sure that UM and Cura can do better and it is my mistake. Could anybody please help?

 

UMS5_bridge_test.ufp bridge_test.stl

[Slashee_the_Cow 268]

If you could provide the Cura project file (.3mf, get it ready to print then go to File > Save Project) that would really help.

 

The line distance: I can say with 78% surety that your Flow setting (in the Material section) is way too low, at least for the top/bottom, but possibly all of them.

Slicing the model myself, this is 100% flow:

And this is 50% flow:

 

It also doesn't help that your circle is really not very circular.

By my count, it's... a tricontahexagon (or in non-smartarse language... has 36 sides).

It's so bad that when you're slicing it the width varies so much that for the inner circles (at least), it varies the number of lines to fill them (from your screenshot, highlights added):

Edited March 20 by Slashee_the_Cow
replaced second screenshot to make it more clear

[JRE 0]

Thanks a lot for the reply!

 

First point: Thanks for educating me - I simply used the default setting for .stl export from NX and wasn't aware. I changed the angular tolerance setting from 18° (default) to 2° and it looks much better. I also deleted the smallest channel from the testpiece. I think the curvature looks much smoother now.

 

I also changed every flow setting I could find to 100%. In fact, most of them were, only initial layer bottom flow and inner wall flow was at 95%. I disabled the experimental bridge setting (but the flow setting was also at 95%). This time I added the project file (sorry for that).

 

As you can see from the photo, the result has not really changed yet. While printing, I can see through the individual lines  of the first layer covering the channels. Furthermore, the channel depth is has a depth of pretty accurately 1.3 mm which is exactly one layer less (0.1 mm) than the targetet 1.4 mm.

test_02.3mf

[Slashee_the_Cow 268]

According to the slice, you should be getting the correct height, since there's 13 layers before the top starts printing, and you have an initial layer of 0.2mm - it shouldn't be the problem here, but when you're going for accuracy, it's always good to have it the same as all the others (that's my way of saying "set it to 0.1mm". It can affect the printed height in that because there's no filament already on the bed, it has a chance to spread outwards instead of holding its shape, and a thicker initial layer allows for the that effect to become more significant (I usually set my initial layer flow to 105%, it also helps with adhesion).

 

As for how easily you can see the lines: partially, it's just the process works. You can get a smooth bottom because it's against a smooth surface, and a smooth top because it can take extra care printing it (especially if you enable ironing - now that's smoooooth), but you can't really get a smooth underhang because there's not much it can do about it. If you printed it the other way up and turned on ironing you'd get pretty much every surface smooth, but you've said you can't print it the other way up and I learned long ago to stop asking questions like "why".

 

Partially, it's also how bridging works: Hot filament wants to adhere to other filament. Preferably other hot filament. That means lines after the first one are going to want to stick to the one next to them instead of falling down (good!) but that also means they're not always completely straight (bad 😞). It's also not great that some of your lines cover fairly wide expanses. Even so, some sag is inevitable (could also explain your height not being exact), especially with the other part I'm getting to:

 

Partially #3: Line width!

Wider lines are more visible. Fact of life. Technically it's not the width of the line itself you notice generally, but that the gap between the tip at the top/bottom of each line (remember, we're printing tubes here) is wider. But wider lines are also worse for bridging, as a general rule. They weigh more, so they sag more.

So: let's put your lines on a diet! Cura will warn you if you set the line width below 0.26mm on a 0.4mm nozzle (the "general rule" is that your line width should be somewhere between 60-150% of your nozzle diameter, you can try pushing it a bit if you want, but 50-200% is probably the hard limits).

But, Exhibit A: your model with 0.4mm lines:

Exhibit B: your model with 0.26mm top/bottom lines:

Exhibit C : your model with 0.2mm top/bottom lines:

Not nearly as big a difference as the drop down to 0.26mm, but it's worth trying if you want. If it's just too thin, I don't think it'll break the printer or anything, it'll just look like crap.

 

If you can do this and it's looking better but still off by 0.1mm... you could spend hours fiddling with print settings and making test prints, or you can make your model 0.1mm taller 😉 

 

[JRE 0]

No success yet. Have been trying with different settings flow flow and line width but without significant progress. It still seems to me that cura is adding a very coarse mesh in layer 14 (which shouldn't be there) just below the designated bridge layer 15. That at least explains the wrong channel depth and also why my changes to not really make a difference.

Is there any such setting that might build some kind of support mesh below the bridge which might be accidentally enabled?

[Slashee_the_Cow 268]

Not that I know of - there's regular supports, and you can make infill add extra bits to the top few layers to better support a surface, but it should never add supports into an air gap for no good reason (and they'd show up in the preview mode anyway).

 

Have you tried printing one with support (yes I know it can be hard to remove, especially cleanly) to see if your depth is correct after you remove the support? Because if that's any better it does lend credence to the theory that it's those top lines sagging that causes your inaccuracy (and possibly being what makes it look bad).

 

Also when it comes to dimensional accuracy, something worth remembering: like most things, plastic filament contracts as it cools, but that's mostly on the X/Y planes (its what the horizontal expansion settings are for) so your depth shouldn't really be affected by it, and PLA usually doesn't do it too badly.

[JRE 0]

Thanks for the advice. Ok, the additional layer seemed unlikely to me too - an there's no such think visible in Preview mode.

 

If I look at the layer when the channels are closed however (attached images) it makes me think that there must be a wrong setting here. Why are the lines so thin?

 

I can try PVA as support as well. Given the low channel width however, I am quite positive that this should be well possible without support material.

[Slashee_the_Cow 268]

It should only be doing really thin lines like that if you:

a) Have Experimental > Enable Bridge Settings turned on (and you didn't in the project file you posted) and don't have enough top/bottom layers that it just prints the whole top as a solid (and yours has enough layers to make it solid)

b) Have your lines set top be that thin

c) Have any of the Flow settings that will affect that line type set pretty low

 

But b) or c) should affect that whole print, and you'd notice that.

It's worth bearing in mind that while the distance between the rings might be short, you're not printing in a radial pattern (because it's not an option), which would do lines back and forth from the centre to the edge (could be worth a feature request), it's printing lines back and forth, meaning it has to cross the longest part:

That's roughly 20mm, and is a bit of an ask for your average line to bridge, which is why I'm suggesting a test with support, and if you're doing it in a material which you can easily remove like PVA, you should set Support > Support Z Distance to 0mm - we're just checking to prevent the lines sagging at all here.

[JRE 0]

Thanks for checking that. About the distance: Yes I am aware - and would it only be that section (your pink line) where quality is bad I'd completely understand. But it is a homogeneous behavior over the complete channel. The feature you suggest sounds funny! But this is just a test geometry to improve the settings, the actual geometry is more complex (that's why I can't print it upside down). 

 

It must be some other issue ... 

[Slashee_the_Cow 268]

I suppose this is probably well past the point where I should put my money where my mouth is and try printing one myself (it's not like it uses a ton of filament).

[Slashee_the_Cow 268]

Left one is mine, I've copied your settings as closely as I can (barring a couple of changes for the temperature of my filament and the capabilities of my printer). What gives with the difference in filament usage?

@ahoeben this seems like the sort of question you might be able to answer. You'll need my customised version of the E3V3SE definition if you want to actually load the project.

slashee_test_02.3mf e3v3se slashee profile.zip

[Slashee_the_Cow 268]

So how did it go, one might ask? Fair question:

Several conclusions can be drawn from this:

1. I really need to clear all the loose hairs off my desk before I take a photo.
2. I was using a small amount of leftover filament so I didn't want to lose it all purging. That's where the black comes from and quite possibly why it had trouble adhering at the start (you can see around the top right edge).
3. I might need to work on my retraction settings, touching the centre, just look at those... spherical deposits. Although I do find it interesting how there's only one line between the inner ring and the centre but several at the outer. Or it could just be that this thing is so tiny they didn't trigger the retraction minimum travel (but I did set that to the same as @JRE, so, 😕)
4. Those lines that are bridging the top don't look like they're full width and as you can see, somewhat uneven. Better or worse? I would say a little better, from a distance. The macro lens does these things no favours.
5. Edit: Oh, and dimensional accuracy. Didn't test that one. What tool are you using to measure it?

Edited March 26 by Slashee_the_Cow

[Slashee_the_Cow 268]

Test #2: The settings I print most of my stuff with:

1. Retraction is fine. Just a couple of tiny little strings I can take out in about five seconds.
2. Very similar effect with the bridging lines (almost looks like I applied a pincushion filter). I think they're a bit thicker, but we're still not getting the all over coverage like the layer above has. It is a pretty great demonstration of how the filament is inclined to stick to itself rather than strictly follow the printer path.
3. Visible gap where each line in the rings starts/ends.

[Slashee_the_Cow 268]

Test #3: The same as above, but thinner lines - 0.24mm (they do produce better looking results, at least from the top, usually, normally I use 0.26mm as my narrowest lines but I figured I'd try the 60-150% rule)

1. Where it's done it smooth it actually looks pretty decent.
2. Egads, those scars! It looks like someone did a coin flip into the blades of a drone.
3. I'm not sure if it's a problem with the design or the printing (looking at the shape, probably the printing) but it can't even do the bottom surface touching the plate right. Just look at that squiggly line in the bottom left corner which seems to be pushing the rest of it outwards.

Edited March 27 by Slashee_the_Cow
changed which corner I say to look at

[Slashee_the_Cow 268]

Test #4: Well if thin doesn't work as I wanted it to, the next step is obvious: 0.6mm lines!

That's actually pretty damn decent. The grid pattern isn't noticeable to the naked eye, it just looks like a series of lines in one direction (but I'm not going to own a 2x macro lens and not use it at every chance I get).

But as you can see, while it's not perfect, the lines themselves are the most uniform they've been in any of the tests when it comes to thickness and straightness. Main problem here is that using 0.6mm lines isn't ideal for something that requires dimensional accuracy, but with a little effort you can use different line widths for different parts.

[JRE 0]

Thanks for trying so many settings! It is so interesting to see that the results get slightly better or worse, but the general issue of lines noch building a homogeneous layer in this case seems to remain. Interesting!

 

I made some further tests with a non-circular testpiece.

 

First test: Insert chamfers of increasing size into the channels until a smooth surface is reached. Idea was, that if the filament is stretched from its own weight- this should solve the problem at some point. 

Result: Yes...somehow. Concluding: If I don't have a channel anymore, my problem will be gone 🙂

 

Second test: You recommended to try the impact of PVA as support again. Also did that. Result: I need to clean up the print certainly. But the result was not too bad. Especially the combination of chamfer and support would give better results. This is not too surprising, as this is exactly what supports are meant for, right? However disappointing , as I would assume that this can easily be done without supports. 

 

[Slashee_the_Cow 268]

I think what it comes down to is that FDM printers aren't able to get results that look good from below because there's not really much they can do to clean it up. If you need this to look good and if what it's in requires that it's printed that way up, you need to do it as a separate part and attach it (glue is probably best, cyanoacrylate - plain old superglue - works incredibly well on PLA) but then making sure of your dimensional accuracy is harder.

 

*has crazy idea*

 

This probably won't work, but I'll be back in a few.

[JRE 0]

Upside down is not possible, as the upper side has exactly similar issues 🙂

 

But slicing horizontally and using superglue ... sounds damn stupid but may be an excellent idea!

Not the way I wanted to do it, but might actually be the least pain for now! I'll see if that could work.

[Slashee_the_Cow 268]

Your crazy idea isn't the same crazy idea as mine. Mine should be ready in a few minutes.

[GregValiant 1,254]

When I find it to be an advantage to split parts for printing, I put 2.1mm holes (for 1.75 filament) into each part by subtracting small cylinders from the STL or adding the holes in CAD.  That way the holes in each piece line up perfectly.

After printing, I put short pieces of filament into the holes in one part and they become locating pins.  The parts line up exactly.

This is a visor for a Kindle tablet.  If the legs were in place for printing then the entire bottom would need support.  You can see the holes I added to line the legs up so they will assemble in the exactly correct position.  (The legs get printed upside down.)  Super glue holds everything together.

[Slashee_the_Cow 268]

10 minutes ago, GregValiant said:

This is a visor for a Kindle tablet.  If the legs were in place for printing then the entire bottom would need support.

I am curious what a Kindle visor with legs is, although part of me worries about the potential answers.

 

Anyway: My crazy idea didn't come out too bad!

Let us remember that a macro lens with a ring flash will make every flaw obvious when it's invisible to the naked eye. Every flaw.

That's printed with my usual settings: 0.2mm layer height, 0.4mm line width. To the naked eye you can still see the lines, I'll admit, but the flaws aren't too obvious and that bridging is shiny!

[JRE 0]

The idea with the filament holes for fixations is smart 🙂

 

I agree that last attempt looks much better - so do you want to tell us about the crazy thing you tried (successfully)?

[Slashee_the_Cow 268]

33 minutes ago, JRE said:

I agree that last attempt looks much better - so do you want to tell us about the crazy thing you tried (successfully)?

Used a cutting mesh going from the bottom to the first layer with skin (the one you see) and enabled ironing. A cutting mesh makes it basically treat that section as a separate model, so with only the first skin layer in there, that's now the top of this separate part and gets ironed.

slashee_test_02_ironing.3mf

[JRE 0]

Wow! This is really a smart workaround ... didn't even know that feature.

(Chrome wouldn't let me download your file ... says it is dangerous 😉 But I think I kind of got the idea ...)

 

Currently, my printer is still very busy with the supported attempt. After that, I'll give both versios (full cut & superglue as well as virtual cut & iron) a try! I could imagine that by tweaking the ironing options there might even be potential in it still. I need the surface really smooth as to have optimal contact with my sealing material, so I could imagine that the 'full cut' option turns out to be the better one for me... However, this option might be very helpful for others as well!

 

Thanks a lot for all your effort! It will take a while but I hope that I don't forget to provide feedback on the attempts.

[Slashee_the_Cow 268]

1 minute ago, JRE said:

Chrome wouldn't let me download your file ... says it is dangerous

Chrome can get stuffed. I'd be impressed to see a 3mf file which actually is dangerous. I'm 99% sure you can override it somewhere but since I don't use Chrome on a regular basis I don't know where.

 

Maybe converting it to a zip file will make it happier - a 3mf file is actually a zip file with the different parts (models, textures, positioning, anything a slicer wants to include) in it so if you change the extension to .zip you can see inside it. Or in this case, if Chrome lets you download it, change the extension back to .3mf so Cura will load it 🙂

 

Also I upload the 3mf file to VirusTotal (owned by Google) which checks your file with pretty much every antivirus software out there:

So I stand by my "Chrome can get stuffed". Or you could just download it in a different browser.

slashee_test_02_ironing.zip