Getting better cylinder walls

[iccherry 1]

I am constantly struggling with artifacts in my cylinder walls on my UM3. Blobs, zits, random noise. This seems to always happen in the same places if I use the same model to test with, so it's got to be coming via Cura.

 

I have tweaked retractions, coasting, z-hop, speeds... nothing seems to get nice smooth walls. I've cleaned the extruder. Swapped cores. Cleaned and lubricated every rod, and checked for play. Any ideas? I've tried printing down to 45mm/s at 0.2mm layers, and I feel like I was able to get this quality from much worse printers.

 

Here is an illustration. The rough print was sliced by Cura. It's the best one I was able to get after tweaks. The smooth print was sliced with Prusa Slic3r (and then I hacked in the start and end code from the Cura file to get it to work). The Cura slice is full of irregularities. The other is incredibly smooth. Same STL. Same filament. Same layer heights. I believe there is a 5 degree difference in print temp, but that isn't driving the result (I've tried it). I do understand it's difficult to tell with this filament, but I've done the best I can with the photos to show what I'm talking about.

[kmorris1077 0]

I'm very curious to hear what people think about this.  I'm finding the same with Cura 4.2.1 and cylinders.  It would be interesting to compare the g-code the two slicers produce.  Mind posting them?  Probably should post the STL as well.

[Smithy 1,144]

Please use in Cura the Save menu item, not export, to save everything as project, then we have the model and all the settings you use. 

[iccherry 1]

Sure. I'll use an even simpler model, as these are just examples. Here's a 40mm cylinder with 1mm thick walls. That's it. Designed in Fusion 360 with two circles and two extrusions.

 

I used the default 0.40mm "fast" 0.2mm layer height profile with the following two adjustments: 0.80mm line width (because I'm using a 0.8mm nozzle) and I changed the brim to 2mm.

 

Result: hideous.

UM3E_40mm Cylinder.3mf UM3E_40mm Cylinder.gcode

Edited September 30, 2019 by iccherry

[Smithy 1,144]

3 hours ago, iccherry said:

0.80mm line width (because I'm using a 0.8mm nozzle)

 

Just an idea, try to set the line width to 0.50 or 0.49 and check the result in the preview.

When you have a wall thickness of 1mm, 0.8 is hard to print because the printer makes a wall with 0.8 and then he has to print a wall with 0.2 which is hard with a 0.8 nozzle.

 

you could also try to set the line width to 1.00 or 0.99, should be no problem.

[iccherry 1]

5 hours ago, Smithy said:

 

Just an idea, try to set the line width to 0.50 or 0.49 and check the result in the preview.

When you have a wall thickness of 1mm, 0.8 is hard to print because the printer makes a wall with 0.8 and then he has to print a wall with 0.2 which is hard with a 0.8 nozzle.

 

you could also try to set the line width to 1.00 or 0.99, should be no problem.

 

Yes, I'd considered that. However, wouldn't that just get rid of the benefit of using a larger nozzle elsewhere in the model? These cylinders are just simple models to show what I'm seeing in more complicated things. I want to be able to print faster (and I'm willing to sacrifice quality in small details), so I'm using a 0.8mm nozzle. Doesn't narrowing the line width defeat that purpose? I'm not sure if Cura does anything dynamically there, do they? I'm reprinting right now with a 0.40mm nozzle to see if that helps. I'll upload results when it's finished.

 

Also, the same nozzle, speeds, temps, retraction, etc in Prusa Slic3r produces this (attached). The walls of both prints are just a hair over 1mm thick. It's clearly a slicing issue, as the stl, filament, and machine are the same between these two. I'm including the hacked together gcode for the Slic3r version, for reference.

40mm Cylinder.gcode

Edited September 30, 2019 by iccherry

[Smithy 1,144]

Of course it makes little sense to print such thin lines with a 0.8 nozzle, you can also change the wall thickness in the model to .8 or 1.6 and then try the 0.8 nozzle. It is important that after slicing you check in the preview whether only one or two lines are really printed, then the result should also be clean. If the slicer then has to print a very thin line in addition, there could be these problems. If there are not exactly 2 lines (at 1.6 wall thickness) then play with the line width, often a reduction to 0.79 or so already helps.

[iccherry 1]

28 minutes ago, Smithy said:

Of course it makes little sense to print such thin lines with a 0.8 nozzle, you can also change the wall thickness in the model to .8 or 1.6 and then try the 0.8 nozzle. It is important that after slicing you check in the preview whether only one or two lines are really printed, then the result should also be clean. If the slicer then has to print a very thin line in addition, there could be these problems. If there are not exactly 2 lines (at 1.6 wall thickness) then play with the line width, often a reduction to 0.79 or so already helps.

 

Yes, I originally assumed that was the problem also. I have had this problem across many models with various wall thicknesses (almost exclusively using a 0.40mm nozzle). This is an example to show the problem. Honestly, this is just one of the issues I'm seeing (constant extrusion issues seem to be happening, but again, it seems it's Cura, not my machine).

 

Printing with a larger nozzle helps to *show* the problem. It is exaggerating it. It is not causing it. I have just finished printing the same stl with a 0.40mm nozzle and the default 0.35mm line width "fast" profile. So, the wall is approximately 3x the line width. It's harder to see the artifacts this way, especially in a photo on the internet, since the features are half as large, but they are still there. And in person, they aren't difficult to see.

 

Note, the print on the left is the Cura 0.40mm nozzle version. The version on the right is the same one I posted above with 0.80mm nozzle in Prusa Slic3r with the same speeds and temps. Look at the evenness of the lines. The Cura version is quite rough, both inside and outside. There are also some runs in the Cura version layers where there's underextrusion or something.

 

I've been having a lot of problems lately with surface quality, and I'm just trying to figure out how I can do better. I was worried I had a mechanical issue or something, but I figured I'd try a different slicer, just to rule that out, and here is what I'm seeing.

[iccherry 1]

Ok, I realize I might be talking to myself here, but I think I'm on to what's actually happening.

 

With the 0.8mm lines, Cura is stuttering the extrusion.  See the green spots? Those are movements with no extrusion.

 

Here's the gcode for that section. Note the movements with no E values:

 

Now, this introduces fairly visible artifacts in the simulated part. See the vertical lines? Those are the corresponding movements with no extrusion:

 

 

The Slic3r gcode exhibits nothing but the seam, and that's the only place it has movements without extrusion. Here, you can see how clean the simulated extrusion is.

 

 

Finally, I checked the other Cura file that I had sliced with 0.40mm lines. Here's what's odd, Cura starts out doing the middle line by tracing the curve... then it switches to trying to do infill with 45 degree lines. And *then* it tries to go back and fill in the gaps. It makes no sense, when it just needs to plot a circular single fill line, but that's what it's doing.

 

Here, you can see the initial infill is a nice smooth curve. Then it starts to zig zag.

 

 

And here's that same section when it come back and tries to cross hatch/fill it.

 

 

Needless to say, the 3d model of that pattern is as ugly as the actual print.

 

So.... I have no idea. Cura seems to just slice curves in odd ways. It's not extruding consistently, and it's also inconsistent with how it's handling the infill.

 

Any thoughts?

 

[Smithy 1,144]

Great analysis!

but it is what I said earlier that you have to adjust your line width accordingly with the wall thickness of your object. And I know you have to play with the value and check afterwards the preview until you see a clean line. So for a wall thickness of 1.6 you should set the line width to 0.79 or 0.81 depends on the object.

 

I guess it is a bug, but maybe @burtoogle know more about it.

[tinkergnome 927]

AFAIK it's not necessarily a bug, but just the way small gaps between walls are filled by recent versions of Cura.

If i remember right, "gap filling" can be switched off (one can choose between "everywhere" and "nowhere"). You could try both options and compare the generated gcode.

Another difference is probably the fact that CuraEngine always generates two passes on each outer contour, while other slicers may be generate a single, but slidely thicker line.

 

[iccherry 1]

Thank you for the feedback. I didn't post up the photos (just to keep clutter down, and they don't show anything new), but I actually tried a couple things to address those issues.

 

First, I thought, perhaps it's that gap filling and the middle wall causing the issue. So, I reprinted with 0.40 nozzle and default 0.35 line width (stock profile). Then, I set fill small gaps = nowhere, and set infill to concentric, just in case. Result: still the same. The outer wall extrusion gcode still has skips in it, and the print was still rough. The gcode did looks smoother (it was drawing circles), and it wasn't zigzagging in the middle wall, but that didn't matter.

 

Second, I tested the line thickness. I set the 0.8mm nozzle to 1.0mm line width, which is exactly the width of the wall. Result: still stuttering extrusion. Still an ugly print. Gcode still shows the blips in extrusion. But maybe going over the nozzle size has other problems, so....

 

I have a print going right now with 0.40 nozzle, 0.333 line width, no gap fill, and concentric everything. The gcode still shows the extrusion stuttering, but that should help determine how much line width matters, as it should just require three equal lines to make the 1mm wall.

 

/edit/ It's not quite done, but I can see enough of it to say that the line width didn't make a difference. The finish is still rough, with extrusion irregularity. /edit/

 

/edit2/ I found an issue listing on the Cura github that discusses what appears to be this, but it's unresolved. 

 

https://github.com/Ultimaker/CuraEngine/issues/702

 

They mention the possibility that it's "outer wall wipe distance" or "wall overlap compensation" causing the problem. I have a model printing with those set to zero/off. I'll have to check it out tonight, but it did look (at a quick glance) like the gcode didn't have the extrusion stuttering. The wall wipe didn't seem to have an impact on the code, but the wall overlap did. It's like the calculation for the required extrusion when compensating for the overlap is a little off or perhaps "mathematically correct" but non-constant, leading to surface irregularity. Hopefully that's it! /edit2/

Edited October 1, 2019 by iccherry

[iccherry 1]

Well. This is just frustrating. Eliminating the outer wall wipe and wall overlap seemed to make an incremental improvement, but there are still extrusion issues. Further, it almost seems like Cura is turning the curve into a rougher polygon than Slic3r. The vertical artifacts from the interpolation of the curve seem much larger in the Cura version.

 

I may just need to open an issue on the Git for this. I am lost for explanations.

 

For reference, on the left is an earlier Cura print. Middle is the Slic3r print. Right is the new Cura print with 0.333 walls, no overlap compensation or wipe. It looks "ok" in the photo, but again, in person, there are a lot of issues. The layers look like some lines are missing in places.

 

[Reywas 0]

I've noticed some strange behavior on curves as well.  What version of Cura are you running?

 

Edited October 3, 2019 by Reywas

[iccherry 1]

3 hours ago, Reywas said:

I've noticed some strange behavior on curves as well.  What version of Cura are you running?

 

 Cura 4.3.0 with the most current firmware ( 5.2.11.20190503) on the UM3E.

[hazedav 1]

FWIW Tackling the Hungry Robot project, I was getting pretty ugly cylinder walls.

 

After printing about 30 cylinders, I finally figured out what was causing all the cylinder wall terribleness (scarring/banding).  For me it came down to 2 settings:

 

1.  Compensate wall overlaps (must disable).  This had only been enabled for inner wall overlaps, but I found several forums that pointed at this being sketchy.

2.  Line width (Marlin recommends between 1.05 and 1.2 nozzle diameter).  I upped mine from 0.4mm to 0.44mm and noticed instant improvement.

 

Things that were tried and had no effect:

* Outer walls before inner (leaving this disabled)

* Coasting on/off (leaving this enabled)

* Linear advance (leaving this at K=0.8 now that i went through all the trouble to calibrate)

* Outer wall flow (4% lower than flow)

* Z-seam (just dictates where the ugliness happened)

 

Perhaps relevant settings:

* Combing mode (not in skin)

* Retract before outer wall (enabled)

* Outer wall wipe distance (0.4mm)

 

David

[bergie5737 0]

Hi Icherry

Thanks so much for all the hard works and effort you've put into this problem. I was hunting around how to speed up a print for a basically square "ammo style" box when I looked at Cura's rendering of the print. (My old laptop were unable to do it...) I was fascinated to see how many useless travel moves happen. After a lot pfaffing, I got a 23hr print down to 16hrs (layer height was the biggest factor of course) Then as I analysed, I tried various setting to get the printer to move less. The walls is just walls, so it should print around and around, one wall at a time, but in stead, it start at the front, work about 3/4 along one edge, then the print head jumps to the opposite side, does some printing, goes back to the previous position and in general does illogical things. In the end, I managed to get the print done in 11hrs, by turning things on and off. (Man, did I have cobwebs inside that box due to no retraction... retraction added 1hr+ to the print time, light sanding took 1 minute to kill the cobwebs. My daughter had fun pushing her hand into the cobweb box - so double bonus there!)

 

Today I am printing and STL from Thingiverse, thus had not control over wall thickness etc. (I normally design walls to be exact multiples of the nozzle width, in the box case 1.2mm) The print is basically three circles in a triangular pattern. With my speedy settings as above, I noticed odd behaviour in the simulator. As your excellent post has shown, Cura will start with the inner wall, go about 1/4 around the circle, and then backtrack on exactly the same line, stop, go over the line once again, and when it reached the end, a travel move takes the printer to a new section of the print. This was still ok, but for only one circle, the simulation show that the print head prints "a dashed" line although it redoes the whole section, as above, then after it has done another part of the print, I comes back and fill in the gaps in the dashes, three times.... No matter the line width, it is the same result. One of the three circle sections have thousand of little holes ( the one printed with dashes) the others looks ok thanks to the "ironing" of every layer! Oh, and my printer skips when it is printing over the same line, even with "flow compensation where a wall exist" turned on. I am going to try Slic3r again, and if it goes faster and cleaner, I'm in. I have noticed this odd "holy" behaviour since about a year ago and thought my filament is to blame - I live in the tropics now, so humidity is super high. I simply thought that it only happens when "bending" the line as straight lines always and still looks great! New laptop, (and your post) has shown me why I can print a straight line at 65mm/s with no skips, but a small curve at 35mm/s skip steps on the inner wall - it reprints the inner wall 3 times.

[Guest maht]

Are you able to try this in 4.6.1? Some issues relating to 'ugly prints' have been resolved since 4.3.

[iccherry 1]

On 6/22/2020 at 2:27 AM, maht said:

Are you able to try this in 4.6.1? Some issues relating to 'ugly prints' have been resolved since 4.3.

Sorry for the necromancy. I just saw this reply. Here's a 25mm cube and a thin wall cylinder (30mm external, 25mm internal, 2.5mm wall) slices and printed at the exact same time. 0.8mm nozzle in ASA with 0.4mm layer heights. 

 

The cylinder walls are still godawful. The cube walls look fine, other than some corner bulge, so it doesn't appear to be a mechanical issue. Note, I do think the internal wall is probably made worse due to the ASA pulling in a bit. It does have that tendency. But even the external Wal finish is much poorer compared to the flat cube walls.