Subtle Problem (Sort of a bug...)

[gwhite]

I have a small electronics enclosure I've been trying to print for close to a week. I'm using eSun PLA+. It has text inset on the bottom, and I've played with all sorts of settings to get the text to come out nice,, but the background has defects in it. They look like under extrusion in a couple small areas near the text, but everywhere else is fine. It's not a defect in the build plate, because the defects are always in the same places on the model, independent of where the model is located on the build plate. Here's what one of the defects looks like:

The attachment Text Bug.jpg is no longer available

If you just glance at the bottom layer Preview, it looks fine, with nothing obvious going on.

I finally sat down tonight and ran the preview line by line, and I think I figured out what is going on. Rather than doing the background in large sections, it's hopping around. The defect shown above starts with a small section of background being laid down just below the "M":

Once it's printed that small section, it pops over to the right, and starts filling in right next to where it started that area a short time before:

I think the problem is that it's printing right next to traces that haven't cooled down yet, and it's ripping them up. There are plenty of other places it could go and work on while this small section cools down, but instead, it insists on printing next to material that it put down seconds earlier. The material must be in an intermediate sticky state, because if it was cool enough, it wouldn't have an issue, and if it had just been laid down (like an adjacent pass), it isn't a problem.

I have already slowed the first layer down to 50% speed. I don't have cooling turned on to get better adhesion & avoid warping, but I may have to experiment with that. I can also try to play with Z lift & avoiding crossing the model.

I suppose it's also possible that the defects are when the small short section is initially being printed, but when I have things tuned up well, it mostly seems to affect only one or twp passes.

I certainly don't expect a quick fix for the fundamental problem, but it's something that could be improved considerably with changes to the algorithm. I haven't noticed a check box for "don't print next to stuff you put down seconds ago"

If anyone has any ideas for a work-around for now, I'd love to hear them. I'll attach a factory file in a bit. I've been printing multiple models with different processes to find a solution, and there's no need for all the near duplicates.

Thanks!

[gwhite]

Here's the factory file:

1st Layer Text Bug.factory

(38.21 KiB) Downloaded 117 times

[parallyze]

Hi,

I've noticed you have the "Only retract when crossing open spaces" option enabled (Advanced-Tab -> Ooze Control Behavior).

This would explain some traces:

There's no retract here. So although you have Z-Lift enabled the material is "dragged" along and missing when
starting the infill in that corner.

Near the "M" there's only a retract when crossing the open space - but none when switching from the right part of infill to the left one.

[gwhite]

Thanks! I'll play with that next. Given that the problem is entirely within the 1st layer, I'm not sure if "retract between layers" will help. I'm also trying the Retract for Movements > X setting.

I just finished a couple test prints, one with cooling on the 1st layer (as low as I can set it), and one at 30% 1st layer speed. My theory as to what is going on appears to be all wet. Cooling made things worse, and slowing down to 30% made no difference at all. If the problem was the short interval between printing the right side of the section under the "M", both should have helped.

Experimental results: This is why we can't have nice theories...

[parallyze]

Experimental results: This is why we can't have nice theories...

Try disabling the "Only retract when crossing open spaces"-option. This will make the slicer add retracts before changing positions while doing the solid infill parts. If retracts are set up right it should be hard to tell where the start points where while printing.

As far as I can tell all slicers do it this way, they sometimes just have to stop/restart at different points to fill everything at a given angle when there's holes. While this can be avoided it is a bit tricky (openings designed to fit within 45° infill or adding a single solid layer and drilling holes afterwards. Not really an option when printing text... ).

[gwhite]

I had high hopes for shutting off the "Only retract when crossing open spaces" option. It had almost no discernible effect on the defects, and introduced all sorts of issues in the details of the text.

I suspect the retraction and extra restart distance will need adjusting to get the text cleaned up, and maybe that will also fix the defects in the background.

[parallyze]

I had high hopes for shutting off the "Only retract when crossing open spaces" option. It had almost no discernible effect on the defects, and introduced all sorts of issues in the details of the text.

I tried to mimic your settings (0.4mm extrusion width, 0.2mm layer height) and I'm not using "Only retract..."....

I suspect the retraction and extra restart distance will need adjusting to get the text cleaned up, and maybe that will also fix the defects in the background.

It's likely some settings in the retraction area. Lots of tiny segments like that using a small extrusion width on the first layer are tricky....

But depending on requirements there's a few options left:

- If you don't need the text at all, move the object -Z to avoid it completely
- Maybe add two small holes instead of the text and add the text afterwards, like model names on cars
- Depending on strength requirements simply rotating the object might be an option

This removes all the problematic tiny segments from the first layer. But printing it this way might have an
impact on the screw holes and the part would benefit from a chamfer instead of a fillet on the outside, there's
two or three layers sagging a bit because of the steep overhang:

[gwhite]

I'm still fine tuning the retraction & extra restart distance to clean up the text, but without much success. What's more annoying is that the original defects are still there as well...

This is just the lower portion of the housing, which is much larger. I appreciate the suggestion of trying another orientation, but that will introduce a host of other problems. The best results I've gotten so far could be easily touched up with a black filler of some sort, but there's something odd going on, and I'd really like to get to the root of it.

I may also be able to find a 1st layer infill angle that works better. I originally had defects that lined up with the tails on the "R"s, so I set the angle to 30 degrees to match that. All that did was move the defects around a bit. An angle that is closer to horizontal will at least elongate the passes, and allow them to cool a bit more.

[gwhite]

I manged to methodically stumbled across the magic settings! I dropped the Retraction down from 1 mm to 0.7 mm in 0.1 mm steps, and that didn't really do much. Then I dropped the Extra Restart Distance from 0.1 mm to zero, and the difference was astounding. No defects, and the text looks nearly perfect.

If I hadn't seen it myself, I wouldn't have believed that tiny change could make such a big difference.

Here's what it looked like immediately after disabling the "Only Retract When Crossing Open Spaces"

The original defect I was fighting is much worse, and the text has all kinds of problems. Here's what it looked like after reducing the retraction from 1.0 mm to 0.7 mm:

Things are a bit better, but still no where near as clean as they were before I unsuppressed the retractions. Then, I changed the Extra Restart Distance from -0.1 mm to 0, and got this:

The defects are gone, and the text looks as good as I've seen it in a week of experimenting.

I'm still not quite sure what was going on. However, allowing it to retract between features on the first layer, and getting the retraction settings just right has given me a result I can run with.

Thanks!

[parallyze]

Things are a bit better, but still no where near as clean as they were before I unsuppressed the retractions. Then, I changed the Extra Restart Distance from -0.1 mm to 0, and got this:

Thanks for posting the pictures, I could imagine them being linked to often in discussions about retractions... ^^

So the only question would be - how did you end up with a negative extra restart distance anyways?