Reduce one top perimeter offset

[vovodroid]

Reduce offset of internal perimeters inward for one top perimeter. Current value seems to be too large causing excessive shift and artifacts. Fixes #6000 **One wall on top causes wrong wall generation in Arachne ** and fixes #3313 **Perimeters offset too much for one wall on top surfaces **

Current value:

New value:

Both Arache and Classib are affected.

Obviously this value could be configurable, though it seems this value is good enough.

[igiannakas]

I think I had experimented with this a while ago but I found that the offset was needed at least with my printer to “tuck in” the top solid infill under the walls of the next layer. What has your experience been? Admittedly it was before implementing the independent top and bottom overlap % though.

[vovodroid]

With suggested offset it's almost "tucked in"

Not prefect, but I guess it's better than having completely broken perimeter. As I mentioned this coefficient can be configurable, or rather offset formula could be improved in general. For now I prefer reduced offset.

[igiannakas]

Maybe 1x would cover the delta nicely?

Need to try the PR with more complex models with circular tops and also some of the Voron parts like the stealth burner cover that didn’t like one wall top surface before…

[vovodroid]

Maybe 1x would cover the delta nicely?

Yes, but it still breaks the round model from screenshot round.zip

[igiannakas]

Maybe 1x would cover the delta nicely?

Yes, but it still breaks the round model from screenshot round.zip

How does prusas implementation behave? Maybe we can port that over if it’s better?

Edit: as in I think they have rewritten the feature

[vovodroid]

I think they have rewritten the feature

I guess they based on my PR of potting from Orca))) Bamboo also works good.

[igiannakas]

OK I've done a bit of testing on this and I think it's ok to be a bit inset. By the way I've also seen this aberration with the feature enabled - 0.16LH - 0.25 first layer height, layer 199.

Have you ever seen this before? Project file below:

one wall issue.3mf.zip

[igiannakas]

Unfortunately it is end up causing ever so slight perimeter gaps as the overlap of the top surface is affected by it :(

left after - right before

Bottom after, top before

The prusa implementation looks a bit different - haven’t had time to dissect it but it may be worth doing so

[vovodroid]

Here is layer 199:

What's wrong here?

I think they have rewritten the feature

Yes, I took a look on they code - it's different.

[igiannakas]

Here is layer 199:

What's wrong here?

I think they have rewritten the feature

Yes, I took a look on they code - it's different.

It’s generating solid infill instead of the walls. That messes with wall ordering algos like IoI

[vovodroid]

It’s generating solid infill instead of the walls. That messes with wall ordering algos like IoI

I'll try to port code from Prusa.

[vovodroid]

@igiannakas

Here is result of porting Prusa algorithm:

Your benchy:

As a side effect one top wall is always enabled. I hope it could be fixed)))

[igiannakas]

Looks good I think?

[vovodroid]

@igiannakas
would you like to test https://github.com/vovodroid/OrcaSlicer/tree/one-top-wall-fix ?

Changes:

1. Port Arachne code from Prusa
2. Set top perimeter offset for Classic to 0.9. It seems producing good results.

Option "One top wall" also works.

[igiannakas]

Would you mind pushing the changes to this PR so I can cherry pick it into my Dev branch to test together with all the features I've pushed recently too?

BTW this looks excellent!

[vovodroid]

Would you mind pushing the changes to this PR

It's another branch with other meaning. I think you can create PR to yourself and merge to yours.

[igiannakas]

Ah yes indeed! Merged in the dev branch. Running a build now and will test. Doesnt this supersede this PR or am I misunderstanding it?

[igiannakas]

Ok tested - it works but it breaks the top/bottom solid infill - wall overlap for surfaces that are not the topmost/bottom most. It only changes overlap based on the infill/wall overlap value for any intermediate top surfaces.

[vovodroid]

Could you provide an example project?

[vovodroid]

Doesnt this supersede this PR

Yes, it's new different code.

[igiannakas]

Could you provide an example project?

Yeap

How its supposed to work (exaggerated for effect)

vs

Only top most surface adjusted

Project file: Overlap.3mf.zip

[igiannakas]

I think its because this piece of code is missing:

        if (!top_fills.empty()) {
            infill_exp = union_ex(infill_exp, offset_ex(top_fills, double(top_inset)));
        }

But without delving deeper in the code I'm not sure where the new top surface polygons are now held

[vovodroid]

Yes, I see.

because this piece of code is missing:

top_fills was filled in deleted call to split_surfaces, I'll restore it for this purpose and test. Or will try to find polygons in new code))

[vovodroid]

Ok, I found top poligons in Prusa code

New commit added to the same branch. If it seems to be good I'll create new PR.

[igiannakas]

Awesome thank you!! I’ll compile it shortly and let you know but the change seems correct so don’t see a reason it wouldn’t work :)

[igiannakas]

Tested - looks good!

A bit of feedback below:

1. I dont believe the one wall threshold parameter works any more with the prusa code.

The below code is not evaluated (replaced with the prusa code)

    // skip if the exposed area is smaller than "min_width_top_surface"
    double min_width_top_surface = std::max(double(ext_perimeter_spacing / 2. + 10), scale_(config->min_width_top_surface.get_abs_value(unscale_(perimeter_width))));

    // get the Polygons upper the polygon this layer
    Polygons upper_polygons_series_clipped = ClipperUtils::clip_clipper_polygons_with_subject_bbox(*this->upper_slices, last_box);
    upper_polygons_series_clipped          = offset(upper_polygons_series_clipped, min_width_top_surface);

[...]

// get the not-top surface, from the "real top" but enlarged by external_infill_margin (and the
    // min_width_top_surface we removed a bit before)
    ExPolygons temp_gap = diff_ex(top_polygons, fill_clip);
    ExPolygons inner_polygons =
        diff_ex(orig_polygons,
                offset_ex(top_polygons, offset_top_surface + min_width_top_surface - double(ext_perimeter_spacing / 2.)),
                ApplySafetyOffset::Yes);

There is a question whether this is needed any more as I've noticed that this approach is far less sensitive to errors with text embossed on top surfaces
(left old, right new)

2. However I did notice that Prusa's approach will result in nozzle collisions over thin wall surfaces, like letters

Notice that the outline is drawn with an internal perimeter but the top surface doesnt segment around it. This will result in a nozzle collision over the letter.

Prusa slicer has the exact same problem:

Whereas with the old version there were artefacts (tiny holes) but no nozzle collisions.
(0% one wall threshold)

(300% one wall threshold)

I suspect this is happening due to the difference in shrinking/expanding the polygons between the prusa approach and the old Orca approach.

Project file: bbb.3mf.zip

3. However the prusa approach (and the port to orca) fixes another long standing issue, where the first internal perimeter when a one wall top surface was printed, had the wrong line width. See below (left old - right new) - look at the thickness of the first wall.

So in summary it's almost perfect, if only we could figure out the segmentation for the lettering issue. Everything else seems to be working better than the current implementation.

[vovodroid]

Does threshold value prevent nozzle collision in current version? I could try to use it.

[igiannakas]

Does threshold value prevent nozzle collision in current version? I could try to use it.

I think so but I’m not 100% sure

[vovodroid]

New PR One top wall fix pr #6236 open.