Problem with concave surfaces using the DirectSunHours component

[LaFleur]

Dear @chris,
I want to report a possible error that I encountered using the Direct Sun Hours component on concave surfaces.
How to see in the first image, also using a offset of 5 cm, seems that the analyzed surface overlaps with itself.

only by increasing the offset to 25 cm, the problem is solved

When I set a offset distance from 1 cm (the one I normally use) the concave surface shades itself almost completely.

to avoid raising doubts, I don´t use any context for the example.

I attached the simple file.
DirSunHours_problemConcaveBrep.gh (69.5 KB)

Thanks for all and best regards

[mostapha]

Use a smaller grid_size so the mesh faces don’t block the sensor grids.

[chris]

Thanks for trying to answer this, @mostapha, but a smaller _grid_size will only make this situation worse. I think what you were trying to say is “increase the _offset_dist_,” which is how we have traditionally handled the fact that some of the sensor points end up behind the study geometry and get blocked. Increasing the _offset_dist_ is a proven way to get a more sensical result no matter the version of Ladybug that you are using.

With all of this said, @charlie.brooker actually came up with a really clever way to avoid the issue entirely no matter your offset distance a couple of years ago:

Sun Hours Analysis rendering glitch

Thanks Chris, I just had a look at the code for this # create the gridded mesh from the geometry study_mesh = to_joined_gridded_mesh3d(_geometry, _grid_size) points = [from_point3d(pt.move(vec * _offset_dist_)) for pt, vec in zip(study_mesh.face_centroids, study_mesh.face_normals)] hide_output(ghenv.Component, 1) # mesh the geometry and context shade_mesh = join_geometry_to_mesh(_geometry + context_) My thinking is that if you change the above to shade_m…

I was kinda reluctant to make it an official solution at first because it can slightly degrade the performance of the components and I wanted to do some testing to see if this change in performance is actually noticeable. I did a little bit of testing right now and, for the most complex studies, it only seems to add a few tenths of a second. For simpler studies like what @LaFleur posted in the sceenshots, it’s not even noticeable at all. So I just made the change here:

fix(geometry): Remove need to change _offset_dist_ for study components

… and you can get the change on your end, @LaFleur , by running the LB Versioner component.

I should have taken @charlie.brooker 's suggestion sooner since I can see that this would have probably saved the forum from a few extra posts. Just to illustrate the change, I have a study of a concave paraboloid like this go from this:

… to this:

… without the need to play with the _offset_dist_ at all.

Still, thanks again for suggesting this in the first place, @charlie.brooker !

[LaFleur]

Dear @chris,
thank you for your prompt interest in the issue.

I only saw the new input _geo_block later, which changes the “modus operandi” of preparing the context model for the analysis, to avoid having the same error.

How always, thanks!!

Best regards