Solar calculation method in the latest release

[m0r0c0ns]

Hello, has the solar calculation method changed since august 4th?
On august 4th I had this:

And now that:

No settings changes for this cover recently.

[jrhubott]

Yeah, this changed, and it was intentional, not a regression that slipped through. v2026.8.0 went out August 5th, which lands right in the gap between your two screenshots. You went from v2026.7.6 to v2026.8.1.

The change is to how Window Depth (Reveal) works in the vertical engine (issue #1169, PR #1177). The old formula added window_depth x sin(|gamma|) to the calculation, but it only kicked in past a 10 degree azimuth angle, and sin(gamma) is basically zero at solar noon. So around midday it was a no-op, which is why your "before" trace is a smooth decline with nothing window-depth-related happening in it. The new version re-projects effective_distance + window_depth against tan(elevation), and that peaks at solar noon instead of going to zero. If the projected drop reaches your window height, the engine now returns fully open immediately. Below that threshold it does nothing at all. There's no partial zone in between. That's exactly the flat 100% then vertical cliff in your second screenshot. I made the change because the old formula opened the shutter from the bottom and let sun leak past the configured depth. It was wrong in about a third of the config sweep I tested it against.

The Solar Calculation sensor you're graphing is the raw geometry output, before interpolation or inverse. It's not reporting some other handler taking over. A reading of 100% means the geometry itself decided fully open. If the solar handler had declined instead (sun outside your acceptance angle), it would read unknown, not 100, so that's ruled out. The Field of View to Sun acceptance angle rename in the same release is cosmetic only too. No option keys changed, so it's not related either.

One thing I can't confirm without your help: this mechanism only fires when Window Depth (Reveal) is non-zero on that cover. If VR_SO has a value set there, this is almost certainly your answer. If it's at 0.0, this doesn't apply at all, and I don't have another candidate in the diff that explains a jump from 19% to 100%. In that case I'd want a diagnostics file to dig further. What's it set to on your end?

If it does turn out to be non-zero and you don't want the all-or-nothing behavior on that cover, setting Window Depth back to 0.0 gets you continuous tracking through midday again. That's a legitimate choice, not a workaround. It just means the reveal itself won't get modeled as shading.

I also just fixed the wiki page for this. It was still describing the old sin(gamma) formula.

For details, see the Enhanced Geometric Accuracy wiki page.

[m0r0c0ns]

Hi, you are right my this cover has a Windows Depth (Reveal) set there:

Thanks for the explanation!