I was confused why AO was not showing up on a recent (very large) project. It was working, just tiny!!!
Shouldn’t AO work with ANY size model?
1 Like
What is the model size? How big are the faces?
Your model is so large that AO can’t even extend far enough to be visible. You’d have to scale down 4x to get it usable I’d guess.
One dimension shows over one thousand feet.
Yes, I just tested. Within a group, I drew a cylinder 4 feet in diameter.
Then I scaled it up 100x. This was enough to make AO worthless.
off course. AO is emulating the naturally occurring shadows in corners. the size of this naturally occurring shadow is not dependant of the size of your model.
if AO adds a shadow within 15cm of the corner, and your model is 100m tall, it’ll still add a 15cm shadow.
yes, the dimensioned corner of the large box volume is 1000 feet.
This is not necessarily something I need to model very often because the question is somewhat pedantic.
I still think of something 1000 square feet tall it should show some kind of ambient occlusion.
conversely, if I was modeling something that was only a fraction of an inch, it would it also not display ambient occlusion?
well yeah… AO is not placing in shadow a % of the face, it’s creating a fake shadow within a set distance.
yeah, at the very base, for a small distance, not the whole volume.
here are 3 cubes, 10cm, 1m, 10m. you see that the AO is roughly 10cm tall, so the first cube is pretty much all in, when the second is 10% covered and the last only has a thin part at the bottom of the 10m.
it’s how AO works…
Wrong. Ambient Occlusion is a rendering technique that is common in 3d modeling… its just Sketchup’s approach is not scalable.
my bad. I should have written “that how SU’s AO works…”. the SU part was assumed considering the forum we’re on. and I’ve seen a similar behaviour in other softwares I’ve used (even in games)
also, it is scalable. by working the distance and intensity values, smallest AO is about 15-20cm tall, biggest is around 1m - 1m20.
so yeah, with identical parameters, the AO of a very big object will be almost invisible compared to the AO of a smaller one.
you could either downscale the model, or upscale the AO.
Understood… and I should have written “not scalable enough.”
Yes, that is great that you can scale it for certain sized scenes (small ones)
I just think ambient occlusion should work regardless of your model size.
Perhaps it could just scale up more?
AO is inperceptable for this ocean front parcel with a 300 foot tower, which is not an unrealistic size for a building.
I love love love the AO feature but I tend to agree with you. I know its there yet it starts to become imperceivable as you zoom out. It is one of those, ‘it’d be cooler if you did’ things.
I understand you cant expect the AO to scale differently based on different objects at the same zoom level, but something about it just feels like it doesn’t respect the scale of the geometry it is applied to.
I also think so (but I’m not very experienced at rendering science.)
So, I wonder if there is a general explanation for AO out there in “web world” that proves our assumptions incorrect.
basically the technique to calculate AO is that for any point on the surface of your model, you can draw an hemisphere around. then the engine analyses how much of the hemisphere is “outside” (in the light)
this is a crude scheme I found the other day when looking at “how it works”
on the left, everything is on the open, it’s 100% light
on the right, part of it is inside the model, it’s 70% light and 30% shadow.
as far as I understand, if you use a 1m radius hemisphere, that mean that any point that is more than 1m away from a corner will likely be 100% light. it looks like in SU we have a radius choice roughly going from 10-12cm to 120-150cm
the bigger you go, the higher the risk of starting getting AO from context (say a wall several metres away), and the higher the need for calculation. after all, going from a 1m radius to a 2m radius hemisphere means 8x more volume (cube factor).
in your example Hank, while the current max AO seems to have a sphere of about 3’, your building is 100x times higher. even if you only wanted to get an AO shadow to about a third of the building, that’s 33x higher, cubed, it’s 36k more volume than the current AO. per point.
my sources, this one about AO in openGL, from early Crytek to more refined hemispheres
and this one about AO in DX12 (so PC engine)
in the “motivation section” :
The premise of AO is that any geometry directly visible in a hemisphere around a surface point act as an occluder that lower the ambient coefficient, making the surface appear darker. A surface lit with such dynamic indirect lighting, provides us with more cues about its shape and placement among other objects in the scene and, therefore, resulting in a visually more convincing look.
so it seems that basic AO (non ray tracing) within DX12 still works the way it did in openGL.
2 Likes
I’ll add that in the second link, the first image shows an AO on buildings in the background higher than just 1m.
further reading explains that while ray tracing AO allow for better results, it also requires better GPU and more math time.
I’m gonna take a guess here, the upper limit of AO in SU has been set so people don’t need a RTX4090 to use it. While DX12 / Metal2 open a lot of possibilities, I don’t think the devs intend on SU becoming an alternative to softwares like cinema4d or vray with higher rendering capabilities.
it’s SU, it’s about ease of manipulation. That would also explain why SU’s AO looks a bit grid shaped on big images, not as smooth as in vray or twinmotion, and whi the upper limit is only about 4’ high. because it’s live and they had to choose a balance between top quality and speed of use (especially on mid tier machines)
but then again, It’s only speculation on my end based on what I understand of AO and what I know about SU.
Until someone behind the new tools comes here and speaks, it’s all I can do, uneducated guesses… 
2 Likes
Thanks so much @ateliernab for the deep dive on AO!
I agree with you that this decision must have had something to do with graphics processing burden. After all, that Sketchup is so nimble is one of its best features!
Anyway, hopefully someone on the Dev team will weigh in to let us know if larger scales are even possible!
1 Like
I have to agree with you here. Even though it isn’t common to have a model that scale, it is not unrealistic, and having the option to use this feature at that scale not only would be a simple adjustment, it would not hinder anyone else at all. It would simply be making the software better for more people 
2 Likes
This topic was automatically closed 183 days after the last reply. New replies are no longer allowed.