View Full Version : Shadow results: need advice.

06-08-2008, 12:10 PM
I can't figure out if this is looking correct. Those arrow shaped shadows are bugging the crap out of me. Is it just because thats the actual surface contour and the shadows are just exposing it? Or is that an error?

The object was originally displaced with an image. It had the same results, then I saved the transformed object, hoping it was a displacement problem, but same shadows. The renders here show an empty scene except for the 'save transformation' object and 1 distant light. Raytrace shadows.

I tried a point light, spot light, and even shadow maps. The same results. :( Do i just need to crank the tessilation on the original displacement?? The first render is single sided, and just to see what would happen I made the object double sided (shown in the second image) Odd results. The light is just at an 11 angle, so not below the surface at all.

Thanks for any advice...


06-08-2008, 01:02 PM
I think the second image looks pretty acurate. If you look at the three triangles in the lower left corner, you'll see them too in the wire.
Objects with surface normals on just one side give problems with shadows, so making them double sided is better.
I would keep the original mesh a sub-D mesh (double sided) and up the resolution of the displacement.

06-08-2008, 01:02 PM
Spot light with shadow map would work, except increase shadow fuzziness.

06-08-2008, 01:19 PM
Sekhar- Ya shadow map helps because the sharp points sorta fuz out, but still pretty undesirable because I'm aiming for pretty sharp shadows. But in a pinch I think that could help for sure.

Sarford- I sorta see what you mean. I've found toggling d-sided can sometimes fix problems you are having, and sometimes it is the problem hehe. You were right about cranking the tessilation amount. I took the transformed model and just hit sub-d on it, (making poly count from 12 thousand to 1.2 million) The results were more or less flawless. The shadows didn't show any artifacts. Lesson learned i guess. If light is cast near the tangent to a lower poly density surface you run the risk of these type errors. The solution is to increase poly count of the surface by a lot.

Thanks for the ideas guys.