Page 1 of 2

normals or smoothing problem ?

Posted: Mon Oct 08, 2007 6:06 am
by delic
Hi all !
On the following image there are some black edge reflections .
It seems to appear only meshes that were subsurfaced in blender with creasing ... exported in blendigo0.9, rendered in 0.9test7.2 .

Looks like a smoothing pb !

What do you think ?

Image

Posted: Mon Oct 08, 2007 6:14 am
by Kram1032
In most cases, creasing isn't a good idea... adding additional loops, to control the shape, is way better...
Might be a mix of normal- and normal smoothing error...

Posted: Mon Oct 08, 2007 6:34 am
by delic
But the problem appear also when the creased subsurf is applied to the mesh, so the mesh should be clean, and is on other renderers ...

The normals are also correctly pointing outside !

any :idea: ?

Posted: Mon Oct 08, 2007 7:13 am
by Kram1032
hum...
try to recalc the normals, anyway... also try reimport.... (obj export, obj reimport)

and are you sure, you have edges, flatter than 30°? (that should be about the limit)

Posted: Mon Oct 08, 2007 8:24 am
by zuegs
ensure objects are uniform scaled... normals can go wrong on objects with non-uniform transformation matrix. Use Ctrl-A to apply scale and rotation to the mesh.
btw: nice scene !

Posted: Mon Oct 08, 2007 8:33 am
by delic
I recalcutated the normals outside(ctrl-N in blender), and checked the normals by displaying them.

I also applied another subsurf modifier with no creasing to smooth the mesh well, even if is heavy, to avoid smoothing pb, but didn't check if they are les or more than 30deg ...

I will try to remodel the lavabo without creasing, to see if it comes from that.

What I dont understand is that the problem remains when the modifiers are applied !

But I like creased models because they offently are better for homogen displacing than classical boxmodelled subsurf with edge loops.

Posted: Mon Oct 08, 2007 8:43 am
by delic
I also wondered how to get a pefect mirror reflection the the mirror.
In the image on the first post a Cr.nk material is used, and the reflection is blurry, even when more converged.

I also tried with a 1000000 exponent phong white material, but it gets white, not reflective !!!

What is the best setting for pure mirror reflect ?

Posted: Mon Oct 08, 2007 10:30 am
by manitwo
try raising the exponent of your cr.nk material (~10000000 just to make sure :) )
or use this instead:

Code: Select all

	<material>
		<name>mirror</name>
		<phong>
			<diffuse>1 1 1</diffuse>
			<exponent>100000000</exponent>
			<ior>1000</ior>
		</phong>
	</material>
or

Code: Select all

	<medium>
		<name>highior</name>
		<precedence>10</precedence>
		<basic>
			<ior>1000</ior>
			<cauchy_b_coeff>0.0</cauchy_b_coeff>
			<absorption_coefficient_spectrum>
				<uniform>
					<value>0</value>
				</uniform>
			</absorption_coefficient_spectrum>
		</basic>
	</medium>

	<material>
		<name>mirror</name>
		<specular>
			<transparent>false</transparent>		
			<internal_medium_name>highior</internal_medium_name>
		</specular>
	</material>

Posted: Mon Oct 08, 2007 11:45 pm
by Kram1032
as you mean a perfectly sharp mirror, you should use specular :)
phong means blurry reflections, where as specular ignores that possibility. ;)
To get a perfect mirror in terms of reflecting all the light, you need an infinite IoR, but 1000 - 100000 shouldn't give any difference, any more. 100 is VERY hard to notice, but there already is slight darkening. (I tested that, once)

Posted: Tue Oct 09, 2007 10:47 am
by Vanessa07
Hi

I had the same problem, don't use set smooth if your object is too subdivided with the subsurf for exemple, it's not a normal problem

Posted: Tue Oct 16, 2007 11:19 pm
by delic
Really cant avoid this, and when "set smooth" is not checked, with big renders it really noticeable !

Perhaps exporting and reimporting the mesh ...

Posted: Tue Oct 16, 2007 11:27 pm
by BbB
Export into an .obj, which will freeze your mesh and fix the normals issues, and reimport. It works for me in all cases.

Posted: Wed Oct 17, 2007 7:54 pm
by zuegs
still in same opinion:
ensure objects are uniform scaled... normals can go wrong on objects with non-uniform transformation matrix. Use Ctrl-A to apply scale and rotation to the mesh :?

Posted: Sat Oct 27, 2007 10:21 pm
by delic
yesss, export/reimport in .obj works !

Posted: Sun Oct 28, 2007 1:39 am
by CTZn
manitwo wrote:try raising the exponent of your cr.nk material (~10000000 just to make sure :) )
or use this instead:

Code: Select all

	<material>
		<name>mirror</name>
		<phong>
			<diffuse>1 1 1</diffuse>
			<exponent>100000000</exponent>
			<ior>1000</ior>
		</phong>
	</material>
Hi manitwo !

I think you'd better use 0 0 0 for the diffuse, real Cr has a diffuse comp near 0. Think about it: ther is a law called approx. "conservation law of energy" wich states that if a material is 98% reflective then it can't be more than 2% diffuse. Breaking that law just creates physically incorrect shaders.

I hope what I'm saying is correct Indigo wise, ppl please bevel your edges then correct me if I'm wrong ;)