GI headaches - Graphics

Gavin,

"Gavin" <replytogroup@nospam.comm> wrote in message
news:nN79b.147817$JA5.3605098@news.xtra.co.nz...
> Some questions I have:
> I'm drawing in metre units. Does the unit system you use to draw your
> objects have an influence on how your lights and shaders etc will work?

It will affect both. The most obvious result will be the light intensities
will need to be much different between the units. (about 1000 times
different. )

> I modified the 'standard' area light shader an added a decay parameter,
> which seems to have helped, but not nearly as much as I'd hoped. I've
ended
> up using an intensity of 3 and a decay of 0.01 in a room 8x10x3 units with
4
> area lights (6 samples per light) This is the shader:

> illuminate (P, N, 1.5707963 /* PI/2 */) {
> Cl = intensity * lightcolor / pow(L.L,gavdecay);

I'd change the decay part of the line to do it differently. Here's how
RhinoMan's arealight shader does it.

float atten = pow(length(L), falloff);
Cl = atten * (1-shad) * intensity * lightcolor;

> If I increase the intensity of the light, the carpet ends up getting
washed
> out and the ceiling is still rather dark. Is there any way to control the
> instensity of the GI apart from the lights?

Hmm, that kinda confuses me. The arealights actually have nothing to do with
the GI.(for the most part). What I'd do it set up the scene using just the
normal lights and get it close to what you want. Then try adding in the GI
(make sure the RhinoMan ambient light color is set to 0,0,0) and see what it
adds. You'll probably need 64 samples for the arealights themselves. I'd
also make sure the max reflection bounces is set to "1" so that the
irradiance (GI) sampling doesn't go too high.

There's also a tutorial for indirect illumination in that version's help
file that may be of assistance.

> I'm using Rhino 2 and Rhinoman and BMRT 2.6
> (sticking with Rhinoman in rhino2 for the mo' Brian, since I wasn't sure
if
> all of the features were implemented for v3 yet)

There's actually more implemented in the new version. It's even easier (I
*think*). If none of this helps, you can send it to me if it's not too big.
No binaries are allowed on this NG, but you can send a link to it.

Brian Perry

"Gavin" <replytogroup@nospam.comm> writes:

> I'm trying to render a basic four wall room with ceiling and floor and a few
> bits of furniture. Using point lights or cone lights give very unrealistic
> lighting. So I'm trying area lights and global illumination. But I'm
> still ending up with result with too much contrast, as if the GI isn't
> having enough influence on the lighting as I'd like it to.
>
> Some questions I have:
> I'm drawing in metre units. Does the unit system you use to draw your
> objects have an influence on how your lights and shaders etc will work?

You betcha.

> Normally I prefer to draw in millimetres - but I switched to metres back
> when I was exporting to Blender3D.

So area-based units need to be divided by a million, since everything
got 1000 times as close.

> I modified the 'standard' area light shader an added a decay parameter,
> which seems to have helped, but not nearly as much as I'd hoped.

This really shouldn't be needed.

<snip>

> I'm using Rhino 2 and Rhinoman and BMRT 2.6
> (sticking with Rhinoman in rhino2 for the mo' Brian, since I wasn't sure if
> all of the features were implemented for v3 yet)

Hmm. I assume you're using irradiance caching? How did you turn on GI?
I suspect that something just isn't turned on.

Another possibility is that your reflectances are all very low. For
direct illumination, you can compensate by turning up the lights. But
secondary bounces will be attenuated more.

--
-Stephen H. Westin
Any information or opinions in this message are mine: they do not
represent the position of Cornell University or any of its sponsors.

Does your scene include the front wall (the back-facing one)? This wall,
even if it’s invisible in the final image, has a very significant
contribution to the indirect lighting calculations. This is a very
common mistake and even the Cornell Box scene in the BMRT distribution
had the same problem.

If you don’t desire physically correct results, you can use a completely
different set of lights for the direct and the indirect lighting. This
can be achieved easily with the message passing mechanism, as described
by Matt Pharr in the Siggraph 2002 RenderMan course notes (available
from the RenderMan repository - www.renderman.org).

You can also try to raise the maximum reflection bounces, if you don’t
mind the extra time.

Paul Maurides


Gavin wrote:
> I'm trying to render a basic four wall room with ceiling and floor and a few
> bits of furniture. Using point lights or cone lights give very unrealistic
> lighting. So I'm trying area lights and global illumination. But I'm
> still ending up with result with too much contrast, as if the GI isn't
> having enough influence on the lighting as I'd like it to.
>
> Some questions I have:
> I'm drawing in metre units. Does the unit system you use to draw your
> objects have an influence on how your lights and shaders etc will work?
> Normally I prefer to draw in millimetres - but I switched to metres back
> when I was exporting to Blender3D.
>
> I modified the 'standard' area light shader an added a decay parameter,
> which seems to have helped, but not nearly as much as I'd hoped. I've ended
> up using an intensity of 3 and a decay of 0.01 in a room 8x10x3 units with 4
> area lights (6 samples per light) This is the shader:
>
> GAVarealight (float intensity = 1;
> color lightcolor = 1;
> float gavdecay = 1
> {
> #ifdef BMRT
> illuminate (P, N, 1.5707963 /* PI/2 */) {
> Cl = intensity * lightcolor / pow(L.L,gavdecay);
> }
>
> If I increase the intensity of the light, the carpet ends up getting washed
> out and the ceiling is still rather dark. Is there any way to control the
> instensity of the GI apart from the lights?
>
> I'm using Rhino 2 and Rhinoman and BMRT 2.6
> (sticking with Rhinoman in rhino2 for the mo' Brian, since I wasn't sure if
> all of the features were implemented for v3 yet)
>
> Regards
> Gav
> BTW, I can send the rhino file if that'd help and if the newsgroup allows
> binaries.
>
>

Paul Maurides <stud1313@N_O_S_P_A_M.di.uoa.gr> writes:

> Does your scene include the front wall (the back-facing one)? This
> wall, even if it’s invisible in the final image, has a very
> significant contribution to the indirect lighting calculations. This
> is a very common mistake and even the Cornell Box scene in the BMRT
> distribution had the same problem.

Well, I don't think of it as a problem. The real Cornell Box has no
front wall. Trust me.

> If you don’t desire physically correct results, you can use a
> completely different set of lights for the direct and the indirect
> lighting. This can be achieved easily with the message passing
> mechanism, as described by Matt Pharr in the Siggraph 2002 RenderMan
> course notes (available from the RenderMan repository -
> www.renderman.org).

Ouch. Sounds complicated.

> You can also try to raise the maximum reflection bounces, if you don’t
> mind the extra time.

That is something to check; after all, indirect illumination calculated
for one bounce is pretty boring...

<snip>

--
-Stephen H. Westin
Any information or opinions in this message are mine: they do not
represent the position of Cornell University or any of its sponsors.

Stephen H. Westin wrote:

> Well, I don't think of it as a problem. The real Cornell Box has no
> front wall. Trust me.
>
Of course you are right. After all, measuring the real radiosity
solution inside a closed box is difficult if not impossible

In case I wasn't clear, my point was that closing the box would
definitely make the scene brighter, because the energy would be
conserved inside the box.

Paul Maurides

Hi,

Thanks for the suggestions. The room is totally enclosed,
4 walls, ceiling and floor with the camera inside.

Ok, Brian, I decided to go up to Rhino 3, and now I can't
seem to assign arealights. Oh, hang on ... do you use a
Rhino3 rectangular light and assign a custom shader using
the material properties->light dialog box?

Also, In Rhino 2, BMRT would work ok displaying in
a display window, but in version 3, even if I change the
setting from framebuffer to display buffer to display
window, it automatically resets itself to framebuffer.
Also, the rendering manager doesn't have any progress
or time info in it.

Ah damn ... I created a rect light, assigned arealight shader
to it, and then tried to change the shadow type from default
to raytrace and rhino bombed out "the memory could not
be "read""

I'm not sure if I've got the latest Rhinoman, I d/l my version
about 2 weeks ago.

Regards
Gavin

Hi again,

Just one thing ... I never really got the point about
what units to use? Is there a standard that I should
stick to? should I draw in feet, meters, millimetres?

Regards
Gavin

"Gavin" <replytogroup@nospam.comm> wrote in message
news:Wrx9b.149142$JA5.3636932@news.xtra.co.nz...
> Ok, Brian, I decided to go up to Rhino 3, and now I can't
> seem to assign arealights. Oh, hang on ... do you use a
> Rhino3 rectangular light and assign a custom shader using
> the material properties->light dialog box?

You have two options:

1. You can just create a rectangular light in Rhino and RhinoMan will use
it's custom arealight shader for it. Just tweak the settings like any other
light in Rhino. Don't worrry about the shader.
2. You can also turn any object into an arealight by clicking the "object
properties" button for the object and choose the "arealight" tab.
Both options 1 & 2 result in the same thing if you use just a flat surface
so take your pick.

> Also, In Rhino 2, BMRT would work ok displaying in a display window,
This isn't enabled yet. I really hate display drivers....

> Ah damn ... I created a rect light, assigned arealight shader
> to it, and then tried to change the shadow type from default
> to raytrace and rhino bombed out "the memory could not
> be "read""
I'll check on that one.

> I'm not sure if I've got the latest Rhinoman, I d/l my version
> about 2 weeks ago.
It's way too old. I loaded one today that fixes lots of stuff.

Brian Perry

Hi Brian,

>
> > I'm not sure if I've got the latest Rhinoman, I d/l my version
> > about 2 weeks ago.
> It's way too old. I loaded one today that fixes lots of stuff.

I downloaded the latest rhinoman, but now whenever I click
refresh preview in the material manager, it bombs out. I'm
running rhino3 build march 2003, so I will have to download
the SR, but at 30 megs! it will have to wait till tomorrow when
I'm at work and can plug into the adsl.

I'll let you know then if it still crashes with the latest rhino update.

Regards
Gavin