hi

i've been going through this paper from ILM:
http://renderwonk.com/publications/s2010-shading-course/snow/sigg2010_physhadcourse_ILM.pdf [renderwonk.com]

one of the tests i decided to do is the light reflection vs geo reflection on page 24. increasing the specular angle on each grid reduces the light intensity on the more blurry reflections, seems to be doing the correct thing!
but when using area lights, my reflection intensities are quite different from the geo reflection, particularly when comparing it against the sharper reflection on the left grid.
the geo reflection returns 0.5 which is the same as the emit value, but the area light reflection returns a varying amount (top to bottom)from 0.08 to 0.04.

ok so i'm not sure on the scene scale they are using and the distances from source to reflection.
so the question is, is this the correct behavior? maybe i'm just missing something, or something in my scene?

BTW, if you guys/gals havn't read these papers, nows the time:
http://renderwonk.com/publications/s2010-shading-course/ [renderwonk.com]
the ILM and Sony one's are great

jason

Hey Jason,

I have tested your scene and come up with the same results. I would guess it has something to do with the attenuation that's present in lights only and not in geometry. But I have no idea how to fix this easily.

cheers,
Dennis

Yeah, the specular reflection is multiplied by the attenuation… it really shouldn't.

The workaround is to have twin light set-ups, one has “Affect Diffuse” On and “Affect Specular” Off, the other the reverse setting and with attenuation Off.

afaik, in the real world the intensity of a specular reflection (edit: arriving at a perfectly reflective surface) can only be affected by an opaque intermediary medium, like fog or tinted glass, or a surface roughness that causes the reflection to be spread over a bigger area.

Another gripe I have with “physical” “specular” light reflections is how they are shadowed by occlusions along “L” (regular diffuse shadows) rather than occlusions along the reflection vector.

S

And the workaround I suggested above wont work fully unless “Render Light Geometry” of the Specular light (with attenuation Off) is switched Off… leave the Diffuse only lights “Render Light Geometry” to On.

For some reason turning attenuation Off causes the intensity of the light geometry as seen directly by the camera to sky rocket to about 123, whereas the intensity of the light as seen on the reflective surface seems correct'ish.

This is all very confusing and unpredictable.

Correct me if I'm wrong, but, if I set a light intensity to 0.5, I expect to see:
- The Intensity of the light geometry as seen directly by the camera to be 0.5, regardless of attenuation settings.
- The intensity of the light geometry as seen in a 100% perfect mirror reflection to be 0.5.
- The intensity of the a lambertian reflection to be 0.5 at Zero distance from the light and thereafter modified by whatever attenuation function.

cheers
S

hey Serg
thx for adding to this, yeah, the first thing i played with was the light attenuation, i also confused as to why the light intensity jumps to 100+ values. plus i don't se why we should be trying to hack setups together to make this work, especially in PBR.

i'd be interested to here from Andrew on this subject?

thx
jason

To get physically correct results, you should change these 2 parameters:
- Turn off “Normalize Light Intensity to Area”
- Set attenuation to “Physically Correct”

This way, the value you assign to the light intensity will be the value seen in reflections. There should be no difference between a constant material and a light with these setting changes.

Andrew

hey andrew, thx for the reply

on my initial setup iv'e:
turned off “Normalize Light Intensity to Area”
set attenuation to “Physically Correct”

the intensity on rendered light is correct(0.5), but the reflection returns much lower values(avarage0.05) see image above

jason

This seems to be a problem with the use of the color limit in this scene. Bumping it to 100 fixes the issue, though this should not be necessary… I'm submitting a bug.

Andrew

hmmm, this should be fixed so that turning Off “Normalize Light Intensity to Area” isn't a pre-requisite for expected reflections.

I don't like this option, turning it Off only adds to physical correctness in that the output Intensity will change correctly with respect to the size of the light, if you were to think of intensity as some amount of light to be spread over an area… but this is much less intuitive than simply thinking of the value as the expected outcome. The only situation I can think off where this would be useful is if you were animating area size and wanted to actually observe this effect.

It's kind of an option you would turn off if you want to mimic a real world inconvenience! After all, wouldn't a Gaffer love to be able to turn this ON?

S

i must say, i was a little confused by “Normalize Light Intensity to Area” at first, i actually thought it was doing the the inverse, so turning it on would be normalizing the intensity to behave correctly.. my bad, only after this test did i realize it was a good idea to read the help tooltip first

i'd say it, should be off by default.

jason

Well, that feature was screamed for in early testing days, while it's not technically accurate, it makes lighting scenes a helluva lot less work. Every time you want to size up an area light, you had to manually re-fiddle the intensity. I agree that doesn't match the real world, but resizing area lights when setting up a scene you do a lot. I personally like it on by default.

Cheers,

J.C.

this might be ok for non physical renders, but i don't see the point of using incorrect behaviors with Physically Based Rendering…

it's similar to the idea that everyone wanted the surface shader to have separate light and object reflection, man this has opened a can of worms

jason

Option Off is not really any more physically correct than Option On… You can get identical renders with either option (apart from the bug with reflections) just by changing Intensity till they match in output.
It's just a question of convenience and what you want the Intensity parameter to mean.

Whether someone likes it On or Off is actually irrelevant, what matters is that the Light be reflected correctly regardless.

cheers
S

Guys, I just saw this in the Journal and remembered what Andrew Clinton said further up in this thread:

The Journal
Houdini 11.0.486: Fixed a problem with the color limit property that would cause too aggressive color clamping (beyond the specified limit).

… so get downloadin'

thx for the heads up Jason

j

jason|slab
BTW, if you guys/gals havn't read these papers, nows the time:
http://renderwonk.com/publications/s2010-shading-course/ [renderwonk.com]
the ILM and Sony one's are great

This is fantastic, Jason! Thanks for the link!