Manta Interactive Ray Tracer Development Mailing List

Text archives Help


Re: [Manta] missing 1/pi in area light code?


Chronological Thread 
  • From: Peter Shirley <shirley@cs.utah.edu>
  • To: "manta@sci.utah.edu" <manta@sci.utah.edu>
  • Subject: Re: [Manta] missing 1/pi in area light code?
  • Date: Tue, 25 Dec 2007 13:29:32 -0700
Ahh. I think I follow you. The missing 1/pi is from the entire equation so it is Lambertian that is amiss? Will
all materials have an analogous problem?

Attached is the image with the emitted light color being 0.333 and a Lambertian sphere with R=1. So the top
of the sphere should tend toward 0.333 but instead goes to something around 3x that. (the cruft near the top
is not a bug-- uniform area sampling just has high variance when the cosine terms are ignored by the pdf)

Pete

PS-- Merry Christmas POV style: http://hof.povray.org/ChristmasBaubles.html

Solomon Boulos wrote:
I disagree. Look at parallelogram in combination with arealight. Also lambertian happens to not include a division by pi but this is a separate issue.




On Dec 24, 2007, at 9:51 PM, Peter Shirley <shirley@cs.utah.edu> wrote:

I think the color computed in area light is missing a (1/pi) I think. The test is a reflectance 1.0 sphere
tangent to an area source with color 1.0. This is too bright by about 3x and the code seems to be
missing that as well.  Anyone agree?

Thanks.

Pete


PNG image


Archive powered by MHonArc 2.6.16.

Top of page