Irradiance

[Shopper2]

Hi guys, i was just curious.

How different i the quality under IR for

-6,-2 and -4,-1 or -5, -3


Also, when should detail enhancement and antialiasing be turned on?

Thanks




Justin

[dalomar_cg]

Re: Irradiance

What irradiance map does is basically calculates an illumination solution at a different resolution then the actual image. 0 means that the IR solution is calculated at the same size as the image (for this example lets say thats 800x800), -1 means its calculated at half the resolution (400x400), -2 at a quarter (200x200) and so on and so forth. Ultimately the quality is going to be based on the max rate, but the minimum rate will have a play in the end result as well. The thing that having a minimum and a maximum rate allows you to do is sample different areas of the image at different resolutions. This allows big flat surfaces to have fewer samples and more detailed areas to get more samples. I generally shoot for having between 3 and 5 prepasses. Having more isn't any more useful then having 5 and the total solution can be prone to density bias. Having less is not good because its not very adaptive and the high density of samples can lead to splotchiness on large flat surfaces (the exact thing we don't want).

As far as Detail Enhancement goes, that is basically a hybrid IR+QMC method. What that does is leaves "gaps" in the IR calculation that get "filled in" with a QMC calculation at rendertime. The end result is something that is good for large flat surfaces, good for details, and fairly speedy as well. Its very effective and I typically use it for finals.

AntiAliasing is always on...as with IR there are 2 AA algorithms that can use adaptive sampling; Adaptive Subdivision and Adaptive QMC. Adaptive Subdivision is what is enabled by default, and I use that in most cases. Its quite effective and is pretty quick. Adaptive QMC I typically use when I'm doing a QMC render because they work well together.

[Micha_cg]

Re: Irradiance

From my test I have done: multipass was not the great experience, some times the nonmultipass was better at critical parts. Also, if the wrong in/max is set, the multipass is slower. If the multipass is set right, the speed is approx. the same like nonmultipass.
If multipass, than I use -3/-1 or -3/0. This seems to be the fastest setups, more passes cause longer rendertime and no extra quality. So, my test experience, maybe it's not general valid, but I use single pass and -1 most.