Final Gathering / Irridiance Particle Output

[Falam]

I have a question about rendering with with Irrandiance Particles \ Final Gathering.

Is there a difference between indirect lighting with Irradiance Particles / Final Gathering and Global Illumination / Final Gathering ?
In my scene I did a comparison between the two and there was nothing I could notice. Does it matter which two you use, besides that Irradiance Particles are supposedly faster ?

[Daniel Brassard]

Some light reading for you.

http://forum.nvidia-arc.com/showthread. ... -Particles
http://userpages.umbc.edu/~bailey/Cours ... nders.html
http://elementalray.wordpress.com/2013/ ... -ray-3-11/

[Falam]

That was helpful, I had to brush up on my IP knowledge, now that I'm more comfortable. One last thing remains, is a IP/FG a IBL pass, and how do I create a IP/FG framebuffer. I've been toying around with the Photon shaders but I want some reassurance on the correct method, if it's possible to create a FG/IP channel ?

[remydrh]

Irradiance Particles will give you better detail typically in complex lighting scenarios than Final Gather.

Importons are shot into the scene and collect lighting information. Then the IP rays collect information from the more important importon locations (those that are in a lit area as opposed to somewhere dark) FG does not seek out the light, it pseudo randomly samples the scene, this is why IP might be superior in more contrasty light situations. Your indirect diffuse lighting pass contains the information from your FG + IP pass just as it would FG or IP alone.

Additional bounces of the IP rays are in multiples (it's a color), for example: 64 8 4 Would be 64 rays primary, 8 on the next bounce, then 4.

FG + IP uses FG as the primary bounce and then secondary bounces are IP rays.

In 3.10 and 3.11 you can use a string option to improve the result when used together (Sorry about SI and string options, Autodesk has been reminded of the importance). The string option to cause FG rays to go in the same direction as IP rays is:

Name: "finalgather importance"
Value: partial
Type: string

Or you can use importons with FG and omit IP rays (turn on importons and IP off) and use this string option:

Name: "finalgather importance"
Value: full
Type: string

Things to remember:

-IP doesn't currently support specular interaction, it won't pass through glass for example. This is where you would want to use it with FG. However, you might be able to hide glass objects from IP using the finalgather hide attribute, not sure if that's available in SI. For Maya it has to be added manually.
-IP Environment rays are deprecated since the inclusion of Native IBL and user_ibl
-The quickest way for a user to improve the result of IP is to increase the amount of importons in the scene
-Importon optimization calculation is currently single threaded, so lots of importons can be slow. However, importons + FG (no IP) can be fairly quick
-Brute force FG + IP interpolated may work well on some scenes (a la Vray brute force primary) but turn off the string option for importance, currently there are artifacts with brute force FG and FG importance

[Falam]

I used RRay String options add-on. Unfortunately, setting final gathering importance to 'partial' causes MR to fire an error; I don't know if 'always' is the same as partial or if 'always' is the default value before the string ?

// WARNING : RC 0.4 27 MB warn 082142: FG+IP, setting the IP interpolation to "always"

Is a IP / FG pass / channel really only a IBL pass ?

[remydrh]

It means that in your IP settings you might have selected something other than 'always' as the interpolation. IP + FG requires that the IP results be interpolated with importance on. That's probably fine.

I'm not sure what you mean by "IBL" pass. IP, Photons, and FG all store their result in the indirect diffuse buffer.

If you use the Native IBL or the user_ibl, the result is stored in the dirrect diffuse. That's because those methods are lights while Photons, FG, etc are all indirect lighting methods. This post might make the distinction between buffers and lighting more clear: http://elementalray.wordpress.com/2013/01/18/the-native-builtin-ibl/

[Falam]

I never bothered with the user_ibl shaders, most of the time I preferred to add the indirect lighting and proceed

There is one issue that I can't put my finger on. If I turn of all indirect lighting, the scene renders fine. When I render a diffuse_raw & diffuse_level channel, multiple the two, the composition is dark. With those two channels, and only those two channels the image should become brighter, it is the diffuse, all my images are linear. It doesn't matter if I load the images in Ps, or another program the results are the same.

I've tried other blend modes as well, and the composition with those two channels, and only those two channels, is dark. I have rendered out the diffuse_result as well, and composited over the other two diffuse channels, the results are the same, something is not right. Otherwise the scene renders fine, in the render engine. Why is it dark, there is only so far you can go with rendering out passes / channels, unless there is sub-channel/passes I'm unfamiliar with.

[remydrh]

user_ibl or Native IBL will be vastly more efficient at lighting with an HDRI. Take a look at the Native IBL post I linked to previously. You'll see the benefits immediately. In SI I would use user_ibl since the string options for Native IBL are not particularly stable right now.

Do not use multiplicative passes. Output additive passes, otherwise you will have zero hope of matching your beauty render. Multiplicative math in post may also cause edge artifacts.

Additive passes in current terms are:

direct diffuse (reflection and transmission)
indirect diffuse (reflection and transmission)
direct glossy (reflection and transmission)
indirect glossy (reflection and transmission)
direct specular (reflection and transmission)
indirect specular (reflection and transmission)
scatter
emission

Older terms are:

diffuse (WITH shadows)
indirect/irradiance (light)
reflection
refraction
highlight
translucence
SSS
incandescence

Add/Plus these in post software to match the beauty render. Currently all mental ray passes are output as linear regardless of lens shader or tonemapping used in-render. Output typically to EXR 16-half precision and it should work fine.

The IBL pass in SI is a Final Gather pass with an isolated environment, that's technically indirect/irradiance. Native IBL and user_ibl doesn't use FG. That result would be in your diffuse pass (direct diffuse)

[Falam]

You mentioned the additive passes/channels to use, yet I only have these channels to render out, even if I render out the a regular 'diffuse' channel, the channel is dark. The results it seems no matter what route I take are a far distance from the beauty render.

Irradiance
Reflection
Refraction
Spec_Raw
Spec_Level
Spec_Result
diffuse_level
diffuse_raw
diffuse_result
refl_raw
refl_level
refl_result

Edit: I've run into a new issue, rendering in render region has halted due to this error:

// WARNING : RC 0.4 260 MB warn 082054: overwriting tag buffer with density data as required by '-diagnostic photon -x'
// WARNING : RC 0.4 260 MB warn 082051: diagnostic mode photon requires caustics or globillum

I have FG/IP off, I can't see what could be the problem, it is late here, maybe I don't have the frame of mind to solve it.

[remydrh]

1. Only use the result buffers, don't use level and raw
2. Are you viewing the buffers in the correct colorspace when you composite them? Remember, framebuffers are in linear colorspace. That's dark compared to sRGB or even rec 709