MR FG and GI problems .. real use? MI answers

[jinmtvt]

here is an answer i got back from Mental images when i emailed them..
they don't usually do support to users directly..but i think that they found my questions revelant enough to give it few mins..

here is the email ....

---

how are you ?
Fine, thanks.
I'v recently noted that lot of people are using FG without GI for rendering purpose...
after having read the rendering book, this seem awfully non phisically correct ...is it ?
No, it's not correct because later light bounces are discarded, but
it often comes very close. It's usually good enough for entertainment
projects, where absolute correctness is not required. Final gathering
without photons is certainly supported by mental ray.
and then i try and use only FG
i get washed out white result and i have to turn down light intesity
You are probably not using a physically correct illumination model.
Phong is not correct, for example; it reflects more energy than came
in. It's an amplifier. Since final gathering samples and integrates
many light paths, light multiplies arbitrarily. The same problem exists
with light shaders that do not properly implement inverse-square falloff.
This causes final gathering to experience light that gets _stronger_
with the square of the distance.
Try the dgs_material and physical_light shaders from our physics library
instead. You can even find source code on ftp.mentalimages.com/shaders.
isn't FG supposed to only sharpen and help distribute photons by ray means when both FG and GI are turned on ?
It's an extra integration step over a photon distribution one generation
closer to the light, which means you can get away with far fewer photons.
But this won't help any with the shader issue.
Then in the rference book, you are talking about reflection coefficient...what is that ?
The amount of radiated energy divided by the incoming energy. Again, the
Phong model with "specular" and "sharpness" parameters is bogus.
Thomas

---

it came to me as a friend of mine that is using XSI1.X has the same problem as i do within max..

could there be problems in the material shaders of thoose 2 that wouldn't quite fit with MR way of doing ?

ihave tried ot find solutions but didn't ..

and i think that this prooves the point that FG wasn't intended to be used as a standalone rendering method but rather as a complementation of GI photons ...

what do you think ?
i need input from people that have been working with XSI thanks!

[atyss]

Bonjour JinMTVT,

Indeed, Final Gathering is meant to be used in duo with Global Illumination.

But, this is not a necessity. Most projects, I mean at least 95%, do not need physical accuracy in lighting. As soon as it's pleasing to the eye it's sufficient. This means that you can use FG as a standalone to achieve the look you want.

Also, FG has the property of "making objects emit light". In some cases, this is a valuable function. So you can use any object as a light source.

Using FG is extremely costly in rendering time. If you use GI at the same time, it becomes even more costly. FG alone can be a good compromise between realism and rendering speed.

As for the Phong thing, it's the first time I hear about this. You can use a Blinn illumination model instead, wich is very similar.

Salutations - Cheers
Bernard Lebel

[jinmtvt]

Bonjour Bernard

my concerns are more toward the problem of having to drop the lights when using FG ..

i can't understand why even using Blinn or any other material shader if you set up the lights to be working correctly using GI then u turn off GI and turn on FG u get a washed out result..that is simply wrong!
so there is indeed a problem with the interaction between shaders and we need to find what it is to be able to use GI/FG with normal lighting..

example .. we are supposed to be able to set up a scene with few light sources ( phisically correct ) and get almost final direct illimination settings..and then turn on GI/FG and get still a good result only with the additional secondary illimunation on diffus and specs..

though GI works correctly when using a properly set up scene, FG just washed out everything to white unless you try and surrender your scene with a black invesed cube or sphere wich is not the point when trying to do interior scenes ..

i also understand that FG alone can be used in rendering when phisical accuracy is not needed ..
but i still need to understand
hehe

merci pour ton commentaire neway

JIN

[wolfinger]

well, of course your scene will get a lot brighter when you turn on FG or GI.

When they are off, you are only taking into account direct light, but when they are turned on you are taking into account direct AND indirect light. That means more light in total, which may mean washed out brightness. This is to be expected.

[swaroop-1]

Ok, I'm a Maya user (Don't kill me for coming to the XSI forums ), and after testing MR3 in it, I've got quite nice results of GI and FG. And the only time I get washed out images with FG are when the cameras enviroment is set to something other than complete black. I haven't got a clue where this could be found in XSI though, but I'm 99% sure this is the problem your'e having.
As FG seems to use much more things to illuminate things, as opposite to GI which just calculates indirect illumination using exact photons shot from exact light source told to shoot photons, when FG actually illuminates (Atleast) trough surfaces with bumped incascedence (No idea on the spelling... you don't write good english at 3am ) and camera enviroment.
For an example, to make one of those blueish scenes with a model in a light blue enviroment illuminated by a "sky" hemisphere, I would just make a direct lightsource, lower it's intensity to 0.2-0.5, change the camera's enviroment color to dark blue (Light blue, or the color you would like the final image to have makes a 100% sure washed out image) and turn FG on... a Brazil like image ready in a few (dozen) minutes! Voil

[atyss]

In fact, you can even use a single point light placed anywhere under the dome with an intensity of 0.001.

Do a quick test: Put a dome (half-sphere) with a constant white shading, the floor to a light gray lambert, and the point light. The objects should not be washed out.
In the image below, the lighting is too dark. I had to lighten it up in post:

Salutations - Cheers
Bernard Lebel

[jinmtvt]

Bon ..again it is going out of hand..

let me clear few things first ok ..

first of all..i ain't an XSI user..i only use 3dMAX

then..i didn't purschare a mental ray license to do "DOME" renderings..i intend to use it for REAL rendering situations..not just an object in the middle of nowhere..

so please stop reffering to DOMES ..

then..the reason why FG doens't washed out everything when u put the environement to black is that FG shoots rays from the objects up to infinity or until it hits somethings..if there is nothing it usually takes the brakground or environement color ..so that's the reason why in a "object on plate" situation, we need to turn environement to black

i know that i can render good stuff without tunnig the lights down too much if i use a dome with a dark map on it..to get a brazil effect..but again..that is not the point

my clients don't wanna see their interior stuff inside a dome and all open..and i need to recreate real life lighting situations as close as i can ..
that is why i bought MR .. and it works dam good except for that thing with FG

so again..i'll quickly explaoin the problem..

the problem seems to be with the material shaders ( i've tried them all in max ..blin,phong,aniso,nayar blabal..)
seems like they are reflecting a non ralistic amount of energy when used with FG

i can get a perfect lighting situation with GI only using lots ofphotons..the only problem is that GI needs FG to get a real smooth look..i can't afford rendering for a week just cause i need 1 000 000 000 photon map
ahha

just make a test..in maya or XSI

build up a small interior scene where there is only 1 light source..like the window or something like that
then turn on the light square fall off propertie
and set it up to direct light everything like it should
then add only GI with a big photon count and a good sample ( adjust raduis to scene size )

should get a not washed out, with good light dispertion and indirect lighting render..

neway ..so if anyone else knows wass sup with stupid FG please let me know

thanks all for the input
good to see that people using different appz can talk the same in here

[jinmtvt]

just as additional info

hope that link works

so..that was rendered using GI + FG
i had to tune down the light settings though
and i used square fallof on energy settings for GI

i tried to light really low the scene..then used GI to bring the lighting up

there was a big box with a dark grey tint all around the building that absorbed alot of GI and FG and helped out with the washed out parts

pic was photoshoped alot after though to balance lighting and all ..

that was quite simple cause it didn't have to get a 100% realistic lighting look..was only to display building

you can see the GI effect a lot on the right "grille" type cover up in the separations ..

[atyss]

D'accord, the rendering looks nice, no?

What you describe on the cover up separation is also present on the dark parts of the building. This is called artifacts. To eliminate them you'll have to crank down the radiuses (GI and FG). For FG it can even be min 0.01/max 0.1. For GI I'm not sure.

If you decrease radiuses you'll see rendering time increase almost exponentially. This is the price to pay

You should really take a look at the tutorials on these pages. They are for SI3D and XSI, but the mental controls are exactly the same as in 3ds max.

http://www.edharriss.com/tutorials/tutorials\_render\_XSI.html
http://www.edharriss.com/tutorials/tutorials\_render\_3d.html

It should get you started.

Salut
Bernard

[srvargas]

Has anyone tried the recommended DGS material from Mental Images? I downloaded it from Mental's website. but they give only the source - it's not compiled. Does anyone know how to compile it and more importantly - how to create SPDL file for use with XSI (meaning incorporating the shader source and adding all the controls for the render tree)..
10x beforhand for any answers...

[atyss]

Bonjour Alive&Kicking,

I don't know if this will help, but here it is:
http://www.nanomation.co.uk/integrating.html

If this does not solve your issue and that no one on this forum answers your question, you still can ask to the XSI mailing list. On this list there are a couple of scripting gurus. Also, you might find answers in the documentation: Programming XSI > mental ray Shader Integration.

Hope this helps
Salutations - Cheers
Bernard Lebel