Mental Ray render problem

[winterlight]

Hi, I'm new to the forum could really appreciate some help.

I've been asked to do some light evaluations on some industrial vehicles and the only program that I have in 3ds MAX 7. The models are coming from the engineers from ProE and are very detailed, and the lights are Photometric lights using IES files from the light manufactures. The problem comes when I render with mental ray and use the Pseudo Color Exposure Control, I get odd light leaking or bleeding though the mesh. This is most evident when rendering a view from the eye point of an operator, the light leaks in though the front of the cab from the head light mounted there. With the default mental ray settings this is very pronounced, however by upping the amount of GI photons per light in the Light properties in the render menu with is lessened.

Here are the setting that I am using currently.

Lights:

Area, Ray Traced Shadows

Web Distribution

Automatically Calculate Energy and Photons

Global Multipliers

Energy=1.0

Caustic Photons=1.0

GI Photons=1.5

Indirect Illumination:

Global Illumination (GI)

Max Num Photons per Sample 1000

Light Properties

Average GI Photons Per Light: 1,000,000

Global Energy Multiplier: 1.0

Decay: 2.0

Since the nature of this is to try to see how far lights illuminate the ground around them and if there are any dangerous glare, areas over 350 LUX, I don't feel comfortable increasing the Energy Multiplier or the Decay. I've not worked with mental ray much other then my own personal work, which is very different from this.

[AJLynn]

Are the light leaks caused by geometry? Have you looked for overlaps and gaps and fixed any problems? Did you try a Maximum Sampling Radius?

 

Energy multiplier 1 and decay 2 are the most realistic values, so I wouldn't change those.

[winterlight]

I’ve checked for geometry intersections and have not found any the peaces of the model fit together as they would in reality. The leaks are happening around and thought the mesh. I set the Max sample size to be smaller then the thickness of the objects, about .25 inches, and that helped but the results were very grainy with very bright dots around the image.

[AJLynn]

Yeah, the problem is the 2-step process - first it bounces the photons, then it lights each pixel, but since there aren't enough photons for each pixel to contain a statistically significant number it averages the photons from the area around each pixel. If the radius is large, it gets a lot of photon hits, so statistically it averages out and becomes smooth - but since it does this by sampling a large area of the mesh, you can get photons from bright areas affecting areas that should be dark, and photon-less dark areas affecting areas that should be bright. Smaller sampling solves that problem but reduces the number of photon hits per sample, increasing noise - which can only be compensated with more photons, which increases render time and runs you out of memory.

 

Possible things to try would be using Final Gather, or changing the geometry so objects are in the bright part or the dark part but not both. You might also be able to get away with switching to the radiosity renderer and putting lines in your mesh to separate the light and dark areas, and possibly doing a Mesh Subdivide - I never used radiosity much in Max7, so I'm not sure how exactly how to optimize it.

[Thiago de Andrade]

As Andrew said, what about FG. I don't kwon if it's going to ruin your light accuracy, once you're really simulating the effect of lights. I never used the pseudo color e.c., I always use the Logarithmic. Does it affect your simulation? But this is not the cause of your problem. Try also using just the FG, without Pmap.

 

Can't you post us the image?

[winterlight]

Thanks for the replies.

The pseudo color render is just to highlight areas that have a dangerous amount of glare, around 350 LUX. When I render an image of the vehical normaly I use the Logarithmic setup though.

I'll try to see if I can post an image, I have to find out if the vehicle design is an in house only one or not. I'll try the model with Final Gather turned on and will post the results one way or another.

[winterlight]

Right now I can't post an image of the vehicle in question. I do however have some examples of what is happening.

This is a proxy of the cab that I am working with, there are two lights mounted inside of boxes that have be extruded so that the are solid with their walls being .25 inches thick. The lights are area lights that have a IES file attached to them. All the geometry fits together and there are no intersections.

[/img]

Now here I set the number of photons to 1,000,000 and made the minimum size .25 inches. now the leaking is gone but the image has a large amount of noise.

Here I used the same setup as above but turned on Final Gather with its default settings. The leaking is still there but not as prominent.

I've also hit another problem is that the setting I used here to get the best results crash mental ray when used. I fairly sure that it is running out of memory, the same thing happens with radiosity and I got it to run on one of the ProE computers that have 3.5 Gig of RAM. Are there any ways to get mental ray to work with less memory and not have it crash? Right now there is talk about getting new computers here and I'd like to know if there is another option other then more RAM to support upping the photon count.

[nisus]

Hi,

 

If you want to simulate real light, do not set deacy to 2.

 

nisus