Scale and render question

[terraformers]

I know that it is best to stick to cm in Maya because that's the
default value for animation,camera clipping, scene handling,tolerance
and many other reasons.

Now that said, is it better to use real world scale ( 1 meter = 100 cm)
or the 1:10 scale (1 meter = 10 cm) just like it was advised in the
Gnomon dvd digital set 4 where when it comes to rendering in MR?

Because i did a quick test with a 5 meter wide room lit by a
mia_sun/sky system. when the scale was 1:1 (so 500 centimeters) it
rendered in a way and when scale to 0,1 ( so 50 centimeters) it renderd
in another way. I didn't even use FG so what's causing this? and how is
scale affecting my scene?

http://sor.typepad.com/.a/6a00d8350600cb53ef0120a5e998a2970c-800wi

scene:
http://sor.typepad.com/files/scale-and-render-issue.rar

[joojaa]

Scale has no bearing on how the physics of light works. However the calculations might become a bit different because you have different bias values, tolerances etc.

[digital-flood]

Scale might have no bearing on how the physics of light works, but it does affect how it works on the macro scale. Light decays according to , so if something is 1/10 closer together it will have decayed 1/100th as much. I think that's what we're seeing here. I modified the scene to change the direction the light was coming in from, and it changed the way the light bounced around the scene. At 1:1, no bounce. At 1:10, lots of bounce, even though there was 0 reflectivity.
Interestingly, when I switched back to a lambert material, the bounce was lost even at 1:10. So it appears to be some way that mia_material_x interacts with scale. It's a more physically accurate representation.

I plunked an area light into the scene and nuked the physical sun/sky, just to see what would happen. There was a lot of bounce then, too.

[joojaa]

afect how it works on the macro scale.

No it does not. In real life light acts equal in all scales.  You want proof? ts easy to prove.*

Light decays according to , so if something is 1/10 closer together it will have decayed 1/100th as much

yes however the decay works at equal rate, the act of scaling compensates for this theres also more surface area to decay FROM. Theese 2 terms negate each other. So at a distance of 100 the in the  10 larger scene the surface to decay form is also bouncing 10 times more light.

The only thing that does not follow suite is ideal lightsources such and pointlights and spots (but not directional lights that do not decay), which you will need to scale up to match the suite (by scale^2 as seen by light its a bit triky to calculate, because no real lights are point shaped, but in all actuality they simplify a small area light and when you scale area  increases and thus it compensates for this fact. ideal lights need scaling because they dont know this simple fact)

The decay law is propotional to any scale.

So it appears to be some way that mia_material_x interacts with scale. It's a more physically accurate representation.

No if theres a cutback then theres a simplification and assumption in mia material. In fact the mia material IS NOT especially physically correct. it just combines a lot of all the tricks we have learnt so far, and has some physically correct features. But by no means are all of its features correct. tahts a miss assumption. Its fporbably far better than most of you can acheve but not whan one can if one calculates the physics oneself.

So if it decays then theres a assumption in the shader that while motivated is wrong as the renderer sees it. The law of stochatics say that you should assume all samples as represantetion of whole set.

The mentlaray export of maya units is NOT defined. so 1 cm in maya on default config is assumed by mentalray shader to be 1 whatever unit is they assume. Wich may or may not be cm! So you cant really know what the scale is according to mentlaray. But thats beside the point since intensity is scaleless

Tested with a cube with point light at scale 1000 and 1 with and without fg if i scale the nonscaling ideal point light then the end result is the same. So what part do you neglect to scale? Photon emission?

Perhaps the portal light needs to know scale?