[tom]

This is the reference photo. Lit with my monitors (1 LCD and 1 CRT) in the room without additional light.
You see a red glossy candle cup sitting on a white paper over a black glass-top desk.
I took the photo last night at home with 0.5 shutter without tripod and the object is modeled in minutes.
So, this is nothing to do with photo-matching. It's only for roughly showing how frosted glass would work.

----------------------------------------------------------------------------------

Photo reference:





Outputs from Maxwell Render V2 are:

LCD + CRT + Room Ambient


LCD-only


CRT-only


Indirect Ambient


LCD + Room Ambient


CRT + Room Ambient



----------------------------------------------------------------------------------

Photo reference of another pose...



LCD + CRT + Room Ambient


CRT-only



----------------------------------------------------------------------------------

A few other frames from this home lab:

Director's Camera


More angles...



What if it was white and not red...



Again, this is NOT a photo-matching challenge. The purpose of this test is to roughly demonstrate
how does a frosted glass look when rendered in Maxwell Render and rendered in other engines.
Why other engines? Because, this could show you what's Maxwell doing in the bkg while some others are losing quality for speed.

So, here's the object files in a pack for the ones who wishes to give it a try with another engine:
http://www.fileden.com/files/2009/5/30/ ... _glass.rar

Now as this offered as a challenge, less words and more images please.
Good luck!

[JTB]

My question is.... What CPU power do you use for these images, because they are completely clean...and some rendertimes if you can... Other than that, they look perfect..

[tom]

They each took a few hours on Dual Xeon 5430 @2.66 (8 cores) to reach SL23 and most of them are out from ML. Small images were ready in some minutes.

[motopiku]

45minutes, shot (iray) in cpu mode, dual-quad xeon@2.27

[tom]

Very nice attempt motopiku. Except the darkenings on the grazing (both inner and outer sides) and the heavy inner reflection at the bottom, did you notice?

[motopiku]

yes, you're right about the darkening, but there is some imperfection in the maxwell version too (you can see just where you put the bottom circle, it's lacking a caustic in the plane), there is definitively something to change in the iray version, the roughness is not perfect, and maybe they need to improve the shaders, but is not too bad for a 45minutes rendering (and it should be even faster in gpu mode)



I gave you the test is made by the user Dagon, shot in the forums.

I make now the same scene for comparison Mentalray and vray.
alas for iRay 3dsmax and Octane ... I do not think you can do comparison.
iray is still in beta and I have to see how it works with max.
octane it has not yet light materials.

maybe make a test with smallLux could be interesting.
Just recently I took a picture FX4800 and I'm very curious to try the GPU (even if only 192 cores).

Long is the road that leads to good GPU rendering ... but rendering times should always be close e. .. focus on many computers (many CPUs) between consumption and consuming too hot.

hopefully good.

[tom]

but there is some imperfection in the maxwell version too (you can see just where you put the bottom circle, it's lacking a caustic in the plane),

That's because the emitter plane is not giving same directional illumination with the LCD matrix. See, that's why both Maxwell and iRay are alike in that pattern. But, here the darkening on grazing counts. Because, it's the heart of roughness and also Maxwell 1.x was not as successful as V2. OK, so iRay is on the green list anyway, what about others?

[motopiku]

excellent


I started doing some tests with Mentalray and I find the right frame for assessing the reflections.

I doubt however that Mentalray vray and has decent results.


This is another test of Dagon, always 45 min, the same material, it may be useful.

[tom]

My eyes promptly refuse it due to the darkening on this one as well. Quick peeking dagon's website I also found the same issue on a buddha render more heavily. So we may easily say iRay has this problem disqualifying it from the topmost level where I believe Maxwell sits.

Let's remember how above frame has been rendered in Maxwell:

[philip99]

...interesting topic Tom

[Lars Magnusson]

He,he, cool idea to use monitors as rectangular lightsources for photography! I'll have to try that! (easier than lighting paper with a flash)

Even though many render engines (biased and unbiased) can create extremely realistic materials, I think that one of the main benefits with Maxwell is that NextLimit has been able to make one single material interface. (not sure how common that is)

The ones I've tried, Mentalray (in CatiaV5), Lightworks and VRay for trueSpace, all have a multitude of shaders, for metal, glass, frosted glass, SSS etc, and in truespace it was split up in different shaders for color, transparancy, reflectance and bump...and if I wanted mapped specularity...I couldn't get rough reflections... so it was pretty cool (at least for me) when I got Maxwell to have one IOR that's used for both metalreflection and glass reflection/refraction, and that roughness also affects both reflections and refractions.

However...slightly back on topic.
Regarding the frosted glass and how Maxwell 1 and many other render engines get these dark areas...why is that?
Is it because the mathematics used are flawed approximations?
Don't know much about math, but I thought that almost everything was now described with accepted laws ( like the ones for refraction )
Or is it because these laws doesn't include stuff like the fact that frosted glass is clear inside and the frosting is only on the surface?

[tom]

I think that one of the main benefits with Maxwell is that NextLimit has been able to make one single material interface. (not sure how common that is) The ones I've tried, Mentalray (in CatiaV5), Lightworks and VRay for trueSpace, all have a multitude of shaders, for metal, glass, frosted glass, SSS etc, and in truespace it was split up in different shaders for color, transparancy, reflectance and bump...and if I wanted mapped specularity...I couldn't get rough reflections... so it was pretty cool (at least for me) when I got Maxwell to have one IOR that's used for both metalreflection and glass reflection/refraction, and that roughness also affects both reflections and refractions.

That's the idea, thanks!

Regarding the frosted glass and how Maxwell 1 and many other render engines get these dark areas...why is that?
Is it because the mathematics used are flawed approximations?

No, it's not due to approximations for a physically correct unbiased engine.

Don't know much about math, but I thought that almost everything was now described with accepted laws ( like the ones for refraction )

Only for speculars, though... Roughness is a different league and there are different approaches. Darkness in different engines could depend on several factors such as TIR, limited bounces or simply the limitation of roughness model they use.

Or is it because these laws doesn't include stuff like the fact that frosted glass is clear inside and the frosting is only on the surface?

No, this is not related.

[zdeno]

It is not biggest problem in 1.7.1 to get good caustics or reflection on 90 deg.
There is no chance to achieve this lovely saturated red.

Please Tom share with us settings of materials. Mayby i am doing something out of scale.

[tom]

It's not possible to achieve this quality with 1.x due to several reasons including bullets 1.3 Rough Transmissives Improved (BTDF, TIR, Roughness), 1.6 Caustics Improved, 2.2 Reflectance / Transmittance Normalization Avoided, 1.1 Color Accuracy Improved (Spectral <> RGB Conversion), etc
http://www.maxwellrender.com/forum/view ... 97&t=32630

[zdeno]

thx for quick answer.

so I am locked in "as easy as taking desaturated photo world" .

GO SPAIN ! make it 50% !!