V-Ray 2.0: VRayBPTracer

[Bechstein]

As i understood reflective caustics by non blurred reflective faces like mirrors should be the strength for BPT. In a short test with the demo (2.00.01) caustics render clear and fast but could not be seen in the mirror themselves. Did i something wrong or is it a technical restriction?

[Dariusz Makowski (Dadal)]

Humh 2 question BTtracer, can I use it without GI ? Just straight forward render+ caustics? And Controling caustits is standard right? Nothing changed...


DVP3D = its under Helper>Vray if I'm not wrong

Bechstein there is 2.00.02 on chaosgroup www or even higher, upgrade ur version I hear they work on caustics in 02ver...

[lightpixel]

one question:
does BPT work as other unbiased render engines...that means, that it works with "emitter materials" instead of light objects??
second question: is it possible to render BPT over network... with additional rendernodes...?

[vlado]

does BPT work as other unbiased render engines...that means, that it works with "emitter materials" instead of light objects??

It can be done of course, but there is no specific requirement for it. You can use the regular V-Ray lights, including rectangle, sphere, dome, mesh, VRayIES, VRaySun lights (which are all physically accurate lights). Standard lights (omni, spot and direct) are supported provided that they are with physically accurate settings (i.e. inverse square falloff for the Omni and Spot lights). Most options of the lights (e.g. the Directionality of the rectange lights, and texture mapping) are also supported.

second question: is it possible to render BPT over network... with additional rendernodes...?

At the moment, no as there was no time for the release. However it will be added in an update.

Best regards,
Vlado

[anchovy]

Vlado-

Looks great! Share more test images if and when you get the time, its inspiring for sure.

-Michael

[anchovy]

Next youll be cooking up MLT. I see caustics...

-Michael

[anchovy]

I couldn't help playing some more with the PPT in 1.5. Here is a scene I grabbed from the indigo forum and reworked the materials. The light is a vray light mtl and all the walls ceiling and floor are glossy. The Subdivisions are set to 60k the max. Unfortunately there is still a fair amount of noise but that's as good it gets. Also this isnt completely unbiased as I have the sample size set to 0.04 for the light cach. Ill run one here soon with it set to zero for see what it looks like with no bias.



Here is the same scene rendered with the light cache. Not sure the settings matter so much as im sure most of you out there with more familiarity with vray could get a cleaner result in significantly less time but that wasn't really the point. Im more interested in what vray can do to create a more accurate image of the light transport. Its all academic, and there are other renders that are designed from the ground up to work with this approach but many if not most are locked into that strict physically correct approach and coming from the vfx side of the tracks there is always a need to bend whats realistic to the needs at hand.

http://morphographic.com/PathTrace/C...tCache_001.jpg

Lastly here is the diff between the two.

http://morphographic.com/PathTrace/C...T_Diff_001.jpg

You can see where they both have their issues. The cache version has light leeks around the pallet edges and is missing the bright reflections of the light in the glass teapot. The PPT image is considerably noisier and lacks any anti aliasing to help with the edge of the light.

Vlado, Id love to see it run with your Bidirectional version to see what difference that might make. If you have a free min please grab the file and give it a go.

http://morphographic.com/PathTrace/C...hTrace_001.rar

-Michael

[vlado]

I will render a BPT image when I get to the office tomorrow, but one thing I noticed is that you have the "sub-pixel mapping" option enabled in the Color mapping settings - turn it off, otherwise the result from the regular bucket rendering might not be correct (V-Ray even prints a warning about this).

Best regards,
Vlado

[anchovy]

Vlado-

Fantastic Im looking forward to the results. Im not seeing many firelies in the test images toy posted Im assuming you have it using multiple importance sampling and possibly weighting based on shorter connecting paths between eye-path and light-path. It certainly looks like the variance is way down.

Do you see a way in the future to have more than 60k subdivisions? Or do you think this might not even be necessary?

Thanks!

-Michael

[vlado]

Is there a particular reason you are not using a VRayLight? Relying just on GI and the VRayLightMtl material without the "Direct illumination" option is going to be very slow.

Best regards,
Vlado

[anchovy]

Vlado-

My guess was that the light mtl was going to be slower/ and have more variance. I was more of a pressure test of the system than for any real reason.

Heres the old WinOsi test scene with PPT. Again with 60k dubdivs.

[vlado]

Here is the BPT version. BTP only works with an actual light source though, so I'm also rendering a PPT version with a light source, for comparison.

Best regards,
Vlado

[vlado]

To answer your question about the 60k subdivisions - yes, it should be possible to implement more than that in the future.

Best regards,
Vlado

[anchovy]

Vlado-

Wow, looks great, a lot less noise! Yeah Id love to see the PPT version and diff it with the DTP. Do you anticipate that it will work with light mtls eventually? Also any thoughts on what can be done to help the AA especially around the light?

-Michael

[anchovy]

I dug up the Cornell box data And took a stab at converting the spectral info to RGB. The light intensity is off im sure but its still a good test file to play with.


PPT 60K subdivs


The large box is mirrored


Just add water.


Vlado, Im sure you've seen Kelemen's paper on a simplified MLT strategy. http://www.fsz.bme.hu/~szirmay/paper50_electronic.pdf
Id seen it a few years ago but hadn't read it. I looks like some rendered are starting to adapt it do to its relative simplicity.

-Michael