CoolColJ's test pics thread

[CoolColJ]

Getting closer

I keep the same smooth Perlin noise combo, but using mulitplication of the 2 layers and then mulitpled this combo with another layer of noise, but using FBM, which has a rougher look to add some chaos. And I stretched the noise to give it a more wave like appearance.

Perlin noise(layer1 300% x layer2 675%) x (FBM noise with dimensions 100x200%) =

and the result underwater with 5% total noise output to water mesh displacement

100 samples

[OnoSendai]

Getting some pretty nice caustics there!

[CoolColJ]

Works pretty well above the water as well

At 1250 samples, but again the underwater cube doesn't seem to be lighting up. I guess all those irregular surfaces really disrupt Bidirectional MLT. I found Bidirectional path tracing better at forming the light on the underwater cube which is a surprise to me .... but BPT fails at all the other caustics

I'll put it up on the material database once I feel it's ready

[CoolColJ]

Ok I figured out the issue of Indigo rendering indirectly seen caustics really slow. It doesn't have anything to do with the Pool model, or bump map on water or pool surface (up to a certain point anyway). It's because of the texture for colour itself it seems! This might explain why I thought Indigo rendered that water caustic test scene under articifial lighting so much faster than under the sun - no textures...
http://www.indigorenderer.com/forum/vie ... 063#p97063

I set the water mesh noise displacement at 3% for a calmer look. I put a checker tile pattern texture at the bottom of the pool, still a flat diffuse surface, and bam all of a sudden the caustics render considerable slower. They seem to render 1 checker square at a time....

How can this be?!

30 mins each
with texture - funny thing is, with the texture, Indigo's samples per second are higher! So that means the MLT isn't working as hard?

Texture for colour, no bump map

without texture - samples per second drops, but the caustics appear faster, maybe indicating MLT is working better...

no texture

No texturing, but with the tile bump mapping the bottom, 0.003, not much slower, but a whole lot faster than with the colour texture!!

just bump map

[dakiru]

Ahh, I am enjoying this thread! Nice tests.

[Zom-B]

very interesting ColJ find about caustics + texture!

I would suggest you have a final test with diffuse color vs texture with same color.
Should in theory look both the same, but regarding your find, texture should render caustics way slower!

[CoolColJ]

very interesting ColJ find about caustics + texture!

I would suggest you have a final test with diffuse color vs texture with same color.
Should in theory look both the same, but regarding your find, texture should render caustics way slower!

Might have something to do with the colour values themselves.
Like I mentioned, with the tile pattern texture, Indigo's MLT renders one square at a time

I'm trying out some tests with -
texture - white and 50% grey, white and that dark blue, all white and no texture to compare

[CoolColJ]

Ok it has nothing to do with textures - but the colour values themselves!

MLT engine needs to be adjusted.
Basicly MLT is completely attracted to bright surfaces. It just doesn't work on dark colours at all!

The more dark colours you have in the scene, the less MLT works, and so the higher the samples per second, becoming more like path tracing....

I placed a texture at the bottom of the pool with various colours - all rendered for the same time
Black, highest SPP, but zero caustics...
50% blue, small amount caustics
80% white, lowest SPP, but plenty of caustics

Just to prove it - I place a black/white tile texture pattern on floor. Only the caustics on the white areas
Also made the cubes white, and that helped them to light up a bit

Even when directly seen - still no caustics on the black squares... Bidirectional path tracing had the same problem too.

[Godzilla]

I don't know if you've tried this yet, but perhaps you could use a dark gray shade instead of pure black?

[Borgleader]

I don't know if you've tried this yet, but perhaps you could use a dark gray shade instead of pure black?

Yes that would probably help, pure black like pure white does cause issues. Afaik pure black == complete absorption and so it would be normal not to see any caustics on the black squares.

[StompinTom]

If the black is 100% black, then it makes sense that all light that hits it will be absorbed, hence no caustics at all. I think it's a very similar issue as the 100% white issue, where you rarely have a 0% reflecting object, therefore even apparent black surfaces may actually be just a very very dark grey.

[CoolColJ]

Yeah I tried a dark grey of RGB 25,25,25, and while you can see some caustics, directly anyway I haven't tried it from above the water, it still looks a bit funny. I know in real life, caustics on black is just as bright as caustics on white

I still find it odd that 50% grey takes longer to find caustics than 80% white! Why should it really matter?
I know, traditionally, MLT does tend to love bright areas, but a caustic on a grey is still bright.

[ENSLAVER]

I think it's looking great. Through on black, no diffuse light reflects, no matter how much you shine on it. You may get some caustics coming back off it if it is reflective enough. At 25,25,25 you can see them pretty well, specially if you up the exposure. Perfectly black items rarely exist in real life, though they are easy to create on a computer.

re.

[zeitmeister]

Guys, really nice tests and gorgeous bumpmap!

Just returned from 1 week holiday in turkey, and I miss that pool water that much!!!

[Pibuz]

Just returned from 1 week holiday in turkey, and I miss that pool water that much!!!

...you bast*rd...
BAN HIM!