Fog or refract?

[simmsimaging]

Robert: I gave up on using photon mapped caustics as I found them too hard to setup and too hard to control. Following a number of threads I opted to use BF GI so I could get caustics that way. Slower, but supposedly accurate. It does not have quite the same look as Maxwell/Fry type caustics right out of the gate, at least not to me.

Glyph: This test was not meant to be a great lighting setup, more to just experiment with a particular problem I was having on a recent job. In any case, the issue was that no matter how I lit the scene the stacks of bottles were going totally black very quickly, seemingly too quickly. The bottles could be better lit for sure, but that might have hidden the problem I was looking into. Thanks for the thoughts though.

b

[vlado]

Robert: I gave up on using photon mapped caustics as I found them too hard to setup and too hard to control. Following a number of threads I opted to use BF GI so I could get caustics that way. Slower, but supposedly accurate. It does not have quite the same look as Maxwell/Fry type caustics right out of the gate, at least not to me.

Accurate caustics are hard to achieve with the current algorithms in V-Ray, at least until we implement bidirectional path tracing.

Best regards,
Vlado

[simmsimaging]

For me accuracy is only really important in that the more they act like 'reality' the easier it is for me to intuitively setup materials and lighting. In the final result though, as long as they look good I am happy.

Good caustics is one of the things that brought me to Maxwell/Fry type rendering in the first place though as glass/liquids just 'sparkle' in those, at least with so much less hassle.

Looking forward to that new method though.....

Thanks Vlado.
b

[simmsimaging]

So I spent yet more time and made some lighting and material tweaks and I'm getting much happier with the material. I still think there are too many controls/knobs for fog though

In an attempt to get a bit closer to what I saw in Maxwell I opted to render with DOF (VrayPhysicalCamera) and I ended up with a ton of noise, but also some big splotchy areas. I think these are caustics that have not quite cleared, being exaggerrated by DOF, but I'm not really sure. Any guesses?

IMages below: clip of the hi-res first, and then a scaled down version of the whole thing. Once it's resized down it looks okay, but that's a cheat. Bit of post colour/contrast work on these, but no sharpening/noise etc.

Thanks /b

[simmsimaging]

Forgot:

Renders settings were pretty high. BF, no secondary, at 15 subd, Adaptive DMC 1/40 .009 threshold, min samples 6. I raised the DOF samples to 12. It's sort of a quasi-Universal settings approach.

If those grainy blotches are caustics, and I wanted this cleaner, especially in there, which settings should I be cranking up?

Thanks /b

[simmsimaging]

Hey Vlado - could you have a look at the hi-res image I recently posted and let me know if you think those really noisy spots are the GI caustics, and I could adjust in the file to avoid/reduce them?

Thanks /b

[vlado]

They are hilights seen through the glossy refraction (each time the hilight goes through a glossy refraction, it ends up being noisier). You can notice these in photos above too (although, of course, they are not noisy there).

Best regards,
Vlado

[simmsimaging]

Thanks for taking the time to look at it.

So to clear them up do you think it makes more sense to raise the DMC to higher than 1/40, or should I crank up the subdivisions of the glossy refractions in the glass? This took about 5 hours to render so experiments are very slow and it's hard to tell which would be more effective overall without running the whole file.


b

[vlado]

It is difficult to figure out what settings will get you a faster result, since there is also DOF involved. Perhaps do a small crop just on that portion with a few different sets of parameters and see which one is best.

Otherwise, using the universal method (1/100 DMC AA subdivs) will certainly work, but it will also probably not be optimal.

Best regards,
Vlado

[simmsimaging]

Thanks Vlado. Sometimes I wish CG could just be simpler. There are just too many parameters and possibilities to keep on top of, at least for me.

I did a quick test with glossy interpolation and that cleared up the noise very well. Thanks for that - I might have never thought of that parameter in this case.

b

[mosetti]

Hi Brett,
can you tell a little bit more on how you you tweaked the material? Looks really nice now!

Ciao
Marco

[simmsimaging]

It was really just playing with some of the values a bit more, and balancing out the thickness of the glass.

Here is the material I ended up with (without maps). I used some grunge maps (fingerprints etc) in the reflection and reflection gloss slots to break things up a bit. There is a falloff map in the refract slot which I used to darken the edges of the glass in some renders as well.

b

http://www.simmsimaging.com/uploader...eer_bottle.mat

[sv]

Forgot:

Renders settings were pretty high. BF, no secondary, at 15 subd, Adaptive DMC 1/40 .009 threshold, min samples 6. I raised the DOF samples to 12. It's sort of a quasi-Universal settings approach.

If those grainy blotches are caustics, and I wanted this cleaner, especially in there, which settings should I be cranking up?

Thanks /b

Bret - why not use Light Cache as secondary instead of none? Surely it wouldn't affect the rendering times to much, and might help to get a better light distribution into the bottles?
.....Perhaps?

Looking good though dude.

[simmsimaging]

Thanks Steve.

The reason for no LC is that I was working with settings from the job scenes, which had many many more bottles and it was definitely bogging down to use LC there. I had tested it on 3 scenes and really did not see much happening with it anyway though (had to put it in the RAM Player to even spot the differences). Might be more pronounced in this scene so I will go back and give it a try when I have a few minutes.

For sure upping the glossy samples and/or interpolating them took care of the noise splotch issue though - it was clearly not caustics in this case.

b

[simmsimaging]

No appreciable difference with LC off or on. Just the refraction glossiness parameters, and at least in this case it was really the interpolation that solved it. Even with 100 subdivisions it was still looking pretty noisy, with or without secondary bounces.

Thanks for the idea though - worth a try