Understanding DMC Sampler

[RockinAkin]

So here's the full 'elegant' formula to see the exact MIN and MAX values of your DMC Secondary Samples:

( Secondary Subdivisions / Primary Subdivisions Max)^2 = Maximum # of Secondary Samples per Primary Sample * ( 1.0 - Adaptive Amount ) = Minimum # of Secondary Samples per Primary Sample (gets affect if less than Min Samples value or higher than Max Secondary Samples per Primary Sample value.)

And just for fun and helping me understand, I whipped up a VRay DMC Calculator where you can input your settings and it'll use the above formula to do the math for you:

http://www.cggallery.com/tutorials/dmc_calculator/

Might be helpful for others as a learning tool.
Cheers!

[lllab]

thats a really great calculator!

thanks
stefan

[joconnell]

Not a bad idea at all Akin, if even to drive home how much "damage" your aa settings are doing to the other subdiv settings in your scene!

[peterguthrie]

ooooh! look at this lovely thread i missed. Going to read through later, once I've got some pesky work out of the way.

[joconnell]

Quick, everyone edit all your posts in case he gets even further ahead of all of us

[peterguthrie]

I'll be happy just to catch up. My brain doesn't work in the right way to be able to understand these concepts!

[joconnell]

To be honest I'd rather be in your position than making really technically clean, average looking renders

[fewlo]

if its any consolation, my brain doesn't work in the right way to be able to understand these concepts, nor can I make technically clean, average looking renders!!

[cheerioboy]

how do you decide when to dial up the min/max subdivs vs. playing with the noise threshold?

currently I'll keep my DMC sampling to around 1/4, 1/6 - and then dial the noise threshold lower till the sample rate pass looks good, and even then when all my scene lights are on I'm finding some noise in the shadows is actually an image sampling issue and not an issue with the lights subdivs. if that makes any sense...

[joconnell]

Most of the time I'm looking at the quality of the edges. Motion blur is also influenced by anti aliasing so that's a factor. I haven't had too many shots with leaves in them recently but they'd be a case where you need heavier aa values - I've found for a lot of vehicle shots on the film I'm on that I'm around 1/8 - 1/12 for the majority of it. Clr Threshold wise I'm rarely going below 0.01 for what it's worth.

Dialing the noise threshold might be dangerous as a way to manipulate the sample rate, you can make the sample rate look like anything you want it to be so what I'd be inclined to do is get your edges and bump detail to a good enough level first with a threshold of 0.01 and changing your max aa up and down, then work on the other aspects of the scene in sequence.

[joconnell]

Akin - just to say that here's a post that Svetlozar from chaos made on the adaptive idea - http://www.chaosgroup.com/forums/vbu...748#post550748 - it's much more comprehensive than mine.

[RockinAkin]

Akin - just to say that here's a post that Svetlozar from chaos made on the adaptive idea - http://www.chaosgroup.com/forums/vbu...748#post550748 - it's much more comprehensive than mine.

Ah, that's great. So that formula / calculator are indeed correct. Sweet!

[cheerioboy]

Still struggling with this, I've had renders that had a nice gradient of red along the edges in my scene within the sample rate pass. But then I've had lots of noise in reflective materials and more recently been battling blotchy GI. When I crank up the DMC sampler, my sample rate pass turns mostly blue but I get cleaner reflections while the GI is still an issue - these are interiors with much of the lighting coming from windows from a domelight... dunno some scenes are easier then others

[joconnell]

Which method of GI are you using? And which dmc sampler are you cranking up in this case?

[cheerioboy]

sorry I should have included that -

I'm using IRR + LC for the GI
&
Adaptive DMC sampler, set at the moment to 4/16, before it was 1/4 and I got some nice red only around the edges of my sample rate pass.

The GI issue has to be something to do with my lights & geometry, since all the views are in a single file and some look cleaner then others.. but getting the glossy reflections smooth by cranking up the DMC sampler seemed off from what I was trying to follow in earlier in this thread. Maybe having the samples of the materials around 16/32 isn't high enough?

Just a matter of trial and error!