Sun study - best practice for irrmap calc?

[peterguthrie]

I'm trying to render a little sun study sequence for fun (when I really should be getting on with fee paying work), and was just wondering if anyone has any advice on what the best set up would be for something like this.

here is a test that I let run overnight: http://www.vimeo.com/670873

lots of crazy shit going on as you can see...

for this i used the new animation (prepass) & then (rendering mode) at medium preset with 2 frames mixed together. LC at 500 subdivs and being used for glossy rays.

I guess i should try it again on a higher preset, but what I can't get my head around is how the light cache should be used. Should i tell it not to store direct light? Should I use it for both the prepass stage and the rendering stage, and if so should the settings be different? Or, should I just use it to help calculate the prepass stage and then turn it off for the rendering stage? (in that case not using it for glossy rays)

Maybe I shouldn't be using the animation mode at all, but I thought that the vraysun would probably count as a moving object...

Any suggestions very welcome!

[tct70]

Link says the video is protected................!

[peterguthrie]

oops! fixed now..

[simon123]

peter did you find out about the sun study im after the same thing..

cheers

si

[simon123]

peter did you find out about the sun study im after the same thing..

cheers

si

[sv]

guys, the best way to do this wouldn't be with Irradiance map, but would be with DMC & LightCache. 500 subdivs is too low.

Typically, for moving lights & objects animations, i use about 1400 - 2000 Light Cache subdivs, don't use the LightCache for the Glossy Rays, and don't use the Adaptive Tracing. I pre-filter it normally between 50 - 500 depending on the scene.

Interior scenes can require more samples as the light is doing all those indirect bounces.

Unfortunately i find that using the LightCache for moving objects is one of the least documented features in Vray and i've had to find this out the hard way - perhaps some people on this forum way more knowledgable than me could step in and advise? chris nichols anyone?

[peterguthrie]

Thanks sv I'll give that a shot. I tried another one with similar settings to the 1st but on the High Irrmap preset: http://www.vimeo.com/673536

I'm tempted to try another one without Detail enhancement and with brute force as the 2nd bounce just to see what happens... and I'll do one using brute force as primary and higher LC settings as secondary like you suggest.

[Subburb]

hey ! nice experiment.. your scene is really nice...
why not a very subtle camera move while progression ?

For answer, yes, as your lighting is changing, you need to calculate a new light cache AND irrmap for each frame (not walktrough). that's the good way.
Store direct light can stays on, you'll speedup things a little.

can you give us your LC sample size please ?
(Lc samples results are strongly tied with sample size)
Remember, if your sample size is 0.02 with 500 subdivs, you need to use 1000 subdivs if sample size is 0.01 to get same quality. (correct me if i'm wrong)

Perhaps can you render some consecutive frames with LC only, to see if your splotches troubles comes from LC or Irrmap. (for this test, use store direct light OFF if you need to see crisp shadows)

I agree with SV, 500 Subdivs can be too low.
LC prefilter value depends of Lc sample size. if you use a large sample size (i.e 0.03), then prefilter at 500 is way too big imo... just try 64 if you're at 0.01 and 32 if you're at 0.02, it's a "not optimal but correct" starting.

[peterguthrie]

Thanks Jérôme,

There actually is a little subtle camera movement, but perhaps too subtle!

LC sample size is 0.02, i normally use at least 1500 subdivs for stills but in this case i was trying to keep frame render times down. I currently don't have any prefiltering going on because, well, I dont really understand what it does to be honest! Is it a good idea to prefilter for animations only, or does it help for stills too?

EDIT: seems i'm guilty of not having read the manual closely enough:

"If you want to use the light cache for animation, you should choose a large enough value for the Filter size in order to remove the flickering in the GI."

[Tommy_TGCG]

guys, the best way to do this wouldn't be with Irradiance map, but would be with DMC & LightCache. 500 subdivs is too low.

Typically, for moving lights & objects animations, i use about 1400 - 2000 Light Cache subdivs, don't use the LightCache for the Glossy Rays, and don't use the Adaptive Tracing. I pre-filter it normally between 50 - 500 depending on the scene.

Hello!

Are you using LC as the primary, or secondary GI? And apart from the pre-filtering, are you setting the regular filtering type to nearest, or fixed? World or Screen sample size?

I can't get a good animation when I set the LC as primary, regardless of the filter and subdivs settings, and it's just a simple outdoor test with just the default Vraymat...

Perhaps you could share some experience, because I'm about to start making an animation with moving objects and until now I've been doing precalculated stuff only.

[^Lele^]

An easier - and precise - way to evaluate light cache sample sizes and distribution is to use my LC/PPT calculator.
Instead of expressing sample size as normalised image fraction (0.02 / 1) it allows you to choose a sample size in pixels (it takes into account image size, and calculates what that pixel size equates to), and instead of a subdivision number, allows you to choose the shading rate, ie. how many times the LC samples should cover the image (it will find out how many pixels the image has, and how many subdivs are necessary to cover it, ie. for an image of 1k x 1k pixels, you'll need 1000 subdivs to reach shading rate 1).

Me, I can't do without it, but I am fairly biased here .

[Tommy_TGCG]

Me, I can't do without it, but I am fairly biased here .

Hmm, just tried the calculator... I think I'm about to get biased as well.
Thanks for the tool!

[Subburb]

Lele, you're the best !

[^Lele^]

I'm humbled
And of course happy as a pup that someone may find the results of my time-wasting habits somewhat useful

[peterguthrie]

Here's another attempt: http://vimeo.com/684784

This time I tried brute force for the primary and light cache for the secondary bounces. The light cache is set at: 1200 subdivs (11.11 shading rate), sample size 0.01 (3.6px), prefilter size 64.

Thanks to Lele for his work on his excellent LC/PPT calculator. While I see the benefit of expressing the subdivs as a sample rate and the sample size in pixels (esp when rendering at different resolutions), I still don't have much of a clue what I should be aiming for

So on this latest anim, do i need to go higher still in terms of subdivs to get rid of those light flashes (which I presume are caused by the light cache) and/or should I try a higher value for the prefilter size?

thanks