Need help pls!!!

[jerry]

Hi there! It's my first time postin here. Desided to render in mantra and everything was fine until i switched to Raytrace engine. As u can see in attached .mov some strange dark"ness" as cam moves around simple geo. As i discovered it only happens in raytrace with motion blur and AO/ or full Irradiance presenting in the scene. In micropoly everything is ok.

I tried to minimize that dark"ness" increasing the raytracing bias value in Shading tab in manra rop op, but the problem still remains/visible on the narrow places.

I just want to find out is it some mantra spesific issue that i dont know, and is it possible to solve?

....Anyway sorry for the newbie question. Hoping for soon answer/// probem.mov

[DaJuice]

Could you upload the file Jerry? Hard to say why that's happening just from the video.

[jerry]

Yes shure, here it is...

test.hip

[jerry]

forgot to mention --- it's Houdini 11, build 491, and in later 11' builds even in 10 the same problem. Woohhh -- there must be something....

[jerry]

Does enybody help me with that? --- or it was a bad idea to render in mantra

Have to render large amount of instanced geomenry and its preferable to use raytrace engine. So if it will be unsolved i would switch to mray again...

Come on houdini masters dont let me go to autodesk's buttons.... sphere.mov

[jerry]

--- the recipe is simple: but Box(give it a Facet sop) - Rotate Cam arount it(ins at least 2 keys) add AO through Vex GI shader switch to Raytrase turn on MB, and render ---- as cam mooves under the certain angle to the cube f..ing "darkness" atarts to emerge. --- fighting with it for 3 days....

[jerry]

Maybe this will explain wthat the probl is----

micropoly.mov

raytrace.mov

[Macha]

Can't open your hip file (makes my H crash)

Do you have a distant light in there by any chance?

[jerry]

this is a hou 10 file

houdini10.hip

[Macha]

I don't know the answer to this and it is a little tiresome to test but my gut feeling is that this has to do with the angle the light creates with the surface normal and the camera. As you may know there is a particular shape of intensity with which the light scatters from a surface; a kind of a lobe shape. I guess what you observe is this change of intensity as you move the camera through this lobe. Why this happens only in raytrace, I don't know.

You can esily observe this effect in real live on matte, flat metallic surfaces.

Anyway, this is my guess. See if you can get rid of it by using an environment light, multiple lights, or separate (or switch off) specular light.

Edited September 7, 2010 by Macha

[Mario Marengo]

Looks like a biasing problem to me.

If the shadows are ray traced, then try increasing the tracing bias (Mantra ROP, Properties>Shading>Raytracing Bias). If they are depth-mapped shadows, then try increasing the light's Shadow Bias. But since you also mention AO, I'd increase the tracing bias regardless of how the direct shadows are being produced (since AO is traced).

Good luck.

[DaJuice]

Hi again Jerry,

I deleted your light with the vex_gilight shader (old method), and replaced it with an Environment Light to do the ambient occlusion. That fixed it. Also Raytracing Bias can be decreased a bit for this scene's scale, to something like 0.001.

[jerry]

Hi all and thank for reply!

Environment light doesnt do the trick because:

-- in Direct lighting mode it acts like big area light, plus it doesnt have those distance treshold like in AO, --- too slow

-- in Ambient Occlusion mode it uses the same vex_gilight shader --- same probl

As for rays Biasing:

-- decreasing it will sharpen darkness,

-- increasing will minimize it and You will LOOSE details

-- MORE that, when you have a noisy geometry like trees its almost impossible to get rid of it

I wonder if somebody here didnt faced with it. Does enybody here rendered in raytrace with MB and AO? --- Is it so rare case?

Edited September 8, 2010 by jerry

[Macha]

Are you sure you have a problem there? I get the feeling it is slightly unusual but normal, expected behaviour resulting from your particular choice of lighting.

Edited September 8, 2010 by Macha

[jerry]

---- well, maby its normal BUT it wasnt expected, and at least it "controls" the quality of the final image.

--- rendered simple sequences with diff raytracing biases in raytrace with and without MB, also in micropoly with MB --- just for comparison.

just notice how .1 bias is flatening AO. I wish that Raytrace with MB look like in micropoly with MB and with .001 bias !!!

----- have to render a 50000 those inctanced trees ---- its not real in micropoly

raytr_noMB_bias.001.mov

raytr_MB_bias.001.mov

raytr_MB_bias.01.mov

raytr_MB_bias.1.mov

micropoly_MB_bias.001.mov

[burgess3d]

Hi I posted a possible solution over on the Sidefx forums. Hope this helps

http://www.sidefx.com/index.php?option=com_forum&Itemid=172&page=viewtopic&p=96832#96832

[Mario Marengo]

Hi I posted a possible solution over on the Sidefx forums. Hope this helps

Burgess beat me to it, but I'm posting anyway just in case.

Attached is a quick test using the shader-based AO calculation (the "Trees_TracedOcclusion" object), and a plain anvironment light based irradiance version (the "Trees_EnvLight" object).

The shader-based AO has the maxdist control that you like, but I'd be careful with that -- it usually sounds more useful than it actually is.

In any case, I didn't come across any problems in this test with 100 tree instances and motion blur (which renders in just over 1 minute on my machine using the simple envlight irradiance -- aka: single-bounce diffuse -- version).

houdini11_AOtest.hip

I don't have H10 handy so this was done in H11.0.484.

HTH.

[jerry]

A Huge THANK goes to burgess3d!!! ---- Probl SOLVED. Now i have a proper AO in RAytrace with MB-- voila....