View Full Version : Polygon cracking in Maxman and BMRT

09 September 2002, 09:46 PM
I'm testing out Maxman and BMRT right now. I have a model that I built completely in max, originally as a poly mesh and then converted it to an edit mesh so BMRT could render it out. When BMRT rendered out the model it came out with huge cracks in it. I also used AIR to render out the image and there were little to no cracks in the image. Does anyone know what I need to adjust in Maxman/BMRT to get rid of those cracks? Thanks.

09 September 2002, 09:47 PM
Here's the Air image for contrast. If anyone know how to get rid of those long thin rectangle artifacts in AIR, that would be a great help too.

09 September 2002, 09:56 PM
is it modelled with quad polys?
could you post a wireframe
maybe try edit patch..

09 September 2002, 10:04 PM
hi again,
did a quick test
think meshsmooth creates those cracks?!
if i put MM model modifier after meshsmooth and switch
output triangles on cracks disappeared!:)

09 September 2002, 01:56 AM
Fex, the MM "output triangle" helped out a great deal. Thank you for the help. Now I have a new problem. The smoothing still isn't there. Each triangle has a significant amount of shading difference from its neighboring triangles. I am shooting for a uniform shading like the AIR rendering. Here is the new problem image.

09 September 2002, 02:00 AM
Here is the problem image

09 September 2002, 02:02 AM
Here is the wireframe you requested. Mostly tri's and quads. There are a few poly faces w/ more than 4 sides.

09 September 2002, 01:38 PM
Triangulation solves the cracking for a single reason: poylgons have to be planar, according to the RI spec. Since BMRT is a raytracer it can use sudden optimizations by assuming the RIB input conforms to the spec when it's polygons. Scanline renderers like AIR or PRMan don't have any advantages by conforming to the spec here, they plain ignore the non-planarity.

The shading discontinuities mean that normals don't line up. Since I don't use MaxMan though I, can't help you with this one export-wise.

I have a suggestion though: never use polygons when your target renderer is RMan compliant. These renderers are optimized for anything but polygons. Use SDS, NURBS, patches or quadrics but don't use polygons. The only exception is indeed AIR which seems to deal quite well with polygons.

Just one example: take a rough control polyhedron, apply meshsmooth a number of times until the density is sufficient for your rendering output resolution and render that image using e.g. BMRT or Entropy. Now render the same image with no meshsmooth at all but make the control polyhedron a SDS beforehand -- renders at least twice as fast plus looks good (silhouette-wise) regardless how close you get.

Additionally, RIB export times will be exponetially shorter with the second example.


09 September 2002, 06:17 PM
when using a rman renderer which supports subdivs (most of them do support subdiv except of 3delight i think) you don't have to use meshsmooth on your mesh except you need the subdivisions for further modelling. just add a mm model modifier and activate render as subdiv.


09 September 2002, 09:33 PM
If you read my post carefully you will find that this is exactly why I gave the xample since the op used polygonal geometry.


09 September 2002, 10:19 PM
Thanks for your help Mauritius, Flinch and Fex. (Man! germans really know their Renderman). Ok, I'm back up to speed now. The meshsmooth and the subd modifier both solved the problem I was having. Those subd's did render out a lot faster than I thought they would Mauritius. AIR had much faster render times with the model that had meshsmooth applied. AIR timed in at 27 secs, BMRT was over a minute. (Pixel Samples were 2,2 and Shading rate was 1.0 for both renderers. I'm running 500mhz PIII w/ 128mb sdram and Diamond Stealth III S540 Extreme graphics card). I couldn't compare the BMRT/AIR subdivision times because I couldn't get BMRT to render out my mesh with subd mm. It's strange because BMRT was able to render out a simple hedra primitive in subd. Thanks again for all the help. The mesh looks good in both renderers now. Here are the NEW IMPROVED IMAGES. (Left BMRT, Right AIR) :)

09 September 2002, 07:30 PM
If you experience problems rendering an sds, this may be due to connectivity.
A sds must be manifold. This e.g. means that any edge may only be shared by two faces. Check your geometry for such problems and retry.


CGTalk Moderation
01 January 2006, 05:00 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.