PDA

View Full Version : what IS this???!!!



illusory
06-25-2005, 11:30 PM
Lately I have been getting this UNFURIATING problem in interior corners, where right angled surfaces meet, with Roundered transitions. Maddening, because NOTHING i try affects it.

It happens:
with or without radiosity;
at any ray recursion ;
with a 'black box' surrounding the outside of the object, to block light leaks;
with specular turned on or off for the light;
on any antialiasing or reconstruction filter setting;
all the points are merged;
any type of light;
remodelling the SIMPLE GEOMETRY from scratch;

These are rounded corners, which I have been using for some time with success.
It happens whether I use 2, 3, 4 or 5 rounding polys.

I can't think of anything else to try!!! This is REALLY putting a damper on my workflow! :mad:

please, any ideas, has anyone seen this/know what it is???
It seems to be a recent, post LW 8 upgrade problem....

thanks,
NJ

illusory
06-26-2005, 12:27 AM
OK, I found the problem. I had smoothing turned on for the rounded edges. When I turned it off, the white dots went away. Don't know why this would be a problem, but there you go. At least I found it.

NJ

illusory
06-26-2005, 09:51 PM
Wel....

The ???!!! white dots are back again, in force --this time no smoothing to blame it on, unfortunately.

If anyone has any idea what's going on here, please help...

thanks,
NJ

UnCommonGrafx
06-27-2005, 06:27 AM
Have you tried turning on "Ray Trace Shadows"?

Looks like light going through your object. Or, at least it looks like what I've had in the past without shadows on...

jgutwin
06-27-2005, 01:00 PM
Looks like a 2 point polygon. Use 'w' polygon statistics to select 2 point polys and see if you have one or more there.

illusory
06-27-2005, 02:01 PM
I wished it was a light leak or a 2 pt poly, or unwelded points, but no...

I even put a box around the thing to cut off any unexplainable light leak (shadows are on), and really checked the geometry.

Thanks for trying though.
This has only been happening to me since one of the LW8 upgrades.

Any other ideas?
NJ

hrgiger
06-27-2005, 03:06 PM
I'm not sure why this happens but I can confirm that it has happened to me several times. I fixed it by putting more polygons behind the wall to keep the light from leaking through there (just make sure they're facing the light, not the problem area) and it was fine after that. It must be a error in the lighting/surfacing algorithim or something....

illusory
06-27-2005, 07:35 PM
I fear so....it doesn't ever do this using the Kray demo.

and i do have a very snug black cover (polys facing the light) over everything but the openings.

NJ

toby
06-27-2005, 07:53 PM
Can post the model and scene?

I like puzzles :)

IgnusFast
06-27-2005, 09:46 PM
I wished it was a light leak or a 2 pt poly, or unwelded points, but no...

I even put a box around the thing to cut off any unexplainable light leak (shadows are on), and really checked the geometry.

Thanks for trying though.
This has only been happening to me since one of the LW8 upgrades.

Any other ideas?
NJ

I've had some odd leaks like that, especially with ray-traced shadows going through the outside of an object. If adding blocking polys doesn't work, have you tried setting the surface to double-sided?

MonroePoteet
06-27-2005, 10:09 PM
If you're using Shadow Maps, bump up the shadow map size by a factor of 2, 4, 8, etc. For scenes where I have critical edges and Spotlights, I usually use a shadow map size of 2048, 4096 or 8192. The jaggy edges of shadow maps leak through when you least expect it.

mTp

illusory
06-28-2005, 02:58 PM
well this is very interesting. Yes, the model is set to double-sided, but the outside light-blocking poly shield was not. So I tried setting the 'blocker' to double-sided also. Guess what? That gives me black dots in place of the white ones!

So there's something going on with the renderer -- I never used to have to go to such lengths to block light leaks --all I used to need to is have clean geometry.

No, there are no shadow maps used, but I've tried every kind of ligh source on this.

Pics below show the white dots vs. the black dots, exact same frame and settings except toggling double-sided on and off for the blocker only. Hope you can see the black.

Maybe I should report this to the bug sticky thread?

thanks for advices,
NJ

MonroePoteet
06-28-2005, 03:14 PM
I can't reproduce this with a simple box using Rounder and flipped. Can you post a set of steps to reproduce this, including building the model and a capture of the settings window for Rounder?

Thanks,
mTp

illusory
06-28-2005, 04:55 PM
This was done by boolean add-ing two identical boxes, merging points, deleting polys at the intersection. Then apply rounder at the settings below.

HOWEVER, I just found something that got rid of these dots, which is very mysterious (at least to me). I had a morph going on at this frame, on another layer, of a wall opening to reveal the hallway (inside, where the halls intersect)-- behind that is where the dots appear. The polygons which formed the opening became transparent, then the points attached to those polys spread out, making the (transparent) opening wider. So in the frames where the dots show up, the transparent poly is covering the opening. When I killed those polys (instead of making them transparent), the dots went away.

attached below is an avi wireframe anim of the opening i tried to explain..

But I don't know why the transparent poly would cause this. AND its not the only instance where this is happening. There are other random white dots in other corners where nothing like that is happening.

count me confused,
NJ
PS sorry Toby i can't post the model, but many thanks for offering -- but though this puzzle is tracked down (if not solved), it happens in random other places as well

toby
06-28-2005, 05:24 PM
Could it ( in this case ) have something to do with ray-traced on not traced transparency, and/or ray recursion?

I don't trust boolean-ed geometry, is it possible to build it another way? Does this happen with non-boolean geometry?

illusory
06-28-2005, 05:39 PM
well the whole thing is ray-traced, nothing is excluded from that. I've tried ray recursions from 8 to 24, doesn't affect it.

As for the boolean, the resulting geometry of this is easy enough to check, and i have seriously checked it because of this. I don't see how the boolean could be it if the resulting geometry is clean. Yes, I did have the same thing happening on a simple room object, no booleans, but i did use rounder on it. I was beginning to wonder if too-small rounding divisions could have something to do with it, if that would bother the renderer. But with larger roundings it did still occur.

NJ

toby
06-28-2005, 06:05 PM
In my experience, it's when you do booleans that you get things like smoothing errors that don't go away until you rebuild the polygons, three-point polygons that are somehow 'non-planar', weird things like that.

Maybe it has something to do with the geometry that Rounder creates.