Jump to content

problem with shadows


giraffe
 Share

Recommended Posts

I was making test renders for lighting study and a strange thing happens. When I use shadow map, there is a part that is lit at the top cprner of the wall as in the picture. when the same scene, same light, same camera and materials but only changing thing is the raytrace shadow (instead of shadow map) then scene render correctly.

 

can anyone help me? I didnt understand why this happens.

Link to comment
Share on other sites

What software are you using?

 

Assuming its Max, maybe its just a shadow bias setting. Set it to .01 or better yet 0. If that's already the value, then I think we really do have a problem. If that's the case, post the wire, or the model. :)

Link to comment
Share on other sites

I remember having the same prob like yours. What i did is make all walls as boxes, ceiling and floors were planes. Its easier for me to adjust mapping individually, but the result was just like that. To work around with this kinda prob. I made the room as 1 box invert it turn it into poly, and turn the lights as 2 sided shadow.

Link to comment
Share on other sites

Hi Elif,

 

If you want to use a shadow map, you need to decrease the bias as suggested already, and increase the size of the map. If you don't have gobs of system memory, you might be stuck with raytraced shadows. You could have a completely tight exterior shell and still have light leaks if the shadow map is too small. 512 kb is generally too small - you might have to go to 3000 or higher.

 

As Alicja suggested (sort of :)), you need to supersample your floor material.

Link to comment
Share on other sites

What software are you using?

 

Assuming its Max, maybe its just a shadow bias setting. Set it to .01 or better yet 0. If that's already the value, then I think we really do have a problem. If that's the case, post the wire, or the model. :)

 

yup, it looks like a RAYBIAS problem, i had the same light leak, and it was because of the RAYBIAS.

Link to comment
Share on other sites

@Fran&Alicja- how would supersampling the floor material help fix the ceiling problem? ;)

 

has no one ever heard of light leakage? it's down to bad meshing. pure and simple.

 

Stephen,

 

The light jaggies on the floor was just an aside. I think I was pretty thorough in my explanation about the leakage.

Link to comment
Share on other sites

Alicja - what makes you think it's vray he's using? besides, even though he's not AA'd the image thats not what he's complaining about. it's not an AA problem he's got, it's a light problem. it's either light leakage or a bias problem. i'd put money on light leakage.

 

Sorry Fran, didn't mean to undermine you there. i know you know what light leakage is :)

Link to comment
Share on other sites

first of all I want to thank you all for your concern.

I am posting the render with the bias value 0-and there you can still see the problem on the ceiling.

I also thought this could be because of a wrong mesh and tried few alternatives such as

1. making a one box for the room-editable poly- flippnig normals

2. making box for one wall- editaple poly-extruding other walls

3. making seperate boxes for each wall

 

bythe way I tried plane for ceiling and the foloor and then I drew a box for each with height of 0.5.

 

I send u the wireframe too.

Link to comment
Share on other sites

there's your problem - inadequate meshing. i've circled the general problem areas -

 

 

shads.jpg

 

it has nothing to do with bias or normals, but more to do with the mesh.

 

look at all those edges and corners i've circled. those walls and floor and ceiling planes all overlap each other. they dont exactly meet up perfectly.

 

this is the problem. every edge should meet up 100% PERFECTLY. this is what i keep banging on about in other posts. sloppy meshing.

 

im not having a go m8, just pointing out areas to fix for your next model :)

 

to fix this make sure you have a single nice clean joining edge where the floor/ceiling and walls meet up together. no overlapping or over extending.

 

it's a strange bug in the system i know, but when you have a perfectly clean mesh the light leakage wont happen, i assure you :)

Link to comment
Share on other sites

I dont understand where the mesh gets wrong. here I attach the images where I have created the image from o single box and flipped the normals. again there is a light leak problem but this time from other places.

but I cant understand where the mesh gets wrong.

I created 2 planes for the ceiling and the floor and I used vertex snap so they are 100% in the right place.

can you please tell me where and how my mesh is wrong?

Link to comment
Share on other sites

this is where your mesh is wrong. notice in the red circles, the verts for where the ceiling/floor and the walls dont match up. they should -

 

 

shads1.jpg

 

 

i've indicated here in red the kind of structure you could have where all corresponding verts match up. this is just an example, you could model it different of course, but the result should be the same, matching verts -

 

shads2.jpg

 

again, if you're still suffering problems then try Fran's modifier idea.

also, like Fran mentions, render that same scene with full AA and a higher shad map size.

 

or post up your file and we'll see what we can do about it :)

 

(btw, did i call you a bloke? :p )

Link to comment
Share on other sites

Okay, going by the style of the transform gizmo, I'm going to conclude that Elif is using Max 6. Now, if she is using radiosity, then the way in which she has modeled her wall for the window opening is preferred for better rad processing - at least, that was how I was brought up.

 

With max or viz radiosity, you can create a box primative, flip the normals, set up a daylight system, process radiosity, and render. You will have light leaks with a single thickness wall. See attached. If the vertices don't line up on a box primative, then it isn't worth the extra modeling trouble to make them so. I find that by producing a model that corresponds relatively to real-world conditions, these problems don't come up.

 

Incidentally, I tried to reproduce light leaks in a double-thickness model, and couldn't do it. I think that was happening only with mental ray. Mental ray doesn't like me. :cool:

Link to comment
Share on other sites

thank you very much again :)

I will work-study and try the both solutions.

I am using max 5.1

 

this scene was only for testing purposes. till now, I always used boxes for walls and I wanted to try different modeling technique and try lighting mainly the shadow maps.

 

no v-ray, or radiosity or anything but just a scanline renderer with one direct and one ambient only. I faced across this problem when setting up lighting so not finished yet.

 

when I work enough on both of the solutions I will post the results.

thank you again :)

Link to comment
Share on other sites

no, i mean an understanding of how a model should be created and how verts should meet up is needed.

 

i've just drawn in those red lines to illustrate the point how the faces should actually be built up. they're just hidden face lines. anyway, even with editpoly, every face is a tri anyway at the end of the day.

Link to comment
Share on other sites

Hi Elif,

 

Shell is just a convenience. The important thing is that you know that having a double-thickness structure solves your light leak problem. Also, you need to have the edges of your surfaces meet with no overlapping. For example, your walls shouldn't extend into the ceiling plane, but meet with it at the edges, just as they do with a box primative.

Link to comment
Share on other sites

ok-thank you all of you :)

it has been great help to me- it is the first time I am experimenting with a interior which is a single mesh.

 

before I used to build every wall from a singla box and didnt have such problems.

and this problem appears only with the shadow map- and there is one more thing I found out during this experimentation is that when I decrease the falloff and the hotspot of the light, this leak problem lessens.

 

anyway it was a very helpful experiment for me :) I would like to re-thank all of you for your support and interest :)

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share

×
×
  • Create New...