Angus Dorbie (dorbie++at++sgi.com)
Tue, 07 Apr 1998 11:37:47 -0700
This is a percentage closef filtering artifact because the
reconstruction filter is after the depth test, by definition
there is no depth reconstruction between samples in s&t before
the depth (otherwise edges & soft shadows would be broken), the
price you pay is that when you _want_ sample reconstruction ie
on a steeply sloping polygon in z you don't get it, there are
three things you can do:
1) increase the resolution of the shadow map.
2) Your problem is a lighting problem in theory the 'moire' effect
is only bad as your diffuse term L.N approaches 0 so this should hide
artifact if you have the lighting correct.
3) You can offset the depth buffer values in the shadow map be
adjusting r texture coordinate w.r.t the depth buffer info
since this is essentially a shadow map z fighting issue this
is similar to polygon offset in more conventional circumstances.
> 2. Glitches in the shadow itself (only seems to happen with
> certain pieces of geometry)
This might be ralated to LOD switching, consistent LOD is important.
> I thought the first of these effects might be the "self
> shadowing" problem mentioned in the man pages; however, playing
> with the PFLS_SHADOW_DISPLACE_SCALE and
> PFLS_SHADOW_DISPLACE_OFFSET parameters results in something that
> looks wrong (the part of the shadow which should appear on the
> geometry seems to "slide off"), and the moire is still visible.
>
> I've illustrated these problems in the following images:
> http://www.media.mit.edu/~kbrussel/shadow0.jpg
> http://www.media.mit.edu/~kbrussel/shadow1.jpg
> http://www.media.mit.edu/~kbrussel/shadow2.jpg
>
> Could someone who is using shadowed lights please take a quick
> look and tell me whether or not this is normal behavior?
>
> This is on an Onyx2 with iR graphics, but not the latest patch
> set; the latest graphics patch we have is 2326. It doesn't look
> like graphics rollup #4 addressed this, though.
>
> Thanks very much,
>
> -Ken
>
> __________________________________________________________________________
> Kenneth B. Russell Synthetic Characters Group, MIT Media Lab
> kbrussel++at++media.mit.edu http://www.media.mit.edu/~kbrussel
> =======================================================================
> List Archives, FAQ, FTP: http://www.sgi.com/Technology/Performer/
> Submissions: info-performer++at++sgi.com
> Admin. requests: info-performer-request++at++sgi.com
-- "Only the mediocre are always at their best." -- Jean GiraudouxFor advanced 3D graphics Performer + OpenGL based examples and tutors: http://www.dorbie.com/ ======================================================================= List Archives, FAQ, FTP: http://www.sgi.com/Technology/Performer/ Submissions: info-performer++at++sgi.com Admin. requests: info-performer-request++at++sgi.com
This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:57:13 PDT