Subject:
|
Re: New Part, 99% done: 6203.dat - Container ???
|
Newsgroups:
|
lugnet.cad.dat.parts, lugnet.cad
|
Date:
|
Sun, 5 May 2002 23:36:52 GMT
|
Viewed:
|
71 times
|
| |
|
|
Well, I know this discussion happened a while back, but I think I have
solutions to all of the stated problems, thanks in part to an observation
made by Ross Crawford in another thread:
http://news.lugnet.com/cad/dev/?n=7170
In lugnet.cad.dat.parts, Don Heyse writes:
> In lugnet.cad.dat.parts, Tore Eriksson writes:
> > Well, what should we call it?
> >
> > 0 Container
> > 0 Name: 6203.dat
> > 0 Author: Tore Eriksson
> > 0 Un-Official Part
>
> I don't know what to call it, maybe "Bellville Luggage"?
>
> Anyhow, I gave it a spin in a few of the CAD programs and
> noticed a few odd things I'd like to discuss. First, the
> use of half cylinders separated by quad spacers seems to
> produce some anomalies with the optional lines. The optional
> lines at the top of the handbag where the cylinders meet the
> spacer quad are missing.
>
> http://www.brickshelf.com/cgi-bin/gallery.cgi?i=135820
This wouldn't happen if the half-cylinder primitive were constructed in the
same way as the quarter-cylinder primitive. Specifically, it is missing the
final type 5 line. This appears to be the case for all of the cylinder
primitives except the quarter cylinder.
> The optional lines at the bottom of the handbag where the
> half cylinders meet the spacer quad appear too soon. You
> can hold down the ctrl key and spin the piece with the mouse
> in ldglite to see the optional lines popup out of thin air
> just a bit earlier than they should.
>
> http://www.brickshelf.com/cgi-bin/gallery.cgi?i=135821
>
> I suspect this is because the cylinder primitive expects to
> placed next to another cylinder primitive which would have a
> sharper curvature than the spacer quad. I guess the optional
> lines at the top were supposed to be provided by the matching
> half cylinder, if it were there.
This seems logical, and the control point is definitely placed like that.
It turns out, however, that the problem can be corrected by changing that
one control point. Instead of assuming the cylinder will continue, it can
be placed with the assumption that it WON'T continue. At first glance, this
would cause problems if the cylinder does continue. However, as long as the
next face has a comparable type 5 line, one or the other will always be
visible. This is extremely difficult to visulize, but it does work. In
fact, the quarter cylinder primitive is already constructed in this way.
> The other odd thing I noticed may be a peculiarity of ldview.
>
> http://www.brickshelf.com/cgi-bin/gallery.cgi?i=135818
>
> In this picture the (specular?) lighting comes out looking odd.
> Is this a result of the new 2 sided lighting scheme? I don't
> remember seeing this particular effect before.
This is because there is only one quad in the middle. Since the lighting is
calculated per-vertex, the middle of the half-disks get its own lighting
calculated. This can easily be fixed by breaking that quad into two quads.
This really should be done anyway, as it guarantees there won't be any tears
visible. Due to round-off error, the two edges might not meet in all
situations. If the quad is broken in half, there can't be any tears,
because the shared edges are made up of exactly the same points.
Note that this is a general requirement for real-time 3D computer graphics:
all shared edges MUST share all vertices exactly. If they don't, you can
get artifacts, such as bad lighting as shown and/or gaps.
--Travis Cobbs (tcobbs@REMOVE.halibut.com)
|
|
Message is in Reply To:
13 Messages in This Thread:
 
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
