Subject:
|
Part building advice - too many triangles/quads?
|
Newsgroups:
|
lugnet.cad.dat.parts
|
Date:
|
Thu, 25 Nov 2004 05:12:28 GMT
|
Viewed:
|
4230 times
|
| |
|
|
I’ve been working on 6898.dat, the cook’s hat, and I think I nearly have a
finished part to submit. The hat of course is in reality a more “sculpted” piece
without true 8-way symmetry, but I decided for simplicity to make it
symmetrical. The bottom of the hat is also not supposed to be flat, so this I’d
like to change. For completeness I threw in a homemade ring along the bottom to
join the 14-radius outer edge with the 13-radius inner edge, but I’d have to
redo that ring to fit any kind of slope I might add. (I’m thinking a 1/10
gradient, since it scales easily and is darn close to the real thing.) Overall I
like the way it looks right now.
There’s one problem that nags me: It has a ton of triangles, quads, and
lines--more so than many curved parts. Without manually counting out the
surfaces (I don’t know if any tool I have actually does that), I figure the
“puff” of the hat--the rest is mostly primitives--has the following counts:
- 72 solid lines
- 64 triangles
- 336 quads
- 676 conditional lines
Because of the hat’s symmetry, I split the top into 4 quadrants and used a
subpart. I didn’t want to use 8 pieces, in spite of the angular symmetry,
because the ends wouldn’t join properly. (Of course if I really wanted to trim
down file size, I could probably pull out the lines along quadrant edges and
diagonals and just chuck those back into the main file, then do some sort of
diagonal mirror matrix with a BFC invert. Naturally I’m not itching to do that.)
The original file without any subpart use was around 111K. The current files are
about 2.2K for the main part, 29.2K for the subpart.
Anyway I’m wondering if the sheer number of surfaces is just too many. Right now
the hat’s puff is done with longitudinal lines at 7.5 degree increments. I could
probably reduce the number of vertices close to the puff’s vertical center
(y=-6) by taking out the lines closest to 0, 45, etc., which doesn’t seem to
sacrifice much quality, but that doesn’t work near the peak. (When viewing
edge-on, the top looks like a bunch of spikes if vertices up there are removed.)
Overall the thing just plain looks better the way it is, and since it’ll
probably be rendered only once in a scene, I’m inclined not to change it. (Plus,
I’ll run into the same issue if I ever work up enough masochism to tackle a
female hairpiece.)
You can take a look at the (very very unofficial) part files in progress here:
http://lummoxjr.byond.com/lego/6898.dat (parts/)
http://lummoxjr.byond.com/lego/6898s.dat (parts/s/)
The files are riddled with comments I made during the build process. The image
below I made in LDView with cook.dat, which is available through the image
link.
Forgive me if I sound like I’m rambling here; I do a lot of thinking out loud
when I post. Can any more experienced part builders offer their thoughts?
Lummox JR
|
|
Message has 2 Replies:
 | | Re: Part building advice - too many triangles/quads?
|
| (...) I think the filesize:quality ratio is good. I notice that almost everything is coded to 4 decimal places. Except when (and maybe not even then) a surface joins to a circular primitive, 3 DPs should be sufficient in parts (0.0001 LDu < 0.00002 (...) (20 years ago, 25-Nov-04, to lugnet.cad.dat.parts, FTX)
|
15 Messages in This Thread:
  
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
