PDA

View Full Version : Why aren't spheres spherical?



stoehr
12-22-2003, 03:20 PM
I ran into a job recently where I had to make a logo with a spinning sphere. Now, I've known for awhile now that a low poly sphere (12 sides and 8 segments) is oblong. Yet, it wasn't until this weekend that the problem really got in the way, and my only way around the problem was to make a very high poly count for the sphere (72 sides and 48 segments). But this seems to defeat the purpose of subpatches. Right? Subpatches are used to create a highpoly "look" with low poly counts. If I increase the poly count, I really don't need subpathes, because I could just turn on smoothing in the surface editor.

All in all, what I'm trying to say is; I'd like to make a non-oblong, perfect sphere from a low amount of polys, with subpatches turned on, but I can't.

HowardM
12-22-2003, 03:26 PM
yep, thats it, cant.
hehe happy holidays buddy!

no really, thats an issue with Subds....kinda lame eh?
id love someone to come on here and tell us how....

art
12-22-2003, 03:27 PM
I'm not too familiar with obolong spheres issue (I just tried it, and with low polycount, the sphere looks obolong indeed), but maybe this will help: try making a box with some segments on x,y and z and then spherizing it (modify->spherize)
It will give you sphere made out of quads only. Try subpatching it. The resulting sphere might not be perfect either, but maybe it will require less polys to get the results you want.

art
12-22-2003, 03:30 PM
I also pondered whether spacing of all the points of spherized cube equally would help the issue. Maybe I'll write a lscript to do that and see the results.

stoehr
12-22-2003, 03:54 PM
Originally posted by art
I also pondered whether spacing of all the points of spherized cube equally would help the issue. Maybe I'll write a lscript to do that and see the results.

That'd be cool. Let me know what you come up with, if you can.
Thanks.

stoehr
12-22-2003, 03:54 PM
Originally posted by HowardM
yep, thats it, cant.
hehe happy holidays buddy!

no really, thats an issue with Subds....kinda lame eh?
id love someone to come on here and tell us how....

Howard! Haven't seen you in a while, thanks for stopping by!

Aegis
12-22-2003, 03:56 PM
Why don't you use a low tesselation sphere?

stoehr
12-22-2003, 04:25 PM
Originally posted by Aegis
Why don't you use a low tesselation sphere?

Tesseslation has similar problems. In order for the sphere to be spherical, the poly count must be increased, and the poles of the trianlges are very visible, and would leave shadow artifacts.

Also, in my particular scenario, I wouldn't have clean U and V edges, which I needed for the project.

stoehr
12-22-2003, 04:26 PM
Kretin gave an interesting solution here:
http://www.cgtalk.com/showthread.php?s=&threadid=111326

art
12-23-2003, 07:39 AM
I have found this interesting link. Although most of the article went over my head, the end result (HELAPix sphere) looks interesting. Its similiar to the spherized cube solution given at CGtalk and myself above, but looks more equally spaced. I'll play around and see what I can come up with.

http://www.eso.org/science/healpix/content/HEALPix_Documentation_html/intronode3.htm

3D|Dave
12-23-2003, 11:11 AM
In mathematical terms, polygons can only approximate a curved surface. No matter how many polygons are used, it is still an approximation. To accuratly describe a curved surface NURBS must be used. NURBS are mathematically accurate.

It would be nice if there was some support for NURBS, this would be one instance where it would come in use.

stoehr
12-23-2003, 11:34 AM
Originally posted by 3D|Dave
In mathematical terms, polygons can only approximate a curved surface. No matter how many polygons are used, it is still an approximation. To accuratly describe a curved surface NURBS must be used. NURBS are mathematically accurate.

It would be nice if there was some support for NURBS, this would be one instance where it would come in use.

That's how I feel, too.

AngelDream
12-23-2003, 04:06 PM
Two things.
1) Subdivision surfaces are better than NURBS in the sense that they have a better continuous topology and are simpler. They don't suffer if you slice them and some other nice stuff.
2) what is all discussion about??? can someone PLEASE explain it to me???
3)here you have an interest place about subdivision surfaces
www.subdivision.org (http://www.subdivision.org)
check the interactive demos.

art
12-26-2003, 06:48 AM
Angel, I wasnt really aware of the problem myself until I've read this thread. Look at the cgtalk link given by Stoehr above. It will give you idea what he asks for.

Stoehr, my results from the "weird" stuff called helapix werent too great so far :) I was getting some pinching here and there, but the polycount was fairly low :)

stoehr
12-26-2003, 12:24 PM
Originally posted by art
Angel, I wasnt really aware of the problem myself until I've read this thread. Look at the cgtalk link given by Stoehr above. It will give you idea what he asks for.

Stoehr, my results from the "weird" stuff called helapix werent too great so far :) I was getting some pinching here and there, but the polycount was fairly low :)

Thanks Art, keep me posted.

hunter
12-27-2003, 10:50 AM
Even subpatched objects use a high polycount in layout. Just look at your low poly supatched object in wire frame mode as you change the patch level. Unless it's a huge problem to have high poly counts in modeller for your sphere why noy not use it. It's gonig to render with just as many. In my experience subpatches don't render any faster than a frozen counterpart.

riki
12-28-2003, 06:55 AM
Freezing a Sub Patch should greatly reduce render time, even if the polycount is higher.