nurbs and subdivs, a history

General Forums Blender and CG Discussions
41

Well, personally, I would consider that a limitation rather than an advantage.

They are definitely not the same. They’re just closely related mathematically.
For instance, as you pointed out, subdivs allow arbitrary topology whereas Nurbs do not.
Also, at the moment I don’t think there’s really an agreed upon way to do weighting with subdivs (I could be wrong about that, so take it with a grain of salt), although that will probably change in the future. Nurbs, on the other hand, do currently have agreed upon weighting (it’s built into the maths).

In my opinion, subdivs have the potential to kick Nurbs ass in almost every respect, it’s just that the technology hasn’t been developed yet. I’m pretty sure subdivs are capable of mathematically sound weighting, and looking at the research that Matte linked to, it definitely looks like they’re capable of trimming and such. These things just need to be developed and implemented.

42

well it could be a disadvantage but at the same time it could be a big advantage.
connecting high res ears to a head is an example where nurbs do better.

next to the hole projection features NURBS has comes the boolean and trimming functions. I am not sure if that is possible with polygons and subdivs.

as far as I know what we see as a 3d mesh is nothing else than proxy of the original NURBS surfaces. inside it is this construction of curves based on the math functions.

subdivs are an interpolation of the polygon cage. i am not so sure if you could project a curve onto the subdivided surface and cut it out …

unfortuntately my math is not that good for it.

43

And a nurbs surface is an interpolation of it’s cage. There is no fundamental difference. Seriously, by definition Bezier patches, B-spline patches, and nurbs patches are all interpolation schemes.

Matte demonstrated in his last reply that you can. Albiet his way is a bit round and about, but it works.

Well, the hole projection is trimming. But I digress.
Matte’s method for the hole projection was to convert the subdivision surface to nurbs, and then use its features. With a bit of simple deduction, it should be obvious that that also means that booleans can be done as well.

(And it is definitely possible with polygons.)

44

mh, that would not be too bad.

I like working with both.
I like the flexibility of polygon mesh modeling and the precision of nurbs.

what bothers me with subdiv is when you create a plane and turn on the subdiv
modifier it will turn into a disk. this always forces to add additional geometry
for edges.

well i am not sure if you worked with Rhino or Cobalt. their functions are just
amazing while i prefer cobalt for its modeling tool set. rhino is a toy compared to that.

but getting some low end tools into blender would help me a lot when i
model a product as i do recently more and more

45

@ brecht: Did you see the PM I sent you?

46

eeshlo, didn’t see it, but replied now.

47

Tesselation in NURBS means nothing. It doesn’t change the geometry or the smoothness of it. You tesselate it for the purpose of editing it. If you don’t do something after the so-called tesselation (like editing it) it is still the same exact, mathematical shape you started with. As opposed to poly/subd, tesselating the object changes the geometry significantly.

48

The power of poly/subd lies in the fact that you are able to edit the object in its bare bones DIRECTLY. Which means verts/edges/faces could be moved around with or without affecting the neighboring verts/edges/faces. Then be able to apply Catmull-Clark subdivision levels. Therein lies the power. You mentioned that subd levels in the “new technology” you are describing have no relevance and it behaves much like NURBS. If that is the case, then why use it at all? Why don’t you just use NURBS? It doesn’t make sense.

1 Like
49

Another one of those blanket statements of yours. “Finale 3d” may mean different things to different people. Whatever it is it is still visualization in 3d.

sketchup is great when you have some nice boxy house models.
You mean architectural models are not generally boxy? What did I miss here? Even if the design is an arc-shape structure, it will still be viewed kilometers away. I wasn’t necessarily recommending SketchUp for any kind of architectural visualization. C’mon, man, use your common sense sometimes.

50

if there were other blanket statements which needed some clarifications why did you not tell me it earlier? writing in a foreign language isn’t that easy. what makes sense to me doesn’t mean it makes sense to others. is that clear to you?

with finale 3d model i meant what you do after you have done your sketching.
i teach my Interior Design students to start in sketchup and after they are done there they bring it into autocad. of course can sketchup cutout - do a basic Boolean
. but with curved surfaces it has its strong problems. that is where autcad, archicad or others come in much stronger.

we often bring in architects to showcase their work here. and i am not talking about cookie cuter houses. they have a mixed workflow and they also use nurbs because not everybody wants to learn how to program surfaces with the build in math visualizations like archicad has. it is the tool set which makes the good result.

i would not say that people in the industry uses parametric only because NURBS are a pain. solids are a pain as well because you cannot control special tools with solids as good as with surfaces you add later a thickness to or turn them into a solid.

mate,

when did that feature come in? when i used subdivs in Maya they never allowed my anything else than 4 side polygons! can you show a bigger screenshot or post the maya file? you make me very interested in that.

51

Maybe I’m just terrible at explaining things, but I am a bit baffled that you’re continuing to discuss that point as if it didn’t apply to subdivision surfaces.
The whole point of what I’ve been saying is that the same thing is true of subdivision surfaces: it’s subdivided to a finite level for the sake of editing, just like tessellating nurbs (or any other mathematical curve or surface, for that matter). The mathematically defined smooth surface that it represents is still there.

It’s not relevant in the “old technology” either. Or rather, it’s relevant only in the same ways that tesselation is relevant to nurbs.
If you read the original catmull-clark research paper (which has already been linked to for you earlier) you will see that. They didn’t start out by saying, “Hmm… lets try this subdivision scheme and see what surface is creates.” They started out with a very specific, perfectly smooth surface that they wanted to create through infinite-series subdivision, and then derived the subdivision scheme based on what would achieve that.

I would use subdivision surfaces instead of nurbs because nurbs are a pain in the ass to model with. Especially for organic models.
You have to keep everything in quad patches, you have to explicitly stitch patches together to make them continuous… it’s just generally a pain to model in a free-flowing, flexible workflow. It’s very rigid. And the topological restrictions aren’t always a good thing (especially during modeling).

BlendDoodler, I seriously get the feeling that you’re arguing here not so much for the sake of gaining a better understanding, but for the sake of winning the argument. I’ve read the research papers on the topic, and I suspect that the original poster (matte) has as well–or at least he looked them over. Unless you’ve done the same, I don’t think you’re in a good position to be arguing with us.
Please go back and re-read our posts with a mind for gaining a better understanding of the technology rather than with a mind to try to refute us to win the debate.

52

as far as i know they’ve always been like that. i did that test in 7.01, here’s the scene file:

http://www.tokeru.com/matt/misc/subdivTrim.mb

53

so cessen

you and matte seem to understand this a better than I do.

There are few things I wonder about.

So you say that the math behind NURBS and Catmull-clark is very similar?
But where are the differences?

I can see the point that NURBS and Subsurf both use a CV to span the geometry inside.

However I know NURBS tools where in the interpolation through the CVs is different.
They can be inside, go through the point, or even be outside. Similar Maya has a pen tool which does not draw the cage CV but draws points the curve goes through. Maya than places correctly the CV points.

From my understanding, NURBS are 100% math based. Thus what we see as the visual is the 3D interpolation. The math enables you you accurately calculate a point or isoprame along u or v of the NURBS patch.

With polygon when you turn on Subsurf you see an interactive subdivided polygon object. You need to turn on smooth to see an interpolation. Even when you set high Subsurf you still would have linear edges, even when so small you cannot see it.

I just have a hard time to understand how those two technologies could be combined into a new one.

It seems logical to me that with the approach NURBS has you can easily calculate points, project curves, and perform those actions like trimming without any further fix work.

When I imagine to do that with a Subsurf object I can only see the moment I remove a face a change in the geometry because of the way the surface is interpolated inside the polygon cage. this does not happen with NURBS at all.

When you create NURBS rectangle and fill it you will have linear edges.

With Subsurf you need to crease the edges to get a similar effect.

This as an example.

I also noticed that this also influences than the way how hu how should I say it, the UVs are stretched? compare the two top patches and compare the edges and the uv ISOs.

You see what I mean?
http://www.jrkuhnen.de/fip2/blender/NURBS.png
http://www.jrkuhnen.de/fip2/blender/NURBS.blend

From what I know NURBS spans a surface between a 4 sided edge FROM edge to edge.

Interestingly Blenders NURBS circle is visually the same way distorted like the Subsurf circle. This wonders me why because I never found that to be true with Rhino.

Huf I hope I make sense :wink:

I can understand why you would prefer NURBS for organic objects. And in most cases I even agree. A Subsurf tool which allows trimming would be a dream. I really love trimming and also stitching patches together with different subdivision levels.

What what Matte showed with Maya is quite great. I need to try it out tomorrow when I am back in my office.

54

Matte

pfff in 7 already? Did I miss something? When I got always the non-fourside polygon error message! Thanks for the file.

55

to bad I do NOT have access to latest Lightwave version because as what it seems we are looking here for LW has already.

http://cg-india.com/Tutorials/SE_T610_1.html

i read through the tutorial and functions like drilling etc works together with Subsurf WITHOUT changing the geometry.

Check it out. I think LW has a demo version as well. I am not sure.

But the modeling functions/possebilities made me quite jealous. There is some Blender can improve :wink:

56

It would be nice to have that type of feature in Blender with Blenders already sweet modeling workflow and all. :smiley:

57

oh ya! those features are rocking amazing and would make mesh modeling a dream!
but i am not sure of the use planar trims or if the trims would also work on curved objects!!!

but it looked like it works in LW!

also lightwaves other subsurf functions and displacement solution are very powerful.

58

blendDoodler, may I try and offer an explanation to what they’re saying, to attempt to put an end to the confusion.

The idea behind what matte and cessen are saying, is that although most application’s idea of subdivision surfaces is just subdividing to a number of levels, conceptually (and in practice in some cases), subdivision surfaces are just as mathematically pure as NURBS are. It’s possible (for example in Maya or renderman or whatever it is) to find an exact point on the subdivision surface, it’s just that that’s not the way many applications are doing it right now. According to this, subdivision surfaces are just another way of representing a smooth 3d surface, and in theory it’s possible for subdivision surfaces to have similar kinds of tools to what are available for NURBS right now. It’s just that in practice, not many applications are doing it (yet?) and I guess the maths are probably more difficult.

In any case, the precise definition of a subdivision surface is the limit surface. In practice at least in Blender, you get an approximation.

(you may also notice that in Blender, NURBS aren’t rendered directly either, in both OpenGL and the software renderer. They’re tesselated to the level specified in the ResolU and ResolV)

59

mh

broken do you know that Rhino desperately tries to include good polygon tools into it.
quite funny to see that.

60

i want those drill functions! i am modeling holes, skrews, etc and Blender is a pain to do it with polygons. argh …