View Full Version : When to use sub-patching?

07-14-2004, 01:25 PM
When is it a good idea to use sub-patched surfaces? Is there a good rule of thumb? For instance if I were to design an airplane or a computer monitor should sub-patching be involved with these objects?

jin choung
07-16-2004, 02:24 AM
good question!

the advantage that subpatches have over a simply high density poly model goes something like this:

1. you get a smooth surface with few actual verts (control points).

2. your mesh is NOT a fixed resolution... it can change depending on distance, etc.


because of 1, it is an advantage to use sds (subdivision surfaces aka subpatches... it's faster to type sds) on characters because when you are 'skinning' your mesh to a skeleton, it is far more preferable to think about the relationship of a FEW verts to the influence of different bones than MILLIONS of verts to the influence of bones.

because of 2, if you need an asset that needs to be able to withstand scrutiny at indeterminate distances, it would be wiser to keep your model sds.

also because of 1, if you find that you need to alter your model later, it is faster to change an sds model than a highly subdivided one.


if you need roundy shapes, you'll probably use it to start off with... but whether you stay with sds is up to you.


there are compelling reasons to 'freeze' to polys or just use polys to begin with... if the model is largely composed of flat surfaces, then it's just overkill to use sds... why explode the overhead if you don't need to?

also, unlike NURBS, there is no way to carve up a smooth surface with cuts or booleans. sure, you could model in such details by modeling them into the sds control cage but sometimes, that is just not possible.

at these times, it is best to freeze and then slice and dice and boolean and such on the polys. some would disagree with me. but i don't.


and even if you have a polygonal but roundy model, there are ways of using METAFORM later get a finer resolution should you need to unexpectedly bring your camera in closer. so just vanilla polys should never be dismissed out of hand.

in many cases, vanilla polys are used far more than one would expect.


remember the rules of CG (and for that matter, all art and science):

1. make it good and right.

2. make it good and right and as ECONOMICALLY AS POSSIBLE for given purpose.

do not expand entities without reason.

good luck.


07-16-2004, 06:32 AM
Nice post Jin

07-16-2004, 09:03 AM
This helps a lot. I'm trying to teach myself on my own and there always seems to be a lot of different approaches to modeling objects.

Can you elaborate on freezing polys and Vanilla polys? I've done some reading on freezing polys but haven't performed any tutorials yet. Is it a way of combining sub-patches with regular geometry?

Which renders faster? It seems to me subpatached objects take longer to render. But I supposed that depends on how much a sub-patched object is split up?

And when is it good to use something like metaform?

jin choung
07-16-2004, 11:53 PM
howdy again,

actually, when i say "vanilla polys", that's just my way of saying 'just plain ol' polys'.... the lowest common denominator of 3d graphics is the polygon (well, you have the vertex too but... nevermind).

anyway, when you hear people talking about 'freezing' in lw, it refers to simply converting an sds surface to polys.

it doesn't merge or combine like you were saying. it simply converts.

after you freeze, you sds model is plain ol' polys. so, make a cube - hit tab - you just converted a polygonal cube into an sds roundy dealy. hit CTRL-D and you turn that sds roundy dealy into a roundy dealy that simply consists of lots of plain ol polys.

anyhoo, you're on the right track and asking the right questions:

- there is a time and place for everything.

- there is a time for polys, a time for sds, a time for NURBS. but for lw, you only have the option of the first two.

good luck and keep on keepin' on.


07-17-2004, 07:13 AM
I think the question is when NOT to use sds....

My general rule is I try to use subpatches whenever possible. Because

A) you can always freeze it if you need to but it's not easy to reverse a freezing operation. You can subpatch a frozen object but chances are, it will be so high resolution at that point, it will be almost unusable and

B) You can change the resolution at will for rendering and of course

C) It's easier to work with a sds surface that has less points then a high detail frozen model.

You can use booleans with a subpatch object, I do it all the time, but there is usually a lot of clean up and some rebuilding of the polys in that area and you have to really understand the shape you are trying to achieve and really think about the object you are "cutting" with.

There is a general rule that "vanilla" polys as Jin calls them, will render faster then SDS because there is nothing to convert at rendertime.

But my biggest rule about using sds is that I always model with SDS turned OFF. If you can get your model looking pretty good in polygon form, it will look superb with sds on. Also, it's an easy way of avoiding non-planar polys. I just turn it on occasionally to check the progress and also see where the model may need tightened up since sds relaxes the surfaces and edges.

07-17-2004, 07:52 AM
I do a lot of industrial work and find that I use it even for boxy non-organic models. It means you don't end up with that I-am-a-com-pu-tah-gen-er-at-ed-im-age infinitely sharp look.

Can mean more work though. A subpatch cube is harder to do than just a plain ol six polygon cube.

07-17-2004, 10:06 PM
Jin`s showing amazing self-restraint: he didn`t mention edges once! :D


jin choung
07-18-2004, 12:57 AM

i would disagree that subpatches should be used for flat surfaced models...

sure, edges shouldn't be razor sharp but it is more economical to bevel in edges to catch hilites rather than go with the added trouble and expense (processor time) of sds.

in my mind, there is no reason to NOT make most television or LCD monitor chassis simple polys. or most furniture. or keys. a circuit board. most VCRs. ww2 era mechanized units. star wars (first trilogy).

to detail them up to the point where they can withstand extreme scrutiny without resorting to multires sds is just too easy to justify going sds.

even if the camera is brought to within an inch of a surface that's SUPPOSED to look flat... well, that's just plain good enough. and a trap of sds 'flat surface' objects is that while they may avoid the pitfall of razor sharp edges, they can start resembling the roundy playschool toy versions of whatever objects they are trying to depict. for the unskilled/unexperienced, that is indeed a huge trap.

also, cg for films/tv is NOT about true verissimilitude.... it's about mere FACADE. all anything need do is withstand just as much scrutiny as screen time demands and not an inch/second more.

some people take great pride in making their cg corvette as true to life as possible and that's great and a valid endeavor that provides their host with hours of enjoyment... kinda akin to plastic model building i guess. but unless you have a passionate interest in doing such things, you'll unlikely require such 'more than skin deep' detail in most production settings.

hollywood FAKES everything... this is not a recent trend. it has always been true. and you should do the same digitally if for hollywood type work.

again, a big part of my personal pet peeves involve economy. i BELIEVE in it.

but that's just my opinion.


07-18-2004, 04:24 AM
Good point Choung, I know I've spent ages modelling stuff that flashes past the corner of the frame for a fraction of a second. It's satisfying, but doesn't keep the clients happy.

I like modeling with sds though, it's more like working with clay or something.