View Full Version : Am I on the right track regarding SubD?

01-02-2012, 06:53 AM
got friendly help in the "general discussion" to find some info about SubD modeling.
Now tried something, but before proceding would like to know, if I got the idea about SubD right or if I just had luck that the cylinder looks nice and smooth.

01-02-2012, 08:46 PM
Hi again
Well there are some things to consider (a bit generalistic):

A - The more sides the cylinder has when frozen, the more smoother it will be.
In your example you have an 8-sided (with 5 divisions) and a 16 sided (with 3 divisions). 8x5=40 and 16x3=48, so the 16-sided wins here. If you divided the 8-sided by 6 then both would be 48.

B - The object's cage (before subdivision) must have the detail you want
This means you shouldn't rely on increasing the Subdivision Level of the whole object to get just a specific area of the object smoother or with more polygons. In your case the 8-sided has a higher Render poly-count because there are lost of useless Segments on the cylinder when it is subdivided. The cylinder needs Sides to be smooth, not Segments.
So when you model the cage you have to take in account that subdivision goes 2 ways and in one those ways extra divisions might be useless, so it's better to model higher and subdivide less.

C - Does the cylinder has to have precise size and circularity or it just needs to look good.
If it just needs to look good, precision is not much of a problem, otherwise SubD might be dropped all together or you have to consider that SubD Bulges the circles inward and it also doesn't divide those circles into circles, but to something approximated.
The lower the number of sides in the cage, the worse the Bulge and the Approximation will be. Meaning, the smaller will be and the less circular will appear.
The Bulge can be handled with sizing the circles up. The Circularity, by increasing the number of sides of the cage.

http://i153.photobucket.com/albums/s202/animatics/Lightwave/th_circles.png (http://i153.photobucket.com/albums/s202/animatics/Lightwave/circles.png)

http://i153.photobucket.com/albums/s202/animatics/Lightwave/th_Blending-Circles-for-SubD.png (http://i153.photobucket.com/albums/s202/animatics/Lightwave/Blending-Circles-for-SubD.png) Numbers on the top used to size up the Circles

D - Will you have a lot of geometry loaded over Modeler or Layout, that will affect performance?
If navigation performance is an issue, then having a lower poly-count cage with low Display Level and higher Render Level of subdivision level might be better even if you end up with a higher poly-count at Render time, like you have in your image.

In conclusion
It's a balance between display performance, precision quality and final polycount. In this particular case 16x3 looks just fine. Less bulge, better aproximation to the circle, performance is not an issue, and you will have less useless polygons when compared to 8x5.

As for the caps. You did one thing that is good, I think. You've made extra edge loops for the control edges. That will give you better rounded edges and will prevent from inner topology to affect the edges.

There's a plugin (http://www.lwplugindb.com/Plugin.aspx?id=fed182e8) (Thank you Carl Merrit) that also gives you more instant options.


More about Subpatch at http://lightwiki.net/wiki/Fundamentals_of_Subpatch_Modeling


01-02-2012, 11:58 PM
Just remember that the subdivision level in modeler is simply a display setting. When you bring a subD model into layout you have the same options there with an additional render subD level setting. If your models are not round enough when rendered just turn up that setting in layout. I model with the least amount of polys possible - that should be your mantra as well. It used to be that you would need to model with as few polys as possible because of hardware limitations, that generally is no longer an issue. The reason you want to model with as few polys as possible is for flexibility and ease of manipulating your model. In your tube example I would use an 8 sided disc to start with. But if you are going to add some details you may need to start with a disc with more. The attached images so two different models that where started with the disc tool. The faucet started with 8 because thats all I would need for the final result. The cake stand was made with 128 because I wanted those little details at the top. Granted I could have modeled it with fewer and worked those details in, but it would have taken time and for what I was doing those extra polys did not hurt me one bit. http://forums.newtek.com/attachment.php?attachmentid=100463&d=1324568039 Just think and plan ahead a little and you'll do fine.

01-03-2012, 06:12 AM
Thank you for your input Probiner and Snosrap,

I start get the idea of SubD in general and that there is no wrong or false.
What I really like about SubD, (beside that it reduces the number of polys) is, that I can round the edges and later on decide to change the radius just by moving the control edge loops. I used rounder on fixed geometry before and did not found a way to do this without starting from scratch.

01-03-2012, 09:16 PM
Thank you for your input Probiner and Snosrap,

I start get the idea of SubD in general and that there is no wrong or false.
What I really like about SubD, (beside that it reduces the number of polys) is, that I can round the edges and later on decide to change the radius just by moving the control edge loops. I used rounder on fixed geometry before and did not found a way to do this without starting from scratch.

Also keep in mind that when you're trying to work on a area that is perfectly flat you have a little more leeway as things will render nice even if you have some poles or a bunch of tris, etc. Any curved surface will be much more difficult to keep really smooth when you have non-ideal geometry.

I think Probiner had some cool charts showing issues with curved surfaces and different types of geometry. Not sure where it was though....

01-04-2012, 03:27 AM
If we are talking cylinders and rounded shapes, one thing to keep in mind - reiterating what has been said - is your final detail. It is easier to know that in advance than to switch midstream.

The idea with Subdivision Surface modeling is to control the movement of Points along a 3 Dimensional dynamic curve. It can be thought of in terms of 2D space like so:


But it is more dynamic than that.

There are weak joints:


And strong:


But in 3D dynamic space so everything more or less effects everything.

To have complex shapes week joints are not avoidable. So you use a denser mesh to reduce the movement of those points.

Low Poly:



And Higher poly:




01-04-2012, 11:45 PM
Played around and found out that the 1/6th rule requires to have one "stable" point, that does not move. In that case the "non stable" point moves exactly 1/6th. Thanks for that hint. Helps to predict the SubD effect.

Tried the Carl Merrit PlugIns. They work in LW 11 beta 64 Bit Win.

01-05-2012, 04:33 AM
On a single flat plane, the 1/6th rule still stands. You could likely take that image and put it in the background and size up a 4 point flat polygon and see it when you press tab.

But this is getting far too technical.

The point of that "rule" (which falls apart very quickly once you get into 3D as exact math) is to illustrate the effect placements of edges have on adjacent edges.

Have a look at my tutorial on that page for a better understanding of what I am talking about as well as the practical use for that information. The math is only for illustrative purposes and you'll see what I mean in the tutorial.