What Is SubD Anyway?

I’ve been writing a little series of articles on various types of non-NURBS CAD data over on Engineering.com and EngineersRule.com. As CAD users we have been conditioned to recoil in horror whenever confronted with incoming data like STL, OBJ, XYZ, or a number of other types. I remember back in the late 90s when the first push for VRML was such a huge disappointment. The fact that it was so slow and you really couldn’t do much with it probably set back augmented reality by years. They took another shot at it pre-2010 with better success, but it was still not really ready for prime time.

As far as engineers would get involved would be a mesh in FEA or the export of an STL file – a set of points connected by lines to make polygonal shapes, tetrahedrals or quadrahedrals.

The more recent explosion of 3D scan and 3D print really pushes this data type right up in your face. You can’t ignore it any more. It’s like 3D CAD challenging 2D CAD in the mid 1990s.

SubD, or sub division modeling, is a set of surfaces based on a cage of points. You can tug and pull on the cage to change the shape of the surface. It’s kind of like a 3D equivalent of a spline, where you move control points.

SubD has traditionally been the domain of applications like 3dsMax, Maya, Blender, Cinema4D, Zbrush, Mudbox, modo, etc. Think Pixar. All of those animated characters have a similar roundy bulbous look because they are made with the same tools and techniques. The characters wind up in games and movies, and in the imaginations of weekend CGI artists.

CAD users have scoffed at that type of data because it is cheap, fast, organic-looking, and worst of all imprecise. Exactly the opposite of most of what we do. I think mostly we scoffed because we couldn’t use that type of data in our work, and because we didn’t have the tools to work with it. Geomagic is the one tool that serious engineers can use to manipulate point-based data, and that’s expensive – like used car kind of expensive. You almost need to specialize to be able to afford stuff like that.

Well, it’s time to stop scoffing. A lot of types of data that have been misfits and off limits to us as engineering CAD users are quickly becoming required reading. Subdivision modeling tools have been in Siemens NX for some time, and are now in Autodesk Fusion 360, and are coming in Rhino 7. I even heard a rumor that another mid-range player is getting involved (and it’s not SW, although DS does have an offering in this space, it isn’t part of the current SW software).

Why is subd modeling important? Well, it’s important to guys like me, who design a lot of swoopy or organic stuff. Say you model a Corvette in your favorite NURBS-based history CAD. Say that takes a week with all the parts, and maybe 80% detail, plus sketches and lofts and boundaries and fills, and all these features.

Now say you were to do that same car in 3dsMax. It would take much less time, especially the hard stuff like the body. The body might take a couple hours if you know what you’re doing instead of a couple of days.

You can go on to harder stuff, like a baby doll. A baby doll was the only project I ever started in SW that I actually had to give up on. In NURBS feature-history CAD, doing a really good a baby doll or any human-ish figure is not very realistic. But using a subd modeler, it’s done all the time.

Product designers have to do this kind of thing very frequently. This blog, and to be totally honest, much of the last 15 years of my life, has been dedicated to training people, writing about, and griping about how hard it is to do advanced surface modeling in a mechanical design software that was never really meant to do it.

That’s where subd comes in. It’s easy. There’s no history. There’s no feature tree. There’s just a bunch of shapes you push around on the screen until it’s right.

The downside? Well, it’s maybe not 100% dimensionally accurate. It’s squint-one-eye sort of accuracy. Or sketch-over-a-sketch level of accuracy. Organic shapes are usually not mainly about accuracy anyway, they’re about shape. Subd stuff is naturally blended. Can you imagine what it might be like not to worry about how to blend separate features anymore?

What if there were a system that used different tools for what they were good for? Subd for shapes, NURBS for engineered features? This is the future of CAD. And not just the future, this is happening now. It’s a real innovation, not like moving the platform back to a mainframe-like cloud. To me, this is the obvious extension of synchronous technology (Siemens) that is already in NX. It is also part of their Convergent technology which brings these different data types together with traditional NURBS.

But Matt, doesn’t embracing subd throw away everything you’ve been doing on the Dezignstuff Episodes? Well, that’s one way of looking at it. Another might be that you might see parallel episodes using different techniques.

There are thousands of Youtube videos showing how to build things in subd modeling. There are many different tools you can use that cost from tens of thousands of dollars to free. It’s a completely different world from the NURBS CAD most of us live in, although there are some parallels. The concept of spline sketching and the control of curvature in splines has some parallels in subd. It’s just a much more intuitive technique. Which is good and bad. If I hear one more Youtuber say “vertisee” (singular of the plural of the singular “vertex”), I’m gonna start putting Geritol in someone’s avocado toast. Lowering the bar for entry into the world of design for complex shaped products is bound to create some rough patches, but this bar needs to be lowered.

Most of the subd tools you see today are driven by the Pixar OpenSubdiv library API which has its roots in a game created in 1996. But everyone implements the tools differently and creates new ways of manipulating the geometry. It’s a completely different way of working, compared to the standard sketch-and-feature method of history-based CAD. Instead of thinking of a product as a series of processes, we will be more concerned about the actual shape because we can control that directly.

If you have read any of my previous rants on the shortcomings of history-based software, you might have seen this coming. There are some parallels between subd and direct edit. I believe some people will maybe resist this, or maybe not understand the editing tools. I was surprised when people stayed away from synchronous technology in droves, and the same may happen here. But I for one am going to jump in. I’m ready for this change.