A Little Range Wouldn’t Hurt You

It has been a while since I’ve posted anything, and there’s a reason for that. Not a good one, really, but it’s a reason. I’ve been working on subdivision modeling. Partially just to work on some different workflow, but also as a bit of a comparison. I’ve long believed we need a more direct approach to surface modeling. I kept wishing Synchronous Technology would come to surfacing. And maybe it has.

A recurring theme in my life has been rejecting something hard, then coming back to see that it really wasn’t that bad after all. I think this is also true of engineering in general when it comes to certain technologies that have been more connected to art than engineering. The engineering world has rejected hard some of those geometry techniques that may have seemed more art than science. That crack between art and engineering has been where I’ve lived since my first career (as a musician). And it’s a crack, or maybe a rut, that I just can’t get out of. Just when you think you’ve committed to one side, something on the other side starts to look attractive.

As a musician, I knew that math is inherent in art, and especially music. I’ve met a lot of engineers who are amateur musicians. The two disciplines really go together. I know there are people who think engineers can’t think off a straight line, but I’ve seen too much of this overlap to fall for that one.

In the same way, I think our products, and yes, even our tools and careers need to be more multi-faceted. Focusing on a single discipline has advantages, and I know that getting too spread out can be a real distraction, but there’s a reason we as humans have two eyes and two ears. Additional point of view brings better perspective, and more true understanding.

Why am I waxing philosophical? When my career starts to branch out and take on a wider scheme, without becoming too diluted, it tends to be more satisfying, and I think I create more well-rounded results. Laser focus is satisfying for a time, but just feels limiting and myopic in the long run.

Without straying too far into the political, and without the intent to create a LinkedIn-worthy click-bait meme, I think people in general now tend to identify themselves with a single idea to the point that they don’t even have the desire to see multiple points of view. This is a real problem, and it extends far beyond which channel you get your news from.

That’s why I’m glad that I’ve started experimenting with subdivision modeling.

When Solid Edge came out with Synchronous Technology about 12 years back, I was dismissive. It took two years for the tech to develop to a point where I thought it had some real potential as something other than a Sketchup-type concepter. History-based modeling has some real flaws that a lot of people just excuse without giving it any serious thought.

Eventually I saw that modeling and design really don’t have as much to do with programming as you want to think. Design and modeling are more like clay than like Legos. Legos predetermine the final structure to some extent, but clay leaves the options open. You need flexibility, especially up front in the development process. For all the talk you hear about Design Intent from history adherents, it really only works within a small window.

So I did a test. With more than a decade of experience using advanced history-based surfacing, and about 2 weeks of self-taught experience with subdivision modeling, I took on similar tasks using the two processes. I made the outside shape of a medical hand-held device model in both methods. For a similar part, the subd part took me about 25% of the history-based part.

That must have been a fluke, right? I’m an “expert” surfacer, and a novice subd pusher. So I did another pair of similar parts. Different CAD packages, different workflows altogether, but I got similar results: 1:4 time spent, subd vs history/feature-based modeling.

And then a third part with similar features. The truth is that 3D shapes take longer to develop when you rely so much on 2D sketches, and a process of sketch-feature-shetch-feature…

And the kicker is that you get an even bigger difference when you start editing. History-based modeling blows up when you make changes bigger than say 20%. I can’t tell you how much time I’ve spent rebuilding surfaced models just from making a simple change, much less a big one. Design intent is fragile, even with the best intentions. But subd modeling is not susceptible to rebuild errors or rebuild time delays. So up front modeling is a big time saver, but edits are where the real money lies.

Sure, there are some arguments you can make about accuracy, or about maintaining analytical shapes, or a bunch of nit-picky stuff that doesn’t come close to negating the 1:4 time differential in initial modeling time, not to mention the absolutely huge editing delta. Hybrid methods can bridge any real gap, and just like I learned with Synchronous Technology, you use tools where they are an advantage to you.

Subdivision modeling belongs in your engineering and product development workflow.