Wednesday, August 12, 2015

Raygun Postmortem

So what have I learned from this project? How did it go, overall?

In overview, the original commission was much simpler. I talked the client up to more than I should have asked them for, and brought it into new methods which I knew going in were untested. What I hadn't anticipated or evaluated properly is the constraints of grinding poverty through critical parts of the build; being unable to afford components or tools and having to spend weeks at a time unable to work on the project (either waiting on parts, waiting on money to get those parts, or working to earn that money).

The first round of conceptualization focused on better-understood processes and materials, but that didn't support the design directions the client wanted to move in. Those directions, when we zeroed in on them, were essentially open to a smaller number of directions; either new fabrication methods with unknown costs, or traditional fabrication that would be (because of the kinds of forms involved), forbiddingly labor-intensive.

To explain this; I was prepped to do wood profiles and lathed metal, but the shape we wanted was streamlined and compound curves. Those can be hand-modeled in clay et al, but take a long time to achieve those smooth curves. Or there are 3d fabrication methods.

The big negative discovery is that both CAD and cleaning up after CAM machining take a lot of time -- significantly more time than hand-modeling would have. Some of this is learning curve. (The bigger impact is I didn't have the materials or the funds at those intervals when I had time, and when I was able to work on the gun, I was far too busy at paying work elsewhere. So I have to consider the idea of 100% CNC machining not entirely tested yet.)

I've also grasped more about what shapes can be CAM, what the tolerances are, where clean-up is needed and how to plan it into the build. I think I'm up to being able to do a Jubal Early now. A PPG is still probably beyond me, though.

The big positive discovery is that 3d printing is an affordable option for a project of this scale. A simpler prop might not justify the price of printing, but given the fifty bucks of electronics alone (and not to mention all the hours involved) the cost of printing isn't out of line with the basic budget of the build.

Another discovery is how to efficiently prep those prints. I had it down to a relatively fast pass of sanding, sealing with superglue, and a minimal number of primer-and-sanding passes with essentially no patching (except where there were errors in the CAD and I had to make changes to the form.)

Less easy to put in words is a better sense of where the tolerances lie and what adjustments need to be made to fit prints to prints (or other measured parts).

On the electronics side, I don't feel I gave a proper test of the surface transducer. I still feel that idea has potential. But I think it likely I could swap it out for a traditional speaker in the current prop. I am still on the quest for props that are properly loud, as well as props that are properly bright. This prop may have suffered from power issues -- the LED seemed a lot brighter on the desk, even if it didn't work well with the tinted acrylic I'd chosen.

My feeling is still that clever strategies are the way to go, because the next increases in sheer power (speaker power or LED power) bring significant costs in weight, power, control circuitry, etc. The essential problem is that human perception is roughly power law. It takes ten times the wattage-into-speakers to sound twice as loud, and a similar ratio is in operation with lights.

Take the latter. At the current low end are 30-60ma LEDs, such as super-brights, or the new neo-pixels (with drivers included in the package). These are the basis of a number of props now, and do decently. The next step up is large-die LEDs on individual heat sinks; these range from 1 watt to 10 watts, but since that is 350 - 700 ma for the typical (vs 60 ma for the neo-pixels) they look about twice as bright as the above. These are what are found in lightsabres these days.

At this size, you can get away with passive heat sinks and simple current limiters (like serial resistance, even). Bump up to 15-30 watts and you need a proper current-controlled power supply and a fan and it is a major engineering project to stuff all that into a portable prop. And audio is similar; up to 2 watts or so can be a chip driver so compact it is available in surface-mount. At 20 watts (sounding twice as loud) you need discrete power transistors with heat sinks and big capacitors; the amp is already larger than will fit in an Altoids tin. And at 100 watts -- this is the range of a powered monitor speaker with a 6" driver that plugs into the wall and costs in the hundreds of dollars. And it won't fit in a raygun.

The positive electronics discovery is I was able to achieve everything I really needed on-chip, with no more than the predicted programming time (two full days -- everything after that was tweaking). One thing that no inexpensive audio package seems capable of delivering is multi-timbral audio that will work properly with an RTOS environment; that is, an audio system that seamlessly plays all the sounds regardless of whether events are overlapping.

I also discovered my prototype DuckNode is not a flexible as I hoped. I cut and re-soldered a half-dozen traces on the PCB to make this project work. Increasingly, although I think there is application for a cheap and minimal RGB controller, the kind of things I want a DuckNode to do it should be configured as a Trinket or Nano shield instead.

On the fabrication side, over the course of this project I matured vacuum-forming in PETG and the making of quick, cheap forms for same, laser-cutting, small-scale CNC parts, and personal 3d printing in PLA. I also tried out lathing acrylic on the metal lathe, stitching pleather, free-hand milling, laser-cutting wood, vacuum-forming ABS, and a few other machines and techniques I've forgotten about already.

This project used a lot of different technologies. It really emphasized two connected ideas; micro-fabrication and TechShop membership. Which are closely related, as TechShop is a way to afford access to the kinds of machines that can be used in a micro-fabrication pipeline.

More to the point, I began to use and became comfortable with a kind of agile just-in-time micro-fabrication; knowing I had access to a variety of machines that make a variety of approaches to a specific part of a model possible, being able to set those up on a drop-in basis, and also being able to do the prep work -- generally CAD -- very quickly or even on-the-fly.

This is the kind of work process I was used to in a well-equipped scene shop. The difference being I can now move metal and plastic and put in toughness and close-up detail.

On the meta side, I learned once again I need to work on my ability to take perspective of props projects. Despite both attempts to chart and estimate and otherwise take long-range overviews, and only partially excused by the known unknowns -- by the fact that I was knowingly pioneering techniques that might or might not work out -- I still didn't track the break points and properly weigh the options, leading to being committed on a path that delivered just barely under deadline and well over budget.

I don't have a good solution for this yet. One thing I am hoping; I've developed most of a suite of new fabrication methods now, over the past couple years and dozen props. This time was also getting up to speed at TechShop (which involves time and cost of all the necessary classes). So the next projects should have a much smaller component of experimentation, either because that particular prop seems to require it or because I'm itching to learn a new technique.

With a basic suite better parameterized, I should be able to plan better how long and how costly a build is going to be, and evaluate better if I need to go in different directions with it. I hope!

No comments:

Post a Comment