Retro Raygun: Realizing a Prop Through CAD

It is entirely possible, and often appropriate, to realize a prop through hand-carving of various materials, either for the finished product or as a master to be used for molding and casting.

It is also possible to do significant parts of the shaping with some of the new computer-aided methods which are filtering down to where a home workshop can economically access them.

This prop, in fact, was originally intended to make extensive use of CNC milling. However (and fortunately for my deadline) the CAD files I created for that process could easily be turned to a faster though less sexy process; 3d printing.

This still means the majority of the work was completed on the computer, with primarily sanding, painting, and some assembly done by "hand." This made it possible to achieve the various curved shapes with a little less of the familiar round of sand, prime, spot putty, sand, lather rinse repeat.

The spec for this from the client was Atom Punk, Raygun Gothic, 50's-60's SF of the whimsical lean; Barbarella, The Jetsons. Except it wanted to have some grounding in reality; to be as fantastic as the rayguns of Forbidden Planet, but to have (to in fact share with those excellent props) the kinds of details and treatments of a real object.

It is not correct that form follows function. But the two are related, as are form and fabrication. Indeed, fabrication can follow design intent; real metal has a verité that resin casts may not. But, alas, we could not find any way to mix metal parts which would be suitable for lathing with wood furniture that could be CNC routed (or cut with scroll saw if need be).

Instead, we narrowed over countless sketches into electric-drill like shapes and finally into the classic bulbous pistol shape, the "dirigible body" on a grip with rings around the business end. And a fin; the client loved the fin.

This also decided a novel and risky approach; to abandon the usual tools of clay sculpt and slush cast and to design the entire thing in CAD, to be realized by either subtractive (CNC milling) or additive (3d printing) techniques. It would be designed in the same clamshell sections as a real manufactured version might have, up to and including using bolts and screws to connect everything together.

Design-wise: the final shape was almost directly inspired by a very cartoony poster. But I'd already zeroed in on dirigible shape, boomerang curves for the grip, Cadillac/jet fighter fin, and of course insulator rings. All that really needed sketching and approval after that was the side swooshes, whether to put jewels or other markings on the power settings dial, and the grip lines of the hilt -- which was actually a happy accident, as I had originally drawn them true to the horizontal but accidentally modeled them orthogonal to the front edge of the grip instead.

It wanted to be basically metal...either brushed or chrome. And have some plastics and some pinks in the design as well, to fit both the period aesthetic and the retro atomic Tiki culture aesthetic as well.

That was actually the biggest design challenge; being both wild and grounded, having both a prop-like quality and a real item of technology quality.

The visible seams and fasteners, I did a design refresher by going down to Urban Ore (a local recycling place) and looking at old power tools.

The time to iterate heavily is when the costs (in materials and time) are low. So I iterated on the approved design to nail down proportions before moving on to the full CAD model.

The very first thing I did was blow up the sketch. I traced various pistol grips and used soup cans and other household objects to get a sense of the dimensions, and created a full-sized sketch. I then transferred that sketch to foam-core and cut it out.

And it didn't look or feel right.

So second iteration; derived a new foam-core off the first, adjusting those dimensions which were necessary. This felt better in the hand. To make a full check, I glued white foam on to the foam core to give it dimension and sanded it into a semblance of the desired shape.

Primed and painted, it provided not just a visual sanity check on the massing, scale of details, grip comfort and so on, it also provided something to show the client and try out for scale and comfort in her hand.

And it works quite well. The grip is odd; at first glance you would think you were supposed to choke up on it with the gun body resting against the meat of your hand. But it feels utterly natural to hold it lower on the curve, and when you do so, not only is the trigger in a comfortable position, but the "sight" (or whatever the design element on top of the fin represents) lines up with your vision in a way that assures you that the gun is aligned with the direction of your hand.

With a full scale model, I could take dimensions directly from it. But there is an even better technique; as a combination of reference and sanity check, take the flattest image you can (I stick my camera as far away as possible and zoom in) square that (I use Gimp these days) and import that as a reference image.

Fusion 360 has a very handy function to where a reference image can be normalized to the scale of the workspace just by clicking on a couple of reference points.

Of course there are other CAD applications out there. You can even do this in a polygon modeler. CAD has certain advantages in approach, however. The methods in CAD are generally well-suited towards making holes for bolts, centering objects on each other, mating faces...and more technical tasks such as checking free play. And Fusion 360 is a very friendly program towards the amateur and the small business.

One rule I have; model everything. It doesn't have to be in a lot of detail if you aren't going to manufacture it yourself, but including all the parts means you can really check to see that everything will work together.

Fusion 360 has the ability to import from various libraries and sources. I sourced most of my electronics from Digikey or Adafruit, meaning I could get 3d models for some of the parts by searching around. By the time I got to hardware, however, I was in the middle of a delayed OS upgrade that meant I could temporarily not use Fusion 360's cloud services. Which means I couldn't import standard screws.

So I purchased the exact hardware I intended to use, measured it carefully and modeled it myself.

Electronics are always a tight fit. I have in my mind a project where I'll pick the components first, and design the shell around them. For this one, alas, I had an approved and iterated shell, and I had to adjust the parts somehow to all fit inside.

Another helpful technique for these moments is paper dolls; cut out exact scale shapes from paper and move them around on a fresh sheet of graph paper. It is faster than grabbing parts one by one in most 3d or CAD interfaces, and allows you to be more creative with what you try. And if you get a plausible fit, then you check it in the CAD.

One thing I had to check in more detail; I needed to see how the trigger felt as well as how it looked when it moved, so I built another mock-up in foam core and attached a trigger with the same spring and switch I intended to use. This, also, required a few adjustments and iterations to get right!

There are two related problems to solve in the CAD; one is how the parts can be assembled. The other is how they can be manufactured.

Take the latter. The intent was to fabricate via CNC milling, This means that dead pockets are not possible. To make the potentiometer that sits behind the settings dial work, then, I have to cut two vertical pillars (vertical from the point of view of the milling machine), then add a separately machined part on top; a strap which screws down to those pillars and, in turn, has the potentiometer bolted to it.

Even design for print has many issues of printability. A print service like Shapeways (where I've gone often, and where the majority of the parts for this ray gun were eventually manufactured) has on their materials pages various rules for minimum wall thickness, clearance holes, etc.

Personal 3d printers generally do not print supports in a separate material, so they have to be even more carefully considered in what kinds of overhangs can be safely printed.

In the case of CNC machining, the main constraint was not making pockets so deep the end mill couldn't reach. After that, it was avoiding as much as possible making sharp inside corners necessary, as those would require either more convoluted milling paths or extensive clean-up.

Related generally to this is materials strength and design tolerance; how thick do parts need to be for strength, how close will the fabrication process come to the desired dimension, how much play has to be allowed for in order that errors in machining or even small irregularities in surface prevent the parts from going together properly.

The other aspect touched on above is design for assembly. It is very easy to design a set of screws that can't be accessed after the part that they secure is in place! You have to really think through which assemblies go together first, in what order, in order that you can actually, you know, assemble it.

I knew I'd assemble the electronics in place, so the primary order of assembly was that all the functional pieces attach to one side of the clamshell; meaning I could assemble the circuitry, test it, then close the shell up.

I made nearly a hundred sketches trying to come up with the best way to approach the stack of parts on the front; originally the dish and nozzle were to be machined separately in order to achieve the best finish, assembled for final finishing, would trap the separately printed insulator rings, and the acrylic emitter would slide in from the front thus be easily removable for replacement if it got broken.

It would been possible to do the entire thing in aluminium. However, I didn't have the time. First off, the tighter the detail, the smaller the end mill for your finishing pass, and the smaller the mill, the longer the milling. Each part would have taken multiple hours each.

And worse. As can be seen from the test mill above, like a 3d print, a contour milling operation will have stair-steps. You can get clever with the g-code in order to limit these to the unavoidable, but they are still there and need to be cleaned up. As well, inside corners and small pockets may remain that will take hand-work -- Dremel, files, sand paper -- to clean up.

Don't be mistaken; those ridges on the first image are very small and come off easily with hand sanding with 220 grit. But it needs to be done, and it adds to the construction time. It also effects the final part dimensions, meaning you have to anticipate and adjust for that as well.

And oh yes; the clever g-code I mentioned above means understanding some of the better software. I was unable to get Fusion 360 to spit out useable g-code during the time I had to work on this project, and what little milling I did for it was generated through the baby program Cut3d.

A last little note. You want to mill the whole thing? Do yourself a favor and print a test version in 3d first!

Fortunately, parts designed to be milled in aluminium have no problems at all in being printed in Nylon; they pass all of Shapeways' design rules. And ten days later I had my parts for the next stage.

However.

Since this was not aluminium, I couldn't tap the bolt holes. So I had to overdrill the holes and glue in brass thread inserts. Fortunately, most of the interior hardware works just fine with sheet metal screws driven directly into the print.

This is another advantage of a proper CAD; since all the hardware was present in the original file, all the holes could also be included. For CNC milling purposes they would have to be broken out to be done in their own operations, but for print purposes they can simply be printed in. And that means the holes are all straight and line up properly.

I also discovered a few places where I hadn't problem-solved the CAD quite right. That required a certain amount of trimming the print, and gluing in a few extra bits here and there. But this was only a day's work. Much better than trying to do the same to metal.

Of course, there was one part I really hadn't designed right, and I ended up trying to solve it on the printed gun...

The shape of the gun precluded having the size of battery I wanted. In fact, after the speaker and trigger mechanism and main electronics board were in, there was no more room for the battery in the main body!

So the CAD put the battery in the grip.

I contemplated sticking the battery in permanently and having a hatch on the bottom of the grip to access a USB charging cable. But that felt wrong; what if you needed to remove the battery to take it on an airline or something? So I was seduced by the idea of a battery inside a machined or printed "clip" that would slide into the grip, and as of the files being sent off to Shapeways, the user would simply have to fish the cable out of a small hole and detach it.

Well, that didn't work. When I had all the parts on the table, it swiftly became obvious you'd never be able to stuff the wire up behind the "energy cell" in its current design. So I modified; clipped the cable short, epoxied the JST to the top, epoxied the matching connector into the grip, and added an ejection spring to make it possible to extract the battery case.

That case being printed at TechShop, by the way, on a MakerBot using a reel of bright pink PLA.

Paint wears. The big advantage to real materials (like raw aluminium) is that they can take abuse and still look the same. Chrome paint, particularly, does not do well. The better the chrome, the more fragile it is, and unlike colors you can't protect it behind a few layers of sealer.

In any case. Although I was concerned with how well it would hold, I went for Krylon Select for this project. On top of a black epoxy spray for strength. And the real trick for cleaning and sealing the Shapeways nylon print was to brush it with superglue before sanding and priming.

In the image above, left is chrome spray paint, middle is Rub 'n Buff, and right is natural aluminium (buffed with 0000 steel wool and corrosion inhibited with a light coat of sewing machine oil).

The first thing to understand (besides the surprisingly small interior spaces I had to work with) is that human senses are roughly power law in their sensitivity.

Which is to say; if you want something to sound twice as loud, you need an audio package that will handle ten times the power. And that is why prop-makers have a lot of difficulty in achieving brighter lights and louder sounds than are present in your basic child's toy.

You can upgrade from the 60 milliamp LED most toy manufacturers will reach for with the kind of 3-7 watt LED that goes in high-end flashlights. That's within the range of what a good battery, a little basic thermal design, and simple current limiters can handle. And the rule of perception says that will look about twice as bright. To double that again...you need a proper driver circuit and a cooling fan.

The log scale for audio is similar; to get a sound that is perceived as twice as loud, you'd need to upgrade from an one-chip amplifier to large solid-state components with individual heat sinks; the footprint of the amplifier circuit doubles in every dimension in order to handle the necessary power. Which is why essentially all those consumer iPod speakers are about the same volume. Unless you get very clever -- which I tried but, in this prop, failed.

This prop had an even worse problem for sound; no sound holes. So my bright idea was to not even bother with a conventional speaker, but to couple a surface transducer directly to the shell of the prop. And I think this could work some day. Especially if I combine that with another bright idea; to create a prop with elements in the shape of a Helmholz cavity, and tune the sounds to that natural resonance. In any case, the shell ended up 3d printed and the sound is...underwhelming.

The surface transducer is vibration isolated from the rest of the gun, with the pusher plate placed firmly on a 1/8" piece of machined aluminium that acts as a sound board. This in turn couples to the rest of the gun...but since the rest of the gun is sintered nylon, it doesn't vibrate well.

I backed up the emitter with a 5-watt "pink" LED. This was also a mistake. As the image above shows, the pink extruded acrylic picks up red LED light very well. The pale-purple color of the "pink" LED did little for it -- and there was no time to change over to a red LED instead.

The trigger mechanism is a CNC milled piece of aluminium that pivots against a leaf switch, and is tensioned with a spring. The other functional element is the dial on the side, which turns a potentiometer with switch; the latter switching the main power on and off.

The trigger action itself is applied via a microprocessor. I rather dislike the feel of having light and sound directly connected to a switch so they go on while the trigger is being pulled and go off when it is released. This gun was programmed so it plays a complete sound and flashes the light once each time the trigger is pulled, and it has to be released to reset it for the next shot.

All the sounds were generated from an ATtiny84. There was no space for another board. And the ATtiny was there anyhow, reading the trigger and selector dial and sending PWM to the LED drivers.

In fact, I leveraged a board I've been working on for this; it has an ATtiny84, an ICSP programming header, and four LC135 constant-current drivers. The latter are a cool little chip that will supply a regulated 350 milliamps to a high-power LED, will take PWM, and can be stacked (or, rather, run in parallel).

The board had to be heavily modified for this prop, of course.

It is relatively simple to do tones on an ATtiny. It will run (with some important caveats) the tone() library for Arduino. It is also capable of wavetable synthesis. The theory is simple enough. One of the on-board hardware timers is set up for analog voltage output; fix frequency, variable cut-off (PWM, basically). A second timer is used to generate a hardware interrupt. When the interrupt service routine is called, it picks up the next value from a wavetable and adjusts the set voltage output of the first timer.

This trick requires a couple of things. It requires the second timer be operating at the desired frequency X the length of the wavetable. It requires the first timer be running several orders faster (otherwise there will be funny aliasing going on). And it requires the program not spend a lot of cycles doing anything other than being ready for the next interrupt to be processed.

That latter proved difficult. I was running software PWM for the LED (there were no more hardware timers available on that chip), I had to take trigger readings and analog readings off the dial, and the kinds of sounds I wanted needed to evolve over time. So I gave up on the full wave table, and generated simple square or sawtooth waves. And that meant I could spare the program cycles to add or subtract numbers to the frequency (that is, the base rate as which changes are made to the requested analog output of timer1.)

It still meant that the sounds changed radically every time I put in a new "If" statement, and they were also strongly dependent on the base frequency chosen for the PWM. But with a fair amount of trial and error I was able to get some chirps, warbles, and clicks for the various shots and function sounds (power up, selector dial motion, etc.)

In fact, the aliasing and general "grit" was a benefit. Using similar techniques to the wavetable synthesis above one can make white or pink noise of various flavors, but that proved unnecessary. If I was using a faster and more powerful chip, such as found in an Arduino Nano, I would probably be digitally combining various wavetables and processing before sending that to the final analog output.

All the programming was nominally done in the Arduino IDE. To leverage the hardware timers in the way I had to, I was addressing them directly without the help of any Arduino libraries. But the benefit of programming on a prop like this is you can think "sketch," throw together whatever messy code works, and call it good enough.

The body, grip, fin, hilt, and dish/nozzle assembly were all 3d printed at Shapeways. The trigger and trigger guard were milled from aluminium for strength. The emitter itself was lathed from pink acrylic for looks.

This wasn't a lot of fun. By making a lot of shallow passes I was able to mill down the acrylic rod to diameter on a metal lathe, but it flexed a lot in the process and cutting the nose shape proved impossible. So that was done by holding a file against the acrylic while spinning it in the lathe. Fortunately, a diffuse surface was what was wanted (to better show the light) so there was no need to go further than 400-grit paper in cleaning it up.

You can achieve optical clarity, but it requires wet-sanding to as high as 2000 grit in progressive stages, then acrylic buffing compound.

The donuts and the energy cell were printed at TechShop in pink PLA, cleaned up and gloss-coated without painting. The side details were also printed by me, but those got the same chrome paint at the other parts. This was part of the "make it look real" philosophy, as well as make it easier to sand flat the sides of the gun; instead of gluing then painting, they were intentionally done to have a visible seam between gun and flashes.

A last-minute inspiration was to use a 1950's "Atom" motif, with a tiki head for a nucleus. The latter was the only part modeled outside of Fusion 360, as it seemed easier to do it in a polygon modeler.

As it happened, I was unable to quickly combine these different meshes, so I printed them separately and glued them together post-print.

That finishes the gun. It also needed a holster for easier and more convenient display at a gathering. The inspiration was, among others, the clear holster worn by Wilma Deering on the 1970's Buck Rogers television series.

Clear vinyl and white pleather did not work by themselves. The ray gun simply has too many parts that stick out (like the dish on the front) and it catches on a soft holster. So I had to move to the standard; a hard-shell holster. Which, once the shell is made, can be covered in whatever way desired. I was able to stick with my original intent of transparent window just by revealing the underlying structure.

First step was fitting. I made mock-ups with cardboard (and paper towels, when I thought a soft holster would still work). One of those mock-ups became the base of the real thing. Or rather, of the buck.

The key tool here was the vacuum forming machine. A handy tool in the kit of many prop makers, and one of the tools available at TechShop. A buck has to be solid enough to take the crush of 16 PSI, plus the touch of plastic heated near the melting point.

The buck in this case started with cardboard strips held together with duct tape. This was then filled with wadded-up cloth for stability and coated with Bondo. Several sanding steps and re-coat later, the buck was smooth enough and strong enough.

This is a very fast method. The major flaw is since you are working from the desired dimension out, the final item will be larger than desired. A better approach is to reinforce the outside of the reference shell (duct tape, pepakura, or whatever) then fill that with plaster (or better yet hydrocal, which does better in the vacuum forming machine). This makes you master closer to dimensional.

In any case, the holster shell was pulled in PETG. The curved side only; the flat body-facing side was cut out of scrap and the whole thing glued together.

Then my Bernina 830 and I struggled for nearly two days to make a neat job of pulling white pleather around the thing. If time had been tighter, I would have just used hot glue. As it was, some of the stitches are decorative and hide a hot glue solution, but the fabric is actually seamed to make it drape around the form.

The final stage was to create padding for the inside. This was white shirting, hand-quilted on the Bernina. With, of all things, shredded cotton balls; I didn't have time to make another fabric store run for proper batting.

To finish it off I made a white pleather belt with a CNC'd aluminium buckle (false buckle; the actual sizing of the belt is via velcro). And it went out the door too fast to get proper pictures, other than the ones above taken in the just-finished holster.

As I mentioned upstream, the LED is underwhelming and so is the sound. If I upgrade it or make a new one, it will swap out a standard 8-ohm speaker and add a couple tasteful sound slots at the bottom, and replace the LED with a red or RGB -- possibly of 7 watts or higher.

All the files are at my Shapeways store and it would be simplest to print another in the same way; although there are some possible modifications for better 3d printing, there were even more discoveries that would have to be corrected before I made another attempt to CNC mill the entire thing from metal.