Beam: a Kinetic Sculpture

Beam is a kinetic sculpture created at Pier 9, during my Autodesk Artist-in-Residency.

The idea is the spinning mirrors make a single line of light twist and turn in erratic patterns. In all my work I hope to inspire inquiry through perceptual manipulations. The viewer can explore and inspect the piece further by changing perspective to get different secondary, tertiary, etc reflections between mirrors.

I used the Omax waterjet to cut the discs, to make them accurate slices of a cylinder (i.e. ellipses), and the Epilog laser cutter to the make the support structure.

Even though I had this idea in my head for a few months, I find it hard to simulate reflections in my head. Anything past a flat, single reflection mirror, and my head explodes. That's why I rely on computers.

Fusion + Maya can both do some decent renderings with ray-tracing. That means, that it will simulate photons moving from the light source and bouncing in the right way throughout the entire model.

If you look through some of the gifs, you can see how I explored this idea in simulation before even building anything. In parallel, I had sketched out a mechanism that could make this seemingly impossible sculpture work.

In one of the gifs, I'm playing with slices of a cylinder (by definition, ellipses) so that the profile of the entire sculpture stays as a cylinder.

In another, I keep the ellipses, but make it regular spacing and angles. Although a little more mesmerizing, this is less interesting to me, because you don't get a chance to explore the piece.

In another, I make the mirrors circles (subtle difference), but it creates a wobbling effect. Although interesting, it is too similar to an exhibit at the Exploratorium, and the effect detracts from the effect I want to highlight – that the line is no longer a line.

Here's a sub-instructable on how to render an animation in Fusion 360 (because it's not obvious).

https://www.instructables.com/id/How-to-Render-an-Animation-in-Fusion-360/

I had this idea that I could precisely cut clear polycarbonate tube and then sandwich laser-cut acrylic mirror in between the tube sections. This was a silly idea. There are too many degrees of freedom, which means that I had to (1) glue things together and (2) align things by hand. This can never leads to good things.

So that was that. This prototype led me to getting on the waterjet and using the 5-axis head to cut out mirrored acrylic with angled edges.

One cool thing that came from this was a way to cut polycarbonate tube at somewhat precise angles on different planes. I 3D-printed a cap that indexed the tube every 45º. It was press-fit on the end and allowed me to know what plane I was on when I was cutting it with the coldsaw.

Intelli-cam is a misnomer. It does CAM, some kind of CAM.

Now that I was going to cut things in 5-axis, I had to use software to generate the 5-axis toolpaths. As you can see from the gif, the .omx file that Intell-CAM spit out was an ellipse with a constant bevel around the the ellipse, rather than what my model was telling it to do, which is a constant angle cut. It was doing the inner ellipse correctly, cutting it at a certain angle, but not the exterior cut. (In the photo, the top disc is incorrect, and the bottom one is the correct way it should be cut).

Long story short. I had to modify the .omx directly in a text-editor. Instructable for how that worked here:

https://www.instructables.com/id/Hacking-the-Omx-Files-From-OMAX-Intelli-CAM-for-th/

After 5-axis cutting ellipses with the waterjet, I bought some bearings, some gears (to drive from a different axis of rotation), and a cheap low-RPM motor off Amazon.

The mirrors are single-sided acrylic mirrors that are then placed back to back. They are cut accurately enough that they slip onto the polycarbonate tube (that is sanded) and stays there due to friction. To avoid using a slip ring, I am spinning the polycarbonate tube (on which the mirrors are attached) while the light stays in place. The tube spins via a gear train connected to a very slow motor. Some pseudo-mechanical-engineering later, the whole thing spins!

As soon as I finished it, it was obvious what to improve:

• Motor noise – the motor I got makes a terrible whirring noise. It's a regular DC motor with an internal gear reduction. As was later suggested to me, I should've used a synchronous motor, which uses AC current.
• Mirror quality – The back-to-back acrylic mirror makes for some thick (0.25") discs. They are second-surface mirror, so they have a small gap before the reflection actually happens. You can see this at each joint. Time to use first surface mirrors (hella expensive!) or make my own mirrors (hella time-consuming!)
• Scale – Right now, it's cool as a table lamp, but for it to be a real Sculpture with a capital S, I need to scale this puppy up.
• Return path for wire – the connection to the top of the fluorescent bulb is a wire that hangs to the side. No bueno! Scale might help by making it so tall, I run it to the ceiling, but maybe I can cleverly tuck the wire away.

Some cool things I didn't notice from the renderings

• Two 45º mirrors, in parallel, will create a neat staircase infinity, and twisting one of the mirrors will cause it to curve. It's really hard to spot this though, unless you're really looking for it.
• The edges of the mirrors get lit up, because the mirrors are acrylic. This should have been predictable, but didn't think about it. It's clear I need to think about it, because it draws a lot of attention.
• You can create seemingly closed loops of the light line within the mirrors.

With the goal to create more interesting perspectives to explore, I bought some larger components and cut larger discs to create a model that was scaled up. I also addressed all the previous issues.

• I bought a AC motor (aka synchronous motor) to run it at a slower rpm
• frosted the tube by hand-sanding it to make it appear as if the discs were directly on the light
• tested out a few different mirrored surfaces: (mirrored acrylic, mirrored glass, clear glass, dichroic acrylic)

Biggest issue was that at 8ft tall, the tube holding the mirrors bent enough to make it look not straight (but not enough to topple it over), and the heavier discs slid down the tube. Tolerance was okay, but they needed a dab of glue to stay in place.

The next step is to create another version with better proportions (larger diameter discs), a stiffer tube, and a single material (e.g. glass and black acrylic).

The first of two. I decided to go with two columns: one with black acrylic for the discs, and one with clear glass. The new materials provide the reflective quality I want, but in a much more subtle and surprising way. The proportions have been modified to more closely match the smaller model. The spinning discs create an 18" diameter cylinder, giving more room for the reflections to breathe and allowing for the discs to be more spaced out.

Some improvements: I found an 8ft fluorescent tube from Universal Electric Supply in SF (without having to buy 25 at a time), so there's no break in the middle. And, I also found a way to get a more even frosted finish on the polycarbonate tube: Sandblast a long tube in a small sandblaster.

The last challenge was waterjetting glass. It's surprisingly hard to find glass in the Bay Area that isn't ridiculously expensive. I ended up buying a bunch of 3/32" and 1/8" thick replacement window glass from Home Depot. Note: don't trust the number of "in stock" on their online store.

Water-jetting glass was it's own thing. Here's what worked for me: https://www.instructables.com/id/Tips-for-Waterjetting-Glass/

The final pieces are much better than what I envisioned when I created the renderings. Each prototype was successful in that it lead to improvements for the next version. I wouldn't have been able to get to the final model without all the valuable help everyone offered, all of which was spurred by the physical prototypes created.

The two pieces are titled Beam 1 and Beam 2 and was exhibited at the Artists-in-Residence Spring 2016 show at Pier 9, Autodesk in San Francisco.