I made a scaled down, reasonable proof-of-concept of the magnetically driven pixel drawing machine. About a week ago I knew there was a lot of work ahead of me; if you’ve not read it, here’s the post about the steps leading up to the developments shared below. I ended it with a little work plan:

  1. Work on prototyping at my chosen scale in acrylic instead of cardboard.
    • Play with the pivoting mount; perhaps glue straws onto the discs but there may be a better way
    • Investigate engraving the acrylic surface to provide a shallow trough for the axle to sit in
    • Tighten tolerances to match the flatness and uniformity of the material
  2. Design and test the mechanism that returns pixels to their dark state.
  3. Link these parts up with the plotter mechanism I’ve already built so that a magnet is driven by control knobs.


I was able to do all these things in a marathon work session, and fortunately for both of us, I took pictures to make the moment last longer. I ended up making a 6×4 array at scale to serve as a proof of concept.

Acrylic construction

I cut pixel pieces using the same file I’d made the cardboard ones out of, albeit with one minor modification: I added a centerline to the pixels to serve as a guide for gluing on the straws (bushings). The alignment of these is not super fault-tolerant, so having a nice straight guide built in to the object itself is very useful.

(My DXF cutfiles are on GitHub.)

pixel with centerline

Note that I also added a registration mark on to the board—this matches up with the rod that the pixels ride on, and ultimately also matches the pixel centerline, minus some wiggle introduced by the loose fit the straw has on the wire it’s riding around.

acrylic pixels with straws

I carefully glued straw pieces onto the pixels, taking care to leave a tiny overhang on both sides, like 1/64” or so if possible, just to help keep the pixel’s body away from brushing against the wall of its cutout. These are, after all, just turning under a fairly weak magnetic force, so any added friction is likely to cause trouble.

For the axles that run along the rows, I used a piece of TIG filler wire we happened to have around the lab, which is really just steel wire that comes in straight sections (that’s better for me because I don’t need to unbend it to start with).

For expediency’s sake, I’m just using hot glue to attach everything to everything—metal axle to paperboard backing, and plastic straw to acrylic pixel. It’s definitely not the best adhesive for these very different applications, but this is a prototype and fabrication speed is the name of the game.

Here’s the assembled and glued array:

unpainted 6 x 4 array

Note that it’s living inside of a sort of cardboard frame I made for it. This is so that it can stand on its own at the proper spacing from the pixel resetter frame, which needs to be about 3/8” in front of the pixel board and able to slide up and down vertically.

Resetter frame design

I ended up designing a simple resetter frame; it’s just an array of rectangles, each of which fits one of the pixels. The first time I tried to cut out a version of the resetter, the cardboard apparently got pushed around a bit by the head of the lasercutter, causing weird misalignment:

resetter frame with misalignment

Since I cut the later version out of clear acrylic, it’s hard to see the shape of the actual piece. (The DXF is on GitHub, as linked above.) Here is the plan:

dot reset frame plan

The way it works is that the bottom of each rectangle in the resetter sits under the overhanging flipped pixel edge. Moving the resetter about 1” straight upwards is enough to push the pixel back to its unflipped position. The bump at the center bottom of the resetter is meant to protrude out of the bottom of the frame it’s sitting in, so that it can be pushed up from below.

A gif is worth 10,000 words:

dot reset movie

How did the dots and frame get colored black but only on one side? Through the magic of spray paint.

Plotter mechanism and putting it all together

Last semester I built the XY gantry and with only some minor adjustments that prior design has been working pretty well. Here’s what it looks like:

XY gantry

Pulling on either chain independently manipulates the X and Y motion:

moving gantry

Jen’s on the left, controlling motion in X and I’m on the right controlling motion in Y. Instead of gripping a chain, of course, both users will simply have a knob to turn, but that knob will be driving the chain like we are.

What’s that thing riding around at the intersection of the axes? Well, the dots can be flipped with a smallish magnet passing close behind them:

flipping perspective

…or with a really big magnet passing a bit farther behind. This is preferable for me, because it means that I can give things a little bit more mechanical clearance, and if somethings bends or warps hopefully it won’t hit something else.

I mounted the aforementioned big magnet (it’s a neodymium cylinder that’s 1” in diameter and 1” tall) on temporary wooden crossbars on the plotter mechanism. To allow it to be able to slide in both dimensions but always be properly oriented, I built a really simple linear bearing for it out of rolled up paper. I then (naturally) hot glued the magnet to it. Looks pretty silly, but it works!

paper linear bearing

Putting it all together

Here’s a movie of me manipulating just one of the axes manually, which moves the magnet from left to right. Holding the dots in front of it, I used the resetter to return them all to black at the end. This illustrates the motion mechanism of the plotter doing what it’s supposed to, and the dots flipping like they’re supposed to, and the resetter returning them to unflipped.

It’s a squeaky machine right now! There’s no lubrication anywhere so that makes sense. I also really like the ka-thunk the pixels make when they hit the paperboard.

Next steps

The beauty of digital fabrication is that making more acrylic pixels will take only a couple minutes in software, maybe 10–20 minutes on the laser, and then lots of unavoidable assembly time. That’s the good news.

But before I go cutting all those pieces out, I’ll need to finalize the actual mechanism of the enclosure, taking account of the stack of sometimes-overlapping things inside the box to make this system work properly. That’s a bit harder to manage, and in order to avoid discovering major problems during assembly, I’m going to try to do some careful drawings before to better understand the best way to put this together.