I think of this project as a kind of lighthouse, even though it has nothing in common with a maritime navigation aid except that it's lit up at night and it rotates.
A lighthouse has an intense, tightly focused beam. This project uses a 13 watt fluorescent tube whose light will be as diffused as possible.
A lighthouse beam rotates, shining through fixed clear windows. This project has a stationary lamp and rotating colored glass.
A lighthouse is a tall building so the light can be seen miles from shore. This project will sit about a foot off the ground and will be visible for maybe a hundred meters.
Before getting into the construction story, let's begin with the "beauty contest" views, namely short movies of the (mostly) finished lighthouse by day and by night.
Click to view a short movie of the rotating lighthouse, installed beside the front door.
Building the lighthouse was not a straight-line progression from idea to finished goods. No, there were a few missteps and detours along the way. In the photos below, you may notice details that appear or disappear, and not all of them will be mentioned in the text. That's to be expected with a one-off development project.
Click an image to view its enlargement
I began by making the plywood housing. The concept was that a rotating glass cylinder would be supported top and bottom by turntables that ride on “lazy susan” bearings. (Quite good bearings are about $6-7 from your local building center). It would be illuminated by a fluorescent light at the center, and all other electrical stuff would be housed inside the box.
First attempt at a drive system. This didn't work very well.
Turntable. The turntable would be driven by a small DC motor that I salvaged from a dead ink-jet printer. A belt would transmit drive power to the lazy susan. Direct drive — gears or idler wheels — was ruled out because I knew the turntable would be imperfectly round and might be mounted a tad off-center.
Initially, I tried a single belt made from two strips of duct tape stuck together, inspired by a YouTube video. As I learned after the whole thing was built, this was a weak point in the design — but not because the belt failed. (In fact, scientific tests have shown that duct tape is quite good for just about everything except taping ducts.) Rather, there were two other problems.
One was keeping the belt at the correct tension to pull the weight of the glass-laden turntable. Too little tension and the belt would slip, too much tension and the motor would stall; I could not find a "Goldilocks zone" where it was just right. Tension was being set by an idler wheel that was bolted in place, and this was doubtless a wrong choice. Pushing the idler wheel against the belt with some sort of spring might work better, but my mechanical skill wasn't up to that.
The other problem was that I wanted the glass to rotate at a very slow pace. With the motor that I had on hand, at full voltage the motor spun the turntable too fast. Reducing the voltage slowed the motor, but gave too little power to reliably overcome friction in the turntable.
So, on to Plan B.
Second attempt at a drive system. So far, this seems to be working nicely.
Plan B was to use two drive belts: one from the motor to the large wheel of an intermediate pulley, and the other from the pulley's small wheel to the turntable. The pulley would act to gear down the final speed, allowing the motor to run faster (thus putting out more power) while moving the turntable quite sedately. And using rubber bands as drive belts solved the whole tension problem — or anyway, transformed it into a search for rubber bands of the right size.
I made the pulley by gluing a 1-inch plastic wheel (actually a gear stolen from a FischerTechnik construction set — it's like a Meccano or Erector Set, but aimed at younger children) onto a piece of 5/32-inch brass tubing. The axle was a 1/8-inch brass rod, which happens to fit nicely inside the 5/32 brass tube. With these dimensions, the pulley yields a 6:1 step-down in speed, with a corresponding step-up of torque.
My wife solved the rubber-band search problem at the supermarket, where she found a band designed for holding garbage bags in place. (Conveniently, it was made of white rubber, as will be discussed below.) It was labeled as a "15 inch band", but the actual circumference of the loop was 30 inches — just right for the 34-inch outer circumference of my turntable. The small amount of stretch provided a good grip on the intermediate pulley, but no fear that it would break under the strain.
The belt from motor to intermediate pulley was a recycled Post Office rubber band.
Inside the box. Although the main function of the box is to provide rigid support for the fluorescent lamp and rotating glass, it also provides shelter for the electronic parts. In the photo above, you can see:
What's not visible, because I added them after the photo was taken, are:
The stained glass. From the foregoing, you might think this was a woodworking project. On the stained glass side, things were much simpler. Cutting the glass sheets into strips and the strips into panels, a repetitious task, was eased by using jigs. Edging each panel (all 240 of them) with copper foil took many hours, but there's no story to be told; I just slogged through it. Where things got interesting was assembling all the panels into a cylinder.
The natural springiness of sheet metal kept the curvature smooth, but it tended to sag into an oval shape. To prevent this, circles cut from cardboard were stuffed inside.
Forming the glass. As with the jack-o-lantern project, the glass pieces for the window were assembled on a scaffold, in this case a cylinder. The math said that I needed a 32-inch circumference, so the form was made of a 36-inch piece of sheet aluminum, with two inches at each end bent inward to hold the cylinder closed. To keep the cylinder round, the inside was braced with several cardboard circles. On this form I laid up the glass pieces, holding them in place with painter's tape, and then soldered them all together.
The glass pieces are laid against the scaffold and held in place with tape. Unfortunately, as soon as you apply soldering flux to the glass, it dissolves the tape's glue. Fortunately, you only need the tape for the first couple of rows, which are nearly vertical. After that, new rows become more horizontal and gravity starts to help more than it hurts.
When the outside seams were all soldered, it was easy to disassemble the scaffolding and pull it out from the center. As I'd hoped, the solder holding the glass pieces together did not stick to the aluminum. (Zinc-plated sheet metal is much cheaper, but solder sticks to zinc.) In fact, the gooey solder flux made the aluminum slippery so it slid right out.
That's when I discovered that, soldered only on the outside, the glass cylinder was not rigid enough to hold its shape. It sagged into an oval. So I quickly slid the scaffolding back in and restored circularity. Sliding the scaffold partway out, I soldered half the inside seams. Then I turned the work around, slid the scaffold out the other end, and soldered the other half of the seams. After that, the cylinder stayed circular when the scaffold was removed. But for extra strength, I added a thin strip of brass all the way around the top and bottom edges.
The almost-finished project, assembled for a dress-rehearsal round of testing.
First tests. Initial testing revealed a few design errors and construction mistakes.
The top and bottom turntables weren't perfectly aligned and, with the glass cylinder in place, they would jam at various points in the rotation. The simplest cure was to remove the top turntable. That left a gap of about 1½ inches between the top of the cylinder and the roof of the enclosure. Not only was this gap ugly, but bugs drawn to the light could fly right in. The cure was to suspend a curtain of black polyethylene from the roof into the cylinder.
Amost nothing sticks to polyethylene, so I clamped the plastic between the halves of an embroidery hoop.
Turntable rotation was jerky. Near as I could tell, this was due to the smaller rubber band trying to climb off the motor driveshaft. Switching to a narrower band seemed to fix it.
The motor made a loud singing noise whenever it was running below full power. The pitch was close to a middle-B (490 Hz), which indicated that it was caused by the controller's pulse-width modulation of the motor voltage. (It was probably some sort of resonance with the motor mounts and case.) Fortunately, it was easy to change the controller's frequency. Not knowing whether to go much lower or much higher, I opted for the lowest available frequency, 31 Hz. That worked well enough that I didn't try other options.
I was concerned that the turntable rotation could be stalled, due to weather or wild animals or malign spirits. Even if I couldn't prevent this, I could detect it. Then I could try to shake off the jam and resume operation; or failing that, shut off the motor to prevent burnout.
With a black permanent marker, I inked stripes on the white rubber band at quarter-inch intervals. Then I aimed a reflection photodetector at the alternating white and black stripes, feeding the signal into one of the controller's analog-input pins. The controller could then measure how much time elapsed from one stripe to the next. If too much time went by, the controller could invoke a "panic" subroutine of some sort.
The stalling detector worked, but created a new problem: The black-and-white striped rubber band was too visible; it distracted from the colored glass. So I installed a strip of sheet metal as a visual barrier.
These changes entailed several new holes in the plywood case and made other holes redundant. So I took everything apart again, sealed the surplus holes with wood putty, and varnished the new holes and wooden pieces that got added along the way.
Status as of October 9. The lighthouse is undergoing second-round testing: I've installed it outdoors to run non-stop for a few days.
The lighthouse is now placed in its intended outdoor location, next to my front door.
While that's happening, I will contemplate whether to paint the woodwork with a dark brown enamel. It's not strictly necessary; everything has two coats of spar varnish already. But it might look better.
Update for October 11. Around dawn, the weather changed to a misty rain, interspersed with heavy downpours. Both rubber bands got wet and the lighthouse stopped rotating. The stall detector did its thing, but wet rubber just slides over wet brass, and the machine's attempts at restarting — one try per minute — were in vain.
Coating the brass with nail polish and heat-shrink tubing (chosen because I had them on hand) proved ineffective. So I glued on another gear (from the same FischerTechnik set) in the hope that, with its spikey surface, it will provide some friction even when wet. I'll find out after the glue has dried.