Fountain of Steel

This project started late one night, a Wednesday night, February 20 2002, just after Junkyard Wars (the remote control car episode) and too much caffeine. I was thinking of making something nifty and mechanical, but not something too big, expensive or difficult. I wanted something physical (not just a piece of software). I also like a bit of complexity, where the device reacts to changes, meaning feedback loops plus possibly some computation. I though up most of the design of a flying mouse ball clock with xylophone chimes, then finally got to sleep. The next day, I decided it was interesting but not worth building. However, at the ex-Amiga programmers meeting a week later, KenR suggested that we do a project rather than playing video games - a robot of some sort. Lots of RobotWars et al stories were brought up, we looked on the web at the Lego Mindstorm kit, the MIT smart brick kits, robotic Palm PDAs and various other things. Ideas were tossed around such as an autonomous device, like a river boat or airplane (solar powered) or just a smart bottle in the sea, with a GPS and wireless link (even just a cell phone) could be cool, but too expensive. Well, while the other guys are thinking up things, maybe I can actually get something going with my simple project.

Initial Plans

An electromagnet to move around iron seemed to be a good fit - cheap, easy to make, can be visually interesting, and you can fiddle with timing to make it do different things to moving lumps of iron. Other people do rail guns or coil guns, but I decided to make a fountain. Partly because it's more interesting to see hundreds of flying objects, and partly because it can run continuously, perhaps as a work of art / conversation piece in the corner of a room. Maybe I could even make money making copies of it, much like the Nixie tube clock guy does - but that could end up being too much like work.

Mouse Ball Clock

[Diagram 1]After several hours mulling over many ideas instead of sleeping, I decided to start with one solenoid magnet and a few dozen iron or steel balls. The solenoid electro-magnet would throw a ball to a certain height and move it over sideways (tilt the solenoid a bit). The ball would then fall down into a funnel, through some chutes and back to the bottom where it would be picked up again. A photo-transistor and infrared light emitting diode would sense the presence of the ball at the entrance and turn on the magnet. A 555 timer would keep the magnet on for an adjustable time delay. The 5 volt signal from the timer output would go through an opto-isolator which would turn on a power transistor that would connect the magnet to the DC power. A second magnet and sensors could be used for fine adjustments of the speed (needed if the balls are of different weights).

For bonus points, a PIC microcontroller based clock circuit could increase the pulse duration on the hour and fling a few balls over to the side, where they would hit a xylophone and chime out the hour. To make it less disturbing, rubber coated balls (mouse balls are a good choice) would be nice, or padding (mouse pads) on the chutes could be used. A solenoid operated gate could also stop the noisy flow of balls, except when the clock needed to strike the hour. If you could get a really big flow, you could make a raster display of balls (with X and Y deflection magnets), perhaps showing the current time.

Physics and Power Requirements

I also had an old power transformer in mind (used in a home-made carbon battery recharger), but after digging it up, I see that it puts out 25 volts alternating current at 1 amp. 25 watts of power probably isn't enough. I should figure out the energy needed to raise an uncertain amount of iron an uncertain distance each second, then design the system.

Sam Barros's Gauss Gun page describes the theory involved in making a solenoid and driving circuitry. Basically, the magnetic field strength (in units of Gauss) is proportional to number of turns and current. But if you have more turns, you have a longer wire and thus more resistance and less current. More turns also increase the inductance from the magnetic field resisting change, and that cuts down the current, at least initially. He also recommends using cylindrical projectiles, spheres being the worst shape. Guess which one I have to use, plus I may want to hide the steel inside a wooden shell (acorns?) to make the whole thing look like a magical wooden clock. Another factor is that, at least for his high power short pulse system, the total efficiency is at best 3%, typically around 1.5%. But I'm not using short pulses, so things may be better (and the equipment won't self destruct after one use). Another site with more actual info about coil winding physics and the associated electrical switching circuits is the Magnetic Gun Club. I'll probably use a variation of their designs.

The height depends on the speed of the balls and the local gravitational field. The speed defines the minimum separation distance between balls - if it is too small, activating the launch magnet will slow down the nearby previous ball. Say the balls are 1 cm in diameter. Magnets should be 1 to 2 times the projectile length (2 cm long in this case) according to Sam Barros. If I want a 6 cm separation, and I want 100 balls per second, that's a speed of 6 m/s. That's equivalent to a drop in Earth's gravity field (9.8 m/s²) for 0.6 seconds, which is ½ a t² = 0.5 * 9.8 * 0.6 * 0.6 = 1.76 metres.

The balls will be made of steel; density is 7850 kg/m³. If they are 1 cm in diameter, each ball has a volume of 4 π r³ / 3 = 4 * 3.1415926 * 0.01 * 0.01 * 0.01 / 3 = 4.19e-6 m³. Multiply by the density and you get a mass of 0.033 kg or 33 grams. Oddly enough, a Logitech mouse ball of about that size weighs 30g. The work done to lift the ball is mass times gravitational acceleration times height, 0.033 * 9.8 * 1.76 = 0.57 joules. So, to add 0.6 joules of velocity, at 1.5% efficiency (more if you add a steel can around the coil and end-caps (washers) to concentrate the magnetic fields), the magnet sucks up 40 joules of power. Doing 100 balls in a second, that's 4000 joules per second, or 4000 watts. Wow. Most of that goes into heating the coils and circuitry - expect things to get hot. That's a bit much, about 1000 watts is the maximum continuous safe level for household outlets, 100 watts is a more reasonable power consumption level for a work of art.

But I can still do it. With 100 watts, that's 100 balls with 1 joule to allocate per ball, of which 0.015 joules actually accelerate the ball. That would lift the ball 0.033 * 9.8 * x = 0.015, giving x = 0.046 m. We can lift 100 balls by 4 cm. That could be enough. Why? Because the device doesn't have to launch balls from a standing start, it only has to make up for the loss in energy as they circle around. If the ball dropped down the 1.76m tall chute only rises back up to 1.70 m, it will work. If not, just design for more power, perhaps 500W is enough?

So, the first test is to get some balls, drop them into a chute from 1.8 m height and see how much energy is lost.

Hunting and Gathering

The dropped ball test was inconclusive, except for pointing out that a really strong chute would be needed, possibly with a large diameter (the size of a bicycle wheel). That brings up the point that the balls will be spinning as they roll along the chute, which will induce eddy currents as they go through the magnetic field and perhaps slow them down a bit.

Ball Holding Facility

[Steel Balls in bags and in hand]KenR found out that Lee Valley Tools sells chrome plated steel balls, which would be great for a fountain of steel. They are 9/16 inches or 14 mm in diameter, thus have a volume of 4 π r³ / 3 = 4 * 3.1415926 * 0.014 * 0.014 * 0.014 / 3 = 1.15e-05 m³. Multiply by the density and you get a mass of 0.090 kg or 90 grams. Three times as heavy as the mouse balls. So, I'd need a 12000 watt power supply, and larger magnets (diameter and length). I'll pass on using them for the first prototype, but they'd be worth trying in a later model if there seems to be excess power in the first prototype. I also found another source of mouse balls at Cables Online, with a pretty good price for 100. Unfortunately they are too large at a 2cm diameter. Of course, the ultimate would be a cannon ball juggling machine.

[Diagram 2]RogerC asked about how the balls would be stored. This ties in with the release rate. I'm imagining a box with a slightly sloping bottom, and a wide exit hole in the wall. The hole is tall enough for a ball to pass through, and maybe 5 balls wide. On the other side will be a very slightly downwards sloping ramp, with curved sides, and decreasing width (down to 1 ball wide). This will encourage the balls to slowly move forwards and merge into a single file. At that point, a gate could be placed to cut off the flow if you don't want it running continuously.

The slope past the single file form-up point will be parabolicly increasing downwards. That ramp turns into the 1 ball wide main chute, which might need to be made of plastic (easier to shape than wood). For artistic points, the ramp and chute look a lot like a tongue, so a head shape could be put around the box. Oh, the purpose of the ramp and curvy things is to have a continuous stream of balls, evenly spaced apart, before they fly into the launch magnet.

Musical Inclinations

I still need to find a toy xylophone to see what kind of balls sound the best, along with impact energies. Someone looked for door bell chimes when I mentioned what I was looking for - they'd work too. Or even a bell might be nice. For that matter, a range of percussion instruments (drums and others) could also work. This seems to be turning into a musical instrument (that "Musical Inclinations" title has a nice double meaning too). Anyway, I'll add some extra timer circuits running in parallel, each adjusted to make the ball hit something else, and select from among the outputs to run the magnet. The selector could be the clock PIC, or could be a keyboard for live performances.

First Parts Acquired

[Sprague Extralytic 7700uF 50V Capacitor]No luck at the St. Vincent de Paul, just printer cables, no mouse balls, or even xylophones. I continued down the street to ADD Electronics, which I think is an offspring of Addison Electronics from Montreal. Anyway, they have lots of surplus stuff as well as a good selection of new parts. Unfortunately, while they had lots of wire, they didn't have any magnet wire thicker than the thread-like 31 gauge, I need somewhere around 20 gauge. They did have lots of capacitors, I got a surplus one (it's kind of old) rated for 50 volts. I also picked up a pair of bridge rectifiers, rated at 25 amps at 200 volts, with nice metal heat sink cases. They should stay cool, with at most 1 amp at 35 volts going through one of them on the way to the capacitor. I'm planning on using that old 25 volt transformer (35 volts DC) for a few initial experiments. If it isn't powerful enough, I'll get something bigger (ADD had many power transformers, but they cost real money), but for now it's good enough to charge up the capacitor.

[A Pair of Bridge Rectifiers]

[Coil from 1960s Stock Broker's Terminal's Transformer]I spent the rest of the afternoon taking apart a transformer from an old Bunker-Ramo 1960s stock broker's terminal (7 by 5 character CRT display), looking for magnet wire. It took the whole afternoon to break apart the shellacked together ("lac" is a resin) iron plates (about 30 to 50 of them, interleaved E and bar shapes) to get the windings off. I'm going to test the windings, maybe I can use them as-is, but they're probably too large (golf ball sized balls). Now, which wires go together? And anyone got a use for a pile of transformer plates, plus a nice big laquered copper cylinder that fits around the whole thing?

Coil Size Calculations

Time for a few calculations. What wire size do I need and how much of it?

Coil Resistance

The coil will be operating 1/3 of the time, on for the 2 cm length of the coil (as the 1cm diameter ball approaches the center of the coil) and off for 4 cm (as the ball leaves, slightly faster). So, running at full power, this burst could use up 3 times the energy available (assuming that the transformer recharges the capacitor the other 2/3 of the time). But since I want it to survive a failure with a 100% duty cycle (in case the control circuits fail, or I'm doing manual testing), I'll just use 25 watts (far short of the 4000 watts calculated earlier, maybe not enough to even do anything, but at least it won't blow up spectacularly). Lower power also means that the capacitor voltage won't drop so much and thus the acceleration of the balls will be more constant (rather than being heavily influenced by the AC power cycle timing). At 37 volts DC, with 25 watts, a resistance of V²/P = 37 * 37 / 25 = 54.76Ω will work. The current will be V/R = 37/54.76 = 0.676 amps.

Wire Thickness

How thick should the wire be? The coil design web page has equations for the temperature increase due to a certain amount of current in a certain wire diameter. We want to keep it cool, so I'll say 1°C per second temperature increase with no air cooling in case air cooling isn't enough, then I'll have time to turn things off before it burns up. It will actually be putting out around 25/3 = 8.3 watts of heat when running. I guess that really depends on the air flow. More heat energy is removed either by letting it get hotter (limited by insulation melting point) or by making the coil have more surface area (make it longer). Of course, longer coils will require more space between balls, which isn't good. Instead, I could have several smaller lower powered coils, though this would require more control circuitry but would run cooler. With a pulse width of 1 second (assuming it has failed, otherwise it would be 1/3 second), 0.676 amps of current, and a delta temperature of 1, their equation suggests a copper wire diameter of 0.00975 inches. The wire chart lists 30 gauge wire with a diameter of 0.0100 inches (0.0109 with insulation). Hmmm, not that far off from the available 31 gauge wire which is 0.0089 inches (0.0098 with insulation) in diameter. So, if we use 31 gauge wire, reversing the equations, it's temperature will rise 1.44 °C per second. 31 gauge wire has a resistance of 130.1Ω per 1000 feet, so for 54.76Ω, we need 421 feet.

Coil Physical Size

What will the coil look like? The coil parameters page will help with finding the size of the coil, given an inside diameter of 1.2cm (the ball is 1cm thick, plus the cardboard coil tube, plus spacing), the coil length of 2cm and the 421 foot length of wire which is 0.0098 inches thick. Each layer of coil is 2cm wide (0.8 inches), which means there are 0.8/0.0098 = 81 turns per layer. The length 421 feet * 12 inches/foot = 2 * π * total turns * average radius. Total turns is approximately 81 * number of layers. Outer radius is inner + number of layers * 0.0098. Inner radius is 1.2cm/2 = 0.236 inches. Average radius is (0.236 + Layers * 0.0098) / 2. So, 421 * 12 = 2 * π * (81 * Layers) * (0.236 + 0.236 + Layers * 0.0098) / 2. 5052 = 120 * Layers + 2.49 * Layers². 27 layers is the result. That's 27 * 0.0098 = 0.265 inches thick, so the outside radius is 0.501 inches, making the total diameter one inch. That's a coil thicker than it is long. Anyway, the length of the wire is what matters, the actual coil shape will depend on the size of the balls and length, possibly a bit longer than 2cm (which would also give a longer acceleration distance for the ball and fewer thermal problems).

Ball Bearings Purchased

[Plastic Baggies of Ball Bearings with Bill and GBS Paper Bag]

As you can tell from the picture above, I finally got around to buying the ball bearings, 100 of them. They are 7/16 inch (1.11 cm) diameter steel balls, mass 6g each (a lot lighter than the 30g I'd calculated, due to using diameter rather than radius, though that means they'll work better with my lower power system), bought from General Bearing Services in Ottawa, cost $0.48 Canadian each, plus taxes. They are FAG Steel Balls, which turns out to be the name of a German factory Fischer AG, apparently the first ones to make large quantities of uniform steel balls in 1883.

[Coil Core and Spool of Wire]

Wire and Coil Core Purchased

On Saturday March 22 2002, I went to Canadian Tire looking for a coil core. I found some flexible plastic tubing, but none big enough to let the balls move freely (a closed loop would reduce the problems with balls bouncing around the room). They didn't have magnet wire at all (but if you want thick power cable, they have lots of that for home construction). A look at the plumbing section turned up some plastic pipes and a hose coupling section that was almost the right size, though slightly too large in diameter and with thick walls. Bought that for $0.50. Since the core fit in my pocket, I decided to continue on and walk up the street to ADD-Electronics. There I got the spool of 31 gauge wire I had spotted before (but rejected as being too fine), 500 feet of it for $12.50. The coil components are now ready. But I wasn't, coming down with a migraine headache and nausea, which wrote off the whole afternoon and only permitted TV watching during the evening. But I was fine the next day, though I spent most of it making my own little world with Microsoft Train Simulator - a mountain with 3% grade tracks downhill, kind of a roller coaster for trains.


I had to get cracking on Monday, to have something to show on Tuesday night's meeting.

Power Supply

First I desoldered my old battery charger circuit (which took most of the construction time), keeping only the power transformer, cord, switch and a resistor. The new power supply circuit has the 12.5 or 25 volt secondary (depending on whether you use the center tap or not) feeding into the 25 amp 200 volt bridge rectifier (the metal cube), which is then connected to the output DC bus, which has a 7700uF capacitor and 13KΩ resistor (to discharge the capacitor over a minute or two when the system is switched off) in parallel. I found some nice red and black push-connectors for the output too.

[Power Supply]

Coil Winding

Then I spent a bit of time watching TV and winding the coil. Oddly enough, the resistance of the wire on the spool was 35Ω, I'd expected more like 60Ω, it seemed like there was less than 500 feet. Well, after making some end pieces from the bottoms of some plastic containers, and winding it by hand, I found that the spool actually had two layers of wire. I soldered the break together and continued winding. The resulting coil has probably one or two thousand turns, and a resistance of 70Ω. Looks like I got more wire than advertised.

[Coil made of
500 feet of AWG 31 wire for 70 ohms]

First Test - March 25 2002

OK, time to fire it up. First with the system at 18 volts DC. Nothing much happens, except that the capacitor discharges faster when the coil is connected. A bit of soldering and it's running at 4 times the power - 37 volts DC. The coil draws power, dropping the level down a few volts when it is on. But the metal ball doesn't even move when dropped through the coil or left at its entrance. But the coil is starting to get warm, so something's happening. I tried it out on a screwdriver, and yes, it is actually pulling on it, with an ounce or two of force. The inside of the coil is noticeably hot after a quarter minute of running (fortunately I didn't run it long enough to melt the core). Oh well, back to the drawing board. I figure that less resistance and a smaller coil diameter needed. But first I need a new core tube, that's just the right size. Maybe KenR can make me one from wood - since he doesn't have a lathe (yet), I'll ask him if he can drill out a dowel to make something with with a slightly larger than 7/16 inch inside diameter, and maybe 1/2 inch outside. It would also need end caps 2 inches in diameter, spaced 1.25 inches apart. Fortunately wood doesn't melt. If it gets real hot, a ceramic core could work, though I'd expect the enamel insulation would melt before then.

Attraction Distance

[Tube with 4 Coils]It's now April 25 2002, and I finally got around to trying out the steel can around the coil and the steel disk sides. They make a little bit of a difference, but not much. The ball can now be held by the field when the core is a few degrees from horizontal. The end of the screwdriver through the core grabs metal things quite strongly, and it pops up to the center of the core quite nicely when power is applied. I tried a metal cylinder inside the core, and it held the ball even when the coil was vertical, but only at its ends - the ball wouldn't go through (essentially the same as a solid thing through the core like the screwdriver).

The one useful thing I found out was that the strongest attraction to the ball was from the entrance of the coil to near the center. There isn't much attraction even a slight distance from outside the entrance. This implies that the power control photo sensor has to keep the power on when the ball is past the sensor, until it is near the center. That means it requires a timing circuit, or a sensor inside the coil.

I also unwrapped the Radio Shack wire collection, and found that the spool of 200 feet of 30 gauge wire has a resistance of 20Ω. That's 68 watts continuous power (at 1.85 amps), about right for the 25 watt power supply if a 1/3 duty cycle is used. I'll be winding it on the 1/2 inch inside diameter copper pipe, with steel disc ends, so it will be smaller diameter, be able to run at higher temperatures (unlike the previous plastic discs and core), and cool down better due to conduction along the pipe. But first I have to get a new soldering iron.

I envision a 1m long copper pipe with around half a dozen (quantity depending on how wimpy they are) coils wound on it. Besides the cooling advantage, the copper pipe also keeps the inside dark, making it easier for the photo sensors to work. I'd put little LEDs on each coil so that you could see them firing, and thus trace the movement of the ball. The long tube also acts as an excellent ball guide, so that they'll come out in a consistent direction.

Newer, Stronger, Hotter

[Angled view of Coil 2]I got a new soldering iron in the afternoon (they're pretty cheap) and started winding the new coil. This one is 200 feet of 30 gauge wire wound on a 0.5 inch inside diameter, and 5/8 inch outside diameter copper pipe. The tops of a steel can were cut down and used as end-discs. For bonus points, a grommet protects the magnet wire as it goes through the end disc. The first two layers of wire went on evenly (due to the smooth pipe, and pushing the loops together while winding by hand under tension), but the rest were chaotic. If they were all smooth, heat conduction might be better.

Coil #2 Performance

Performance was much better than before. It could even hold the ball while the tube was vertical. Placing the ball at the entrance and giving it a short burst of power (by tapping the big red button) reliably moves the ball upwards by 2 cm, sometimes a bit more. Not a lot, but enough to be useful. It looks like I'll need to use the stack of coils technique. There is also a decent force on the ball when you push on it with a plastic straw (a good technique for finding the starting position, which is at the coil entrance). Of course, lots more force is felt when moving a much larger steel drill bit through the core.

Heat Problem

The drawback is that the coil gets very hot quickly, after about 5 seconds it will be hot, 10 will make it too hot to touch and it will likely fail after that (due to insulation melting, electrical tape coming undone etc). The copper core does help a lot by conducting the heat out. When it is continuously on, the power supply voltage drops to 27 volts, implying that it is using 50 watts. That little power supply is twice as powerful as I thought. I tried simulating normal operation by tapping the button rapidly, giving it a less than 50% duty cycle (voltage dropped to 30), but it still got too hot, too quickly.

To avoid meltdown, there are several things I could do. Perhaps fins soldered onto the long copper pipe for cooling. Definitely a lower duty cycle (not hundreds of balls a second, more like one per second). Shorter pulses could help, if it turns out that my manual button pushing was too slow.

The next thing to do is to automate it by having a photo sensor trigger a time delayed electronic switch. That means some electronics design (mostly guesswork on my part since I don't have much experience there), and finding a suitable power transistor. Then I'll see if the shorter pulse will work, and might even be able to make a small ball chute loop, if it only loses 2cm in height on each pass.

[Big Red Button (formerly for Commodore 64 reset), Coil #2 Frontal View, Size Reference 9V Battery]

$Id: index.html,v 1.15 2003/07/22 19:57:20 agmsmith Exp $