I have always been fascinated by the concept of a player piano, perhaps because I have no playing skill of my own. Of course, such a thing would be totally impractical to build in the basement. The mechanical linkages would demand serious carpentry chops; once built the thing would be rather large; and finding perforated paper rolls of music would be challenging. So instead of a player piano, I chose to build a player xylophone.
Well, technically a xylophone is supposed to use tuned wooden bars — xylo comes from the Greek word for wood. Ideally, one would use rosewood or a fine-furniture hardwood. But that’s expensive. I chose to use copper plumbing pipe because it is cheap, easily worked, and gives a nice bell-like tone when struck. So it’s a player glockenspiel, glock being the German word for bell. But glockenspiel is even harder to spell than xylophone, so phooey to that. It’s a xylophone.
Here’s the initial design in broad strokes:
An array of tuned copper pipes spanning two octaves or so — most tunes fit within a two-octave range;
A small solenoid for each pipe, positioned so it just taps its pipe when energized;
A low-cost microcontroller to interpret a stream of MIDI (Musical Instrument Digital Interface) codes and trigger the appropriate solenoid(s).
The device is a simple one, really. The only complexity (that is, the only complexity that I anticipate, going into the project) is that there’s two dozen pipes, solenoids, driver circuits and so on.
You need one pipe for each note you wish to produce. The only tricky bit is getting the pipe to be the right length so it’s in tune with the others.
I’ve cut pipes for almost the full two octaves, from B4 to G6. (The piano's Middle C is C3.) The mounting holes are placed 22.4% of the pipe’s length from each end, which is why there’s a calculator in the photo.
And what, you may ask, is the right length? There’s a formula for transverse vibration:
frequency × length ² = a constant
And what’s the constant? I found a physics journal article with a complete formula that includes the thickness of the pipe wall, the speed of sound in metal, and several mysterious fudge factors. But nuts to all that! What you do is cut one pipe to a convenient size, perhaps half a cubit, and measure its resonant frequency and its length. Plug the length and frequency into the formula to get the constant. Then the same constant will work for all your other notes.
Provided, that is, that they are all cut from the same type of pipe. My local building center carries two types, called Type L and Type M, and Wikipedia mentions two others. The difference seems to be the metal thickness. The moral is, pick one type and stick with it.
Measuring the frequency is best done with a guitar-tuning app. You can probably find a free app for your phone or computer online. I found one for my Macintosh called Talking Tuner, by Hot Paw Productions, that works pretty well. In addition to showing the frequency in Hertz, it tells you whether you are sharp or flat with respect to the nearest standard pitch.
The tuning test jig. Not shown is a notebook computer running a guitar-tuning app.
Set up a spreadsheet with the standard pitches (which you can find here), plus the constant you have worked out, to calculate a bunch of target lengths. Cut each pipe just a bit long, though. Measure its frequency and, if it’s flat, shorten it a bit with a file. (It’s really hard to lengthen a pipe if you cut it too short!) Measure, file, repeat until it’s as close as you want.
Bear in mind that temperature does affect the pitch of a metal pipe, and filing or grinding will heat the metal up a bit.
Mounting the pipes. The optimum position for a support is 22.4% of the pipe’s total length from each end. This is a null point in the pipe’s vibration pattern, so attaching a support at that spot will dampen its tone the least. I “attached” the pipes by resting them on foam-rubber weatherstrips.
The almost-complete musical instrument. (The right-most pipe, a B-flat, hadn't been cut when I took the photo.)
To keep them from sliding out of position, I passed a string through holes drilled at the null spot in each pipe. (I suspected that drilling a hole in a pipe might affect its frequency, so I drilled before I tuned it.) Then I put blobs of hot-melt glue on the string on each side of the pipes.
I found some suitable solenoids at All Electronics. “Suitable” means they are tiny — a mere half-inch in diameter and one inch long — and not too expensive: $2.50 each if you buy 10 or more at a time. (I will confess, however, that this cost was one factor in my decision not to build a three-octave instrument.)
The All Electronics site discloses two problems with this solenoid, though. First, the plunger can fall out the back of the cylinder. Second, there’s no spring to retract the plunger when power is released. The solution to both problems is simply to mount each solenoid vertically below its pipe (let gravity work for you) and put a board underneath.
Running at their nominal 12-volt rating, the solenoids aren’t very powerful. But that’s okay; you don’t really need to strike a copper pipe very hard to get a nice tone. And if need be, you can run the solenoid at higher voltages, because the power is applied in very short pulses and so the solenoid has lots of time to cool down between notes. I found both a 12-volt and an 18-volt supply in the junk box; I’ll do some tests with both.
Test jig to fire a solenoid with various voltages and pulse durations.
At 12 volts, each solenoid draws about 300 milliamps. (At 18 volts, it’s 450 mA.) The Arduino controller puts out 5 volts and can safely supply at most 10 milliamps per pin. Thus, a power transistor is required for each solenoid. All Electronics was selling beefy TIP112 transistors for 50¢ each, so that’s what I used. If power mosfets had been cheaper when I was shopping, I’d have used them instead.
The solenoids are mounted in holes drilled in a sheet of polycarbonate plastic. Each solenoid is wired to its power transistor mounted on perfboard. Full info on the driver circuit is in this Adafruit tutorial.
During testing, I found that striking a pipe with the solenoid’s bare metal plunger gave an annoying clank. So I put a blob of hot-melt glue on the tip of the plunger; it cooled to a rubbery ball that improved the tone a lot.
You’d be ill-advised to imitate my design, but I’m happy to share some notes about what I did and why.
This view shows the height-adjusting screws, which allow placing the solenoids at the proper distance below their pipes. The proper distance will be determined by adjusting the screws, then firing the solenoids and listening to the resulting tones.
The array of solenoids, awaiting installation of the pipes.
With the solenoids and pipes in place, it was time to write some software for the Arduino Mega controller. (It would certainly be possible to use a different device: a Raspberry Pi, perhaps. But my local computer store frequently offers Arduino clones at deep discount.) The first test script simply ran up and down the scales to make sure that each note was being sounded.
The next test was to code up a few simple tunes. I picked three rounds — Row Row Row Your Boat, Dona Nobis Pacem, and White Coral Bells — because those would also verify that multipe notes can be played at once.
A recording of Dona Nobis Pacem, a three-part round. (Click to play movie.)
This test quickly revealed that the system was working. Oh, the pipes weren’t perfectly in tune, but close enough for now. And one pipe was giving a distinct “clank” when struck; probably that solenoid had lost its hot-melt blob. But there was one serious flaw: All the solenoids were adding a noticeable “thump” sound to each note.
This will have to be investigated and fixed somehow. More to come….
Let’s suppose that the initial problems can be overcome somehow. There’s a couple of possible directions to extend the device:
MIDI. Right now, tunes for the xylophone have to be hand-coded (tedious!) and stored in the controller’s flash memory. An interesting alternative would be to give the controller a MIDI interpreter, because there are lots of MIDI tunes freely available from the Internet.
Fortunately, (FortySevenEffects)[http://fortyseveneffects.com/] has already written a (MIDI code library)[https://github.com/FortySevenEffects/arduinomidilibrary] that does most of the work.
Robot-human duets. The xylophone pipes are being struck from the bottom, and a human player can tap the pipes from the top at the same time. The robot could be programmed as a continuo plus rythm section, upon which the human could improvise ad libitum.
Add more sounds. The controller has 54 output pins, and the xylophone uses only 26. Other pins might be assigned to external devices such as drums.
Back to home page.