I had also seen Wintergatan's then recently released marble machine video. The machine is magnificent. I watched all of the prologue videos documenting the process of building the machine, and I was inspired. This and the stepper motor ideas were all floating around in my head at the same time, and eventually the ideas collided. I decided to build a musical instrument.
My initial concept was to have one motor for each note, and each motor would have a spool wrapped with thread. On the end of the thread is a weight. When the note is played, the weight is released and would pull down on the spool, spinning it, and thus spinning the motor. The higher notes would have bigger weights so they would spin the motors faster and generate a higher pitch, and the low notes would have smaller weights.
The other issues with my design were that I would need some sort of rewind to pull the weights back up when they weren't being played, so I wouldn't eventually run out of string. Or alternately, just make the instrument very tall so that the notes do not need to be rewound. Either way, I think it would have been very messy.
So I fell back on my plan B: a series of spinning disks that each stepper motor can engage with, where each disk is sized so that they move at a different speed on the outside edge such that they backdrive the stepper motors at the correct speeds to generate pitches that are in tune with each other.
I considered cutting the shape out of a solid block by hand or on a giant lathe if I could find one, but no lathe I could ever hope to get access to would have been big enough. Instead, I decided to make each disk a separate piece. This ensures that the width of each section of the cone is constant, and it also allows me to replace an individual disk from the stack if it gets damaged instead of having to remake the entire thing.
I sent some emails around the engineering department on campus, pitching my project, and eventually was allowed to use a CNC to cut the disks and later also a machine shop for other parts of the project, although it felt a little reluctant. I think people are protective of their tools, and are skeptical of independent student projects that aren't affiliated with engineering classes.
The CNC was very important, because it would be able to cut perfect circles that are the perfect sizes, and do 49 of them quickly. It's crucial that the disks are correctly sized relative to each other, or the instrument would be out of tune. I used this note frequency chart to determine the correct size ratios of the disks, and I made this spreadsheet using those values to calculate the final sizes. From there, I laid out the circles on 4x4' sheets in Inventor, and those drawings were used to cut the MDF on the CNC.
For the keys, these tasks included the 98 pegs arranged on the key board to keep the keys aligned, and the corresponding 98 channels cut in the bottom of the keys.
For the levers, I cut 48 lever bases and 48 unique lever arms, all at different angles and lengths. The high C is just a straight piece, so the base and arm are combined into one piece for this note. Each lever pivots on a steel rod with a bearing cut out of a steel tube, ground flat against the wood. These also are held in position by 98 pegs, and are supported by 49 springs. The lever bases and lever arms are connected by 96 brackets cut from sheet metal: one long bracket fastened with four screws on the top of each lever and one short bracket fastened with two screws on the bottom of each lever.
The angle and lengths of each unique lever arm was calculated in this spreadsheet, the same one from the disk stack. To make them the correct sizes, I used a ceiling mounted downward facing projector to project my layout, and traced each arm onto the plywood. This is not as precise as laser cutting or printing a template, but it was sufficient. I would be able to make slight adjustments to the lengths of each arm when mounting the upper lever arms to the lower lever arms.
| || |
Getting the unique lower lever arms all installed correctly was also very time consuming. They all had to be straight and aligned so the motors properly engage with the disks and so they don't rub against adjacent levers in any of many possible ways.
Results and Persisting Problems
- Occasionally, at the bend in the lever arms, one of the wires may slip off of the arm, which would cause the arm to catch on the adjacent arm when the key is played. This could be easily fixed with more glue, but I have not yet done this.
- Sometimes the rubber wheels on the stepper motors will slip off when being played.
- The disk stack isn't built perfectly; it actually wobbles a little bit off axis. Because the pitch of any note isn't actually determined by the radius of the disks, but rather the instantaneous radius at the current point of contact, this causes most notes to slightly fluctuate in pitch as the disk stack rotates. This adds an interesting character to the sound.
- The rubber wheels on the stepper motor have a similar issue, where not all of the bores are drilled perfectly centered. This also causes the pitch of the notes to oscillate, but at a much higher frequency than the problem above.
- Especially when spinning at higher speeds, it may be difficult to get the higher notes to fully engage with their disks. You have to push the higher keys quite hard to get full engagement, otherwise the note may go up to a half or a full step flat.
- It is very difficult to get the disk stack spinning to at a speed to reach a specific key so that it can play in tune with other instruments.
- When many notes are played at the same time, the sound of the instrument quickly gets very muddy and cluttered. Some chords are acceptable, but others don't work very well.
Here is some extra video of the instrument being played:
Features Never Implemented
The first is adding a tachometer that shows how fast the disk stack is spinning. I got this really cool vintage tachometer and flexible shaft from eBay.
I came very close to finishing this, but I had two problems. The first was that I could not affix the pulley end of the flexible shaft securely enough that it could handle the belt tension against the disk stack pulley. The other reason is that the alignment between the other end of the flexible shaft and the tachometer had to be extremely precise, or else it would catch periodically or the tachometer would not spin without excessive resistance.
The other big feature I wanted was the ability to drive the disk stack by pedal power rather than by a motor. This would involve building an entire base for the instrument to sit on instead of resting on a table. Inside of that base would be the pedals and some gearing. The disk stack would be disengaged from the motor, and then engaged with the pedals on the new base. Since the disk stack is so heavy, it acts like a flywheel, so I think it would be possible to spin the stack at a pretty constant speed under human power, with the help of the tachometer to monitor the speed.
I didn't build this because I haven't yet had time to and I also have not decided if the quality of the sound of the instrument merits the work of this addition.