The concertina is a hard instrument to digitise, and the reason is instructive. Its expression lives in the bellows: how hard you push and pull, how you change direction, how you maintain or break pressure across a phrase. The buttons only pick the notes. Everything that makes a phrase yours is in the air.
Xcott Craver built an electronic anglo concertina that attacks exactly that problem, covered by Hackaday this week.
The build
- Load cells measure pressure on the wooden chassis sides — the player still squeezes and pulls, but no air moves. The force is measured directly.
- Cherry MX mechanical switches with custom keycaps, mounted on PCBs, for the button array
- A Teensy 4.0 doing the sensing and sound
- A USB extension lead used as strain relief on the connector, which is the kind of unglamorous detail that separates instruments that survive being played from prototypes
Using load cells rather than an air-pressure sensor is the notable choice. It means the instrument works without a sealed air path, which removes the hardest mechanical part of concertina construction — the bellows themselves are the expensive, skilled, failure-prone component.
Cherry MX is the other pragmatic decision: keyboard switches are cheap, available in a dozen weights, and have a known feel. For an instrument where the buttons aren’t the expressive element, that’s an efficient place to save effort.
The objection worth taking seriously
In the Hackaday comments, an experienced concertina player raised the real issue: they were sceptical of the load-cell approach, because as a player the physical movement itself is so important to the dynamics of the music.
That’s the crux of the whole project, and it’s a distinction that anyone building an electronic version of an acoustic instrument should sit with. A load cell measures force. A bellows involves force and displacement — how far you’ve opened, how much air remains, how the instrument’s resistance changes through the stroke. A player is controlling a continuous physical quantity with inertia and limits, and the phrase is shaped by where in the stroke you are, not just how hard you’re pressing.
The suggested fix in the thread — adding inertial measurement or other sensor data — is right in principle: combine force with position and acceleration and you can reconstruct much more of the gesture.
The deeper point is one this beat keeps returning to. The sensor you choose defines which part of the performance survives digitisation. Pick force alone and you get dynamics without travel. Pick position alone and you get travel without effort. Most digital instruments feel thin next to acoustic ones not because the synthesis is bad but because somebody picked one channel where the real instrument had three.
Building this and then hearing directly from players is how that gets worked out, which is why the comment is as valuable as the build.