The premise: instead of one driver per ear trying to reproduce the whole audible spectrum, use several smaller drivers each responsible for a narrower band, with a crossover routing each frequency range to its own driver. Standard idea for speakers; uncommon at headphone scale. I wanted to see if it would work.
The build
Inside each earcup, several drivers salvaged from broken headphones I’d collected, each fed through its own op-amp-based filter — high-pass, band-pass, low-pass — with hand-picked capacitor and resistor values. The enclosure was the weird part. I didn’t have the means to machine or 3D-print a proper earcup, so I used disposable earbud shells as a skeleton, wrapped them in rope, and poured glue over the rope in layers until it set into a rigid outer shell. It looked exactly like what it was: a boarding-school experiment with available materials.
What I learned
They sounded OK — not great. Over-multiplying drivers at small scale caused the same interference issue I’d later see on the Keynote V1 speakers: overlapping frequency bands cancelling each other at the listener’s ear. That was a real physics lesson I didn’t get from any textbook. Also: rope-and-glue is not a good enclosure material, but it’s an educational one.
The project wasn’t a product, wasn’t really a success as audio, and I’m not sure where the actual headphones are anymore. But they’re one of the first builds where I designed a full signal chain from principles rather than copying a schematic.