2026-03-27
TL;DR: I built out the remainder of the second octave, sanded down the solenoids a little, and took the system for a more challenging test drive. There are still some issues, but I think the system is really coming along.
The first thing I did was to snap off the wide parts of the plunger hats so they wouldn’t get stuck on the solenoid wires. This wasn’t too difficult because the hats are FDM printed, so I just had to overcome the adhesion between two layers. The “brims” of the hats weren’t even necessary anyway because the parts that go around the steel are wide enough to stop the plungers from falling through. I modified the CAD model for this part to remove the brims, so future plungers won’t look as rough.
Last time I had some problems with keys getting stuck, so I sanded down the inside walls of the solenoids as well as some of the plunger extensions. This helped a lot, and now stuck keys are pretty uncommon. I still might try to add some kind of spring arrangement to provide a small return force for the plungers, but for now this is good enough. It seems like the biggest cause of stuck keys at this point is poorly-aligned plungers, which are relatively easily fixed by snapping off the existing extension, reaming out the hole in the steel, and gluing in another extension with better alignment.
I salvaged a few diodes from previous experiments and installed them on the most recent controller board. This doesn’t look as nice as it would if I’d used new parts, but it seems to work just as well, so I don’t really care. I did have to debug a few cold solder joints, but that was a result of my own skill more than anything.
After I got everything set up, I ran through my usual chromatic scale tests and everything was alright. At some point I realized that I could just pull some MIDI from the web and feed it in to see what would happen, so I grabbed a file almost at random and played it. The result is in the video above. I deliberately chose a piece in the key of F because I thought that would have the best chance of sounding okay given my machine’s limited range. It was surprising to see the machine, which had until that point been a mostly soulless robot, suddenly trying to reproduce real music. I let it run through the whole movement (about 10 minutes), and after it was finished I touched some of the solenoids and noticed that they were only slightly warm.
While the machine was running, I saw that the quietest notes weren’t sounding at all because the solenoids weren’t pushing hard enough. The plungers in those cases would move down slightly but get stuck after engaging more of the internal key mechanism. Next time I want to do much more to understand the keys and what force profile they need so I can map MIDI velocities more effectively. One approach I’ve seen is to start every pressing sequence with a short “kick” where the solenoid is on at full power for maybe 10 milliseconds. This is meant to give the key enough vertical speed to overcome the mechanism. I would like to try to measure the position of the key over time during many different attempts, but I haven’t come up with a good way to do it yet. In the worst case I suppose I could just resort to trial and error.
I intend to keep the system at this scale for a while so I can dial in the motion model and anything else that needs adjustment. Then maybe over the summer I’ll continue expanding. Eventually I should also improve my solutions for power delivery and circuit board mounting.