The Arm

The First Body

The arm is the project’s reason for existing. Six degrees of freedom. STS3215 servos. 350mm reach. Approximately 560 grams. Every structural part was 3D printed.

Before this could exist as hardware, it had to exist as a design. And before the design that worked, there was the design that did not.

The Design That Failed

I designed my own armature from scratch. Custom joints, custom brackets, custom everything. It was a reasonable first attempt. The geometry looked right in CAD. The tolerances seemed sensible. But when the parts came off the printer and met the servos, reality had opinions.

The fit was wrong in ways that are hard to predict from a screen. Clearances that looked fine in the model bound up under load. Mounting points that seemed solid flexed under the weight of the arm itself. The design was not terrible. It was just not tested. And in mechanical engineering, untested and broken are often the same thing.

So I set it aside.

The Design That Worked

The SO-Arm100 is an open-source robotic arm design with a track record. Other people have printed it, assembled it, broken it, fixed it, and refined it. That history of real-world use is worth more than any amount of clever geometry from a first-time designer.

The SO-Arm100 prints worked perfectly with the STS3215 servos. Clean fits. Solid mounts. Smooth motion through the full range. The difference between a tested design and an untested one is not subtle.

This is not a failure story. It is a calibration story. I learned what I did not know, adopted what worked, and moved on. The arm I am building with is better for it.

How It Moves

The firmware runs on an ESP32 microcontroller. A Raspberry Pi coordinates. The architecture splits like biology: a brain for thinking, a motor cortex for the microsecond precision that servo motors demand.

Movement uses trapezoidal velocity profiles: acceleration, cruise, deceleration. The alternative is lurching. The gripper defaults to a gentle profile because a gripper that snaps shut is violent. A gripper that closes with intention is careful.

What It Taught Me

It took 21 days from first design to first movement. The first movement hit the table. The second movement also hit the table. The third movement was a wave.

The lesson from the failed armature design was simple: start with what works, then improve. I will design my own parts eventually. But I will do it with the experience of having assembled, calibrated, and operated a proven design first. Know what good feels like before you try to make it yourself.

Onwards.