This is one big catch-up log covering nine recordings. I worked on the initial design, rebuilt parts of the belt assembly, chose and mounted the encoders, dealt with the wiring, and started learning how to control the system with a state-space model and LQR.

Starting the CAD

I began by researching different mechanical options while building the model in Onshape. Before I could get far into the full assembly, I needed to settle details such as the shaft, bearings, supports, and how everything would fit together.The first recording was only 57 minutes because recording my screen was making everything lag. In the next session, I continued developing the pivot assembly. The shaft passes through a bearing mounted in a shaped support plate, with another support behind it.The following 2h 52m session continued the CAD work. I spent more time on the plates, their hole layouts, and the surrounding geometry. At this stage, I was still developing individual parts rather than working with a finished mechanism.

Reworking the belt assembly

I spent another 38 minutes remaking the belt-related parts. I moved between editing the plate and checking how everything fit in the full assembly. This made it easier to see how the rail, carriage, shaft supports, and belt system would work together.This is why the update contains several separate CAD sessions. I kept returning to the same mechanism as the design developed: first the pivot, then the plates and belt assembly, and eventually the encoder mounts and smaller hardware details.Choosing the encodersThe next problem was deciding which encoders to use. I researched different options and mounting arrangements, including AMT-style encoders, before deciding to use the encoders I already had. That took another 1h 1m.Choosing the encoder was only half of the problem. I still had to determine how to mount it and connect everything reliably.

Encoder wiring pain

The appropriately named “Encoder Wiring Pain” session lasted 3h 40m. I worked through the encoder connections and looked at mounting references while figuring out how the sensor would fit into the rest of the mechanism.After that, I spent another 1h 52m making smaller mechanical changes in Onshape. By this point I was working on details around the encoder housing, connector, fasteners, and nearby parts.These changes were individually small, but they mattered because the sensor needed to fit around the existing hardware without interfering with the moving assembly.Learning state space and LQR

The longest of the first eight sessions was 5h 19m. I spent it coding, learning about state-space control and LQR, and continuing some of the CAD work.I worked through notes for a state-space model and LQR controller for a cart double-pendulum system. This included defining the system’s state vector and understanding how the state-space equations describe its motion.This part was mainly development and learning. I haven’t established a successful balancing test from these recordings yet, so the controller still needs to be implemented, tuned, and tested on the real mechanism.Hardware work and final CAD changes

The last recording covers hands-on hardware work alongside a few final CAD changes. I worked on the wiring at my desk while using references such as the Raspberry Pi Pico pinout.Again, this recording ran much longer than the actual work. Please count it as roughly 4–5 hours rather than the full 7h 41m shown by Lapse.