Design Rationale¶
Why V3 looks the way it does.
| Decision | Why |
|---|---|
| Start from an Ender 3 | cheap or free donor machines: extrusion, motors, belts and hardware for one build |
| CoreXY instead of roller X/Y | both motors stay on the frame, so the moving head is light: higher speed and less wobble than V2's rollers |
| LumenPnP-style toolhead, CP40 tips | a proven open design; standard, cheap, interchangeable tips |
| Compressor + venturi instead of a diaphragm pump | stronger, steadier vacuum; the valve switches it in milliseconds |
| Cameras instead of precise mechanics | a printed machine can't be built to 0.05 mm; cameras measure where things really are and the software corrects for it |
| Visual homing on a printed datum | sensorless homing alone lands 0.1 - 2 mm off; one camera look fixes it to ~0.01 mm without endstops |
| Board found from its own pads | no fiducials needed on student boards; any distinct pads work |
| Our own software, not OpenPnP | runs headless on a 1 GB Pi next to Klipper, and the touchscreen can drive it; OpenPnP needs a desktop computer |
| Job file, not G-code | placing needs cameras, feeders and retries; the job file stays independent of where the board lands |
| Klipper with a second MCU (Uno) | servos and LEDs on a cheap board; the Octopus endstop ports would not pass LED data |
| Servo feeders (0816) | simple and cheap; one servo per feeder |
| Raspberry Pi 4 instead of 5 | touchscreen compatibility |
| Software emergency stop for now | the touchscreen e-stop stops every motor and output at once; a hardware switch in the 24 V line is planned |