How to build Qubit: the whole process
43 steps to build Qubit, from buying parts, printing and assembly to software, calibration and training, with the bill of materials, shop links and the official sources.
Qubit is a dual-arm desktop robot made from two SO-100 arms with expressive 16×16 LED eyes, ambient RGB lighting and a Raspberry Pi brain. The repository includes 3D-printable structural parts, a full BOM, a wiring diagram, print settings and Python code for teleoperation, vision, emote animations and a web dashboard.
Bill of materials · bill of materials — per SO-100 arm
| Part | Qty | CN | US |
|---|---|---|---|
| STS3215 Servo (C001, 1/345 gear) · $14 · €13 · ¥97.72 | 6 | ||
| Motor Control Board · $11 · €12 · ¥27 | 1 | ||
| USB-C Cable 2-pack · $7 · €7 · ¥23.90 | 1 | ||
| Power Supply (7.5V DC) · $10 · €13 · ¥22.31 | 1 | ||
| Table Clamp 2-pack · $5 · €8 · ¥7.80 | 1 | ||
| Screwdriver Set · $6 · €10 · ¥14.90 | 1 | ||
| per arm · $123 · €128 · ¥682 | |||
| 2 arms total · $246 · €256 · ¥1,364 |
Buy parts
- Buy Two SO-100 Arms
Each Qubit uses 2× SO-100 arms, so multiply the arm quantities by 2.
- Choose Leader-Arm Servos
The optional SO-101 teleoperation leader arm needs 6× STS3215: 3× C046, 2× C044, and 1× C001. The bundle is available on Alibaba.
- Verify Gears and Power
Verify gear ratios before ordering; the follower arm uses all 1/345 (C001). Each arm’s power supply must output 7.4–7.5V DC at 5A minimum.
- Source the Full-System BOM
Source the remaining components from the full system BOM. The estimated full build cost in the US is about $700–$800, depending on sourcing.
Print and make
- Print structural parts
Print all structural parts on a Bambu Lab A1 using PLA+ or PETG.
- Set print parameters
Set the layer height to 0.2mm, infill to 40% Gyroid, and enable supports for arm joint brackets.
- Choose material for stressed parts
Print the body frame and shoulder brackets in PETG because they take the most torque stress. PLA+ is fine for cosmetic shells.
Assemble
- Connect the Raspberry Pi Peripherals
Connect the Raspberry Pi 5 over USB-A to the left and right arm motor boards and over USB-C to the 5V/5A supply. Connect GPIO/BLE to the 16×16 LED matrix, GPIO PWM to the left and right WS2812B tubes, and CSI/USB to the optional Pi Camera.
- Wire Motor Boards and Servo Bus
Connect each motor board’s USB-C to a dedicated 7.5V DC supply and connect the serial bus to 6× STS3215 servos. Never power the servo boards from the Pi’s USB ports, as this can cause brownouts and servo damage.
Software
- Install Raspberry Pi OS
Prepare Raspberry Pi OS Bookworm 64-bit, Python 3.10+, and Git. Download Raspberry Pi OS from the linked Raspberry Pi website.
- Clone the qubit repository
Clone the qubit repository from GitHub and change into the qubit directory.
- Create and activate a virtual environment
Create a .venv virtual environment with Python 3 and activate it in the current terminal.
- Install project dependencies
Install the dependencies listed in requirements.txt in the activated virtual environment.
- Install teleoperation support
Optionally install lerobot for teleoperation support.
- Create and edit the configuration file
Copy config/config.example.yaml to config/config.yaml and edit the configuration file with nano.
- Run the idle loop at boot
Run the script on the Raspberry Pi to start the idle behavior loop in the background.
- Edit or preview an LED emote
Use the built-in emote editor in the web dashboard, or run emote_preview.py locally to preview emotes.
- Enable Bluetooth
Enable the Bluetooth service.
- Install OpenCV
Install the headless OpenCV Python package.
- Enable the camera
Run `sudo raspi-config`, open Interface Options, and select Camera to enable the camera.
Calibrate and test
- Configure arm servos
Set the serial ports and servo IDs in the configuration file: the left arm uses /dev/ttyUSB0 with IDs 1–6, and the right arm uses /dev/ttyUSB1 with IDs 7–12.
- Configure the LED matrix
Set the LED matrix type to idotmatrix, size to 16, and connection to bluetooth, then replace mac_address with the actual address.
- Configure LED tubes
Set the LED tube left pin to 18, right pin to 19, and number of pixels to 30.
- Configure the camera
The camera is disabled by default; its device is 0 and its resolution is 640×480.
- Map the arm serial ports
Run ls /dev/ttyUSB* to find serial ports, then plug and unplug each arm separately to map them.
- Check servo port
Run `ls /dev/ttyUSB*` to check the port assignment.
- Verify power supply
Verify that the supply is 7.4–7.5V at 5A+ per arm.
- Run servo diagnostics
Run the diagnostics script for all arms.
- Scan servo IDs
Run the scanning tool to find servo IDs.
- Pair the LED matrix
Open `bluetoothctl`, scan for devices, and pair using the LED matrix MAC address.
- Verify the configured MAC
Verify that the MAC address in `config.yaml` matches the LED matrix.
- Recalibrate the left arm
Run the calibration script to recalibrate the left arm.
- Check mechanical binding
Check whether the arm is mechanically bound; first-print tolerances commonly cause this.
- Lower servo polling rate
Set `teleop.hz` to `50` in `config.yaml`.
- Confirm Python and venv
Confirm that Python 3.10 or newer is being used and the virtual environment is active.
Data and training
- Record a right-arm wave
Run the recording script and save the right-arm movement to demos/wave.json.
- Move the arm during recording
Physically move the arm while recording to capture the motion.
- Replay the wave three times
Read demos/wave.json and play the wave demo in a loop three times.
Use it
- Launch the web dashboard
Run dashboard.py to launch the web dashboard.
- Open the local dashboard address
Open http://qubit.local:5000 in a browser.
- Play a single LED emote
Use led_matrix.py to play the single emote named happy. Emotes live in /emotes and describe 16×16 pixel frame sequences.
- Play an LED emote sequence
Play a sequence of LED emotes in the order startup, happy, blink.
- Loop an LED emote indefinitely
Continuously loop the LED emote named sleep.