Overview
Stylus is a cutting-edge medical device designed to evaluate movement disorders through precise sensor technology. I was responsible for the firmware development, creation of calibration tools, and testing of the firmware, ensuring a robust and reliable solution from start to finish.
Interactive Stylus model


My Contributions
Firmware Development: Designed, implemented, and tested firmware to integrate sensor data from an accelerometer, magnetometer, gyroscope, and force sensor. The firmware ensures accurate, reliable data collection.
Bluetooth Low Energy (BLE) Integration: Developed robust BLE protocols for seamless data transfer to an iPad, delivering an intuitive user experience for healthcare providers.
Embedded Systems Expertise: Leveraged a deep understanding of embedded systems to meet strict medical device standards.
Precision Data Handling: Ensured high reliability in real-time data acquisition and communication, critical for assessing movement disorders accurately.
Why Calibration Is Critical for Good Data
- Heading accuracy — A 50 mG (≈5 µT) hard-iron offset can introduce more than 10° of heading error at mid-latitudes.
- Sensor fusion — IMU filters (Madgwick, Mahony, EKF, etc.) assume a constant magnetic-field magnitude; uncalibrated bias corrupts not only yaw but also roll and pitch estimates.
- Repeatability & traceability — For overlaying data on maps, comparing devices, or detecting anomalies you need readings in absolute Earth-field units.
- Drift compensation — Bias varies with temperature and supply voltage; regular calibration (or live monitoring) keeps measurements trustworthy.
The three coloured clouds you see at first are raw magnetometer readings. Each cloud sits inside its own sphere because the sensor is still carrying hard-iron offsets and soft-iron distortion—so every orientation of the device reports a slightly different “magnetic field length” and a different centre point.
Click the Calibrate button to run the calibration routine. Watch as the three clouds collapse into one perfectly centred sphere: that’s the sensor after both the offsets and the scale/skew errors have been removed. Once all the points share the same origin and radius, the magnetometer can deliver accurate, repeatable heading data to the rest of the system.
Below are some 3D renders of the Stylus that I created in Blender.



