Overview
From January to April of my 2nd year, I interned at Accelovant, a burgeoning figure in the semiconductor industry making precision temperature probes, for my first ever co-op term. There I led 2 major projects: creating a temperature/phosphor decay simulator to test sensors without placing them in a thermally controlled environment, and an optical fiber detection algorithm to very quickly detect colour contrasted circles in microscope images and deliver statistics about the optical fiber bundle (repo). In prototyping the simulator I used Altium for PCB design, Fusion for CAD, and Python PID and VISA for controls - later I manufactured via 3D printing and soldering. OpenCV Python was the core ingredient for fiber detection.
Design
- Designed a thermally & optically controlled LED enclosure in Fusion 360, using a thermoelectric cooler and an Arduino-controlled stepper motor attenuator to maintain constant LED output
- Prototyped an LED current driver PCB in Altium incorporating optical negative feedback and external signal input
- Used Python to implement optical amplitude PID program and PID tuning, wavegen VISA control, and sensor logging
- Used OpenCV library to preprocess images, isolate fibers using different masks, then perform the distance transform and calculate radii + fiber centroids
Results
- Achieved constant simulated temperatures within 0.2deg Celsius over multiple days through effective optical amplitude & temperature feedback
- Successfully integrated wavegen control with main Python program to allow for simulated temperature cycling runs, representative of the full sensor's operating range
- Redesigned existing LED driver PCB for faster, immediate photodiode feedback and much smaller form factor
- Shortened fiber counting time from ~5 minutes to under 2 seconds with increased accuracy through total algorithm overhaul
- Semi-automated the process of updating the company's fiber statistics database
