Overview
Capstone project for ETC 423: Microprocessor Interfacing at SUNY Polytechnic Institute, completed December 2025. Built with Rejhan Karajkovic, Admir Munjakovic, and Ater Lay.
The idea came from watching predictive maintenance teams in industrial environments. They carry thermal readers, ultrasonic probes, and various specialized devices, all separately and each requiring their own setup. The portable sensor bay combines those capabilities into a single platform built around a Raspberry Pi 5 and a 7-inch touchscreen. Sensors connect via I2C and are hot-swappable: no rewiring, no reconfiguration between swaps.
Hardware
| Component | Qty | Cost |
|---|---|---|
| Raspberry Pi 5 (16 GB RAM) | 1 | $132.00 |
| Pi Foundation 7" Touchscreen Display | 1 | $79.95 |
| AMG8833 8×8 IR Thermal Camera | 1 | $35.59 |
| Mini Keyboard with Touchpad | 1 | $22.99 |
| Official Raspberry Pi 27W USB-C Power Supply | 1 | $14.04 |
| Raspberry Pi 5 Active Cooler | 1 | $13.50 |
| Adafruit MAX98357A I2S 3W Class D Amplifier | 1 | $5.95 |
| Speaker, 40mm 4Ω 5W | 1 | $4.95 |
| Total | $308.97 |
System Design
Sensor Interface
All sensors communicate over I2C, routed through a dedicated port on the enclosure. Hot-swapping works because I2C supports multiple devices on the same bus; each sensor is addressed individually, so removing one and attaching another requires no changes to the wiring harness or power routing. DuPont connectors handle the connections for now; the plan is to swap them for watertight Deutsch connectors in a future revision.
Audio Feedback
A MAX98357A I2S Class D amplifier drives a 40mm speaker for audible alerts and status cues. I2S was chosen over PWM audio for cleaner output and lower CPU overhead on the Pi.
MAX98357A I2S Amplifier Wiring
| Pin | Connects To |
|---|---|
| VIN | 5V |
| GND | Ground |
| DIN | I2S Data |
| BCLK | I2S Bit Clock |
| LRC | I2S Left/Right Clock |
| SPK+ | Speaker + |
| SPK− | Speaker − |
I2C Sensor Wiring
| Pin | Connects To |
|---|---|
| VDD | 3.3V or 5V |
| SDA | SDA pin |
| GND | Ground |
| SCL | SCL pin |
Thermal Camera Software
The AMG8833 produces an 8×8 grid of temperature readings. The Python script reads the array, maps each value to an ironbow-style RGB color, builds a PIL image, and upscales it to 300×300px for display in a tkinter window. The loop runs every 50ms (~20 FPS).
- Temperature bounds are recalculated each frame so the colormap autoscales to the current range
- Ironbow palette: high temp → red, mid → green, low → blue
Image.NEARESTupscaling keeps the pixel grid sharp rather than blurred- Built with: Python · tkinter · PIL/Pillow · NumPy · adafruit_amg88xx · busio/board
Enclosure
3D-printed in PLA. The enclosure houses the Pi, active cooler, amplifier board, and routes sensor cables to the front-facing port. The design required mid-build modifications to accommodate the heat sink clearance; the first print didn't account for the cooler's height above the board.
Future revision: reprint in TPU for flexibility, shock resistance, and better field durability. Also planning to add cable management channels and make the enclosure watertight.
Demo
Results
The system deployed and performed as intended. Real-time thermal imaging worked reliably through the touchscreen interface, sensor swaps worked without rebooting, and the Wi-Fi + Bluetooth connections held up throughout testing. The AMG8833's 8×8 resolution is low by camera standards, but more than sufficient for detecting heat signatures on machinery surfaces.
Future Work
- Reprint enclosure in TPU for better durability and shock resistance in the field
- Replace DuPont connectors with Deutsch connectors for watertight, secure connections
- Add ultrasonic and vibration sensor modules to the swap library
- Improve the UI with better graphs and customizable display options per sensor type
- Enable wireless data logging to cloud storage for remote monitoring
- Reduce cost by sourcing components in bulk or finding equivalent alternatives
Documentation
Full project report and presentation deck from ETC 423.