UV Wearable Tracker
A wrist-worn sensor that tracks your real-time UV Index exposure and alerts you before your skin burns — designed, built, and tested as part of Engineering Capstone.
UVI 0–11Sensing range
BLEWireless sync
5-dayBattery target
±0.4UV Index accuracy
#GUVA-S12SD#nRF52840#LiPo-60mAh#TPU-Wristband#Bluetooth-LE#FR-4-PCB
Section 01
The UV Shield wrist device prototype demonstrates functional UV Index sensing, wireless data transmission via BLE, and a wearable form factor sized for daily use.
UV INDEX7HIGH ● PROTECT NOW2:34 PM ☀ Partly cloudy
Prototype v1.0
GUVA-S12SD photodetector · nRF52840 BLE SoC · 60 mAh LiPo · polycarbonate housing · TPU wristband
Technical Specifications
UV sensor:GUVA-S12SD (240–370 nm, SiC photodiode)
MCU:Nordic nRF52840 (ARM Cortex-M4, BLE 5.0)
Battery:60 mAh LiPo, USB-C rechargeable
PCB:2-layer FR-4, SMT assembly
Enclosure:Polycarbonate (PC) housing
Wristband:Medical-grade TPU (Shore A 85)
Display:4× RGB LED UV Index indicator
Wireless:BLE to companion smartphone app
Accuracy:±0.4 UV Index units (calibrated)
Target battery life:5 days continuous
Live UV Index Output
7.2
⚠ High — Apply SPF 30+ now
Insert project demo video here — upload to Google Drive or YouTube and embed link
Section 02
01
Identified the problem: people don't know when they've had too much UV
02
Researched sensors, components, and existing wearable UV devices
03
Designed PCB layout, enclosure, and firmware architecture
04
Built and tested partial prototypes; ran calibration and material experiments
The UV Shield is a wrist-worn device I designed and engineered from the ground up to solve a real public health problem: most people have no accurate, real-time way to know when their personal UV exposure has reached a harmful level. Generic weather-app UV Index readings are calculated for an open, cloudless sky at noon — they don't account for shade, reflective surfaces, your body's orientation, or whether you're sitting under an umbrella. My device puts a calibrated UV photodetector on your wrist, continuously measuring the actual UV radiation reaching your skin throughout the day.
The core engineering challenge was translating UV radiation — invisible electromagnetic energy — into a reliable, human-readable number that could trigger a timely alert, all within a package small enough to wear comfortably for five days without charging.
The device integrates a GUVA-S12SD silicon carbide UV photodetector, which converts UV photons into a small analog current. A transimpedance amplifier (TIA) circuit converts that current to a measurable voltage, which is read by the 10-bit ADC on the Nordic nRF52840 Bluetooth Low Energy system-on-chip. Firmware applies a calibration factor I derived experimentally (CF = 2.14) to produce a UV Index value that matches laboratory reference measurements within ±0.4 UV Index units. The BLE module broadcasts UV data every second to a paired smartphone app, where users can set personal exposure thresholds based on their skin type.
For the Step 4 detailed design phase, I conducted two formal experiments: one validating the sensor's accuracy and linearity across UV Index levels 1, 5, and 10 using a UV-A lamp at calibrated distances, and one testing three candidate wristband materials (silicone, TPU, and nylon-silicone) for dimensional stability after artificial sweat exposure and subjective comfort during wear. The experiments confirmed TPU as the optimal wristband material and produced the calibration correction factor now embedded in the device firmware.
Section 03
Every major concept from the unit is applied directly in the UV Shield. Here's how each one shaped a real design decision.
Electromagnetic Spectrum
UV light occupies wavelengths of 100–400 nm — just beyond visible violet. The GUVA-S12SD is tuned to detect UVA (315–400 nm) and UVB (280–315 nm), the two bands that penetrate the atmosphere and damage skin DNA. The WHO UV Index is a weighted integral of these wavelengths using the erythemal action spectrum, which maps to biological harm at each wavelength. In the project: choosing a sensor with the correct spectral response (not just any photodetector) was essential to measuring meaningful health risk rather than arbitrary light intensity.
Electromagnetic radiation
Circuits & Electronics
UV photons with sufficient energy eject electrons from the SiC semiconductor layer in the GUVA-S12SD (Einstein's photoelectric effect). This generates a tiny photocurrent proportional to UV irradiance. A transimpedance amplifier (op-amp with feedback resistor) converts the current to a measurable voltage: V_out = I_photo × R_feedback. In the project: I had to account for the TIA gain when deriving the calibration factor — the raw Arduino formula underestimated UV Index by 2.14× because the breakout board's resistor value differs from the one assumed in the datasheet formula.
Photoelectric effect · Op-amps
Digital Systems
The nRF52840 MCU's 12-bit ADC samples the TIA output voltage 1× per second and maps it to a digital integer (0–4095). The resolution is V_ref / 2^12 = 3.3V / 4096 ≈ 0.0008V per step. In the project: I chose 12-bit ADC resolution over the Arduino's 10-bit (≈0.005V/step) to reduce quantization error at low UV Index levels (near UVI 1–2), where accurate readings matter for Vitamin D monitoring. Understanding the Nyquist theorem also guided the 1 Hz sampling rate — UV changes slowly enough that this avoids aliasing while conserving battery.
ADC · Digital signal processing
Wireless Systems
BLE operates in the 2.4 GHz ISM band using frequency-hopping spread spectrum (FHSS) across 40 channels. The nRF52840 advertises UV data packets at 1-second intervals; the smartphone central device connects and receives notifications. In the project: BLE was chosen over Wi-Fi specifically because its average current draw (~7 mA active vs. ~170 mA for Wi-Fi) allows a 60 mAh battery to achieve the 5-day target. The tradeoff is a 10–30 meter range limit, which is more than sufficient for wrist-to-pocket BLE pairing.
RF communication · FHSS
Materials Science
The wristband material must balance flexibility (low Shore A hardness), sweat chemical resistance, and skin biocompatibility. Shore A hardness measures a polymer's resistance to elastic indentation: lower values = softer. Silicone (Shore A 50) is very soft but has high surface friction and absorbs slightly more water. TPU (Shore A 85) is stiffer but shows superior dimensional stability (only +0.33 mm width change after 60-min sweat soak vs. +1.50 mm for nylon-silicone). In the project: experimental data from Experiment 2 directly drove the TPU material selection for the prototype wristband.
Materials selection · Polymer science
Energy Systems
The LiPo battery stores energy electrochemically (lithium-ion intercalation) at a nominal 3.7V. Energy capacity = V × Ah = 3.7V × 0.060Ah = 0.222 Wh. The nRF52840 in connected-mode with BLE active draws ≈4 mA average; total runtime = 60 mAh / 4 mA = 15 hours. To reach 5 days (120 hours), firmware uses aggressive sleep modes (System OFF: <1 µA) between 1-second measurement cycles, reducing average current to ≈0.5 mA. In the project: every hardware component was evaluated for its sleep-mode current to validate the 5-day target before ordering parts.
Electrochemistry · Power management
Biology / Health ScienceThe erythemal action spectrum (how each UV wavelength causes sunburn per joule) informed how I weighted UVA vs. UVB sensitivity in the sensor selection. Understanding skin cancer biology motivated the entire project problem statement.
ChemistryLCA research required understanding lithium-ion battery electrochemistry (cathode intercalation), PCB wet-etching reactions (cupric chloride), and the toxicology of cobalt and NMP solvent — all applied in the materials selection and sustainability analysis.
Statistics / MathBoth experiments required calculating mean, standard deviation, and percent error across 3 trials. The R² value from the sensor calibration regression (R² = 0.97) was used to statistically confirm the sensor's linearity — a math concept applied directly to engineering validation.
Environmental ScienceThe full Life Cycle Assessment (LCA) applied ISO 14040 methodology to quantify greenhouse gas emissions, conflict mineral sourcing risks (cobalt, tantalum), and e-waste regulations (EPA RCRA universal waste rule) across the device's complete lifecycle.
Section 04
An honest evaluation of what went well, what I'd do differently, and what this project taught me about myself as an engineer.
Peaks · What went well
I learned how to derive a calibration factor from scratch
Before this project, I assumed sensors were plug-and-play — you connect them and the reading is correct. Running Experiment 1 taught me that real sensors almost always require calibration. When my GUVA-S12SD was reading 2.14× lower than the reference radiometer, I didn't panic or order a different sensor. I analyzed the circuit, understood why the TIA gain was different from the datasheet assumption, and derived a correction factor from my own data. That process — observe, hypothesize, measure, derive — is real engineering, and I'm proud I worked through it without just copying someone else's calibration value.
Project example
CF = Reference UVI mean ÷ Calculated UVI mean = (1.2 + 4.83 + 9.8) / 3 ÷ (0.43 + 1.62 + 4.58) / 3 = 5.28 / 2.21 = 2.14. This factor is now hard-coded in firmware.
I made data-driven material decisions, not gut-feeling ones
When I chose TPU for the wristband, I didn't just pick it because it looked good or because another wearable uses it. I ran a controlled experiment with three candidate materials, measured dimensional change with digital calipers to 0.01 mm precision, collected comfort ratings from five subjects over three time points, and used the combined data to justify the decision. This gave me confidence in the choice and also taught me how to design a test that answers a specific engineering question rather than just "trying things" and guessing.
Project example
TPU: +0.33 mm swelling (1.3%), comfort 4.2/5. Silicone: +0.80 mm (3.2%), comfort 3.8/5. Nylon-silicone: +1.50 mm (6.0%), comfort 2.9/5 — clear winner.
Pits · What I'd improve
I underestimated how long component sourcing takes
I originally planned to have a working breadboard prototype by Week 3 of Step 4. In reality, the GUVA-S12SD breakout board took 11 days to arrive, which compressed my experiment timeline significantly. As a result, I could only conduct 3 trials per condition instead of the 5 I'd planned, which reduced my statistical confidence — especially in the comfort study (n=5 subjects). Going forward, I will always order components 3 weeks before I need them and maintain a small inventory of common parts (resistors, capacitors, breakout boards) so delays don't block experiments.
What I'd change
Order all components during Step 3 (concept selection), not Step 4. Build a parts checklist into the project timeline as a dependency gate.
My experimental setup wasn't controlled enough for Experiment 1
The UV lamp I used for sensor calibration was a consumer blacklight, not a calibrated UV source. Its output varied slightly depending on warm-up time and ambient temperature, introducing systematic error I couldn't fully quantify. I also didn't use a positioning jig, so the sensor's angle to the lamp changed slightly between trials. These uncontrolled variables are why my mean absolute error (±0.4 UVI) is higher than what lab-grade calibration would achieve (±0.1 UVI). For the next iteration, I'll build a simple plywood jig with a fixed sensor holder and use a longer lamp warm-up period (10 minutes rather than 5).
What I'd change
Design a laser-cut acrylic positioning jig. Source a calibrated UV lamp with a NIST-traceable output spec. Document ambient temperature for every trial.