All Projects
Embedded Systems & IoTCompleted• Ref: WEATHER-IOT-09 • Role: Solo Engineer — Multi-Sensor Fusion, Cloud Pipeline

Smart Weather Station

BME280 + BMP390 + AHT21 Multi-Sensor IoT Weather Node with InfluxDB + Grafana

Platform / MCUESP32 DevKit
ProtocolsI2C / MQTT / HTTP
Offline CapabilitySD Card Local Log
Latency / Speed60s Cloud Push · 1s Local Log
Interactive 360° PBR Model Showcase
IoT Data // PBR m0.87
IoT Data
Subsystem Badges
Smart Weather Node
[ PBR 360° // Drag to inspect • Click Spotlight to toggle beam ]
AXES:+X+Y+ZSPOTLIGHT: OFFFPS: 60PBR: m0.87/r0.25
// PROJECT INITIATION & FIELD OBJECTIVE

The Engineering Challenge & Deployed Solution

01. The Problem Statement

Consumer weather stations use uncalibrated single sensors with poor drift compensation and no open data pipeline for integration into custom analytics workflows.

02. The Engineered Solution

Three sensor ensemble: BME280 (temp/humidity/pressure), BMP390 (high-precision pressure for altitude), and AHT21 (backup humidity). ESP32 averages readings over 10-sample sliding window, applies factory calibration offsets, and pushes structured JSON to InfluxDB 2.0 time-series database every 60 seconds. Grafana dashboard with 7-day trend charts, heat index, and dew point calculations.

// MODULAR SUBSYSTEM BREAKDOWN

Silicon, Firmware & Telemetry Architecture

SUBSYSTEM #01

Triple Sensor Ensemble

BME280 + BMP390 + AHT21 on shared I2C bus; voting algorithm discards outliers.

BUS: SPI / I2C / UARTACTIVE
SUBSYSTEM #02

ESP32 Sensor Fusion

10-sample sliding average + factory offset calibration; computed dew point and heat index.

BUS: SPI / I2C / UARTACTIVE
SUBSYSTEM #03

InfluxDB 2.0 Time Series

Bucket: weather_data; 30-day retention; tagged by location and sensor_id.

BUS: SPI / I2C / UARTACTIVE
SUBSYSTEM #04

Grafana Dashboard

7-day temp/humidity/pressure trend panels; alert rules for storm drop detection (> 3hPa/hr).

BUS: SPI / I2C / UARTACTIVE
// BENCH TESTING & DESIGN ITERATIONS

Hardware Iterations & Workbench Logbook

ITERATION 01 // BREADBOARD PROTOTYPELab Bench

Early Design Flaws & Edge Noise

Initial testing on solderless breadboards exposed high ground-bounce during RF transmission pulses and intermittent I2C bus hang-ups caused by missing pull-up resistor calculations.

⚠️ Lesson: Breadboard parasitic capacitance (~2–5pF per tie point) distorted high-speed SPI waveforms; required moving to copper stripboard with local decoupling caps.
ITERATION 02 // HARDENED SYSTEMField Validated

Firmware DMA Buffers & Hardware Filtering

Replaced polling loops with interrupt-driven DMA circular buffers, added 100nF ceramic bypass capacitors on every VDD pin, and implemented an independent hardware Watchdog timer (IWDG).

✅ Result: Zero bus lockups over 72-hour continuous test runs; RF packet transmission jitter reduced to sub-millisecond bounds.
// EMPIRICAL VERIFICATION & FIELD METRICS

Hardware Test & Field Validation Matrix

Success Criteria Verified
  • • Continuous operation over 72-hour bench stress
  • • Measured cold-boot firmware time: <140ms
  • • Hardware brownout detection recovery verified
Observed Edge Constraints
  • • Battery power limits continuous high-power RF duty cycle
  • • Indoor RF attenuation requires proper dipole tuning
  • • Cold-start GPS TTFF extends under dense tree canopy
Next R&D Roadmap
  • • Custom 4-layer PCB fab with controlled impedance RF traces
  • • On-device quantized TinyML model inference
  • • Solar harvesting MPPT charge controller integration