IoT-Based Cold Storage Monitoring Solution for Agri + Pharma Using EFM32HG210

The cold chain worldwide for food and drugs is continuously threatened. Based on industry statistics, when temperatures are exceeded while products are stored, 20% of those temperature-sensitive products fail. Further, door left open incidents can generate energy loss in commercial refrigeration of up to 35%.

Traditional monitoring systems often rely on manual log sheets or expensive, proprietary hardware that lacks scalability.

This case study presents FrostGuard, a wireless, low-power cold storage monitoring system designed for agri warehouses, pharmacies, and vaccine cold rooms. The prototype combines Silicon Labs Si7021/Si7006 humidity and temperature sensors with EFM32HG210 ultra-low-power MCUs and Si4438 sub-GHz wireless transceivers.

Four essential parameters are tracked by the system - ambient temperature; relative humidity; status of the door (open or closed); and spoilage risk trend. The data is sent to a cloud platform where it is then displayed via a mobile application developed using .NET MAUI Cross-Platform Framework.

The unique value proposition lies in its predictive spoilage alerts, calculating dew point to warn users of condensation risks before product degradation begins, and its compliance-ready logging for FDA 21 CFR Part 11 and similar regulatory frameworks.

Market Gap

Existing solutions for cold storage are largely either wired expensive building management systems that cannot be used by small farms or pharmacies, or simple Bluetooth temperature loggers requiring an actual visit to retrieve the temperature data and not allowing real-time alerting.

Our Solution:

This system bridges that gap by offering an affordable, multi-parameter wireless mesh specifically optimized for battery-powered deployment. Unlike competitors that treat temperature and door monitoring as separate systems, FrostGuard integrates both on a single hardware platform.

Moreover, the system provides predictive alerts by using the dew point calculated by the Si7021 to alert users of impending condensation inside a vaccine refrigerator or grain silo based on temperature and humidity trends.

The Microcontroller EFM32HG210F64-QFN32 (Silicon Labs) is the primary component in our FrostGuard Prototype. Multiple factors contributed to the decision to utilize this microcontroller including:

3.1 Primary Components and Selection Rationale

1. Microcontroller: EFM32HG210F64-QFN32 (Silicon Labs)

• This is an ARM Cortex-M0+ architecture (32-bit) microcontroller, with very low active power consumption (114 µA/MHz) and having a low-power USB interface to program firmware over the USB interface.
• The recommended operating temperature for this microcontroller is between -40°C and 85°C making it qualified to operate in a freezer environment.
• The EFM32HG210F64-QFN32 microcontroller offers 64KB of Flash memory and 8KB of RAM to allow for the polling of sensors, wireless stack, and local buffering of data.

The FrostGuard prototype is designed around three core Silicon Labs components selected for their ultra-low power consumption—critical for multi-year battery operation in cold environments.

2. Humidity & Temperature Sensor: Si7021-A20-GM1R (Silicon Labs)

• Accuracy: ±3% RH (0-80% RH range), ±0.4°C temperature accuracy (-10°C to +85°C).
• Power consumption: 150 µA during conversion, 60 nA standby current—ideal for battery-powered nodes.
• Key feature: Factory-calibrated I²C interface eliminates need for external calibration equipment. The integrated on-chip heater allows the sensor to "dry out" after exposure to condensing environments, preventing drift.
• Alternative option: Si7006-A20-IM1R (industrial grade, -40°C to +125°C) for deep freezer applications.

3. Wireless Transceiver - Si4438 M2X (Silicon Labs)

Sub-GHz vs BLE

Sub-GHz frequency range (425-525 MHz) penetrates walls of freezers and metal doors much better than 2.4 GHz does.

Power

RX current = 14 mA, TX current = 20 mA (13 dBm). Support maximum output power of 20 dBm for longer range.

Sensitivity

-124 dBm (Highly reliable); can receive reliably in RF noisy environments (i.e., commercial kitchens).

4. Door Contact Sensor:

Magnetic reed switch (normally open) connected to GPIO with pull up. When the door opens, the magnet moves away from the reed switch and the MCU detects a rising edge.

3.2 Supporting Components

Component	Part Number
Power Regulator	TPS62743 (TI)
Battery	CR2477 (3V, 1000mAh)
PCB Antenna	Meandered Inverted-F
External Temperature Probe (Optional)	DS18B20

Board Design:

• Measurement: 50mm x 50mm (2 layers of FR4, 1.6mm thick)
• Protection from environmental deterioration: Conformal coating (acrylic) used to reduce the incidence of shorts due to moisture and condensation.
• Antenna: An SMA external connector/feed with a quarter wave whip (16.4 cm long for 433 MHz) for maximum range in areas with lots of metal, such as walk-in freezers.

Sensing Plan:

Prototype has 2 different ways to measure temperature:

• Ambient (Si7021 on PCB): Measures the temperature of the air inside the cold storage unit
• External probe (optional): Wired via 3.5mm audio jack and measures the temperature of the core of the product (important for storing vaccines, because the difference between the air temperature and the product surface temperature could be significant).

Power Management Circuit:

A TPS62743 step-down converter maintains a stable 1.8V supply for the Si7021, which improves measurement accuracy across the battery's discharge curve (3.0V down to 1.8V).

Firmware & Logic: Firmware written in Simplicity Studio for Silicon Labs is implemented as a duty cycled state machine:

State Machine

State 1: Deep Sleep (EM2 Mode)

• MCU Current ~ 1.5 µA (RTC is active, RAM is retained)
• User Configurable Time Default = 60 seconds (when in Deep Sleep)

State 2: Measurement Cycle (Wake)

• Wake from sleep.
• Power on Si7021 (wait 10 ms for stabilization).
• Initiate temperature measurement (10 ms conversion time).
• Initiate humidity measurement (12 ms conversion time).
• Read door contact status (digital input).
• Compute dew point using the Magnus formula:

Td = (b * α(T,RH)) / (a - α(T,RH))

where α(T,RH) = (a * T)/(b+T) + ln(RH/100). Constants a = 17.625, b = 243.04°C.

State 3: Decision Logic for each Alert Condition:

Alert Condition 1: If temperature is outside a configured threshold (e.g., temp should be between 2 degrees C & 8 degrees C for vaccines), transmit wirelessly without delay

Alert Condition 2: If the door is open for more than 60 seconds, then transmit alert

Alert Condition 3: Dew Point is ≤ 2 degrees C from ambient temperature, then send condensation risk warning to app. Advise user to lower humidity in the area or check the seals on the refrigerator doors

State 4: Transmit Data

• Packet Format: 32 bytes (ID of node, timestamp, temp, humidity, dew point, door, and battery voltage)
• Transmitter Configuration: Si4438 radio set up for 10 Kbps GFSK 433.92 MHz modulation at 10 Kbps

Estimated for prototype quantities (100 units).

Component	Part Number
MCU	EFM32HG210F64-QFN32
Temp/Humidity Sensor	Si7021-A20-GM1R
Wireless Transceiver	Si4438-C2A-GM
Power Regulator	TPS62743YFPR
Matching Network	0402 caps/inductors (Murata)
Crystal	32 MHz (NX3225SA)
Crystal	32.768 kHz (for RTC)
Reed Switch	MKA-20101
PCB	2-layer, ENIG, 50x50mm
Battery Holder	Keystone 1067 (for CR2477)
SMA Connector	Right-angle (Johnson)
Passive Components	Resistors, capacitors

The user interface is an application that is cross-platform and will be developed using .NET MAUI (Multi-platform App UI) which will run on both Android and iOS devices. We chose this Framework because of its performance like that of a native application, it supports the MVVM architecture and integrates well with the Azure Cloud Service.

7.1 Architecture Overview

Back-End Services (Microsoft Azure):

IoT Hub

Ingestion of telemetry data from the gateway nodes (via MQTT protocol and TLS)

Cosmos DB

Store time-series records of temperature and humidity with TTL (Time-To-Live) policies for compliance retention (e.g. 3 years for the pharma industry)

Azure Functions

A trigger-based alerting engine (e.g. if temperature exceeds 8°C for longer than 5 minutes → send push notification)

SignalR Service

Push updates to dashboards in real time (no polling)

Mobile App Architecture (.NET MAUI):

• Pattern: MVVM (CommunityToolkit.MVVM)
• Navigation: Shell-based routing
• Local Storage: SQLite (for offline logging of critical data, in cold-storage facilities where cellular coverage is poor)
• Push Notifications: Firebase Cloud Messaging (Android) / APNs (iOS), called via Azure Notification Hubs

7.2 Key Features Implemented

Compliance Dashboard:

• The compliance dashboard includes graphs displaying temperature and humidity trends with color-coded alert zones.
• An audit trail log records every parameter change, alert acknowledgment and report generation, including date, time and user attribution.
• All FDA-compliant temperature excursion reports can be generated with a single click by exporting to PDF.

Door Activity Monitoring:

• The system provides real-time monitoring of door open/close events and tracks the duration of each event.
• Alerts occur after the user-defined threshold for "Door left open" (e.g., 2 mins).
• At the end of each shift, the system provides a summary of which doors have not been sealed to prevent unsealed doors from allowing for the possibility of raising the temperature overnight.

3. Spoilage Prediction Engine.

• Calculates the dew point using temperature and humidity data.
• Provides an alert when the dew point is nearing the ambient temperature (indicating that the staff should check door seals or lower humidity levels before visible moisture appears on products).

4. Multi-Site Management.

• Provides a hierarchical view: Portfolio → Facility → Cold Storage Unit → Sensor.
• Provides role-based access by providing technicians the ability to acknowledge alerts, managers the ability to create reports, and administrators the ability to provision new sensors.

The FrostGuard system is based on a star of stars network.

• The sensor nodes (using the Si4438 IC) are set to transmit data every 60 seconds or as soon as they detect an alert.
• The Gateway (using the EFR32MG21 MCU & Si4438 IC combined with the ESP8285 Wi-Fi module) receives and decrypts the sub-GHz packet received from the sensors. Then, it forwards the decrypted packets to the Azure IoT Hub using an Ethernet or Wi-Fi connection.
• The gateway will buffer up to 10,000 readings in local Flash memory until a connection to the internet can be re-established. At that time, the buffered readings will be forwarded to the Azure IoT Hub.

The results of range testing were as follows:

• Line-of-sight (in open air): 500 meters
• Through one concrete wall (commercial/frozen food): 150 meters
• Through two metal doors (walk-in cooler): 75 meters

Regulatory Compliance:

• Compliance with 21 CFR Part 11 by the implementation of audit trails, use of electronic signatures to confirm acknowledgment of alerts (to be accessible via user login), and implementation of Data Integrity.
• Compliance with GDPR by storing user data in Azure data centers located within the EU and maintaining data for the required duration as per data retention regulations.

Security Measures:

• AES-128 encryption for secured transmission of all wireless packets (utilizing Si4438 hardware accelerator).
• TLS 1.2 for all cloud-delivered data (utilizing ESP8285).
• Exceptional Device Authentication by ensuring that every node possesses a factory-original unique ID (to prevent device spoofing).

The FrostGuard prototype exemplifies our ability to develop cost-efficient, enterprise-quality cold chain temperature monitoring systems using Silicon Labs sensors, low power MCUs, wireless connectivity, cloud platforms and .NET MAUI applications.

The system allows organizations to monitor environments in real-time, predict spoilage through analytics, and log compliant data to provide organizations with enhanced product quality, reduced losses, and full visibility into their cold storage and supply chain operations.