Architecture Overview

The Multiflexmeter (MFM) is an IoT sensor platform built around the ATmega1284P microcontroller. It uses LoRaWAN for long-range communication to The Things Network (TTN).

System Overview

┌─────────────────────┐     I2C/SMBus      ┌─────────────────────┐
│   Sensor Module     │◄──────────────────►│   ATmega1284P MCU   │
│   (Address 0x36)    │                    │                     │
└─────────────────────┘                    │  ┌───────────────┐  │
                                           │  │  LMIC Stack   │  │
                                           │  └───────┬───────┘  │
                                           └──────────┼──────────┘
                                                      │ SPI
                                                      ▼
                                           ┌──────────────────────┐
                                           │   RFM95 LoRa Radio   │
                                           │      (868 MHz)       │
                                           └──────────┬───────────┘
                                                      │ RF
                                                      ▼
                                           ┌──────────────────────┐
                                           │   LoRaWAN Gateway    │
                                           └──────────┬───────────┘
                                                      │ IP
                                                      ▼
                                           ┌──────────────────────┐
                                           │  The Things Network  │
                                           └──────────────────────┘

Core Components

ATmega1284P MCU

The main microcontroller running the firmware. Features 128KB Flash, 16KB SRAM, and 4KB EEPROM.

RFM95 Radio

LoRa transceiver module operating at 868 MHz (EU868) using the SX1276 chipset.

Sensor Module

External I2C/SMBus sensor module at address 0x36 providing measurement data.

LMIC Stack

IBM LoRaWAN-in-C library handling LoRaWAN protocol, join procedures, and TX/RX scheduling.

Firmware Architecture

The firmware uses a cooperative job scheduling model provided by the LMIC library. There are no interrupts for application logic; all operations are scheduled as jobs.

Execution Flow

setup()
  ├─ Board initialization
  ├─ Sensor I2C init
  ├─ LMIC stack setup
  ├─ EEPROM config load
  └─ LMIC_startJoining() → EV_JOINING event

loop()
  └─ os_runloop_once()  ← Processes scheduled jobs

Event-Driven State Machine

EV_JOINING
  │
  ▼
EV_JOINED
  │
  ├─► job_pingVersion()
  │     └─ Sends HW/FW version on FPort 2
  │
  └─► (45s delay)
        │
        ▼
      job_performMeasurements()
        └─ Triggers sensor via I2C CMD 0x10
              │
              └─► (10s delay)
                    │
                    ▼
                  job_fetchAndSend()
                    ├─ Reads sensor via I2C CMD 0x11
                    ├─ Transmits data on FPort 1
                    └─ scheduleNextMeasurement()
                          │
                          └─► (interval delay) → job_performMeasurements()

Module Structure

Module	File	Purpose
Main	src/main.cpp	LMIC event handling, job scheduling, downlink processing
Sensors	src/sensors.cpp	Sensor module interface wrapper
SMBus	src/smbus.cpp	Low-level I2C/TWI protocol (80 kHz)
Config	src/rom_conf.cpp	EEPROM configuration storage
RPM	src/rpm_calculation.cpp	Pulse count to RPM conversion
Board	src/boards/mfm_v3_m1284p.cpp	Board-specific initialization
Watchdog	src/wdt.cpp	MCU reset mechanism

Design Principles

Sensor Passthrough Architecture

The firmware operates as a passthrough system:

• Raw sensor data is transmitted without firmware-level interpretation
• The sensor module at address 0x36 encodes data in its own format
• Data decoding happens in the application layer (TTN decoder + backend)

This design minimizes firmware complexity and allows flexible sensor integration.

Cooperative Multitasking

All operations use LMIC’s job scheduler:

• No blocking code in the main loop
• Jobs are scheduled with delays using os_setTimedCallback()
• The watchdog ensures recovery from hangs

Configuration Abstraction

All runtime configuration is stored in EEPROM:

• LoRaWAN credentials (AppEUI, DevEUI, AppKey)
• Measurement interval
• Sensor-specific settings (wheel teeth count)
• Configuration can be modified via OTA downlinks