How do I choose an RTOS for my MCU?

Choosing the right Real-Time Operating System (RTOS) for your microcontroller (MCU) involves evaluating technical constraints, project requirements, and ecosystem support. Here’s a step-by-step guide with actionable criteria:

---

1. Assess Hardware Constraints

Factor	Questions to Ask	Common Options
MCU Flash/RAM	Does the RTOS fit within available memory?	≤32KB: FreeRTOS, Zephyr (minimal config)
		≥128KB: RT-Thread, NuttX
CPU Architecture	Is the RTOS ported to your MCU core?	ARM Cortex: All RTOSes
		RISC-V: FreeRTOS, Zephyr
		8-bit (e.g., AVR): FreeRTOS (limited)
Peripheral Support	Does it include drivers for your hardware?	Zephyr: Broad driver support
		FreeRTOS: Relies on vendor HALs

---

2. Evaluate RTOS Features

Requirement	Solutions	Best Matches
Hard Real-Time	Needs deterministic scheduling	FreeRTOS (with preemption), QNX, VxWorks
Low Latency	Fast context switching (<10µs)	Zephyr, RT-Thread
Power Management	Sleep modes tickless idle	Amazon FreeRTOS (with low-power extensions)
Networking	TCP/IP, Bluetooth, LoRaWAN stacks	Zephyr (built-in), FreeRTOS (add-ons)
Filesystem	FAT, LittleFS, or NAND support	RT-Thread (with Elm-Chan FATFS)
Safety Certification	Need ASIL-D/SIL3?	SafeRTOS (certified FreeRTOS variant), QNX

---

3. Development Ecosystem

Tooling	Key Considerations	RTOS Examples
Debugging	Trace tools, stack overflow detection	FreeRTOS (Tracealyzer), Zephyr (SEGGER)
IDE Integration	Eclipse, VS Code, Keil, IAR support	All major RTOSes
Community	Active forums, GitHub issues	FreeRTOS (large community), Zephyr (Linux)
Documentation	API references, tutorials	Zephyr (excellent docs), RT-Thread (Chinese-heavy)

---

4. Licensing & Cost

License Type	Implications	RTOS Examples
MIT/BSD	Free for commercial use, no royalties	FreeRTOS, Zephyr, RT-Thread
GPL	Requires open-sourcing derivative code	RIOT OS
Proprietary	Paid licenses, vendor support	ThreadX, VxWorks, QNX

---

5. Industry-Specific Needs

Domain	Recommended RTOS	Why?
Automotive	AUTOSAR OS, QNX, SafeRTOS	ASIL-D compliance, CAN FD support
IoT Edge	Zephyr, Amazon FreeRTOS	Built-in BLE/MQTT, OTA updates
Medical	SafeRTOS, VxWorks	FDA/IEC 62304 compliance
Industrial	FreeRTOS, RT-Thread	Modbus/PLC support, deterministic control

---

Decision Flowchart

mermaid

graph TD
  A[Start] --> B{Memory < 64KB?}
  B -->|Yes| C[FreeRTOS/Zephyr minimal]
  B -->|No| D{Need networking?}
  D -->|Yes| E[Zephyr/RT-Thread]
  D -->|No| F{Hard real-time?}
  F -->|Yes| G[FreeRTOS/QNX]
  F -->|No| H[RT-Thread/NuttX]

---

Top RTOS Options Compared

RTOS	Best For	Memory Footprint	Key Strength
FreeRTOS	General-purpose, small MCUs	5–10KB ROM	Largest community, Amazon-backed
Zephyr	IoT, networked devices	10–50KB ROM	Linux Foundation, modern tooling
RT-Thread	Rich middleware needs	20–100KB ROM	Built-in filesystem/GUI
ThreadX	Commercial products	2–20KB ROM	Azure RTOS, certified for safety
QNX	Automotive/medical	500KB+ ROM	Microkernel architecture, hypervisor

---

Quick Selection Guide

1. For ARM Cortex-M0/M3 (≤64KB Flash): FreeRTOS

2. For BLE/Wi-Fi IoT devices: Zephyr

3. For China-market products: RT-Thread (strong local support)

4. For ASIL-D systems: SafeRTOS or QNX

5. For legacy 8-bit MCUs: Nano-RK or cooperative scheduler

---

Next Steps After Choosing

1. Benchmark: Run the RTOS on your MCU with a blinky test (measure context-switch latency).

2. Check Drivers: Verify support for your peripherals (UART, SPI, ADC).

3. Profile Memory: Use FreeRTOSHeapStats() or Zephyr’s mem_usage shell command.

Example FreeRTOS Config for STM32:

c

// FreeRTOSConfig.h
#define configTOTAL_HEAP_SIZE ((size_t)16 * 1024)  // 16KB heap
#define configMINIMAL_STACK_SIZE ((uint16_t)128)   // Idle task stack
#define configUSE_PREEMPTION 1                     // Enable preemption

---

When to Avoid an RTOS

• Ultra-low-power devices (use event-driven superloop).

• Extremely tight deadlines (bare-metal with interrupts).

• Projects with <32KB Flash (consider protothreads).

By matching your MCU’s capabilities with the RTOS’s features and community support, you’ll optimize both development speed and runtime performance.