An embedded system’s architecture is the arrangement of its processor, memory, peripherals, software layers and power supply, and the way they connect to do one dedicated job. Understanding the block diagram, and why each block is there, is the starting point for designing, programming or debugging any embedded device.
The embedded system block diagram
At the hardware level every embedded system has the same blocks, whether it is an 8-bit thermostat controller or a multi-core automotive computer:
- Processor: a microcontroller, microprocessor, DSP or SoC that executes the firmware.
- Memory: non-volatile memory (flash) for the program and configuration, and RAM for data and stacks. Microcontrollers integrate both; larger systems use external DRAM and flash or eMMC.
- Input devices and sensors: buttons, keypads, temperature or motion sensors, cameras, microphones, communication receivers.
- Output devices and actuators: LEDs, displays, motors, relays, speakers, transmitters.
- Peripherals and interfaces: timers, ADCs and DACs, PWM, GPIO, UART, SPI, I2C, CAN, USB, Ethernet, wireless radios.
- Clock and reset circuits: oscillators, PLLs, watchdog timer, brown-out detection.
- Power supply: regulators, battery and charging circuit, power sequencing; often the block that decides the whole design.
- Debug interface: JTAG or SWD for programming and debugging.
Sensors feed the processor through ADCs or digital interfaces; the firmware decides; the processor drives actuators through outputs. A detailed look at each block is in components of an embedded system.
Processor architectures
| Type | Characteristics | Typical use |
|---|---|---|
| Microcontroller (MCU) | CPU, flash, RAM and peripherals on one chip; no MMU; runs bare metal or an RTOS | Appliances, sensors, motor control, automotive body electronics |
| Microprocessor / application processor | CPU with MMU and caches; external memory; runs Linux or Android | Gateways, infotainment, HMI panels, robots |
| Digital signal processor (DSP) | Optimised for multiply-accumulate and streaming data | Audio, radar, motor control, communications |
| System on chip (SoC) | Several of the above plus accelerators (GPU, NPU, video) on one die | Phones, cameras, ADAS, edge AI |
| FPGA / FPGA-SoC | Programmable logic, often with hard processor cores | Custom interfaces, high-speed processing, prototyping |
ARM Cortex-M and Cortex-A cores dominate, with RISC-V growing quickly. See microcontrollers in embedded systems for how to choose one, and VLSI vs embedded systems for how these chips are designed.
Harvard vs von Neumann
Von Neumann architecture uses one memory and bus for instructions and data; Harvard architecture separates them so instruction fetch and data access happen in parallel. Most microcontrollers are modified Harvard: separate buses to flash and RAM inside the chip, presented as one address space to the programmer.
Software architecture layers
- Boot code: initialises clocks, memory and the stack, then jumps to the application or loads an OS. On larger systems a bootloader (for example U-Boot) loads Linux.
- Hardware abstraction layer and drivers: code that configures and operates peripherals, so the application does not touch registers directly.
- Operating system (optional): an RTOS or embedded Linux providing tasks, scheduling and services. See embedded operating systems.
- Middleware: protocol stacks (TCP/IP, BLE, USB), file systems, graphics libraries, security libraries.
- Application: the device’s actual logic: the control loop, the state machine, the user interface.
In a bare-metal design, layers 3 and 4 shrink to a main loop and a few libraries; the division between drivers and application still matters for maintainability.
Timing and execution models
- Super-loop: a
while(1)loop that polls inputs and updates outputs; interrupts handle urgent events. Simple and predictable for small devices. - Interrupt-driven: the processor sleeps until an interrupt (timer, sensor, communication) wakes it; saves power.
- RTOS-based: several tasks with priorities, scheduled pre-emptively, communicating through queues and semaphores.
- Time-triggered: tasks run on a fixed schedule from a timer; used in safety-critical control for predictability.
Design considerations that shape the architecture
- Real-time requirements: worst-case response time decides the execution model and processor speed.
- Power: battery life decides sleep modes, radio duty cycles and even the processor family.
- Memory: code size and RAM decide the MCU part number and whether an OS fits.
- Cost and size: integration (one SoC vs several chips), PCB layers, enclosure.
- Reliability and safety: watchdogs, redundancy, error-correcting memory, certified software processes.
- Connectivity and security: which radios, which protocols, secure boot and key storage. See embedded systems and IoT.
- Upgradability: bootloader design for firmware updates over the product’s life.
A worked example: a battery-powered temperature logger
Processor: a low-power Cortex-M0+ MCU with 64 KB flash and 8 KB RAM. Sensor: a digital temperature sensor on I2C. Storage: the MCU’s flash for logged samples. Interface: BLE radio to sync with a phone. Power: coin cell with a low-dropout regulator. Software: bare metal; a timer interrupt wakes the MCU every minute, it reads the sensor, stores the sample, advertises over BLE briefly and returns to deep sleep. The architecture is driven entirely by the power budget: everything that can sleep, sleeps.
Learn embedded architecture by building
Our embedded systems course starts from this block diagram and builds up through peripherals, RTOS and embedded Linux with hardware in hand; the embedded internship applies it to a real project. For interview preparation see embedded systems interview questions.
Frequently asked questions
What are the main components of an embedded system architecture?
Processor, memory (flash and RAM), input sensors, output actuators, peripheral interfaces, clock and reset circuits, power supply and a debug interface, plus the firmware layers that run on them.
What is the difference between a microcontroller and a microprocessor in embedded systems?
A microcontroller integrates CPU, memory and peripherals on one chip and runs bare metal or an RTOS; a microprocessor needs external memory, usually has an MMU, and runs a full OS such as Linux.
What is a hardware abstraction layer?
A software layer that wraps peripheral registers in functions, so application code is portable and readable and hardware changes are contained in one place.
What is the difference between Harvard and von Neumann architecture?
Harvard separates instruction and data memory and buses; von Neumann shares them. Most microcontrollers use a modified Harvard design for speed.
Which execution model should I use?
Super-loop for simple devices, interrupt-driven for low power, an RTOS when several tasks need timing guarantees, and time-triggered scheduling for safety-critical control.
Share your question in comments or talk to our mentor team for batch guidance.
Ask the Admin Team
Drop your basic question in comments: eligibility, prerequisites, tools, fee range, and placement support.
Our team reviews and responds regularly.