Contents

ECU classes: from HPC to LIN slave and hardwired I/O

A first-principles tour of the computing nodes in a vehicle — central HPCs, MCU-based ECUs, smart sensors and actuators, LIN slaves and plain hardwired I/O — and why each class exists.

Why there are different classes

A vehicle needs both: huge compute for perception and services, and tiny, cheap, robust nodes that switch a lamp or read a door switch. No single kind of computer fits all of this, so the architecture is a pyramid of node classes.

What the classes are

Class Typical silicon Software Attached via Examples
HPC / central computer Multi-core SoC, DRAM, GPU/NPU, HSM, safety island Hypervisor, Linux/QNX, AUTOSAR Adaptive, services Multi-gig Ethernet Vehicle computer, ADAS computer, cockpit
Zone controller High-end MCU or small SoC AUTOSAR Classic, gateway, power management Ethernet + CAN FD/LIN/T1S Front, rear, left, right zone
MCU-based ECU 32-bit MCU, embedded flash, HSM, often lockstep AUTOSAR Classic or RTOS CAN FD or classical CAN Brake, steering, airbag, BMS, inverter
Smart sensor / actuator Small MCU or ASIC with transceiver Small firmware LIN, 10BASE-T1S, CAN Window lifter, seat motor, smart lamp, ultrasonic sensor
LIN slave Very small MCU or LIN-ASIC State machine or none LIN Mirror, HVAC flap, rain/light sensor
Hardwired I/O No electronics, or a simple sensor None Copper wire to an ECU or zone Door switch, lamp, temperature sensor, solenoid

How they work together

Control loops close at the lowest level that can meet their timing: a brake ECU reads wheel speeds and drives valves itself, while the central computer decides what the vehicle should do. Zone controllers collect simple I/O and power, and LIN or 10BASE-T1S connect smart edge devices. Data flows upward for services and diagnostics; commands flow down.

When to choose which

  • Highest ASIL, hard real-time, actuator nearby: dedicated MCU ECU.
  • Feature-rich, frequently updated: service on central compute.
  • Many simple loads in one area: zone controller with hardwired I/O and eFuses.
  • Cheap, low-data devices: LIN slaves; for more data or diagnostics at the edge, 10BASE-T1S.

Best practices

  • Allocate by timing, safety and update frequency — not by organisational history.
  • Keep the number of software-bearing node types small to reduce update and security effort.
  • Remember that hardwired I/O has no software to update but also no diagnostics beyond what the reading ECU provides.

Common pitfalls

  • Putting security-relevant logic in nodes that cannot hold keys.
  • Underestimating lifetime: MCUs and SoCs must stay available for 10–15 years of production and service.

What it means for your discipline

  • E/E architecture Core

    Chooses which class of node implements each function — the main lever for cost, latency, safety and updateability.

  • Hardware & silicon Core

    Each class has a different silicon, power and lifetime profile: SoC with DRAM, MCU with embedded flash, or a tiny ASIC with no software at all.

  • Base software Partial

    MCU ECUs run AUTOSAR Classic or an RTOS; LIN slaves often run tiny state machines; HPCs need a full platform stack.

  • Networks & communication Partial

    The node class decides the attachment: Ethernet for HPCs, CAN FD or classical CAN for MCU ECUs, LIN for slaves, wires for simple I/O.

  • Cybersecurity Partial

    Only nodes with enough compute and an HSM can run strong cryptography; simple slaves must be protected by their master and wiring.

  • Functional safety Partial

    High-ASIL functions usually sit on lockstep MCUs with dedicated sensors; the class choice shapes the safety concept.

  • Software update & OTA Partial

    Tailors installation and recovery strategies to each node class: HPCs can support partition-based updates, while small MCUs and LIN devices have tighter flash, transport and bootloader constraints.

No direct stake: HPC & application software · Diagnostics & lifecycle

Where it sits in the software stack

Generic layered software of a central HPC (application processor with hypervisor, plus a safety island) and of a classic MCU-based ECU (AUTOSAR Classic). Highlighted modules matter for this topic. Click any module to open its page.

HPC · application processor (MPU) and safety island (MCU)
  • Cortex-A cores · GPU/NPU · LPDDR — Provides the CPU, GPU/NPU and external memory needed for central services, cockpit applications and perception.
  • Type-1 hypervisor — Partitions central compute so workloads with different criticality can share the application processor.
  • Lockstep cores — Provides dedicated lockstep cores and ECC SRAM for safety functions alongside the application processor.
MCU-based ECU · AUTOSAR Classic
  • Microcontroller · HSM — Combines embedded flash, peripherals and optional lockstep cores and HSM for compact, dedicated controllers.
  • OS — Schedules real-time tasks for local control loops with predictable timing.
  • Microcontroller abstraction (MCAL) — Connects local control software to sensor inputs, actuator outputs and vehicle buses.