These techniques add redundant information to detect or correct data corruption.

• Error Correcting Code (ECC) Memory: Memory controllers implement sophisticated algorithms (e.g., Hamming codes, SECDED - Single Error Correct Double Error Detect codes) that generate extra "parity" bits for each data word. During read operations, these parity bits are checked, allowing the system to automatically correct single-bit errors and detect (and often report) multi-bit errors caused by noise, cosmic rays ("soft errors"), or subtle hardware defects. Critical for server, automotive, and aerospace applications.
• Cyclic Redundancy Check (CRC): A highly effective, widely used mathematical algorithm to detect unintentional alterations of raw data. A CRC value (checksum) is computed for a block of data and appended to it. When the data is received or read back, the CRC is re-calculated and compared. If they don't match, data corruption is detected. Used extensively in communication protocols (Ethernet, USB, CAN), data storage, and firmware verification.
• Checksums: Simpler sums of data bytes, less robust than CRC but quicker to calculate, used for basic integrity checks.
• Parity Bits: The simplest form of error detection, adding a single bit to ensure an even or odd number of '1's in a data byte/word. Can only detect an odd number of bit errors.

Redundancy involves duplicating components or functionalities to provide backup in case of failure.

• Hardware Redundancy:
• Triple Modular Redundancy (TMR): Three identical hardware modules (e.g., processors, sensors) execute the same operation simultaneously. A "voter" circuit compares their outputs, and the majority output is chosen. If one module fails, the system continues to operate correctly. Used in ultra-reliable systems like aircraft flight control.
• N-Modular Redundancy (NMR): An extension of TMR with N modules and a voter.
• Active Redundancy (Hot Standby): A primary component is active, and an identical redundant component is also powered on and continuously performing the same task or receiving the same inputs. If the primary fails, the standby can take over immediately with minimal disruption.
• Warm Standby: The redundant component is powered on but not fully active. It can take over quickly but not instantaneously.
• Cold Standby: The redundant component is powered off. It takes a significant amount of time to power up and take over, but consumes no power while idle.
• Software Redundancy:
• N-Version Programming: Developing the same software specification by multiple independent teams using different algorithms, programming languages, or development tools. This aims to reduce the likelihood of common-mode software bugs (bugs present in all versions). The outputs are compared by a voter.
• Data Replication: Storing critical data in multiple memory locations or on different storage devices.
• Replicated Computations: Performing the same calculation multiple times and comparing results to detect transient errors.

These techniques enable the system to detect and respond to failures.

• Watchdog Timers (WDT): A dedicated hardware timer. The embedded software is responsible for periodically "feeding" or "kicking" (resetting) this timer. If the software fails to kick the watchdog within a predefined timeout period (indicating a software hang, infinite loop, or crash), the watchdog timer expires and triggers a system reset, forcing a restart and attempting to recover from the fault. Some systems use "windowed watchdogs" which also require the kick to be within an upper and lower bound, ensuring execution is neither too fast nor too slow.
• Error Reporting and Logging: Implementing mechanisms to detect errors (e.g., via hardware fault flags, software sanity checks) and log them to non-volatile memory or send them over a communication link for later analysis.
• Fail-Safe States: Designing the system to transition to a safe, predefined state upon detection of a critical failure. For example, a motor controller might shut down the motor, or a heating system might turn off the heater.
• Graceful Degradation: Instead of a complete system failure, the system reduces its functionality or performance in a controlled manner upon detecting a non-critical fault. For example, a multimedia system might reduce video quality rather than crashing completely.
• Self-Checking Mechanisms and Diagnostics:
• Power-On Self-Test (POST): Firmware executed at boot-up to check the integrity of key hardware components (CPU, memory, peripherals) before loading the main application.
• Runtime Diagnostics: Software routines that periodically check the health and integrity of hardware components, memory, and software states during normal operation.

Protecting the embedded system from external disturbances is crucial for robustness.

• Electromagnetic Compatibility (EMC) Design:
• EMI (Electromagnetic Interference) Reduction: Designing the PCB and enclosure to minimize unwanted electromagnetic radiation generated by the system itself (e.g., careful routing of high-speed signals, shielding, grounding, filtering).
• EMS (Electromagnetic Susceptibility) Immunity: Designing the system to be resilient to external electromagnetic interference (e.g., from nearby motors, radios, lightning). This involves robust power supply filtering, transient voltage suppressors (TVS diodes) on I/O lines, and proper grounding techniques. Compliance with EMC standards (e.g., CE, FCC) is often mandatory.
• Thermal Management: Ensuring components operate within their specified temperature ranges.
• Passive Cooling: Heat sinks, thermal pads, optimized PCB layout for heat dissipation.
• Active Cooling: Fans, liquid cooling (for high-power systems).
• Thermal Throttling: Reducing clock frequency or voltage (via DVFS) to prevent overheating when temperatures rise.