Replicated Computation Blocks for Fault Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fault detection methods in processing circuitry, such as software test libraries and dual core lock step, are either area and power inefficient or limited in diagnostic coverage, especially in safety-critical systems where accurate and efficient fault detection is crucial.
Innovation Solution
The use of replicated computation blocks within processing circuitry to perform fault detection by latching and comparing output values during stable input periods, allowing for efficient detection of stuck-at faults and other faults without interfering with normal operation, leveraging existing replication for area and power efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If software test libraries are used for fault detection, then diagnostic coverage is achieved, but area and power efficiency deteriorate
Solution Approach 1:
The computation blocks perform self-diagnosis by comparing their own output values with each other. Each computation block's output is used as input to another computation block, creating a self-testing mechanism that eliminates the need for external test equipment, thereby reducing power consumption while maintaining diagnostic coverage.
Solution Approach 2:
The system uses replicated computation blocks that are identical copies of each other. By having multiple copies perform the same calculation and comparing their outputs, the system achieves diagnostic coverage without requiring complex external testing infrastructure, thus improving power efficiency.
2Measurement precision
If dual core lock step is used for fault detection, then diagnostic coverage is improved, but area requirements increase
Solution Approach 1:
The computation blocks serve dual purposes: they perform normal computational functions during operation and simultaneously function as test subjects for fault detection. This multi-functionality allows the same hardware to be used for both processing and self-diagnosis, eliminating the need for separate test equipment and reducing area requirements.
Solution Approach 2:
The system performs self-testing by having computation blocks test each other through cross-comparison of output values. This self-service mechanism eliminates the need for external dual-core lockstep testing infrastructure, thereby reducing area requirements while maintaining diagnostic coverage.
3Measurement precision
If replicated computation blocks are used for fault detection, then diagnostic coverage is improved, but power consumption increases
Solution Approach 1:
The fault detection mechanism operates continuously during normal system operation without requiring separate testing phases. The computation blocks continuously perform calculations and simultaneously undergo self-diagnosis, ensuring that the useful computational action continues uninterrupted while fault detection is performed, thereby minimizing additional power consumption.
4Measurement precision
If fault detection mode is activated, then fault detection capability is improved, but normal operation may be interrupted
Solution Approach 1:
The fault detection function is merged with the normal computational operation. The computation blocks perform both their primary computational task and self-diagnosis simultaneously by exchanging output values as input values for other blocks. This integration ensures that fault detection capability is improved without interrupting normal operation, as both functions occur in parallel within the same operational cycle.
Data Source
AI summary
An apparatus has processing circuitry with at least two replicated computation blocks. Each computation block performs a particular calculation based on a respective input value to produce a respective output value to be used as a result of the calculation. The apparatus also has storage circuitry arranged on output paths of the computation blocks to store the output values of the computation blocks. The processing circuitry is operable in a fault detection mode in which the processing circuitry latches the output values of the computation blocks in the storage circuitry and presents, as the result of the calculations, the stored values. The processing circuitry also provides a first input value for a first computation block as an input value to one or more additional computation blocks of the at least two replicated computation blocks and compare the output values of the one or more additional computation blocks, when performing the calculation based on the first input value, with an output value of the first computation block when performing the calculation based on the first input value. If the processing circuitry detects a mismatch in the comparison, the processing circuitry determines that a fault exists in at least one of the computation blocks.