Processing System Fault Detection Using Segmented Reference Paths
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The implementation of machine learning and artificial intelligence in safety-critical industries like automotive, medical, and aviation has increased the demand for high-performance computing, leading to bottlenecks in computational performance due to the PPA (performance-power-area) trade-off and manufacturability, which existing functional safety practices struggle to address effectively.
Innovation Solution
A processing system is designed with separate control, data, and non-safety critical paths, utilizing error detection codes and low-level register transfer level comparisons to detect faults, minimizing area and power consumption while maintaining high diagnostic coverage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional full duplication methods are used for fault detection, then diagnostic coverage is improved, but area and power consumption increase
Solution Approach 1:
The control path is segmented into multiple comparison locations where comparisons are performed at different hierarchical levels. Instead of duplicating the entire control path, the system divides it into segments that can be compared independently, reducing the overall area required for redundancy while maintaining diagnostic coverage across critical points.
Solution Approach 2:
The patent applies different levels of redundancy and comparison strategies to different segments of the control path based on their criticality. High-criticality segments receive full duplication and comparison, while less critical segments use simplified comparison mechanisms, optimizing the balance between diagnostic coverage and area consumption.
2Reliability
If traditional full duplication methods are used for fault detection, then diagnostic coverage is improved, but power consumption increases
Solution Approach 1:
The control path is segmented into multiple comparison locations where comparisons are performed at different hierarchical levels. Instead of duplicating the entire control path, the system divides it into segments that can be compared independently, reducing the overall area required for redundancy while maintaining diagnostic coverage across critical points.
Solution Approach 2:
The patent applies different levels of redundancy and comparison strategies to different segments of the control path based on their criticality. High-criticality segments receive full duplication and comparison, while less critical segments use simplified comparison mechanisms, optimizing the balance between diagnostic coverage and area consumption.
3Reliability
If functional safety practices are implemented at hardware level, then safety is improved, but computational performance becomes bottlenecked
Solution Approach 1:
The patent implements a dynamic comparison mechanism where the system can adaptively select which comparison locations to activate based on operational conditions and criticality requirements. This allows the safety mechanism to be more aggressive when needed and more conservative when not needed, reducing the average performance overhead while maintaining safety guarantees.
Solution Approach 2:
Instead of applying full redundancy throughout the entire control path, the patent applies partial redundancy only at strategically selected comparison locations. This partial action approach provides sufficient safety coverage for critical operations while minimizing the performance penalty associated with comprehensive duplication.
Data Source
Figure 1
Figure 2~3
AI summary
The present disclosure relates to a processing system comprising at least one control path and at least one data path, and at least one reference path, wherein each of the reference paths is functionally equivalent to and associated with a respective one of the at least one control paths and at least one comparison location for fault detection. The processing system is configured to compare the at least one control path and the at least one associated reference path at the at least one comparison location for fault detection. The disclosure also relates to a corresponding method.