Bit Apportionment for Soft Error Resistance in Digital Functions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in handling multi-bit errors caused by nuclear radiation, particularly in noisy environments, where approaches like triple modular redundancy struggle to maintain data integrity with increased resource requirements and error detection/correction methods are insufficient.
Innovation Solution
A method involving simulations and a genetic algorithm to identify optimal bit apportionments by determining the number of copies for each data bit, combining apportionments with lower numerical errors to generate improved allocations, which can minimize cumulative errors while maintaining resource efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If triple modular redundancy is used to handle soft errors, then data integrity is improved, but resource usage increases
Solution Approach 1:
The patent applies local quality by differentiating the redundancy level assigned to each bit position based on its significance. More significant bits (e.g., most significant bits in a binary number) receive higher redundancy protection, while less significant bits receive lower redundancy. This non-uniform allocation optimizes data integrity for critical bits while reducing overall resource consumption compared to uniform TMR approaches.
Solution Approach 2:
The patent changes the parameter of redundancy allocation from a fixed uniform value to a variable value that depends on bit significance. By dynamically adjusting the number of copies based on the positional weight of each bit, the system adapts the protection level to match the actual importance of each bit, resolving the contradiction between reliability and resource usage.
2Reliability
If multiple copies of each data bit are maintained with voting schemes, then soft error resistance is improved, but system complexity increases
Solution Approach 1:
The patent reduces system complexity by applying local quality to the voting mechanism. Instead of implementing complex voting logic for all bits uniformly, the system applies simplified voting schemes only to the most significant bits that require protection, while less significant bits use simpler or no voting mechanisms. This localized approach maintains soft error resistance for critical data while reducing overall system complexity.
3Device complexity
If uniform redundancy is applied to all bits, then data protection is simplified, but cumulative numerical errors increase
Solution Approach 1:
The patent resolves this contradiction by applying local quality to differentiate protection levels across bit positions. The system identifies which bits contribute most to cumulative numerical errors (typically more significant bits) and applies enhanced redundancy specifically to those positions. This targeted approach maintains simplicity for less critical bits while reducing cumulative numerical errors through focused protection of error-sensitive bits.
Solution Approach 2:
The patent changes the redundancy parameter from a uniform value to a position-dependent value that correlates with bit significance and error impact. By adjusting the redundancy parameter based on the positional weight of each bit, the system optimizes the balance between protection scheme simplicity and cumulative numerical error reduction, applying stronger protection only where it most effectively reduces overall error accumulation.
Data Source
AI summary
A method includes identifying multiple apportionments, where each apportionment identifies numbers of bit copies to be stored in at least one memory for at least some bits of a data value. The method also includes, for each apportionment, estimating a numerical error associated with use of the apportionment with a specified function, where the numerical error is estimated by creating errors in bit copies of multiple data values processed using the specified function. The method further includes combining portions of different ones of the apportionments having lower estimated numerical errors to create multiple derived apportionments. The method also includes, for each derived apportionment, estimating a numerical error associated with use of the derived apportionment with the specified function. In addition, the method includes selecting a final apportionment for use with the specified function, where the final apportionment includes or is based on at least one of the derived apportionments.