Hybrid Solver for IC Diagnostics Using Polynomial Minimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current fault-diagnostic methods for digital integrated circuits are inefficient and costly, especially for large-scale ICs, as they do not scale well and lack mechanisms for automatic diagnosis and test pattern generation.
Innovation Solution
A hybrid solver system that converts diagnostic problems into polynomial minimization problems, using a combination of software modules and hardware-based solvers like Ising machines to generate diagnosis or test vectors by representing logic gates as polynomials and reducing polynomial order through techniques like cube-and-conquer and graph isomorphism, facilitating parallel processing and efficient fault detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional fault-diagnostic methods are used for digital integrated circuits, then diagnostic coverage can be achieved, but computational complexity and cost increase significantly with circuit scale
Solution Approach 1:
The patent replaces conventional software-based SAT solvers with a hardware-based Ising machine solver. The diagnostic problem is transformed into a polynomial minimization problem that maps to an Ising model, where logic gates are represented as polynomial equations and the fault diagnosis becomes an energy minimization problem solved by the Ising machine's physical dynamics.
Solution Approach 2:
The patent changes the representation parameters of the diagnostic problem by converting Boolean logic variables into polynomial variables and transforming the satisfiability problem into a polynomial minimization problem. This parameter transformation enables the problem to be solved using physical optimization processes rather than sequential computational search.
2Reliability
If testing and fault-diagnosis are performed on modern digital ICs, then quality assurance is improved, but testing and diagnostic costs reach up to 40% of total production cost
Solution Approach 1:
The patent replaces computationally expensive software-based SAT solving with a hardware-based Ising machine that solves the diagnostic problem through physical energy minimization. This substitution dramatically reduces computation time and associated costs while maintaining diagnostic accuracy, directly addressing the high testing cost issue.
3Quantity of substance
If feature size of circuits decreases to increase density, then IC capacity is improved, but defects become more common leading to malfunction
Solution Approach 1:
The patent implements a self-diagnostic capability where the circuit design includes augmented diagnostic circuits that automatically detect and identify defects without external intervention. The system uses the Ising machine to autonomously solve the diagnostic problem and generate diagnosis vectors that pinpoint faulty components, enabling the system to self-service its quality assurance needs.
4Reliability
If conventional SAT-based diagnostic methods are used, then fault detection is achieved, but the methods do not scale well to large-scale ICs
Solution Approach 1:
The patent replaces sequential software-based SAT solving with parallel hardware-based Ising machine computation. The Ising machine's physical architecture naturally parallelizes the optimization process, allowing it to handle large-scale circuits with thousands of gates efficiently, thus achieving scalability that software methods cannot provide.
Data Source
AI summary
One embodiment provides a method and a system for computing diagnoses for a physical system. During operation, the system can obtain a design of the physical system, generate a design of a diagnostic system by augmenting the design of the physical system based on a number of fault-emulating subsystems, and convert the design of the diagnostic system into a polynomial formula comprising a plurality of variables. The plurality of variables can include inputs and outputs of the original physical system and a number of ancillary variables. The system can further embed the polynomial formula on a hardware-based solver configured to perform optimization using the polynomial formula as an objective function to obtain a diagnostic vector used for explaining faults in the physical system.