Selective Memory Hardening via SEIP Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current microelectronic designs face challenges in reducing the risk of soft errors caused by Single Event Upsets (SEUs) due to radiation, particularly in smaller technologies, where blanket hardening techniques increase power consumption, real estate, and cost, and are not always effective as not all SEUs inflict soft errors.

Innovation Solution

A method to selectively harden memory elements by computing the Soft Error Infliction Probability (SEIP) of each node in a circuit, simulating SEU propagation, and providing recommendations for specific hardening techniques based on calculated probabilities, reducing unnecessary power consumption and real estate inflation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If blanket hardening techniques are applied to all memory elements, then reliability against SEU is improved, but power consumption increases

Engineering Contradiction:
Improvesoft error resistanceVSAvoidpower consumption
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent applies hardening techniques selectively to specific memory elements based on their individual SEIP values rather than uniformly to all memory elements. The SEIP calculation identifies which memory elements are most susceptible to soft errors, allowing targeted hardening of only those elements that require protection, thereby reducing overall power consumption while maintaining reliability for critical components.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent segments the memory system into individual memory elements with calculated SEIP values, allowing independent analysis and selective hardening of each element. This segmentation enables the system to identify and protect only the vulnerable portions rather than hardening the entire memory system, reducing unnecessary power consumption.

Inventive Principle:
Principle #1Segmentation

2Reliability

If blanket hardening techniques are applied to all memory elements, then reliability against SEU is improved, but silicon real estate increases

Engineering Contradiction:
Improvesoft error resistanceVSAvoidsilicon real estate
Core Design Contradiction:
ReliabilityVSArea of stationary object

Solution Approach 1:

The patent implements selective hardening based on local SEIP characteristics of individual memory elements. By calculating and comparing SEIP values, the system identifies specific elements requiring protection and applies hardening techniques only to those elements, thereby minimizing the additional silicon real estate required compared to blanket hardening approaches.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The memory system is segmented into individual analyzable units with distinct SEIP values. This segmentation allows the system to apply hardening only to specific segments (memory elements) that demonstrate high SEIP, rather than expanding the hardening infrastructure across the entire memory array, thus conserving silicon real estate.

Inventive Principle:
Principle #1Segmentation

3Reliability

If blanket hardening techniques are applied to all memory elements, then reliability against SEU is improved, but device complexity increases

Engineering Contradiction:
Improvesoft error resistanceVSAvoidhardening implementation complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent performs preliminary SEIP calculation and analysis during the design phase to identify which memory elements require hardening before the actual hardening implementation. This preliminary action simplifies the overall device complexity by providing a clear roadmap for selective hardening, avoiding the need for complex runtime decision-making or intricate hardening control logic.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent replaces complex physical hardening implementations with a software-based SEIP calculation and analysis system. By using computational methods to identify and prioritize hardening targets, the system avoids the need for complex hardware control mechanisms, thereby reducing overall device complexity while maintaining effective selective hardening.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

4Use of energy by moving object

If selective hardening is applied based on SEIP calculation, then power consumption is reduced, but measurement precision is required

Engineering Contradiction:
Improvepower consumptionVSAvoidSEIP calculation accuracy
Core Design Contradiction:
Use of energy by moving objectVSMeasurement precision

Solution Approach 1:

The SEIP calculation system incorporates feedback mechanisms that analyze simulation results and trace data to refine the accuracy of SEIP values for each memory element. This feedback loop ensures that the measurement precision required for effective selective hardening is achieved, allowing the system to confidently identify which elements require protection while minimizing power consumption on unprotected elements.

Inventive Principle:
Principle #23Feedback

Data Source

PatentUS11068629B2Circuit simulation using a recording of a reference execution
Publication Date: 2021.07.20 OPTIMA DESIGN AUTOMATION
  • US11068629B2 patent drawing
  • US11068629B2 patent drawing
  • US11068629B2 patent drawing

AI summary

A method, system and product for circuit simulation using a recording of a reference execution. The method comprises obtaining a design of a circuit, wherein the circuit comprises nodes which are assigned values during execution. The method further comprises obtaining a recording of a reference execution of the circuit, wherein the recording comprises recorded values of the nodes in a plurality of cycles. The method further comprises simulating, by a processor, an execution of the circuit, wherein said simulation is performed using the recorded values of the reference execution.