Bloom Filter Memory Fault Diagnosis for Fast SoC Accessibility Checks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In System-On-a-Chip (SoC) devices, diagnosing memory faults and determining memory accessibility is inefficient due to the integration of CPU and memory on a single chip, leading to the need to replace the entire SoC when memory defects occur, and existing comparison methods are time-consuming.
Innovation Solution
A Bloom filter-based memory fault diagnosis device that includes a pattern adaptive Bloom filter, a Bloom filter dictionary, and a second Bloom filter, along with a built-in self-test circuit, to compress test diagnosis data and enable rapid fault detection during manufacturing and operation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Volume of moving object
If a CPU and memory are integrated on a single chip (SoC), then the product size is reduced and power consumption is minimized, but when a memory defect occurs, the entire SoC must be replaced instead of just the memory component
Solution Approach 1:
The patent divides the SoC into functional segments by implementing a standalone memory fault diagnosis device that can independently test and identify defective memory cells without affecting the entire chip. The bit fail map is segmented to track individual memory cell status, allowing selective exclusion of only faulty cells while preserving the rest of the SoC functionality.
Solution Approach 2:
The patent introduces a memory fault diagnosis device as an intermediary component between the CPU and memory. This intermediary performs fault detection and management functions, including maintaining the bit fail map and controlling memory access based on fault status, thereby preventing direct CPU involvement in fault handling and enabling granular memory cell management.
2Measurement precision
If traditional comparison methods are used to diagnose memory faults, then fault detection can be performed, but the diagnosis time is considerably long due to 1:1 comparison requirements
Solution Approach 1:
The patent implements preliminary fault detection during the manufacturing process by performing memory tests before the SoC is shipped. The bit fail map is populated with defect information in advance, so that during runtime, the CPU can quickly reference pre-computed fault data without performing time-consuming 1:1 comparisons of memory cells.
Solution Approach 2:
The patent creates a compressed representation (bit fail map) that copies and stores the essential fault information of memory cells in a compact format. Instead of performing exhaustive comparisons with actual memory cells during operation, the system uses this copied fault map data to quickly determine memory accessibility, dramatically reducing diagnosis time while maintaining accuracy.
3Quantity of substance
If memory area occupies 70% to 80% of the SoC area, then memory capacity is increased, but the amount of memory that fail during testing increases accordingly
Solution Approach 1:
The patent changes the parameter of fault management from binary (pass/fail) to granular (individual cell status tracking). By implementing a bit fail map that tracks each memory cell's status separately, the system can identify and exclude only the specific defective cells among the large memory array, thereby maintaining high memory capacity utilization while compensating for the increased absolute number of failures in large memory configurations.
Data Source
AI summary
A memory fault diagnosis device, a System on a Chip (SoC), and a Bloom filter update method and a memory accessibility determination method can operate to diagnose memory faults in a minimum time by using a Bloom filter to detect failed memory areas and thereby determine memory accessibility in a minimum time.