Semiconductor Memory Training Interval Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor memory training methods are time-consuming due to the need to test each signal line under multiple reference voltages, leading to prolonged system startup times.
Innovation Solution
A method that selects two adjacent reference voltages to determine a target interval for the expected margin value, reducing the search range and improving efficiency by obtaining minimum margin values for each target signal line under these voltages, thereby accelerating the training process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the traditional method of testing each signal line under multiple reference voltages is used, then the training accuracy is ensured, but the training time is prolonged
Solution Approach 1:
The patent segments the reference voltage range into multiple intervals, and for each interval, selects two adjacent reference voltages (first and second reference voltages) to perform margin value testing. This segmentation approach divides the comprehensive testing into smaller, manageable segments, reducing the overall testing scope while maintaining accuracy within each segment.
Solution Approach 2:
The patent performs preliminary testing by obtaining minimum margin values for all target signal lines under the first and second reference voltages before determining the expected margin value. This preliminary action identifies the target interval where the maximum margin value is likely to occur, allowing subsequent focused searching within a narrower range rather than testing all possible reference voltages.
2Productivity
If the search range for the expected margin value is reduced to a target interval, then the training efficiency is improved, but the complexity of determining the target interval increases
Solution Approach 1:
The patent uses feedback from the minimum margin value testing to determine the target interval. By comparing the minimum margin values obtained under the first and second reference voltages, the system feedbacks information about where the maximum margin value is likely located, thereby defining the target interval for subsequent expected margin value searching.
Solution Approach 2:
The patent changes the parameter scope by transitioning from searching the entire reference voltage range to searching only within the determined target interval. This parameter change in the search scope significantly reduces the number of tests required while the interval determination algorithm, though adding some complexity, enables this efficient parameter change.
Data Source
AI summary
A semiconductor memory training method includes: selecting two adjacent reference voltages from a plurality of reference voltages as a first reference voltage and a second reference voltage; obtaining a first minimum margin value for the plurality of target signal lines under the first reference voltage; obtaining a second minimum margin value for the plurality of target signal lines under the second reference voltage, according to a minimum margin value for each target signal line under the second reference voltage; determining a target interval for an expected margin value according to the first minimum margin value and the second minimum margin value, the expected margin value being the maximum one among the minimum margin values for the plurality of target signal lines under the plurality of reference voltages; and searching for the expected margin value in the target interval.