Balanced DBI Coding for Low-Power Memory Refresh
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current memory refresh methods consume high power and generate noise due to unbalanced coding and parasitic inductance, particularly in high-frequency computing systems, where 8B/10B coding methods result in significant overhead and increased power consumption.
Innovation Solution
The method involves receiving data encoded using Data Bus Inversion (DBI), balancing the number of zeros across the data by adjusting the least significant bits, and using additional flags to optimize coding, thereby reducing power consumption and noise during memory refresh.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If 8B/10B coding is used for memory refresh, then data transmission quality is improved, but power consumption increases
Solution Approach 1:
The patent changes the parameter of data representation from standard 8B/10B coding to balanced coding with controlled zero transitions. By modifying the coding parameters to limit the number of zero transitions between consecutive refresh operations, the patent reduces the frequency of driver inversions and associated power consumption while maintaining data integrity through the balanced code structure.
Solution Approach 2:
The patent introduces dynamic selection of coding modes (standard DBI vs. balanced coding) based on the data pattern being transmitted. The system dynamically adjusts the coding approach to minimize power consumption during memory refresh operations while preserving the reliability benefits of coded transmission when needed.
2Device complexity
If data bus inversion is used to reduce coding complexity, then device complexity is reduced, but power consumption remains high due to varying zero counts
Solution Approach 1:
The patent modifies the DBI coding parameter by introducing balanced coding that specifically controls the number of zero transitions. This parameter change adds minimal complexity to the encoding/decoding logic while significantly reducing power consumption by ensuring that consecutive refresh operations do not have excessive zero transitions that would require frequent driver inversions.
3Reliability
If frequent memory refresh is performed to maintain data, then data reliability is maintained, but noise increases due to simultaneous switching
Solution Approach 1:
The patent changes the temporal parameter of refresh operations by using balanced coding to predict and control the timing of driver inversions. By ensuring that consecutive refresh operations have limited zero transitions, the patent reduces simultaneous switching events and associated noise while maintaining the necessary refresh frequency for data retention.
4Measurement precision
If driver current paths vary with data level to represent binary values, then data transmission accuracy is improved, but simultaneous switching noise increases
Solution Approach 1:
The patent changes the parameter of data encoding to use balanced coding with controlled zero transitions. This ensures that consecutive refresh operations do not cause excessive simultaneous switching of drivers, reducing noise while maintaining the ability to accurately represent binary values through controlled current paths.
Data Source
AI summary
A method and apparatus for reducing power consumption needed to refresh a memory may receive data having been encoded using data bus inversion (DBI), the DBI data having a first delta between a number of zeros for different cases between zero and a DBI maximum, balance code the DBI data to balance the number of zeros across the DBI data, and output data having a number of zeros for different cases between a minimum number greater than zero and less than or equal to the DBI maximum and a maximum number equal to the minimum number plus a second delta, the second delta being less than the first delta.