Memory Controller Data Training via Error Detection Codes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data training methods for memory devices are inefficient, as they consume significant time, increase memory device size when using separate resistors, and do not allow direct data error detection, and become complex with increasing controller functions, while also not being applicable if data is already present in memory cells.
Innovation Solution
A method and system that use error detection codes (EDCs) to determine data skew by comparing transmitted and received data patterns, adjusting data delay times to reduce skew, and calculating error occurrence frequency to optimize data training time, applicable to next-generation DRAMs, which allows for error detection and reduced training time without overwriting existing data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional data training process is used to regulate skew, then data error reduction is achieved, but training time becomes excessively long
Solution Approach 1:
The patent applies skipping by directly transitioning to error detection code comparison without performing the complete conventional data training sequence. The controller generates EDCs for both original and read-back data patterns, compares these codes to detect errors, and determines skew based on error patterns, thereby rushing through the training process while maintaining data error reduction effectiveness.
2Reliability
If separate resistors are used for data training, then skew regulation capability is improved, but memory device size increases
Solution Approach 1:
The patent applies universality by enabling the memory device to perform multiple functions using existing components. The memory device uses its standard read/write circuitry and error detection code generation capabilities to participate in the skew detection process, eliminating the need for separate dedicated training resistors. The same data paths and control logic serve both normal operation and skew regulation.
Solution Approach 2:
The memory device performs self-service by using its own error detection code generation and comparison capabilities to detect and report skew conditions. The device autonomously generates EDCs for transmitted data, compares them with EDCs for read-back data, and provides skew information to the controller without requiring external training components.
3Measurement precision
If data pattern is written to memory cell for training, then skew detection is enabled, but existing data is overwritten
Solution Approach 1:
The patent applies preliminary action by generating and comparing error detection codes for the data pattern before actually writing it to the memory cell. The controller generates an EDC for the original data pattern, the memory device generates an EDC for the read-back pattern, and the comparison is performed before any write operation occurs, thus preserving existing data while enabling skew detection.
4Reliability
If conventional data training is applied, then skew of each data pin is regulated, but controller complexity increases
Solution Approach 1:
The patent replaces the mechanical/conventional iterative data training system with an error detection code-based detection system. Instead of writing test patterns, reading them back, and iteratively adjusting delays based on pattern comparison, the system substitutes this with direct EDC generation and comparison, which is computationally simpler and requires less complex control logic for pattern management and iterative adjustment.
Data Source
AI summary
A data training system and method thereof are provided. The example data training system may include a memory controller transmitting a given data pattern to a memory device, the memory controller first determining whether an error is present within the transmitted data pattern based on at least one error detection code, the at least one error detection code based on at least one of the given data pattern and the transmitted data pattern and second determining a data delay time for reducing an amount of skew based on whether the first determining step determines an error to be present within the transmitted data pattern.