Memory Calibration Endpoint Replay for Data Strobe Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current memory calibration methods in memory subsystems often require full-range calibrations, which can be time-consuming and may not accurately determine the endpoints of the data strobe signal, affecting the precision of data transfer and noise susceptibility.
Innovation Solution
A method involving an initial 'light' calibration followed by a 'heavy' endpoint replay calibration, where the second calibration is performed within a restricted range around the initially determined endpoints, using a higher number of bits to refine the delay values and reference voltages for accurate data transfer.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full-range calibration is performed to determine endpoint values, then measurement precision is improved, but calibration time increases
Solution Approach 1:
The calibration process is divided into two distinct phases: a first calibration that performs a coarse search across a full range to identify initial endpoint values, and a second calibration that performs a refined search only within a restricted range around those initial endpoints. This segmentation allows the system to achieve high measurement precision for endpoint determination while significantly reducing overall calibration time by avoiding redundant full-range scans.
Solution Approach 2:
The first calibration serves as a preliminary action that establishes initial endpoint values before the second calibration begins. By performing this preliminary coarse calibration first, the system obtains sufficient starting information to define a restricted range for the subsequent refined calibration, thereby avoiding the need to re-scan the entire range and reducing total calibration time while maintaining precision.
2Measurement precision
If more bits are transmitted during calibration to improve data integrity, then measurement precision is improved, but calibration time increases
Solution Approach 1:
The calibration process segments transmission operations into two phases: the first calibration transmits a first number of bits to establish initial endpoints, and the second calibration transmits a second number of bits (greater than the first) within a restricted range to refine the endpoints. This segmentation allows the system to achieve higher data integrity through increased bit transmission while limiting the time penalty to only the restricted range portion rather than the full range.
3Productivity
If calibration range is restricted to improve efficiency, then productivity is improved, but measurement precision may deteriorate
Solution Approach 1:
The calibration is segmented into a coarse phase that scans a full range to identify initial endpoints and a refined phase that operates within a restricted range around those endpoints. This segmentation enables the system to achieve high productivity in the refined phase by focusing only on the critical region, while the coarse phase ensures that the restricted range is properly centered and contains the true endpoints, thereby maintaining measurement precision.
Solution Approach 2:
The first calibration performs a preliminary full-range scan to establish initial endpoint values and define the restricted range for the second calibration. This preliminary action ensures that the restricted range is accurately positioned around the true endpoints, allowing the second calibration to achieve high measurement precision within the smaller range, thus improving productivity without sacrificing accuracy.
Data Source
AI summary
A method and apparatus for performing memory calibration with endpoint replay is disclosed. A first calibration of a data strobe signal in a memory subsystem is performed. The first calibration includes determining initial values of first and second endpoints indicative of first and second delay values, respectively, applied to the data strobe signal. A second calibration of the data strobe signal is performed around these endpoints, within a range thereof that is less than a full range there between. Based on the second calibration, the endpoints are adjusted.