DFE Tap Training With Hybrid Search for Faster Memory Boot
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As data rates increase in memory systems, signal degradation due to Inter Symbol Interference (ISI) causes the data eye to close, leading to insufficient training time for decision feedback equalization (DFE) taps, especially with multiple memory channels, which significantly prolongs boot time.
Innovation Solution
A hybrid search method is employed to efficiently select tap coefficients for DFE, comprising a modified binary search to identify initial values and a Tabu search to find final values, reducing training time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If data rate is increased in memory systems, then transfer speed between memory device and controller is improved, but signal degradation due to ISI causes data eye to close
Solution Approach 1:
The patent applies preliminary action by performing DFE tap training during the boot-up process before normal operation begins. The hybrid search method (combining binary search and Tabu search) systematically determines optimal DFE coefficients in advance, allowing the system to compensate for ISI effects before they degrade signal integrity during high-speed data transfer.
2Reliability
If DFE training is performed with multiple memory channels, then signal quality is improved, but boot time is significantly prolonged
Solution Approach 1:
The patent segments the DFE training process into two distinct phases: a coarse training phase that establishes initial tap coefficients, and a fine training phase that optimizes the coefficients. This segmentation allows the system to balance between achieving sufficient signal quality and reducing overall training time, particularly important when dealing with multiple memory channels.
Solution Approach 2:
The patent implements partial action by performing coarse DFE training on all memory channels during boot-up, then performing fine training only on channels that require it based on signal quality metrics. This approach ensures adequate signal quality across all channels while minimizing the time spent on training, avoiding excessive training of channels that already meet performance requirements.
3Loss of time
If DFE tap training time is reduced, then boot time is decreased, but training precision may be insufficient
Solution Approach 1:
The patent applies preliminary action by using binary search to quickly establish initial DFE tap coefficients before fine-tuning with Tabu search. This preliminary coarse training provides a good starting point that reduces the search space for subsequent optimization, ensuring both speed and precision in the overall training process.
Solution Approach 2:
The patent implements feedback mechanisms where the system continuously monitors signal quality metrics during training and uses this information to adjust the training process. The Tabu search algorithm uses feedback from the cost function evaluation to guide the search toward optimal tap coefficients, ensuring precision while maintaining efficiency through adaptive search strategies.
Data Source
AI summary
Decision feedback equalization (DFE) training time in a memory device is reduced through the use of a hybrid search to select values of tap coefficients for taps in the DFE. The hybrid search includes two searches. A first search is performed to identify initial values of tap coefficients, a second search uses the initial values of tap coefficients to find the final values of tap coefficients.