Command Bus Training Mode for Faster LPDDR CA Calibration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The existing command bus training methods for LPDDR4 and LPDDR5 memory devices are slow and require significant firmware coordination, limiting the speed and quality of command bus training due to the need for iterative testing of command/address values, which can miss thousands of potential sampling points for observing correctness or errors.

Innovation Solution

The implementation of a command address training mode (CATM) similar to that used in DDR5 memory devices, which simplifies error checking by returning a parity value instead of actual sampled values, reducing the coordination work required for firmware and enabling faster and more efficient command bus training.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional command bus training methods are used for LPDDR4 and LPDDR5 memory devices, then iterative testing of command/address values can be performed, but the training process becomes slow and requires significant firmware coordination, missing thousands of potential sampling points

Engineering Contradiction:
Improvecommand bus training accuracyVSAvoidtraining speed
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent applies the CATM mode from DDR5 memory devices to LPDDR4 and LPDDR5 devices, creating a copied training mechanism that returns parity values instead of actual sampled values. This copying of the DDR5 approach enables faster training by reducing firmware coordination overhead while maintaining sufficient training accuracy through parity-based error checking.

Inventive Principle:
Principle #26Copying

2Reliability

If actual sampled CA pattern values are returned during command bus training, then complete sampling data is available for error checking, but the coordination work required from firmware increases significantly

Engineering Contradiction:
Improveerror checking completenessVSAvoidfirmware coordination complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent extracts only the essential error checking information (parity values) from the complete sampled CA pattern data, removing the unnecessary transmission of full sampled values. This extraction approach maintains reliability by detecting sampling errors through parity checks while dramatically reducing firmware coordination complexity.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Instead of returning actual sampled values and having firmware perform error checking, the patent inverts the approach by having the memory device compute and return only parity values, with the memory controller then performing the error checking based on these compressed parity results.

Inventive Principle:
Principle #13The other way round (Inversion)

3Adaptability or versatility

If command bus training is performed using existing LPDDR4/LPDDR5 methods, then standardization is maintained, but the training process is limited by slow iterative testing and firmware coordination requirements

Engineering Contradiction:
Improvememory device compatibilityVSAvoidtraining time
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent makes the CATM training mode universal by applying it across multiple memory device types (LPDDR4, LPDDR5, and DDR5), enabling a single standardized fast training approach to work across different memory technologies. This multi-functionality reduces training time while maintaining broad compatibility through standardized parity-based error checking.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentUS11675716B2Techniques for command bus training to a memory device
Publication Date: 2023.06.13 INTEL CORP
  • US11675716B2 patent drawing
  • US11675716B2 patent drawing
  • US11675716B2 patent drawing

AI summary

Techniques for command bus training to a memory device includes triggering a memory device to enter a first or a second command bus training mode, outputting a command/address (CA) pattern via a command bus and compressing a sampled CA pattern returned from the memory device based on whether the memory device was triggered to be in the first or the second command bus training mode.