Adaptive DFE Tuning via Oscillating Reference Voltage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing adaptive decision feedback equalization (DFE) tuning methods face challenges in architectures like DDR5, where received and filtered signals necessary for error measurement are unavailable, limiting the ability to optimize tap weights effectively and reduce signal fidelity loss.
Innovation Solution
Implementing adaptive DFE tuning using an oscillating reference voltage and an adaptive tap step size, which adjusts based on error accumulators and a decay rate, to control the adaptation of DFE circuitry and minimize bit error rate (BER) by inducing quantization errors and updating tap weights.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional adaptive DFE tuning methods are used, then error measurement and tap weight optimization can be performed, but received and filtered signals necessary for error measurement are unavailable in architectures like DDR5
Solution Approach 1:
The patent introduces an intermediary oscillating reference voltage signal that mediates between the unavailable received/filtered signals and the error measurement requirement. By oscillating the reference voltage and measuring the resulting bit error rate, the system can indirectly determine channel quality without direct access to received and filtered signals, enabling DFE adaptation in DDR5 architectures.
Solution Approach 2:
The patent creates a simplified copy of the error measurement process by using an oscillating reference voltage instead of the full received and filtered signal path. This copying approach allows the essential error measurement function to be performed through a different, available signal path that mimics the necessary measurements without requiring the original unavailable signals.
2Speed
If the oscillating reference voltage amplitude is increased to induce quantization errors for faster adaptation, then convergence speed improves, but signal fidelity may deteriorate
Solution Approach 1:
The patent applies dynamics by making the oscillating reference voltage amplitude adjustable rather than fixed. The amplitude can be dynamically changed during the adaptation process - increased during initial convergence to speed up learning, then reduced or adjusted as the system approaches optimal performance to maintain signal fidelity. This dynamic adjustment resolves the contradiction between convergence speed and signal fidelity.
Solution Approach 2:
The patent uses periodic oscillation of the reference voltage to induce controlled quantization errors. By applying periodic action at specific amplitudes and frequencies, the system can systematically explore the error landscape and guide the adaptation process. The periodic nature allows for controlled, reversible perturbations that speed convergence while maintaining overall signal integrity through systematic rather than random variations.
Data Source
AI summary
A device of the present disclosure may include interface circuitry and a decision feedback equalization (DFE) tuner. The interface circuitry may be coupled to DFE circuitry by data interconnect. The DFE tuner may control adaptation of the DFE circuitry to a channel associated with the data interconnect using an oscillating reference voltage provided to the DFE circuitry. The interface circuitry may be coupled between the DFE tuner and the DFE circuitry.