Zero Force Equalizer Circuit for Low-Complexity Read Signal Equalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data storage devices face challenges in implementing zero force equalization for read data signals, which is computationally difficult and expensive, and current methods like LMS adaptive techniques are prone to misequalization due to noisy terms.
Innovation Solution
A channel circuit using a zero force adaptive algorithm that adapts equalization filter tap weights based on a known data signal derived from previous read operations, employing a real-time adaptive gradient algorithm without matrix calculations, and combining it with a least mean squared adaptive algorithm for improved equalization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If zero force equalization is implemented using traditional methods, then equalization performance is improved, but computational complexity and manufacturing cost increase significantly
Solution Approach 1:
The patent segments the equalization process into two distinct phases: a training phase where tap weights are adapted using zero force equalization on known data patterns, and an operational phase where the adapted weights are applied to equalize unknown data. This segmentation allows the computationally intensive adaptation to occur only during training, reducing overall computational complexity while maintaining equalization performance.
Solution Approach 2:
The patent performs preliminary adaptation of the equalizer tap weights during a training phase using known data patterns before actual data processing begins. This preliminary action prepares the equalizer in advance, so that during normal operation, only lightweight filtering is required, significantly reducing real-time computational complexity while preserving equalization effectiveness.
2Measurement precision
If zero force equalization is implemented using traditional methods, then equalization performance is improved, but manufacturing cost increases
Solution Approach 1:
The equalizer is pre-adapted during manufacturing or initial operation using known training patterns. This preliminary action allows the device to be manufactured with simpler, less expensive hardware since the computationally intensive adaptation occurs once during training rather than requiring expensive real-time computational resources during normal operation.
Solution Approach 2:
The patent uses known data patterns as copies or references during the training phase to adapt the equalizer. These known patterns serve as templates that guide the adaptation process, allowing the system to learn optimal tap weights without requiring complex real-time computation during actual data processing, thereby reducing manufacturing costs.
3Device complexity
If LMS adaptive techniques are used, then equalization is simpler to implement, but misequalization occurs due to noisy terms
Solution Approach 1:
The patent implements a feedback mechanism where the equalizer output is compared against known training patterns, and the error signal is used to adjust the tap weights. This feedback loop allows the system to converge to optimal weights that minimize equalization error, improving reliability while maintaining implementation simplicity through the use of standard adaptive filtering techniques.
Solution Approach 2:
The patent changes the parameter being optimized from minimizing mean-square error (LMS approach) to forcing the equalized output to match known patterns exactly (zero force approach). This parameter change eliminates the noisy gradient estimation problem inherent in LMS, improving equalization accuracy while maintaining computational tractability through the training-operations separation.
Data Source
AI summary
Example channel circuits, data storage devices, and methods for using zero force adaptation to equalize a read data signal based on known data are described. A known user data signal may be determined during prior read operations and used with a residue term from the equalized read data signal to adapt the tap weights for an equalizer filter using a zero force adaptation algorithm. For example, the known user data signal may be determined by a soft output detector (e.g., SOVA detector) or full or partial decoding by an iterative decoder (e.g., LDPC decoder) and fed back for adapting the equalizer.