NAND Read Threshold Clustering for Storage Compression

The patent segments the read threshold data by clustering similar threshold values into groups. Instead of storing all possible threshold values, the system identifies representative threshold sets from historical data and stores only these clustered representatives, significantly reducing storage volume while maintaining reliability for threshold retrieval.

The patent creates a compressed representation of read thresholds by copying only the essential characteristic data (clustered threshold values) rather than storing complete threshold tables. This copying approach retains the necessary information for reliable reads while dramatically reducing the storage footprint.

The patent applies local quality by analyzing and clustering read thresholds in a multi-dimensional space where different dimensions represent different threshold values. The clustering algorithm identifies local patterns and representative points that capture the essential characteristics of threshold data, preserving precision through intelligent selection of representative values rather than uniform quantization.

The patent performs preliminary action by pre-clustering historical read threshold data into representative groups before actual read operations. This preliminary organization of data into clustered representations allows for efficient retrieval and reduces storage requirements while maintaining the precision needed for accurate reads, as the clustering is performed in advance on historical data.