Skhynix Inc.

Lsi Corporation

Jianqing Chen

Fan Zhang

Meysam Asadi

Xuanxuan Lu

Aman Bhatia

Chenrong Xiong

Johnson Yen

Lei Chen

Madhusudan Kalluri

Sancar Kunt Olcay

Sathiyanarayanan Sampath Kumar

Embodiments provide a scheme for non-parametric PV-level modeling and an optimal read threshold voltage estimation in a memory system. A controller is configured to: generate multiple optimal read threshold voltages corresponding to multiple sets of two cumulative distribution function (CDF) values, respectively; perform read operations on the cells using a plurality of read threshold voltages; generate cumulative mass function (CMF) samples based on the results of the read operations; receive first and second CDF values, selected from among a plurality of CDF values, each CDF value corresponding to each CMF sample; and estimate an optimal read threshold voltage corresponding to the first and second CDF values, among the multiple optimal read threshold voltages.

Systems and methods for non-parametric PV-level modeling and read threshold voltage estimation

Disclosed are devices, systems and methods for providing hardware implementations of a quasi-cyclic syndrome decoder. An example method of reducing the complexity of a decoder includes receiving a noisy codeword that is a based on a transmitted codeword generated from a quasi-cyclic linear code; computing a plurality of syndromes based on the noisy codeword; selecting a first syndrome from the plurality of syndromes; generating a memory cell address as a function of the first syndrome; reading, based on the memory cell address, a coset leader corresponding to the first syndrome; and determining, based on the noisy codeword and the coset leader, a candidate version of the transmitted codeword.

Hardware implementations of a quasi-cyclic syndrome decoder

Disclosed are devices, systems and methods for reducing the latency of a quasi-cyclic linear code decoder. An example method includes receiving a noisy codeword, the codeword having been generated from a quasi-cyclic linear code and provided to a communication channel prior to reception by the decoder; computing a syndrome based on the noisy codeword; generating a plurality of memory cell addresses, a first memory cell address being a function of the syndrome and subsequent memory cell addresses being within a predetermined address range of the function of the syndrome; reading, in a parallel manner to reduce the latency of the decoder, a plurality of error patterns from the plurality of memory cell addresses and computing a checksum for each of the plurality of error patterns; and determining, based on the checksum for each of the plurality of error patterns, a candidate version of the transmitted codeword.

Reducing the latency of a syndrome-based quasi-cyclic decoder

Techniques related to improving a performance related to at least data reads from a memory are described. In an example, a computer system hosts a regression model that includes a neural network. The neural network is trained based on training data that is measured under different combinations of operational conditions and storage conditions. In operation, actual operational and storage conditions associated with the memory are input to the regression model. The neural network outputs a voltage read threshold based on these actual conditions. The computer system uses the voltage read threshold to read data stored in the memory.

Deep learning based regression framework for read thresholds in a NAND flash memory

A method for data sequence detection includes generating a first sample stream, equalizing the first sample stream to generate a first equalized sample stream, and buffering the first equalized sample stream. The first sample stream is interpolated to generate a second sample stream. The second sample stream is equalized to generate a second equalized sample stream. In a first processing path, the samples of the buffered first equalized sample stream are filtered using a first noise predictive filter bank to generate a first set of noise sample streams. In a second parallel processing path, the samples of the buffered first equalized sample stream are interpolated using a second interpolation filter to generate an interpolated sample stream and the interpolated sample stream is filtered to generate a second set of noise sample streams. The first equalized sample stream and the second equalized sample stream are processed to generate adapted filter coefficients for the second interpolation filter. The first and second set of noise sample streams are diversity combined to generate a set of combined noise sample streams. A data sequence is detected using the set of combined noise sample streams.

Growing community of inventors

San Jose, CA, United States of America

Jianqing Chen