Abstract: The present disclosure describes apparatuses and methods for implementing a reinforcement learning-enabled low-density parity check (LDPC) decoder. In aspects, an RL-enabled LDPC decoder processes, as part of a first decoding iteration, data of a channel to generate LDPC state information and provides the LDPC state information to a machine learning (ML) algorithm of an RL agent. The RL-enabled LDPC decoder is then configured with LDPC decoding parameters obtained from the ML algorithm and processes, as part of a second decoding operation, the data using the decoding parameters to generate subsequent LDPC state information. The RL-enabled LDPC decoder provides decoded data of the channel based on the subsequent LDPC state information. By using the LDPC decoding parameters provided by the ML algorithm of the RL agent, the RL-enabled LDPC decoder may decode channel data in fewer decoding iterations or with a higher success rate, thereby improving LDPC decoding performance.

Abstract: A controller estimates optimal read threshold values for a memory device using deep learning. The memory device includes multiple pages coupled to select word lines in a memory region. The controller performs multiple read operations on a select type of page for each word line using multiple read threshold sets, obtains fail bit count (FBC) information associated with each read operation, and determines an optimal read threshold set for each word line based on the FBC information. When optimal read threshold sets for the select word lines are different each other, the controller predicts a best read threshold set using the optimal read threshold sets.

Abstract: The present disclosure describes apparatuses and methods for implementing an adaptive low-density parity check (LDPC) decoder. In various aspects, an adaptive LDPC decoder processes a first portion of data using first parameters effective to change a status of the LDPC decoder. The LDPC decoder selects second parameters of the LDPC decoder based on the status of the LDPC decoder. The LDPC decoder then processes a second portion of the data with the LDPC decoder using the second parameters and provides decoded data of the channel based on at least the processing the first portion of the data using the first parameters and the processing of the second portion of the data using the second parameters. By adaptively altering the decoding parameters based the status of the decoder, the adaptive LDPC decoder may decode channel data in fewer decoding iterations or with a higher success rate, thereby improving LDPC decoding performance.

Abstract: Disclosed is a computer-implemented method for optimizing read thresholds of a memory device using a deep neural network engine, comprising reading, using a set of read threshold voltages applied to the memory device, data from the memory device under a first set of operating conditions that contribute to read errors in the memory device, producing a labeled training data set using the set of read threshold voltages under the first set of the operating conditions, determining, based on characteristics of the memory device, a number of layers, a size of each layer, and a number of input and output nodes of the deep neural network engine, training the deep neural network engine using the labeled training data set, and using the trained deep neural network engine to compute read thresholds voltage values under a second set of operating conditions.

Abstract: A descrambler receives data from a memory device. The descrambler calculates a sub-syndrome weight for multiple bits in each of the plurality of descrambled sequences using a set parity check matrix to generate multiple sub-syndrome weights, one for each of the plurality of descrambled sequences. The descrambler selects a sub-syndrome weight among the multiple sub-syndrome weights. The descrambler determines, as a correct scrambler sequence for descrambling the data, a scrambler sequence corresponding to the selected sub-syndrome weight, among the plurality of scrambler sequences.

Abstract: Disclosed are methods, systems and devices for decoding data read from a memory device, including receiving noisy data from a first memory location included in a word line zone of the memory device, identifying the word line zone and a prior successful decoder parameter associated with the word line zone, decoding the noisy data using the prior successful decoder parameter used in a prior successful decoding with respect to a second memory location included in the same word line zone, determining whether the decoding based on the prior successful decoder parameter has succeeded, maintaining, upon a determination that the decoding has succeeded, the prior successful decoder parameter as a decoder parameter for the first memory location, and decoding, upon a determination that the decoding operation has failed, the noisy data read from the first memory location by using another decoder parameter selected from a set of predefined decoder parameters.

Abstract: A memory system includes a memory device and a controller. The controller determines a target word line group to which a target word line corresponding to a read command belongs. The controller identifies a reference voltage corresponding to the target word line group. The controller controls the memory device to perform a read operation on a target page coupled to the target word line, using the reference voltage.

Abstract: Memory controllers, decoders and methods to selectively perform bit-flipping (BF) decoding and min-sum (MS) decoding on codewords of an irregular low-density parity-check (LDPC) code. Bit-flipping (BF) decoding is executed with respect to variable nodes having relatively high column weights. MS decoding is executed with respect to variable nodes having relatively low column weights. A column-weight threshold is used to group the variable nodes into the higher and lower column weight groups. The two decoding techniques exchange results during the overall decoding process.

Abstract: A controller optimizes a read threshold value for a memory device using model-less regression. The controller performs read operations on cells using read threshold voltage values. The controller measures probability values for the multiple read threshold voltage values, and estimates a threshold voltage distribution curve based on the multiple read threshold voltage values and the measured probability values using a set regression formula. The controller determines a read threshold voltage value corresponding to a set point on the threshold voltage distribution curve, and performs a read operation on the cells using the read threshold voltage value.

Abstract: A controller optimizes a read threshold value for a memory device using model-less regression. The controller performs read operations on cells using read threshold voltage values. The controller measures probability values for the multiple read threshold voltage values, and estimates a threshold voltage distribution curve based on the multiple read threshold voltage values and the measured probability values using a set regression formula. The controller determines a read threshold voltage value corresponding to a set point on the threshold voltage distribution curve, and performs a read operation on the cells using the read threshold voltage value.

Abstract: Memory controllers bit-flipping (BF) decoders and methods that selectively apply a checksum-aided error reduction (CA-ER) scheme to BF decoding of a low-density parity-check (LDPC) code. In decoding a codeword, a hard decision value resulting from decoding a select variable node is changed when a first condition is satisfied to yield an updated hard decision value. Also, when the first condition is satisfied, a current checksum value after processing the select variable node is updated using the updated hard decision value. The CA-ER scheme is applied when the updated checksum value is not reduced to a set minimum and a second condition based on a previous checksum value, calculated after a previous variable node is processed, is satisfied.

Abstract: A controller estimates optimal read threshold values for a memory device using deep learning. The memory device includes multiple pages coupled to select word lines in a memory region. The controller performs multiple read operations on a select type of page for each word line using multiple read threshold sets, obtains fail bit count (FBC) information associated with each read operation, and determines an optimal read threshold set for each word line based on the FBC information. When optimal read threshold sets for the select word lines are different each other, the controller predicts a best read threshold set using the optimal read threshold sets.

Abstract: A controller optimizes read threshold values for a memory device using domain transformation. The controller determines, for the decoded data of each read operation, an asymmetric ratio (AR) and the number of unsatisfied checks (USCs), the AR indicating a ratio of the number of a first binary value to the number of a second binary value in the decoded data. The controller determines a Z-axis such that AR values of the threshold sets are arranged in a set order along the Z-axis. The controller determines an optimum read threshold set using coordinate values on the Z-axis, which correspond to a set AR value and a set USC value.

Abstract: A memory controller optimizes read threshold values for a memory device using multi-dimensional search. The controller performs a read operation on cells using a pair of default read threshold values on a multi-dimensional plane. When the read operation has failed, the controller determines program states of cells and a pair of next read threshold values based on the program states and performs an additional read operation using the next read threshold values.

Abstract: A memory system includes a memory device and a controller. The controller performs multiple read operations on a target block, using a first duster of read threshold voltages. The controller generates a second duster of read threshold voltages using the first cluster when a difference between the maximum number of fail bits and the minimum number of fail bits associated with the multiple read operations exceeds a threshold. The controller splits pages in the target block into a first group of pages for the first cluster and a second group of pages for the second cluster. The controller performs additional read operations on the first group of pages using the first cluster and on the second group of pages using the second cluster.

Abstract: Multiple memory systems with respective decoders employ a low latency implementation of a read recovery level feature in decoding data. The decoding comprises receiving from a host a read request for decoding read data at a first recovery level by a first memory system, a first decoder of the first memory system being set at a second recovery level with a corresponding maximum iteration number when the read request is received by the first memory system; and operating the first decoder, after a set time elapses, to decode the read data at the second recovery level. A second decoder of a second memory system is set at the first recovery level for at least part of the time during which the first decoder operates to decode the read data at the second recovery level.

Abstract: Devices, systems and methods for reducing complexity of a bit-flipping decoder for quasi-cyclic (QC) low-density parity-check (LDPC) codes are described. An example method includes receiving a noisy codeword that is based on a transmitted codeword generated from an irregular QC-LDPC code, the irregular QC-LDPC code having an associated parity matrix, storing, based on a weight of a plurality of columns of the parity matrix of the irregular QC-LDPC code, a portion of the noisy codeword corresponding to the plurality of columns in a first buffer of a plurality of buffers, and accessing and processing the portion of the noisy codeword that includes applying a vertically shuffled scheduling (VSS) scheme that uses a plurality of processing units to determine a candidate version of a portion of the transmitted codeword that corresponds to the portion of the noisy codeword.

Abstract: A descrambler receives data from a memory device. The descrambler calculates a sub-syndrome weight for multiple bits in each of the plurality of descrambled sequences using a set parity check matrix to generate multiple sub-syndrome weights, one for each of the plurality of descrambled sequences. The descrambler selects a sub-syndrome weight among the multiple sub-syndrome weights. The descrambler determines, as a correct scrambler sequence for descrambling the data, a scrambler sequence corresponding to the selected sub-syndrome weight, among the plurality of scrambler sequences.

Abstract: Disclosed are devices, systems and methods for improving a bit-flipping algorithm for an irregular LDPC code in a non-volatile memory device. An example method includes receiving a noisy codeword, the codeword having been generated from an irregular low-density parity-check code, performing a first iteration of a bit-flipping algorithm on the noisy codeword, computing a first syndrome based on an output codeword of the first iteration, determining that the first syndrome comprises a non-zero vector and no bits of the noisy codeword were flipped during the first iteration of the bit-flipping algorithm, flipping, based on the determining, at least one bit of the output codeword, the at least one bit corresponding to a variable node of the plurality of variable nodes with a smallest column weight connected to one or more unsatisfied check nodes of the plurality of check nodes, and computing, subsequent to the flipping, a second syndrome.

Abstract: Decoders are provided for memory systems. A decoder includes a seed generator that generates seeds based on a physical address corresponding to a read request from a host; a descrambling module that receives a sequence from a storage area among, multiple storage areas, corresponding to the physical address, and descrambles the sequence using the seeds to generate multiple descrambled sequences; and a selector that selects one of descrambled sequences based on syndrome weight values of the descrambled sequences.