Micron Technology Incorporated

Marvell Asia Pte., Ltd.

Marvell International Limited

Marvellworld Trade Ltd.

Phong Sy Nguyen

Dung Viet Nguyen

James Fitzpatrick

Kishore Kumar Muchherla

Shashi Kiran Chilappagari

Nedeljko Varnica

Christophe Therene

Sivagnanam Parthasarathy

Jianmin Huang

Xiangang Luo

Ting Luo

Jeffrey S McNeil, Jr

Violante Moschiano

Mustafa Nazmi Kaynak

Patrick Robert Khayat

Zhengang Chen

Jeremy Binfet

Eyal En Gad

Kishore Kumar Muccherla

Systems and methods are disclosed including a memory device comprising a memory array and control logic, operatively coupled with the memory array. The control logic can perform operations comprising causing a read operation to be initiated with respect to a set of target cells of the memory array; obtaining, for a respective group of adjacent cells, respective cell state information; performing a set of strobe reads on the set of target cells; and generating, for a target cell of the set of target cells, semi-soft bit data based on the respective cell state information of the respective group of adjacent cells and on data obtained from a first strobe read and a second strobe read of the set of strobe reads performed on the target cell.

Generating semi-soft bit data during corrective read operations in memory devices

A method includes performing a read operation of a first codeword including first hard data and generating an error vector using a reliability metric of the first hard data. The first hard data and error vector are stored in first and second portions of memory. A first corrected codeword is returned that combines the error vector and the hard data from the first and second portions of memory. A read operation of a second codeword is performed, including second hard data and soft information. The hard data and soft information are stored in the first and second portions of memory. A bit of second hard data is flipped responsive to comparing a reliability metric of the bit of the second hard data to a bit flipping threshold, wherein flipping the bit includes updating the second hard data. The updated second codeword is returned resulting from reading the portions of memory.

Efficient memory use to support soft information in bit flipping decoders

A memory system configured to dynamically adjust the amount of redundant information stored in memory cells of a wordline on an integrated circuit die based on a bit error rate. For example, in response to a determination that a bit error rate of the wordline is above a threshold, the memory system can store first data items as independent first codewords of an error correction code technique into a first portion of the memory cells of the wordline, generate second data items as redundant information from the first codewords, and store the second data items in a second portion of the memory cells of the wordline. If the bit error rate is below the threshold, third data items can be stored as independent second codewords of the same length as the first codewords in the memory cells of the wordline.

Selective and dynamic deployment of error correction code techniques in integrated circuit memory devices

A memory device comprising an array of memory cells organized into a set of sub-blocks and a set of wordlines. Control logic is operatively coupled with the array of memory cells, the control logic to perform operations including: receiving a program command from a processing device, the program command including information indicative of a physical address associated with a retired wordline of the set of wordlines; in response to detecting the information within the program command, generating dummy data that is one of pseudo-random data, all one values, or all zero values; and causing the dummy data to be programmed to memory cells of multiple sub-blocks of the set of sub-blocks that are selectively connected to the retired wordline.

Program command generation with dummy data generation at a memory device

A memory system to store multiple bits of data in a memory cell. After receiving the data bits, a memory device coarsely programs a threshold voltage of the memory cell to a first level representative of a combination of bit values according to a mapping between combinations of bit values and threshold levels. The threshold levels are partitioned into a plurality of groups, each containing a subset of the threshold levels. A group identification of a first group, among the plurality of groups, containing the first level is determined for the memory cell. The memory device reads, using the group identification, a subset of the data bits back from the first memory cell, and combines the bits of the group identification and the subset to recover the entire set of data bits to finely program the threshold voltage of the memory cell to represent the data bits.

Simplified operations to read memory cells coarsely programmed via interleaved two-pass data programming techniques

A memory system to store multiple bits of data in a memory cell. A memory device coarsely programs a threshold voltage of the memory cell to a first level representative of a combination of bit values according to a mapping between bit value combinations and threshold levels. The threshold levels are partitioned into groups, each containing a subset of the threshold levels and having associated read voltages separating threshold levels in the subset. A group identification of a first group, among the groups, containing the first level is determined for the memory cell. The memory device applies read voltages of different groups, interleaved in an increasing order in a sequence, to read the memory cell when a read voltage applied is associated with the first group. The data bits read back from the memory cell are used to finely program the threshold voltage of the memory cell.

One-ladder read of memory cells coarsely programmed via interleaved two-pass data programming techniques

A command to program data to a memory device is received. Target charge levels of a set of memory cells in the memory device for a first programming step are determined based on the data. A first set of indicators are provided to the memory device. The first set of indicators indicate the target charge levels for the first programming step. Target charge levels of the set of memory cells for a second programming step are determined based on the data. A second set of indicators are provided to the memory device. The second set of indicators indicate the target charge levels for the second programming step.

Memory device programming technique using fewer latches

A memory system to generate data with a relation among data groups for reliably storing a predetermined number of bits per memory cell in memory cells. For example, from first groups of date bits, a second group of data bits is generated. Data groups of the predetermined number is formed to have the first groups and the second group and a predetermined relation (e.g., XOR or XNOR) among the data groups. Threshold levels of memory cells in a memory cell group are determined based on a predetermined mapping, where a threshold level of each memory cell is determined to represent one bit from each of the data groups. In the predetermined mapping, bit values represented by any two successive threshold levels differ by one bit. Threshold voltages in the memory cell group are programmed according to the threshold levels to store the data groups with improved reliability.

Data programming techniques to store multiple bits of data per memory cell with high reliability

A method includes receiving a command to write data to a memory device and writing the data to a first memory tier of the memory device. The first memory tier of the memory device is a dynamic memory tier that utilizes single level cells (SLCs), multi-level cells (MLCs), and triple level cells (TLCs). The method further includes migrating the data from the first memory tier of the memory device to a second memory tier of the memory device. The second memory tier of the memory device is a static memory tier that utilizes quad level cells (QLCs).

Memory sub-system data migration

The present disclosure describes apparatuses and methods for artificial intelligence-enabled management of storage media. In some aspects, a media access manager of a storage media system receives, from a host system, host input/output commands (I/Os) for access to storage media of the storage media system. The media access manager provides information describing the host I/Os to an artificial intelligence engine and receives, from the artificial intelligence engine, a prediction of host system behavior with respect to subsequent access of the storage media. The media access manager then schedules, based on the prediction of host system behavior, the host I/Os for access to the storage media of the storage system. By so doing, the host I/Os may be scheduled to optimize host system access of the storage media, such as to avoid conflict with internal I/Os of the storage system or preempt various thresholds based on upcoming idle time.

Artificial intelligence-enabled management of storage media access

Control logic in a memory device identifies a first plurality of groups of programming distributions, wherein each group comprises a subset of programming distributions associated with a portion of a memory array of the memory device configured as quad-level (QLC) memory. During a first pass of a multi-pass programming operation, the control logic coarsely programs memory cells in the portion configured as QLC memory to initial values representing a second plurality of pages of host data and stores, in a portion of the memory array of the memory device configured as single-level cell (SLC) memory, an indicator of the first plurality of groups of programming distributions with which each of the coarsely programmed memory cells is associated. During a second pass of the multi-pass programming operation, the control logic reads the coarsely programmed initial values from the first pass based on the indicator of the first plurality of groups of programming distributions and finely programs the memory cells in the portion configured as QLC memory to final values representing the second plurality of pages of host data.

Double interleaved programming of a memory device in a memory sub-system

Systems and methods for selecting an optimal error recovery procedure for correcting a read error in a solid-state drive are provided. A machine learning model is trained to forecast which error recovery procedure of a plurality of error recovery procedures is most likely to achieve a predetermined goal given a state of a solid-state drive. The predetermined goal is based on at least one of a read latency and a failure rate of the solid-state drive. A current state of the solid-state drive is determined. An error recovery procedure is selected from among the plurality of error recovery procedures by inputting the current state of the solid-state drive into the trained machine learning model, thereby triggering the trained machine learning model to output the selected error recovery procedure. The selected error recovery procedure is executed to recover data from the solid-state drive.

Solid-state drive error recovery based on machine learning

Exemplary methods, apparatuses, and systems include receiving a request for a segment of data. The requested segment data is one of a plurality of segments of data in a stripe of data. A failure to decode the requested segment is detected. Each of the plurality of segments in the stripe other than the requested segment are read. Reading each segment includes reading raw encoded data and attempting to decode the raw encoded data, the result of reading each segment including decoded data when decoding is successful and the raw encoded data when decoding fails. A combined result of each read is generated. The combining includes combining decoded data for segments that were successfully decoded and the raw encoded data for segments for which decoding failed. A statistical model for the requested segment is updated using the combined result. The requested segment is decoded using the updated statistical model.

Growing community of inventors

Livermore, CA, United States of America

Phong Sy Nguyen