Abstract: Method and apparatus for managing data in a non-volatile memory (NVM) of a storage device, such as a solid-state drive (SSD). A circuit measures programming and reading temperatures for a set of memory cells in the NVM. Error rates are determined for each of the reading operations carried out upon the data stored in the memory cells. A code rate for the NVM is adjusted to maintain a selected error rate for the memory cells. The code rate is adjusted in relation to a cross-temperature differential (CTD) value exceeding a selected threshold. The code rate can include an inner code rate as a ratio of user data bits to the total number of user data bits and error correction code (ECC) bits in each code word written to the NVM, and/or an outer code rate as a strength or size of a parity value used to protect multiple code words.

Abstract: Weak erase detection and mitigation techniques are provided that detect permanent failures in solid-state storage devices. One exemplary method comprises obtaining an erase fail bits metric for a solid-state storage device; and detecting a permanent failure in at least a portion of the solid-state storage device causing weak erase failure mode by comparing the erase fail bit metric to a predefined fail bits threshold. In at least one embodiment, the method also comprises mitigating for the permanent failure causing the weak erase failure mode for one or more cells of the solid-state storage device. The mitigating for the permanent failure comprises, for example, changing a status of the one or more cells to a defective state and/or a retired state. The detection of the permanent failure causing the weak erase failure mode comprises, for example, detecting the weak erase failure mode without an erase failure.

Abstract: Method and apparatus for decoding error correction code (ECC) code words. Reference voltages are used to extract a selected code word from a communication channel. The selected code word is processed by an ECC decoder, and an initial syndrome weight is determined indicative of unresolved parity errors. A coarse search operates to concurrently adjust, over a first succession of iterations, each of the reference voltages. A subsequent fine search operates, over a second succession of iterations, to individually adjust the reference voltages. Decoding and syndrome weight determination continues over each iteration until a minimum syndrome weight is obtained, after which a user data content of the code word is decoded. The coarse search may transition the decoder from a saturated operational region to a linear operational region. The decoder may be a low density parity check (LDPC) decoder.

Abstract: A connection between a user device and a network server is established. Via the connection, a deep learning network is formed for a processing task. A first portion of the deep learning network operates on the user device and a second portion of the deep learning network operates on the network server. Based on cooperation between the user device and the network server, a boundary between the first portion and the second portion of the deep learning network is dynamically modified based on a change in a performance indicator that could affect the processing task.

Abstract: Systems and methods are disclosed for open block stability scanning. When a solid state memory block remains in an open state, where the block is only partially filled with written data, for a prolonged period of time, a circuit may perform a scan on the block to determine the stability of the stored data. When the scan indicates that the data is below a stability threshold, the data may be refreshed by reading the data and writing it to a new location. When the scan indicates that the data is above a stability threshold, the circuit may extend the time period in which the block may remain in the open state.

Abstract: Method and apparatus for managing data in a semiconductor memory, such as but not limited to a three dimensional (3D) NAND flash memory array. In some embodiments, the memory has non-volatile memory cells arranged into addressable blocks. Each memory cell is configured to store multiple bits. A program/read control circuit programs data sets to and reads data sets from the memory cells in the addressable blocks to service a sequence of host access commands. The circuit concurrently performs background reads in conjunction with the servicing of the host access commands. The background reads result in the reading of a different data set from each of the addressable blocks over each of a succession of time intervals of selected duration. The background reads condition the memory cells prior to a first read operation associated with the host access commands.

Abstract: Systems and methods presented herein provide for monitoring block, page, and/or stripe degradation. In one embodiment, a controller is operable to scan a first block of memory to identify a failure in a portion of the first block. The controller suspends input/output (I/O) operations to the failed portion of the first block, and tests the failed portion of the first block to determine if the failure is a transient failure. Testing includes loading the portion of the first block with data, and reading the data from the loaded portion of the first block. If the failure subsides after testing, the controller is further operable to determine that the failure is a transient failure, and to resume I/O operations to the portion of the first block.

Abstract: Systems and methods are disclosed for wear-based access optimization. An apparatus may comprise a circuit configured to perform a data access operation at a target location of a memory, and determine a wear value of the target location. The circuit may compare the wear value to global wear value of other locations of the drive, and adjust data access parameters for the target location based on the comparison.

Abstract: Systems and methods for improving data refresh in flash memory are described. In one embodiment, the method includes identifying a first garbage collection unit (GCU) of the storage system, computing a parity function in relation to the first GCU, identifying a data impairment in a first block, the first block being from the N blocks in the first GCU, removing the first block from the first GCU after identifying the data impairment in the first block, and recomputing the parity function when the first block is not cloned.

Abstract: Systems and methods presented herein provide for failure detection and data recovery in a storage system. In one embodiment, a method operable in a storage system comprises locating failures in data blocks in storage area of a storage device, categorizing the failures into block groups, each block group comprising one or more data blocks having failures, and halting input/output (I/O) operations to data blocks in a first of the block groups due to the failures of the first block group. The method also includes detecting additional failures in one or more data blocks of other block groups remaining in the storage area, and determining when to fail the storage area of the storage device based on the detected failures.

Abstract: Method and apparatus for managing data in a semiconductor memory, such as but not limited to a three dimensional (3D) NAND flash memory array. In some embodiments, the memory has non-volatile memory cells arranged into addressable blocks. Each memory cell is configured to store multiple bits. A program/read control circuit programs data sets to and reads data sets from the memory cells in the addressable blocks to service a sequence of host access commands. The circuit concurrently performs background reads in conjunction with the servicing of the host access commands. The background reads result in the reading of a different data set from each of the addressable blocks over each of a succession of time intervals of selected duration. The background reads condition the memory cells prior to a first read operation associated with the host access commands.

Abstract: Systems and methods presented herein provide for monitoring block, page, and/or stripe degradation. In one embodiment, a controller is operable to scan a first block of memory to identify a failure in a portion of the first block. The controller suspends input/output (I/O) operations to the failed portion of the first block, and tests the failed portion of the first block to determine if the failure is a transient failure. Testing includes loading the portion of the first block with data, and reading the data from the loaded portion of the first block. If the failure subsides after testing, the controller is further operable to determine that the failure is a transient failure, and to resume I/O operations to the portion of the first block.

Abstract: Method and apparatus for managing data in a semiconductor memory, such as but not limited to a three dimensional (3D) NAND flash memory array. In some embodiments, the memory has non-volatile memory cells arranged into addressable blocks. Each memory cell is configured to store multiple bits. A program/read control circuit programs data sets to and reads data sets from the memory cells in the addressable blocks to service a sequence of host access commands. The circuit concurrently performs background reads in conjunction with the servicing of the host access commands. The background reads result in the reading of a different data set from each of the addressable blocks over each of a succession of time intervals of selected duration. The background reads condition the memory cells prior to a first read operation associated with the host access commands.

Abstract: Systems and methods are disclosed for open block management. In certain embodiments, an apparatus may comprise a circuit configured to determine an error sensitivity of a last-written page of a block of a solid state memory that is in an open state where the block has not been fully filled with data. The error sensitivity may include a value that represents a susceptibility to developing data errors while in the open state. The circuit may perform a first error mitigation procedure when the error sensitivity is lower than a first threshold, include increasing an open block timeout period applied to the last-written page. The circuit may perform a second error mitigation procedure when the error sensitivity is higher than the first threshold, including copying data from the block to a new location when a first open block timeout is reached.

Abstract: Systems and methods are disclosed for open block refresh management. In certain embodiments, an apparatus may comprise a circuit configured to monitor an amount of time a block of a solid-state memory remains in an open state where the block has not been fully filled with data, and in response to reaching an open block time limit, compare an amount of the block already written with data against a threshold amount. When less than a threshold amount of the block has been written with data, the circuit may refresh data from a last N pages from the block by writing the data to a new location, N being a number of pages less than all pages in the block. When more than the threshold amount of the block has been written with data, the circuit may fill a remaining unwritten amount of the block with dummy data.

Abstract: Method and apparatus for managing data in a semiconductor memory, such as but not limited to a three dimensional (3D) NAND flash memory array. In some embodiments, the memory has non-volatile memory cells arranged into addressable blocks. Each memory cell is configured to store multiple bits. A program/read control circuit programs data sets to and reads data sets from the memory cells in the addressable blocks to service a sequence of host access commands. The circuit concurrently performs background reads in conjunction with the servicing of the host access commands. The background reads result in the reading of a different data set from each of the addressable blocks over each of a succession of time intervals of selected duration. The background reads condition the memory cells prior to a first read operation associated with the host access commands.

Abstract: Method and apparatus for managing data in a stacked semiconductor memory, such as but not limited to a three dimensional (3D) NAND flash memory array. In some embodiments, a data set is written to the memory array by programming a stack of memory cells to a desired set of program states. A first set of pulses is applied to verify the memory cells conform to the desired set of program states. The verified stack of memory cells are subsequently conditioned by applying a second set of pulses to remove accumulated charge from a shared channel region of the stack. The conditioning of the memory cells reduces a step-wise increase in the number of read errors during the first read operation as compared to subsequent read operations on the memory cells.

Abstract: Method and apparatus for managing a solid state memory, such as but not limited to a NAND flash memory. In some embodiments, a storage device includes a non-volatile solid state memory and a control circuit configured to transfer user data between the memory and a host device. The control circuit maintains, in a local memory, a data structure indicative of measured readback error rates associated with memory locations in the memory in relation to erasure counts associated with the memory locations. The control circuit retires a subset of the memory locations identified by the data structure from further availability to store user data from the host device responsive to the measured readback error rates, and responsive to the erasure counts of said memory locations indicating the memory has reached an end of life (EOL) condition.

Abstract: A data storage device receives a write data command and data. The data is stored in a buffer of the data storage device. The data storage device issues a command complete status indication. After the command complete status indication is issued, the data are stored in a primary memory of the data storage device. The primary memory comprises a first type of non-volatile memory and the buffer comprises a second type of non-volatile memory that is different from the first type of non-volatile memory.

Abstract: Systems and methods for low latency acquisition of soft data from a memory cell based on a sensing time and/or a leakage current are described. In one embodiment, the systems and methods may include applying a first read voltage to a word line of a page of memory cells selected by a processor of a flash memory device for a read operation, applying a pass voltage to word lines associated with one or more different pages of memory cells of the memory block, upon applying the first read voltage sensing whether a bit line of a memory cell in the selected page conducts, measuring a side effect associated with sensing whether the bit line of the memory cell in the selected page conducts, and assigning a LLR value to the memory cell as a soft LDPC input based at least in part on the measured side effect.