Abstract: A data storage device includes a memory device and a controller coupled to the memory device. The controller is configured to receive a read command form a host device, collect environment data of the memory device, decode data associated with the read command, determine a bit error rate (BER) of the decoded data, compare the BER to a threshold, and determine whether the data associated with the read command is to be relocated. The environment data includes temperature, number of program/erase cycles, amount of grown defects, number of past relocations and time since last data relocation. The controller is further configured to dynamically adjust the threshold based on the collected environment data and an amount of time that has passed since a last relocation of the read command data.

Abstract: Methods and apparatus are disclosed for managing the storage of dynamic neural network data within bit-addressable memory devices, such phase change memory (PCM) arrays or other storage class memory (SCM) arrays. In some examples, a storage controller determines an expected amount of change within data to be updated. If the amount is below a threshold, an In-place Write is performed using bit-addressable writes via individual SET and RESET pulses. Otherwise, a modify version of an In-place Write is performed where a SET pulse is applied to preset a portion of memory to a SET state so that individual bit-addressable writes then may be performed using only RESET pulses to encode the updated data. In other examples, a storage controller separately manages static and dynamic neural network data by storing the static data in a NAND-based memory array and instead storing the dynamic data in a SCM array.

Abstract: Methods and apparatus are disclosed for managing the storage of static and dynamic neural network data within a non-volatile memory (NVM) die for use with deep neural networks (DNN). Some aspects relate to separate trim sets for separately configuring a static data NVM array for static input data and a dynamic data NVM array for dynamic synaptic weight data. For example, the static data NVM array may be configured via one trim set for data retention, whereas the dynamic data NVM array may be configured via another trim set for write performance. The trim sets may specify different configurations for error correction coding, write verification, and read threshold calibration, as well as different read/write voltage thresholds. In some examples, neural network regularization is provided within a DNN by setting trim parameters to encourage bit flips to avoid overfitting. Some examples relate to managing non-DNN data, such as stochastic gradient data.

Abstract: Methods and apparatus are disclosed for implementing machine learning data augmentation within the die of a non-volatile memory (NVM) apparatus using on-chip circuit components formed on or within the die. Some particular aspects relate to configuring under-the-array or next-to-the-array components of the die to generate augmented versions of images for use in training a Deep Learning Accelerator of an image recognition system by rotating, translating, skewing, cropping, etc., a set of initial training images obtained from a host device. Other aspects relate to configuring under-the-array or next-to-the-array components of the die to generate noise-augmented images by, for example, storing and then reading training images from worn regions of a NAND array to inject noise into the images.

Abstract: Systems, apparatus and methods for generation of XOR signature metadata and XOR signature management are presented. In one or more embodiments, a storage device controller includes a host interface, configured to receive one or more string lines (SLs) of data from a host, the one or more SLs to be programmed into a non-volatile memory (NVM), and processing circuitry. The processing circuitry is configured to, for each of the one or more SLs, generate signature metadata and provide the signature metadata in a header of the SL. The processing circuitry is still further configured to XOR two or more of the SLs with their respective signature metadata to generate a snapshot, and write the snapshot to the NVM.

Abstract: A memory controller includes, in one embodiment, a memory interface, a plurality of decoders, and a controller circuit. The memory interface is configured to interface with a memory having a plurality of wordlines. Each decoder of the plurality of decoders is configured to determine a bit error rate (BER). The controller circuit configured to generate a plurality of bit-error-rate estimation scan (BES) hypotheses for one wordline of the plurality of wordlines, divide the plurality of BES hypotheses among the plurality of decoders, receive BER results from the plurality of decoders based on the plurality of BES hypotheses, and adjust one or more read locations of the one wordline based on the BER results from the plurality of decoders.

Abstract: A storage apparatus includes non-volatile memory cells formed on a memory die, each memory cell configured to hold bits of data, and a control circuit. The control circuit is configured to calculate parity data for data to be stored in the memory cells and program the memory cells to first distributions. The control circuit is also configured to read memory cells in the first distributions, recover the data from results of reading the memory cells in the first distributions combined with the parity data, and further program the memory cells from the first distributions to second distributions to store the data. To improve the accuracy of recovering the encoded foggy phase data, techniques are presented to calibrate the voltage levels used in sensing the foggy state distributions.

Abstract: An integrated memory assembly comprises a memory die and a control die bonded to the memory die. The memory die includes a memory structure of non-volatile memory cells. The control die is configured to program user data to and read user data from the memory die based on one or more operational parameters. The control die is configured to calibrate the one or more operational parameters for the memory die. The control die is also configured to perform testing of the memory die using the calibrated one or more operational parameters.

Abstract: A read threshold voltage can vary over time due to process variation, data retention issues, and program disturb conditions. A storage system can calibrate the read threshold voltage using data from a decoded codeword read from a wordline in the memory. For example, the storage system can use the data instead of syndrome weight in a bit error rate estimate scan (BES). As another example, the storage system can use the data to generate a bit error rate distribution, which can be used instead of a cell voltage distribution histogram. Using these techniques can help reduce latency and power consumption, increase throughput, and improve quality of service.

Abstract: A data storage device includes a memory device and a controller coupled to the memory device. The controller comprises an XOR module, an ECC module, a scrambler, an encoder, and comparison logic. The controller is configured to retrieve data from the memory device, decode the retrieved data, execute XOR protection logic on the decoded data, encode the decoded data, and compare the encoded data to the retrieved data stored in the memory device.

Abstract: A storage system generates a low-density parity check (LDPC) code from a plurality of subcodes. The storage system stores each subcode in a different page of a word line in the memory. The subcode can be stored in one plane in the memory or across multiple planes. When the subcodes are stored across multiple planes, they can be stored in a checkboard pattern.

Abstract: A memory includes, in one embodiment, one or more storage elements; read/write circuitry; and compressed bit circuitry. The read/write circuitry is configured to read a set of hard bits from the one or more storage elements, and sense a set of soft bits while reading the set of hard bits from the one or more storage elements, the set of soft bits having a first fixed size, and the set of soft bits indicating a reliability of the set of hard bits. The compressed soft bit circuitry is configured to generate, with a fixed size soft bit lossy compression algorithm, a fixed size compressed soft bits by compressing the set of soft bits, the fixed size compressed soft bits having a second fixed size that is smaller than the first fixed size, and output the fixed size compressed soft bits to a memory-to-controller bus.

Abstract: A memory includes, in one embodiment, NAND elements; read/write circuitry; and compressed soft-bit circuitry. The compressed soft-bit circuitry is configured to determine or receive one or more NAND conditions and then determine a soft-bit delta and select a compression scheme based on the NAND conditions. The read/write circuitry is configured to read a set of hard bits from the NAND elements and sense a first set of soft-bits using the determined soft-bit delta while reading the set of hard bits from the NAND elements. The first set of soft-bits has a first fixed size, and each soft-bit of the first set of soft-bits indicates a reliability of a corresponding hard bit of the set of hard bits. The compressed soft-bit circuitry is also configured to generate a second set of soft-bits based on the selected compression scheme and output the second set of soft-bits to a controller.

Abstract: A storage system generates a low-density parity check (LDPC) code from a plurality of subcodes. The storage system stores each subcode in a different page of a word line in the memory. The subcode can be stored in one plane in the memory or across multiple planes. When the subcodes are stored across multiple planes, they can be stored in a checkboard pattern.

Abstract: Apparatus, media, methods, and systems for data storage systems and methods for optimized scheduling of background management operations. A data storage system may comprise a controller. The controller is configured to determine a timeout value of an adaptive timeout parameter of the data storage system. The controller is configured to determine whether a first host operation is received. The controller is configured to, when the first host operation is not received, determine whether the timeout value satisfies a threshold value. The controller is configured to, when the timeout value satisfies the threshold value, cause one or more background management operations to be executed at the data storage system.

Abstract: A storage device includes a controller that can dynamically adjust the zone active limit (ZAL) for a zoned namespace (ZNS). Rather than assuming a worst-case scenario for the ZNS, the ZAL can be dynamically adjusted, even after providing the ZAL to a host device. In so doing, device behavior changes due to factors such as temperature, failed or flipped bit count, and device cycling can be considered as impacting the ZAL. The ZAL can then be adjusted over time, and the new ZAL can be communicated to the host device. As such, rather than a fixed, worst-case ZAL, the host device will receive updated ZAL values over time as the device performs.

Abstract: Read reference levels are calibrated by calibrating integration times. An integration time is the length of time for which the charge on a sense node is allowed to change while the memory cell is being sensed. Calibrating the integration time is much faster than calibrating the reference voltage itself. This is due, in part, to reducing the number of different reference voltages that need to be applied during calibration. Calibrating the integration time may use different test integration times for a given read reference voltage, thereby reducing the number of read reference voltages. Hence, calibrating the integration time(s) is very efficient timewise. Also, power consumption may be reduced.

Abstract: The present disclosure generally relate to dynamically changing predictive latency related attributes to increase the deterministic window (DTWIN) of operation. The host device workload characteristics as well as the memory device's current condition provide valuable information for the duration of the DTWIN. If the memory device is near the end of life, then the DTWIN duration will be smaller. Additionally, if the workload from the host device is heavy, then the DTWIN duration will also be smaller. Rather than utilizing a fixed DTWIN duration based upon worst case scenarios for host device workload and memory device condition, dynamically adjusting the DTWIN duration based upon the workload and condition will provide a DTWIN duration that can gradually decrease over time from a much longer DTWIN duration than is currently available.

Abstract: A storage apparatus includes non-volatile memory cells formed on a memory die, each memory cell configured to hold bits of data, and a control circuit. The control circuit is configured to calculate parity data for data to be stored in the memory cells and program the memory cells to first distributions. The control circuit is also configured to read memory cells in the first distributions, recover the data from results of reading the memory cells in the first distributions combined with the parity data, and further program the memory cells from the first distributions to second distributions to store the data. To improve the accuracy of recovering the encoded foggy phase data, techniques are presented to calibrate the voltage levels used in sensing the foggy state distributions.

Abstract: A control circuit on a control die compensates for interference caused by adjacent memory cells on target memory cells on a memory die. The compensation may be based on the data states of the adjacent memory cells. Data latches may be used to store data states of the memory cells. However, reading the target memory cells can over-write the data states of the adjacent memory cells in the data latches. The control die may store data state information for the adjacent memory cells prior to sensing the target memory cells (e.g., prior to a decoding error of a codeword in the target cells). Saving the data state information on the control die reduces storage requirements of the memory die and alleviates the need to sense the adjacent memory cells again if decoding the codeword in the target memory cells fails.