Abstract: Technology for two-sided adjacent memory cell interference mitigation in a non-volatile storage system is disclosed. During reading of target memory cells, the storage system applies a suitable magnitude read pass voltage to a first unselected word line adjacent to a target word line to compensate for interference from adjacent cells on the first unselected word line while applying a suitable magnitude read reference voltage to the target word line to compensate for interference from adjacent cells on a second unselected word line on the other side of the target word line. The read pass voltage may compensate for interference due to charge being added to when programming cells on the first unselected word line after programming the target cells. The read reference voltage may compensate for interference due to charge movement near the target cells that results from charge stored in the cells on the second unselected word line.

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 formed on the memory die. 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 the 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. In some cases, the recovered data may have a high bit error rate. To handle higher bit error rates, the use of soft bit data is incorporated into an encoded foggy-fine scheme.

Abstract: A data storage device includes a memory device and a controller coupled to the memory device. The controller is configured to determine a read threshold on a wordline, adjust a read threshold voltage level associated with the read threshold, determine an adjusted read threshold at the adjusted read threshold voltage level, where the adjusted read threshold is different from the read threshold, compare the adjusted read threshold to the read threshold, and calibrate the read threshold based on the comparing. The controller is further configured to analyze a bit error rate (BER) difference based on the calibrating and/or a previous read threshold voltage level movement, choose a next target read threshold for next calibration, and read a second page at the next target read threshold.

Abstract: Recurrent Neural Networks (RNNs) wherein a non-volatile memory (NVM) array provides a memory bank for the RNN. The RNN may include a Neural Turning Machine (NTM) and the memory bank may be an NTM matrix stored in the NVM array. In some examples, a data storage device (DSD) that controls the NVM array includes both a data storage controller and a separate NTM controller. The separate NTM controller accesses the NTM matrix of the NVM array directly while bypassing flash translation layer (FTL) components of the data storage controller. Additionally, various majority wins error detection and correction procedures are described, as well as various disparity count-based procedures.

Abstract: Recurrent Neural Networks (RNNs) wherein a non-volatile memory (NVM) array provides a memory bank for the RNN. The RNN may include a Neural Turning Machine (NTM) and the memory bank may be an NTM matrix stored in the NVM array. In some examples, a data storage device (DSD) that controls the NVM array includes both a data storage controller and a separate NTM controller. The separate NTM controller accesses the NTM matrix of the NVM array directly while bypassing flash translation layer (FTL) components of the data storage controller. Additionally, various majority wins error detection and correction procedures are described, as well as various disparity count-based procedures.

Abstract: Interleaved ECC coding for key-value data storage devices. In one embodiment, a controller includes a memory interface including a namespace database; an ECC engine; a controller memory; and an electronic processor. The electronic processor is configured to receive a host write command, determine whether write access was setup as a key-value (KV) namespace in the namespace database and is associated with the host write command, and control the ECC engine and the memory interface to perform one or more program operations on the data in the memory using the interleaved ECC coding and based on the host write command in response to determining that the write access was setup as the KV namespace in the namespace database and the KV namespace is associated with the host write command.

Abstract: An adaptive-feedback-based read-look-ahead management system and method are provided. In one embodiment, a method for stream management is presented that is performed in a storage system. The method comprises performing a read look ahead operation for each of a plurality of streams; determining a success rate of the read look ahead operation of each of the plurality of streams; and allocating more of the memory for a stream that has a success rate above a threshold than for a stream that has a success rate below the threshold. Other embodiments are provided.

Abstract: Disclosed are systems and methods for efficient power management for storage devices. A method includes receiving a request to transition a flash memory to a first stand-by mode, wherein the flash memory comprises a plurality of dies. The method also includes causing one or more guard dies of the plurality of dies to transition to the first stand-by mode while causing one or more other dies of the plurality of dies to transition to a second stand-by mode, wherein the second stand-by mode is configured to consume less power than the first stand-by mode. The method also includes receiving an input/output (I/O) request for the flash memory. The method also includes causing the I/O request to be performed on the one or more guard dies that are in the first stand-by mode but not in the second stand-by mode.

Abstract: A method and apparatus for cache management and eviction polices using unsupervised reinforcement learning schemes is disclosed.

Abstract: A method and apparatus for calibrating read threshold for cells of a target wordline (WL) that may be conducted on a die, in a controller connected to a memory die, or both. Voltage values of one or more adjacent WL cells are read, and based on the voltage values of the adjacent cells, cells of the target WL are grouped. A read threshold calibration is carried out on each group. The calibration thresholds are then used for read operations on the cells of each distinct group of the target WL.

Abstract: Methods and apparatus are disclosed for implementing data augmentation within a storage controller of a data storage device based on machine learning data read from a non-volatile memory (NVM) array of a memory die. Some particular aspects relate to configuring the storage controller to generate augmented versions of training 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 and stored in the NVM array. Other aspects relate to controlling components of the memory die to generate noise-augmented images by, for example, storing and then reading training images from worn regions of the NVM array to inject noise into the images. Data augmentation based on data read from multiple memory dies is also described, such as image data spread across multiple NVM arrays or multiple memory dies.

Abstract: Boot code is loaded to the data storage device controller in a flexible manner by being able to receive chunks of the boot code from two separate locations, the host memory buffer (HMB) and the memory device, which may be a NAND device. Part of the boot code may be received from the HMB and another part of the boot code may be received from the memory device. If either the HMB or the memory device can deliver the chunks faster than the other, then the controller can receive the chunks from the faster location and periodically confirm the speed of delivery to ensure the boot code latency is optimized. The controller is configured to track an HMB turnaround latency and derive whether a next request should be sent to the HMB or the memory device when the data is present in both the HMB and the memory device.

Abstract: Calibration of soft bit reference levels in a non-volatile memory system is disclosed. When the soft bit reference levels are to be calibrated, encoded data is read from a group of the memory cells. The encoded data is decoded and error corrected. Therefore, the original data that was programmed into the memory cells is recovered. The group of memory cells are sensed at candidate soft bit reference levels, and possibly other reference levels. For each candidate soft bit reference level, mutual information between the original programmed data and the data for that candidate soft bit reference level is determined. The mutual information serves as a good measure for how well the candidate soft bit reference level will aid in decoding the data. In an aspect, a soft bit reference level having the highest mutual information out of several candidates is selected as the calibrated soft bit reference level.

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: 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: A data storage device includes a memory device and a controller coupled to the memory device. The controller is configured to store XOR parity data in a host memory buffer (HMB) of a host device, monitor a health of the memory device, determine that a threshold corresponding to the health of one or more blocks of the memory device has been reached or exceeded, and copy the XOR parity data from the HMB to the memory device. The controller is further configured to receive a low power mode indication from the host device and enter the low power mode after copying the XOR parity data from the HMB to the memory device. The controller is further configured to correct read failures using the XOR parity data retrieved from the HMB.

Abstract: Calibration of soft bit reference levels in a non-volatile memory system is disclosed. When the soft bit reference levels are to be calibrated, encoded data is read from a group of the memory cells. The encoded data is decoded and error corrected. Therefore, the original data that was programmed into the memory cells is recovered. The group of memory cells are sensed at candidate soft bit reference levels, and possibly other reference levels. For each candidate soft bit reference level, mutual information between the original programmed data and the data for that candidate soft bit reference level is determined. The mutual information serves as a good measure for how well the candidate soft bit reference level will aid in decoding the data. In an aspect, a soft bit reference level having the highest mutual information out of several candidates is selected as the calibrated soft bit reference level.

Abstract: Methods and apparatus are disclosed for implementing data augmentation within a storage controller of a data storage device based on machine learning data read from a non-volatile memory (NVM) array of a memory die. Some particular aspects relate to configuring the storage controller to generate augmented versions of training 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 and stored in the NVM array. Other aspects relate to controlling components of the memory die to generate noise-augmented images by, for example, storing and then reading training images from worn regions of the NVM array to inject noise into the images. Data augmentation based on data read from multiple memory dies is also described, such as image data spread across multiple NVM arrays or multiple memory dies.

Abstract: A storage circuit is configured to store multiple vectors associated with variable and check nodes of an iterative decoding operation. As part of the iterative decoding operation, a processor circuit is configured to retrieve, from the storage circuit, an intermediate value vector, a first estimation vector, a second estimation vector, and a sign vector, and determine an absolute value of the intermediate value vector. The processor circuit is also configured, using the retrieved vectors, to generate updated values for the first and second estimation vectors as part of determining a bit estimate for a check node included in the iterative decoding operation.

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.