Abstract: In some implementations, a controller may perform, on one or more wordlines of a block of a non-volatile memory device, read operations using default threshold voltages associated with two overlapped charge states. The controller may determine, using a machine learning model, a distribution of threshold voltages for the two overlapped charge states based on read errors associated with the threshold voltages. The controller may determine, based on the determined distribution of threshold voltages, a health of the block. The controller may perform a block refresh operation for the block based on the health of the block. The block refresh operation may be performed when the health satisfies a health threshold. The block refresh operation may not be performed when the health does not satisfy the health threshold.

Abstract: A controller may determine, using a machine learning model, reliability characteristic data associated with memory cells of a non-volatile memory device. The machine learning model may be trained using characterization data that identifies different reliability characteristic of one or more non-volatile memory devices. The controller may group, based on the reliability characteristic data, a first portion of the memory cells of the non-volatile memory device in a first management group, and a second portion of the memory cells of the non-volatile memory device in a second management group. The controller may manage, based on the reliability characteristic data, background scanning and logical to physical mapping of the first management group of memory cells, and the second management group of memory cells.

Abstract: In some implementations, a controller may receive a request for an inference. The controller may determine, based on the received request for the inference, a first inference model of a plurality of inference models, to generate the inference. The controller may obtain, from a memory associated with an inference cache, first attribute data regarding first attributes of the first inference model. A location of the first attribute data, in the memory, may be determined using the inference cache. The attributes may include weights associated with the first inference model, biases associated with the first inference model, and a structure of the first inference model. The controller may utilize the first attribute data to generate the inference based on the request.

Abstract: Systems and methods are disclosed including a memory device and a processing device operatively coupled to the memory device. The processing device can perform operations comprising performing a set of write operations on a first block in a first plane of the memory device and on a second block in a second plane of the memory device, performing a program verification check on the first block, responsive to determining that the first block fails the program verification check, incrementing a counter value associated with the second block; responsive to the counter value satisfying a threshold criterion, performing a failure verification operation on the second block, and responsive to determining that the second block fails the failure verification operation, retiring the second block.

Abstract: Read calibration by sector of memory can include reading a page of memory, having more than one sector, with a read level, such as a default read level. In response to an error, such as an uncorrectable error correction code read result, the respective read level can be calibrated for each sector to yield a respective calibrated read level per sector. The page of memory can be read with the respective calibrated read level per sector. The calibrated read levels can be stored.

Abstract: Exemplary methods, apparatuses, and systems including memory self-recovery management to correct failures due to soft-error rate events. The self-recovery manager detects a failure of a memory device. The self-recovery manager retrieves a set of register values from the memory device. The self-recovery manager stores the set of register values from the memory device. The self-recovery manager issues a reset command to the memory device, the reset command including generating a re-initialized set of register values. The self-recovery manager compares the set of register values with the re-initialized set of register values. The self-recovery manager triggering a self-recovery attempt using the comparison of the set of register values with the re-initialized set of register values.

Abstract: A methods and system directed to a wordline ramp rate monitor for early detection of defect activation are disclosed. A memory access directed to a wordline is initiated. Based on an applied ramping voltage, a ramp rate of the wordline is determined. Responsive to determining that the ramp rate satisfied a defect condition, the memory access operation is aborted.

Abstract: Prioritization of VT scans can be performed using particular select gates of a memory device or memory sub-system in the absence of performing such select gate scan operations on all of the select gates of an entire memory die or of all the memory dice of a memory device or memory sub-system. A method for such prioritization of VT scans includes determining quality characteristics of a memory die and altering a threshold voltage applied to the memory die in performance of a select gate scan operation based, at least in part, on the determined quality characteristics of the memory die. Such methods can further include performing the select gate scan operation by applying signaling having the altered threshold voltage to a select gate of the memory die.

Abstract: Various embodiments provide for performing a memory operation, such as a memory block compaction operation or block folding or refresh operation, based on a temperature associated with a memory block of a memory device. For instance, some embodiments provide for techniques that can cause performance of a block compaction operation on a memory block at a temperature that is at least at or higher than a predetermined temperature value. Additionally, some embodiments provide for techniques that can cause performance of a block folding/refresh operation, at a temperature that is at or higher than the predetermined temperature value, on one or more blocks on which data was written at a temperature lower than the predetermined temperature value.

Abstract: Exemplary methods, apparatuses, and systems write data to a first wordline of a partially programmed block of memory. A second wordline of the block is determined to fail to satisfy a first margin threshold by comparing a first voltage threshold of the second wordline to a reference voltage. In response to the second wordline failing to satisfy the first margin threshold, a second margin test is applied to the block. In response to determining the block passed the second margin test, data is written in a subsequent write operation to the block using an adjusted trim setting.

Abstract: Read calibration by sector of memory can include reading a page of memory, having more than one sector, with a read level, such as a default read level. In response to an error, such as an uncorrectable error correction code read result, the respective read level can be calibrated for each sector to yield a respective calibrated read level per sector. The page of memory can be read with the respective calibrated read level per sector. The calibrated read levels can be stored.

Abstract: Methods, systems, and devices for techniques for managing a voltage recovery operation are described. In some cases, as part of performing a write command to store data to a set of memory cells, the memory system may store an indication of the initial time at which the write operation occurred, the temperature of the set of memory cells at the initial time, or both. The memory system may subsequently manage an accumulated value based on a duration from the initial time and the temperature of the set of memory cells during the duration. If the accumulated value exceeds an accumulation threshold, the memory system may identify an indication of degradation of the set of memory cells. If the indication exceeds a degradation threshold, the memory system may perform a voltage recovery operation to modify voltages of the set of memory cells.

Abstract: Memory cells may store multiple bits per cell. For example, three-level cell (TLC) memory stores three bits per cell using eight voltage levels. The number of memory cells at each voltage is approximately the total number of cells divided by the number of voltage levels. The number of memory cells above a certain read voltage is the CFByte value for the read voltage. Based on a difference between the CFByte value and a target CFByte value for the read voltage, an adjustment value is determined. Characteristics of an individual memory device may be determined by finding several CFByte values for a small range of read voltages. Using the gathered CFByte values, a DAC adjustment value is determined for the individual memory device.

Abstract: Methods, systems, and devices for memory pillar selection transistor evaluation are described. A memory system may be configured to monitor threshold voltage characteristics of pillar selection transistors, which may include evaluations relative to certain subsets of the pillar selection transistors. For example, an activation voltage may be applied to the pillar selection transistors to determine whether threshold voltages associated with each subset of pillar selection transistors have shifted. Determining whether the threshold voltages have shifted may include determining whether an access parameter has been satisfied, such as a duration to program memory cells. For example, a relatively long duration may indicate that channels associated with pillar selection transistors have become less conductive for a given activation voltage.

Abstract: Various embodiments provide for performing a memory operation, such as a memory block compaction operation or block folding or refresh operation, based on a temperature associated with a memory block of a memory device. For instance, some embodiments provide for techniques that can cause performance of a block compaction operation on a memory block at a temperature that is at least at or higher than a predetermined temperature value. Additionally, some embodiments provide for techniques that can cause performance of a block folding/refresh operation, at a temperature that is at or higher than the predetermined temperature value, on one or more blocks on which data was written at a temperature lower than the predetermined temperature value.

Abstract: Systems and methods are disclosed including a memory device and a processing device operatively coupled to the memory device. The processing device can perform operations including initializing the memory device; selecting at least one sample management unit on the memory device; performing a calibration operation on the sample management unit to determine a duration value reflecting a duration during which the memory device was powered down; adjusting an accumulator value based on the duration value; determining a read voltage value based on the accumulator value; and performing a read operation using the read voltage value.

Abstract: Aspects of the present disclosure are directed to a memory sub-system using a block family error avoidance (BFEA) scan to adjust read voltages. Three-level cell (TLC) memory stores three bits per cell. Due to variances in manufacturing and degradation over time, the actual voltages stored in the memory cells deviate from the target voltages. As a result, the comparisons between the read voltages and the stored voltages may generate erroneous results. A BFEA scan may be based on a single wordline and single page type. However, determining a single threshold voltage shift to apply to all read voltages may not compensate for all causes of voltage shifting. Accordingly, a BFEA scan may use multiple wordlines (e.g., one for each page) and determine different voltage offset values for each page. As a result, the accuracy of the read voltage applied is increased and the bit error rate (BER) is reduced.

Abstract: Prioritization of VT scans can be performed using particular select gates of a memory device or memory sub-system in the absence of performing such select gate scan operations on all of the select gates of an entire memory die or of all the memory dice of a memory device or memory sub-system. A method for such prioritization of VT scans includes determining quality characteristics of a memory die and altering a threshold voltage applied to the memory die in performance of a select gate scan operation based, at least in part, on the determined quality characteristics of the memory die. Such methods can further include performing the select gate scan operation by applying signaling having the altered threshold voltage to a select gate of the memory die.

Abstract: Apparatuses and methods for determining performing read operations on a partially programmed block are provided. One example apparatus can include a controller configured to apply a read voltage to a word line in an array of memory cells during a read operation on the word line, apply a first pass voltage to a number of programmed word lines in the array of memory cells during the read operation, and apply a second pass voltage to a number of unprogrammed word lines in the array of memory cells during the read operation.

Abstract: Apparatuses, systems, and methods for determining a dynamic erase voltage step. One example apparatus can include an array of memory cells and a controller coupled to the array of memory cells, wherein the controller is configured to apply a first erase voltage to a first wordline and a second wordline in the array of memory cells to perform an erase operation, apply a first verify voltage to the first wordline to verify the erase operation, apply a second verify voltage greater than the first verify voltage to the second wordline in response to failing to verify the erase operation by applying the first verify voltage to the first wordline, and apply a second erase voltage to the first wordline and the second wordline in response to verifying the erase operation by applying the second verify voltage to the second wordline.