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 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: The present disclosure generally relates to data storage devices, such as solid state drives (SSDs), and efficient data storage device operations related to power loss incidents. A controller of the data storage device is configured to periodically pre-encode data that is stored in random access memory (RAM), detect a power loss event, and program the data and parity data to non-volatile memory (NVM) in response to detecting the power loss event. Upon reaching a threshold size, the data in RAM may be pre-encoded and the pre-encoded data can be programmed to the RAM or the NVM. The parity data may be stored in one or more locations of the NVM. Upon detecting a power loss event, any data remaining in RAM that is not pre-encoded is encoded. The data and any parity data not yet programmed to the NVM are programmed to the NVM.

Abstract: Fast verification of data integrity of non-volatile memory cells is disclosed. In one aspect, an estimate is made of a bit error rate (BER) associated with the data to be verified without fully decoding the data. If the estimated BER is below a threshold, then the storage system reports that the data meets a data integrity criterion. If the estimated BER is above the threshold, the storage system may decode the data to determine a BER and report whether the data meets the data integrity criterion based on the determined BER. The estimate of the BER may be based on a syndrome weight of the data, a BER of an XOR codeword formed from multiple codewords of the data, or a BER of a sample of the data. Hence, considerable time and power are saved verifying data integrity, at least when the data is not fully decoded.

Abstract: Fast verification of data integrity of non-volatile memory cells is disclosed. In one aspect, an estimate is made of a bit error rate (BER) associated with the data to be verified without fully decoding the data. If the estimated BER is below a threshold, then the storage system reports that the data meets a data integrity criterion. If the estimated BER is above the threshold, the storage system may decode the data to determine a BER and report whether the data meets the data integrity criterion based on the determined BER. The estimate of the BER may be based on a syndrome weight of the data, a BER of an XOR codeword formed from multiple codewords of the data, or a BER of a sample of the data. Hence, considerable time and power are saved verifying data integrity, at least when the data is not fully decoded.

Abstract: A storage system and method for implementing an encoder, decoder, and/or buffer using a field programmable gate array are provided. In one embodiment, a storage system is provided with a field programmable gate array and a memory that stores sets of instruction code for the field programmable gate array. The sets of instruction code can be for different error decoder implementations, for providing an additional encoder and/or decoder, or for implementing a host memory buffer or a controller memory buffer.

Abstract: A method and apparatus for content aware decoding utilizes a pool of decoders shared data statistics. Each decoder generates statistical data of content it decodes and provides these statistics to a joint statistics pool. As codewords arrive at the decoder pool, the joint statistics are utilized to estimate or predict any corrupted or missing bit values. Codewords may be assigned to a specific decoder, such as a tier 1 decoder, a tier 2 decoder, or a tier 3 decoder, based on a syndrome weight or a bit error rate. The assigned decoder updates the joint statistics pool after processing the codeword. In some embodiments, each decoder may additionally maintain local statistics regarding codewords, and use the local statistics when there is a statistically significant mismatch between the local statistics and the joint statistics pool.

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: Calibration of soft bit reference levels in a non-volatile memory system is disclosed. A set of memory cells are sensed at a hard bit reference level and test soft bit reference levels. The test soft bit reference levels are grouped around the hard bit reference level. A metric is determined for the test soft bit reference levels. Bins are defined based on the hard bit reference level and the set of test soft bit reference levels. A metric may be determined for each of the bins. The new soft bit reference levels are determined based on the metric. In one aspect, the metric is how many memory cells have a value for a physical parameter within each bin. The soft bit reference levels may be established based on a target percentage for the bins. In one aspect, the metric is how many unsatisfied counters are within each bin.

Abstract: A storage system and method for implementing an encoder, decoder, and/or buffer using a field programmable gate array are provided. In one embodiment, a storage system is provided with a field programmable gate array and a memory that stores sets of instruction code for the field programmable gate array. The sets of instruction code can be for different error decoder implementations, for providing an additional encoder and/or decoder, or for implementing a host memory buffer or a controller memory buffer.

Abstract: The present disclosure generally relates to data storage devices, such as solid state drives. A data storage device includes a controller, one or more volatile memory locations, and one or more non-volatile memory locations. Computations, including reinforcement learning algorithms, may be completed by the controller using the one or more non-volatile memory locations. Data associated with reinforcement learning is stored in a table on one or more planes of the non-volatile memory, where the results from the computations update the table with the relevant values. The data in the table are aligned to one or more wordlines, such that sensing the wordline senses all the data stored in the table.

Abstract: A memory controller including, in one implementation, a memory interface and a control circuit. The memory interface is configured to receive a punctured codeword read from a non-volatile memory. The control circuit is configured to determine error probability values for a plurality of check nodes associated with a punctured bit included in the punctured codeword. The control circuit is also configured to determine an error probability value for the punctured bit based on the error probability values for the plurality of check nodes associated with the punctured bit and a variable degree associated with the punctured bit. The control circuit is further configured to determine a log likelihood ratio (LLR) value for the punctured bit based on the error probability value for the punctured bit. The control circuit is also configured to decode the punctured codeword using the LLR value for the punctured bit.

Abstract: Calibration of soft bit reference levels in a non-volatile memory system is disclosed. A set of memory cells are sensed at a hard bit reference level and test soft bit reference levels. The test soft bit reference levels are grouped around the hard bit reference level. A metric is determined for the test soft bit reference levels. Bins are defined based on the hard bit reference level and the set of test soft bit reference levels. A metric may be determined for each of the bins. The new soft bit reference levels are determined based on the metric. In one aspect, the metric is how many memory cells have a value for a physical parameter within each bin. The soft bit reference levels may be established based on a target percentage for the bins. In one aspect, the metric is how many unsatisfied counters are within each bin.

Abstract: The disclosure relates in some aspects to on-chip processing circuitry formed within the die of a non-volatile (NVM) array to perform data searches. In some aspects, the die includes components configured to sense wordlines of stored data in the NVM array by applying voltages on the wordlines serially, and then search for an input data pattern within the serially-sensed wordlines. In some examples, the components of the die include latches and circuits configured to perform bitwise latch logic search operations. In other examples, the search components are configured with under-the-array or next-to-the-array dedicated search circuitry that uses registers and/or random access memory (RAM). Other aspects relate to a separate controller device for controlling the on-chip NVM search operations. For example, the controller may determine whether to search for data using search components of the NVM die or processors of the controller based, e.g., on a degree of fragmentation of data.

Abstract: A method of temperature compensation to read a flash memory device includes determining a state of the flash memory device. An action is selected with a maximum Q-value from a Q-table for the current state during exploitation. A read operation of a code word from the flash memory device is conducted using one or more parameters according to the selected action. The code word is decoded with an error correction code (ECC) process.

Abstract: A method and apparatus for determining the sector size and concomitant host metadata size to determine the difference between total size of the data block to be stored, and using the difference for parity data. This allows an increase in parity bits available for smaller sector sizes and/or data with smaller host metadata sizes. Because the amount of space available for additional parity bits is known, data with lower numbers of parity bits may be assigned to higher quality portions a memory array written with longer programming trim times, and/or written to memory dies with good redundant columns, further increasing performance and reliability.

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: 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: Exemplary methods and apparatus are provided for implementing a deep learning accelerator (DLA) or other neural network components within the die of a non-volatile memory (NVM) apparatus using, for example, under-the-array circuit components within the die. Some aspects disclosed herein relate to configuring the under-the-array components to implement feedforward DLA operations. Other aspects relate to backpropagation operations. Still other aspects relate to using an NAND-based on-chip copy with update function to facilitate updating synaptic weights of a neural network stored on a die. Other aspects disclosed herein relate to configuring a solid state device (SSD) controller for use with the NVM. In some aspects, the SSD controller includes flash translation layer (FTL) tables configured specifically for use with neural network data stored in the NVM.

Abstract: An error correcting code (ECC) decoder for a non-volatile memory device is configured to decode data stored by the non-volatile memory device using a parity check matrix with columns of different column weights. The ECC decoder is further configured to artificially slow processing of one or more of the columns of the parity check matrix in response to column weights for the one or more columns satisfying a threshold.