Patent Application
20230162803
Embodiments of the present disclosure relate to a scheme for determining an optimal read threshold voltage in a memory system.
The computer environment paradigm has shifted to ubiquitous computing systems that can be used anytime and anywhere. As a result, the use of portable electronic devices such as mobile phones, digital cameras, and notebook computers has rapidly increased. These portable electronic devices generally use a memory system having memory device(s), that is, data storage device(s). The data storage device is used as a main memory device or an auxiliary memory device of the portable electronic devices.
Memory systems using memory devices provide excellent stability, durability, high information access speed, and low power consumption, since the memory devices have no moving parts. Examples of memory systems having such advantages include universal serial bus (USB) memory devices, memory cards having various interfaces such as a universal flash storage (UFS), and solid state drives (SSDs). Memory systems may determine an optimal read threshold voltage among read threshold voltages according to various schemes.
Aspects of the present invention include a system and a method for non-parametric optimal read threshold estimation using a deep neural network.
In one aspect of the present invention, a memory system includes a memory device including a plurality of pages and a controller including a neural network. The controller is configured to: perform one or more read operations on a page selected from among the plurality of pages using a read threshold set including a plurality of read threshold voltages; obtain the read threshold set, a checksum value and an asymmetric ratio of ones count and zeros count which are associated with decoding of the selected page according to each of the read operations; provide the obtained read threshold set, the checksum value and the asymmetric ratio as input information to the neural network; and estimate, by the neural network, an optimal read threshold voltage based on the input information and weights for the input information, the weights including a combination of multiple matrices and bias vectors.
In another aspect of the present invention, a method for operating a memory system, which includes a memory device including a plurality of pages and a controller including a neural network, includes: performing one or more read operations on a page selected from among the plurality of pages using a read threshold set including a plurality of read threshold voltages; obtaining the read threshold set, a checksum value and an asymmetric ratio of ones count and zeros count which are associated with decoding of the selected page according to each of the read operations; providing the obtained read threshold set, the checksum value and the asymmetric ratio as input information to the neural network; and estimating, by the neural network, an optimal read threshold voltage based on the input information and weights for the input information, the weights including a combination of multiple matrices and bias vectors.
Additional aspects of the present invention will become apparent from the following description.
Various embodiments of the present invention are described below in more detail with reference to the accompanying drawings. The present invention may, however, be embodied in different forms and thus should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure conveys the scope of the present invention to those skilled in the art. Moreover, reference herein to “an embodiment,” “another embodiment,” or the like is not necessarily to only one embodiment, and different references to any such phrase are not necessarily to the same embodiment(s). The term “embodiments” as used herein does not necessarily refer to all embodiments. Throughout the disclosure, like reference numerals refer to like parts in the figures and embodiments of the present invention.
The present invention can be implemented in numerous ways, including as a process; an apparatus; a system; a computer program product embodied on a computer-readable storage medium; and/or a processor, such as a processor suitable for executing instructions stored on and/or provided by a memory coupled to the processor. In this specification, these implementations, or any other form that the present invention may take, may be referred to as techniques. In general, the order of the operations of disclosed processes may be altered within the scope of the present invention. Unless stated otherwise, a component such as a processor or a memory described as being suitable for performing a task may be implemented as a general device or circuit component that is configured or otherwise programmed to perform the task at a given time or as a specific device or as a circuit component that is manufactured or pre-configured or pre-programmed to perform the task. As used herein, the term ‘processor’ or the like refers to one or more devices, circuits, and/or processing cores suitable for processing data, such as computer program instructions.
The methods, processes, and/or operations described herein may be performed by code or instructions to be executed by a computer, processor, controller, or other signal processing device. The computer, processor, controller, or other signal processing device may be those described herein or one in addition to the elements described herein. Because the algorithms that form the basis of the methods (or operations of the computer, processor, controller, or other signal processing device) are described herein, the code or instructions for implementing the operations of the method embodiments may transform the computer, processor, controller, or other signal processing device into a special-purpose processor for performing any one of the methods herein.
If implemented at least partially in software, the controllers, processors, devices, modules, units, multiplexers, generators, logic, interfaces, decoders, drivers, generators and other signal generating and signal processing features may include, for example, a memory or other storage device for storing code or instructions to be executed, for example, by a computer, processor, microprocessor, controller, or other signal processing device.
A detailed description of various embodiments of the present invention is provided below along with accompanying figures that illustrate aspects of the present invention. The present invention is described in connection with such embodiments, but the present invention is not limited to any specific embodiment. The present invention encompasses numerous alternatives, modifications and equivalents. Numerous specific details are set forth in the following description in order to provide a thorough understanding of the present invention. These details are provided for the purpose of example; the present invention may be practiced without some or all of these specific details. For clarity, technical material that is known in technical fields related to the present invention has not been described in detail so that the present invention is not unnecessarily obscured.
Referring
The host device 5 may be implemented with any of various types of electronic devices. In various embodiments, the host device 5 may be an electronic device such as for example a desktop computer, a workstation, a three-dimensional (3D) television, a smart television, a digital audio recorder, a digital audio player, a digital picture recorder, a digital picture player, and/or a digital video recorder and a digital video player. In various embodiments, the host device 5 may be a portable electronic device such as for example a mobile phone, a smart phone, an e-book, an MP3 player, a portable multimedia player (PMP), and/or a portable game player.
The memory system 10 may be implemented with any of various types of storage devices such as a solid state drive (SSD) and a memory card. In various embodiments, the memory system 10 may be provided as one of various components in an electronic device such as for example a computer, an ultra-mobile personal computer (PC) (UMPC), a workstation, a net-book computer, a personal digital assistant (PDA), a portable computer, a web tablet PC, a wireless phone, a mobile phone, a smart phone, an e-book reader, a portable multimedia player (PMP), a portable game device, a navigation device, a black box, a digital camera, a digital multimedia broadcasting (DMB) player, a 3-dimensional television, a smart television, a digital audio recorder, a digital audio player, a digital picture recorder, a digital picture player, a digital video recorder, a digital video player, a storage device of a data center, a device capable of receiving and transmitting information in a wireless environment, a radio-frequency identification (RFID) device, as well as one of various electronic devices of a home network, one of various electronic devices of a computer network, one of electronic devices of a telematics network, or one of various components of a computing system.
The memory system 10 may include a memory controller 100 and a semiconductor memory device 200. The memory controller 100 may control overall operations of the semiconductor memory device 200.
The semiconductor memory device 200 may perform one or more erase, program, and read operations under the control of the memory controller 100. The semiconductor memory device 200 may receive through input/output lines a command CMD, an address ADDR and data DATA. The semiconductor memory device 200 may receive power PWR through a power line and a control signal CTRL through a control line. The control signal CTRL may include for example a command latch enable signal, an address latch enable signal, a chip enable signal, a write enable signal, a read enable signal, as well as other operational signals depending on design and configuration of the memory system 10.
The memory controller 100 and the semiconductor memory device 200 may be integrated in a single semiconductor device such as a solid state drive (SSD). The SSD may include a storage device for storing data therein. In one embodiment of the invention, where the semiconductor memory system 10 is used in an SSD, operation speed of a host device (e.g., host device 5 of
The memory controller 100 and the semiconductor memory device 200 may be integrated in a single semiconductor device such as a memory card. For example, the memory controller 100 and the semiconductor memory device 200 may be integrated to configure a personal computer (PC) card of personal computer memory card international association (PCMCIA), a compact flash (CF) card, a smart media (SM) card, a memory stick, a multimedia card (MMC), a reduced-size multimedia card (RS-MMC), a micro-size version of MMC (MMCmicro), a secure digital (SD) card, a mini secure digital (miniSD) card, a micro secure digital (microSD) card, a secure digital high capacity (SDHC), and/or a universal flash storage (UFS).
Referring to
The memory device 200 may store data to be accessed by the host device.
The memory device 200 may be implemented with a volatile memory device such as for example a dynamic random access memory (DRAM) and/or a static random access memory (SRAM) or a non-volatile memory device such as for example a read only memory (ROM), a mask ROM (MROM), a programmable ROM (PROM), an erasable programmable ROM (EPROM), an electrically erasable programmable ROM (EEPROM), a ferroelectric random access memory (FRAM), a phase change RAM (PRAM), a magnetoresistive RAM (MRAM), and/or a resistive RAM (RRAM).
The controller 100 may control storage of data in the memory device 200. For example, the controller 100 may control the memory device 200 in response to a request from the host device. The controller 100 may provide data read from the memory device 200 to the host device, and may store data provided from the host device into the memory device 200.
The controller 100 may include a storage 110, a control component 120, which may be implemented as a processor such as for example a central processing unit (CPU), an error correction code (ECC) component 130, a host interface (I/F) 140 and a memory interface (I/F) 150, which are coupled through a bus 160.
The storage 110 may serve as a working memory of the memory system 10 and the controller 100, and storage 110 may store data for driving the memory system 10 and the controller 100. For example, when the controller 100 controls operations of the memory device 200, the storage 110 may store data used by the controller 100 and the memory device 200 for such operations as read, write, program and erase operations.
The storage 110 may be implemented with a volatile memory such as a static random access memory (SRAM) or a dynamic random access memory (DRAM). As described above, the storage 110 may store data used by the host device in the memory device 200 for the read and write operations. To store the data, the storage 110 may include a program memory, a data memory, a write buffer, a read buffer, a map buffer, and the like.
The control component 120 may control general operations of the memory system 10, and a write operation or a read operation for the memory device 200 in response to a write request or a read request from the host device. The control component 120 may drive firmware or other program instructions, which can be referred to as a flash translation layer (FTL), to control operations of the memory system 10. For example, the FTL may perform operations such as logical-to-physical (L2P) mapping, wear leveling, garbage collection, and/or bad block handling. The L2P mapping is known as logical block addressing (LBA).
The ECC component 130 may detect and correct errors in the data read from the memory device 200 during a read operation. In one embodiment, the ECC component 130 may not correct error bits when the number of the error bits is greater than or equal to a threshold number of correctable error bits, but instead may output an error correction fail signal indicating failure in correcting the error bits.
In various embodiments, the ECC component 130 may perform an error correction operation based on a coded modulation such as for example a low density parity check (LDPC) code, a Bose-Chaudhuri-Hocquenghem (BCH) code, a turbo code, a turbo product code (TPC), a Reed-Solomon (RS) code, a convolution code, a recursive systematic code (RSC), a trellis-coded modulation (TCM), or a Block coded modulation (BCM). However, error correction is not limited to these techniques. As such, the ECC component 130 may include any and all circuits, systems or devices suitable for error correction operation.
The host interface 140 may communicate with the host device through one or more of various communication standards or interfaces such as for example a universal serial bus (USB), a multi-media card (MMC), a peripheral component interconnect express (PCI-e or PCIe), a small computer system interface (SCSI), a serial-attached SCSI (SAS), a serial advanced technology attachment (SATA), a parallel advanced technology attachment (PATA), an enhanced small disk interface (ESDI), and an integrated drive electronics (IDE).
The memory interface 150 may provide an interface between the controller 100 and the memory device 200 to allow the controller 100 to control the memory device 200 in response to a request from the host device. The memory interface 150 may generate control signals for the memory device 200 and process data under the control of the control component 120. In one embodiment where the memory device 200 is a flash memory such as a NAND flash memory, the memory interface 150 may generate control signals for the memory and process data under the control of the control component 120.
The memory device 200 as shown for example in
The voltage generation circuit 230 may generate operational voltages of various levels. For example, in an erase operation, the voltage generation circuit 230 may generate operational voltages of various levels such as for example an erase voltage and a pass voltage.
The row decoder 240 may be in electrical communication with the voltage generation circuit 230, and the plurality of memory blocks 211. The row decoder 240 may select at least one memory block among the plurality of memory blocks 211 in response to a row address generated by the control circuit 220, and transmit operation voltages supplied from the voltage generation circuit 230 to the selected memory blocks.
The page buffer 250 may be coupled with the memory cell array 210 through bit lines BL (shown in
The column decoder 260 may transmit data to, and receive data from, the page buffer 250 or may transmit and receive data to and from the input/output circuit 270.
The input/output circuit 270 may transmit to the control circuit 220 a command and an address, received from an external device (e.g., the memory controller 100 of
The control circuit 220 may control the peripheral circuit in response to the command and the address.
Referring to
The memory block 211 may further include a plurality of cell strings 221 respectively coupled to bit lines BL0 to BLm−1. The cell string of each column may include one or more drain selection transistors DST and one or more source selection transistors SST. In the illustrated embodiment, each cell string has one DST and one SST. In a cell string, a plurality of memory cells or memory cell transistors MC0 to MCn−1 may be serially coupled between the selection transistors DST and SST. Each of the memory cells may be formed as a multiple level cell. For example, each of the memory cells may be formed as a single level cell (SLC) storing 1 bit of data. Each of the memory cells may be formed as a multi-level cell (MLC) storing 2 bits of data. Each of the memory cells may be formed as a triple-level cell (TLC) storing 3 bits of data. Each of the memory cells may be formed as a quadruple-level cell (QLC) storing 4 bits of data.
The source of the SST in each cell string may be coupled to a common source line CSL, and the drain of each DST may be coupled to the corresponding bit line. Gates of the SSTs in the cell strings may be coupled to the SSL, and gates of the DSTs in the cell strings may be coupled to the DSL. Gates of the memory cells across the cell strings may be coupled to respective word lines. That is, the gates of memory cells MC0 are coupled to corresponding word line WL0, the gates of memory cells MC1 are coupled to corresponding word line WL1, etc. The group of memory cells coupled to a particular word line may be referred to as a physical page. Therefore, the number of physical pages in the memory block 211 may correspond to the number of word lines.
The page buffer array 250 may include a plurality of page buffers 251 that are coupled to the bit lines BL0 to BLm−1. The page buffers 251 may operate in response to page buffer control signals. For example, the page buffers 251 my temporarily store data received through the bit lines BL0 to BLm−1 or sense voltages or currents of the bit lines during a read or verify operation.
In various embodiments of the present invention, the memory blocks 211 may include a NAND-type flash memory cell. However, the memory blocks 211 are not limited to such cell type, and may include NOR-type flash memory cell(s). Memory cell array 210 may be implemented as a hybrid flash memory in which two or more types of memory cells are combined, or one-NAND flash memory in which a controller is embedded inside a memory chip.
Referring to
An SLC may include two states P0 and P1. P0 may indicate an erase state, and P1 may indicate a program state. Since the SLC can be set in one of two different states, each SLC may program or store 1 bit according to a set coding method. An MLC may include four states P0, P1, P2 and P3. Among these states, P0 may indicate an erase state, and P1 to P3 may indicate program states. Since the MLC can be set in one of four different states, each MLC may program or store two bits according to a set coding method. A TLC may include eight states P0 to P7. Among these states, P0 may indicate an erase state, and P1 to P7 may indicate program states. Since the TLC can be set in one of eight different states, each TLC may program or store three bits according to a set coding method. A QLC may include 16 states P0 to P15. Among these states, P0 may indicate an erase state, and P1 to P15 may indicate program states. Since the QLC can be set in one of sixteen different states, each QLC may program or store four bits according to a set coding method.
Referring back to
Referring to
In the MLC, as shown in
Referring to
In the TLC, as shown in
After a memory array including a plurality of memory cells is programmed as described in
When a read operation is performed on memory cells in a data storage device using MLC technology, the threshold voltage levels of the memory cells are compared to more than one read threshold level to determine the state of individual memory cells. Read errors can be caused by distorted or overlapped threshold voltage distributions. An ideal memory cell threshold voltage distribution can be significantly distorted or overlapped due to, e.g., program and erase (P/E) cycles, cell-to-cell interference, and/or data retention errors. For example, as program/erase cycles increase, the margin between neighboring threshold voltage distributions of different programmed states decreases and eventually the distributions overlap. As a result, the memory cells with threshold voltages that fall within the overlapping region of the neighboring distributions may be read as being programmed to a value other than the original targeted value and thus cause read errors. Such read errors may be managed in many situations by using error correction codes (ECC). When the number of bit errors on a read operation exceeds the ECC correction capability of the data storage, the read operation using a set read threshold voltage fails. The set read threshold voltage may be a previously used read threshold voltage (i.e., a historical read threshold voltage). The historical read threshold voltage may be the read threshold voltage used in the last successful decoding, that is, a read voltage used in a read-passed read operation performed before read retry operations. When the read operation using the set read threshold voltage failed, the controller 120 may control an error recovery algorithm such as for example the algorithm shown in
Referring to
When all read retry operations using the read threshold voltages have failed, the controller 120 may perform additional recovery operations. For example, the additional recovery operations may include an optimal read threshold voltage search (S200), a soft decoding using an error correction code (ECC) (S300) and/or a redundant array of independent disks (RAID) recovery (S400).
As noted above, data retrieved from a memory device (e.g., a NAND flash device) in a data storage device (e.g., a solid state drive (SSD)) usually contains many bit errors due to various noise sources. To protect the data, it is encoded by an error correction code before being written to the memory device (i.e., before being written to a storage medium). A decoder may correct all bit errors by using this error correction code. As illustrated in
For a solid state drive, one source of increased bit errors is the use of sub-optimal read thresholds during the read operation as shown in
Referring to
The controller 100 may include a read processor 910, a decoder 920 and an optimal read threshold determiner 930. Although it is illustrated that components of the controller 100 are implemented separately, these components may be implemented with an internal component (i.e., firmware (FW)) of the control component 120 in
The read processor 910 may control one or more read operations for the memory device 200 in response to a read request from a host (e.g., the host 5 of
In various embodiments of the present invention, the read processor 910 may control a read operation for the memory cells using a select read threshold from a set read level table. In various embodiments, the read level table may include multiple read thresholds and the select read threshold may include a default read threshold. When the read operation is performed for the MSB page of TLC, the select read threshold may include a pair of first and second read thresholds [VT0, VT4] as shown in
In one embodiment of the present invention, it is determined whether the read operation using a read threshold selected from a read threshold set succeeded or failed, depending on the decoding result of the decoder 920. When the read operation using the selected read threshold failed, the read processor 910 may control one or more read retry operations for the memory cells using read retry entries, such as shown in
The optimal read threshold determiner 930 may provide a scheme of an optimal read threshold voltage (Vt) estimation (or prediction). The optimal read threshold determiner 930 may be implemented with one or more deep neural networks (DNNs). In various embodiments of the present invention, the optimal read threshold determiner 930 may include a training component 930A and an inference component 930B. The operations for the training component 930A and the inference component 930B is described below.
Referring to
The neural network 1000 may be a multi-layer neural network that represents a network of interconnected nodes, such as an artificial deep neural network, where knowledge about the nodes (e.g., information about specific features represented by the nodes) is shared across layers and knowledge specific to each layer is also retained. Each node represents a piece of information. Knowledge may be exchanged between nodes through node-to-node interconnections. Input to the neural network 1000 may activate a set of nodes. In turn, this set of nodes may activate other nodes, thereby propagating knowledge about the input. This activation process may be repeated across other nodes until nodes in the output layer 1030 are selected and activated.
As illustrated, the neural network 1000 includes a hierarchy of layers representing a hierarchy of nodes interconnected in a feed-forward way. The input layer 1010 may exist at the lowest hierarchy level. The input layer 1010 may include a set of nodes that are referred to herein as input nodes. When the feature map 1002 is input to the neural network 1000, each of the input nodes of the input layer 1010 may be connected to each feature of the feature map 1002. Each of the connections may have a weight. These weights may be one set of parameters that are derived from the learning/training of the neural network 1000. The input nodes may transform the features by applying an activation function to these features. The information derived from the transformation may be passed to the nodes at a higher level of the hierarchy.
The output layer 1030 may exist at the highest hierarchy level. The output layer 1030 may include one or more output nodes. Each output node may provide a specific value of the output information 1004. The number of output nodes may depend on the needed number of the output information 1004. In other words, in one embodiment, there is a one-to-one relationship or mapping between the number of output nodes and the number of the output information 1004.
The hidden layer(s) 1020 may exist between the input layer 1010 and the output layer 1030. The hidden layer(s) 1020 may include “N” number of hidden layers, where “N” is an integer greater than or equal to one. Each of the hidden layers 1020 may include a set of nodes that are referred to herein as hidden nodes. Example hidden layers may include up-sampling, convolutional, fully connected layers, and/or data transformation layers.
At the lowest level of the hidden layer(s) 1020, hidden nodes of that layer may be interconnected to the input nodes. At the highest level of the hidden layer(s) 1020, hidden nodes of that level may be interconnected to the output node. The input nodes may be not directly interconnected to the output node(s). If multiple hidden layers exist, the input nodes are interconnected to hidden nodes of the lowest hidden layer. In turn, these hidden nodes are interconnected to the hidden nodes of the next hidden layer. An interconnection may represent a piece of information learned about the two interconnected nodes. The interconnection may have a numeric weight that can be tuned (e.g., based on a training dataset), rendering the neural network 1000 adaptive to inputs and capable of learning.
Generally, the hidden layer(s) 1020 may allow knowledge about the input nodes of the input layer 1010 to be shared among the output nodes of the output layer 1030. To do so, a transformation ƒ may be applied to the input nodes through the hidden layer 1020. In one example, the transformation ƒ can be non-linear. Different non-linear transformations ƒ are available including, for instance, a rectifier function ƒ(x)=max(0,x). In another example, a particular non-linear transformation ƒ can be selected based on a cross-validation. For example, given known example pairs (x,y), where x∈X and y∈Y, a function ƒ: X→Y can be selected when such a function results in the best matches.
For example, the neural network 1000 may be a deep learning neural network used for a memory system including a NAND flash memory device. The deep leaning neural network may be created with “K” input nodes and an output node, where “K” is the number of factors (e.g., features) that define input conditions for the memory system. The output node(s) may be used to perform an activation function for a combination of input conditions. The number of layers and size of each layer in the neural network 1000 may depend on the NAND flash memory device and on an amount of the data that this memory can store.
As noted above, the neural network 1000 may be implemented for the optimal read threshold determiner 1030 in
Referring to
The optimal read threshold estimator 1100 may estimate optimal read thresholds for a failed read page based on information collected from previous read attempts of the failed read page. That is, DNN 1100 may estimate optimal read thresholds without necessarily any additional sensing operations and any intermediate parameter estimation, which are required in the implementations above mentioned.
As shown in
The read threshold set Vt in
Each checksum may be an approximation for the number of bit errors, which indicates a noise level in an optimal read threshold voltage estimation. When low density parity check (LDPC) codes are used for decoding a read page as an error correction code (ECC), a syndrome weight may correspond to the number of unsatisfied check (USC) node(s) of a codeword and may be used as checksum.
Each asymmetric ratio/ones count/zeros count (i.e., an asymmetric ratio of ones count and zeros count) may be any one of the three metrics: ones count; zeros count and an asymmetric ratio of ones count and zeros count. Ones count may be the number of bit 1's in a codeword. Zeros count may be the number of bit 0's in a codeword. The asymmetric ratio AR may be defined as a ratio of the number of a first binary value to the number of a second binary value in a codeword. For example, the asymmetric ratio AR may be a value representing the number of 1's (i.e., ones count) divided by the number of 0's (i.e., zeros count) in decoded raw data. In one embodiment in which reverse logic convention is used, AR may be the number of 0's (i.e., zeros count) divided by the number of 1's (i.e., ones count). In one embodiment, a read threshold optimization system in the present invention may use any of the asymmetric ratios described in U.S. patent application Ser. No. 16/927,567, filed on Jul. 13, 2020, entitled “READ THRESHOLD OPTIMIZATION SYSTEMS AND METHODS USING DOMAIN TRANSFORMATION,” the entire contents of which are incorporated herein by reference.
The DNN 1100 may be a single fully-connected network with multiple layers with a set function (e.g., a rectified linear unit (ReLu) function) as non-linearity or any other appropriate DNN structure. The DNN 1100 may estimate (i.e., predict) an optimal read threshold set OptVt based on the inputs, i.e., a read threshold set (Vtread0, Vtread1, . . . ), checksums (CSread0, CSread1, . . . ) and an asymmetric ratio/ones counts/zeros counts set (ARread0, ARread1, . . . ), as shown in
DNN 1100 may be trained on a particular dataset that includes read threshold set (Vtread0, Vtread1, . . . ), checksums (CSread0, CSread1, . . . ) and an asymmetric ratio/ones counts/zeros counts set (ARread0, ARread1, . . . ) as input features and the predicted optimal read threshold voltage as the output. In some embodiments, DNN 1100 may be trained offline on measurements collected from pages of a real NAND flash memory device or a suitable channel model.
For the training of DNN 1100, a set loss function 1150 (e.g., a differentiable loss function) may be defined to measure an error between the predicted optimal read threshold voltage VtDNN, which is the output of DNN 1100, and the expected optimal read threshold voltage VtOpt of data points in the dataset. For example, the mean square error function (VtDNN−VtOpt)2/n (where n is the number of data points in a subset of training data of the dataset) may be used as the loss function 1150. The loss function 1150 may be not restricted to the mean square error function only.
The DNN 1100 may receive the result from the loss function 1150. Further, the DNN 1100 may be trained to find the predicted optimal read threshold voltage VtDNN based on the result of the loss function such that the difference between the true (actual) optimal read threshold voltage VtOpt and the predicted optimal read threshold voltage VtDNN from DNN 1100 is minimized.
In the illustrated example of
In various embodiments of the present invention, weights of DNN 1100 may be obtained. For example, a stochastic gradient descent (SGD) process, its variants and/or other algorithms may be used to obtain weights of DNN 1100 such that the loss function is minimized. The obtained weights (weight values) may be then used to process the input features (vectors) to acquire the optimal read threshold voltage. In various embodiments, input features may have read thresholds, checksums and an asymmetric ratio. For example, DNN's obtained weights may be in the form of multiple matrices ADNN,0, ADNN,1, . . . ADNN,N and bias vectors bDNN,0, bDNN,1, . . . , bDNN,N. That is, DNN's obtained weights may include a combination of multiple matrices ADNN,0, ADNN,1, . . . , ADNN,N and bias vectors bDNN,0, bDNN,1, . . . , bDNN,N.
DNN 1100 may receive inputs (i.e., a read threshold set, checksums and an asymmetric ratio/ones counts/zeros counts set) and estimate an optimal read threshold voltage OptVt based on the received inputs and weights. One example of estimation of the optimal read threshold set by the DNN 1100 is described below.
In one implementation, three reads are used to collect input information such as the checksum and asymmetric ratio/ones count/zeros count. As noted above, since two read threshold voltages are used for the LSB page in the TLC NAND flash memory device, these three reads correspond to six read threshold voltages. As a result, the read threshold set Vt, the checksum CS and the asymmetric ratio/ones count/zeros count AR are represented as below:
Vt=[Vtread0L,Vtread1L,Vtread2L,Vtread0R,Vtread1R,Vtread2R];
CS=[CSread0,CSread1,CSread2]; and
AR=[ARread0,ARread1,ARread2].
In the equations above, Vtread0L represents a first read threshold voltage (e.g., VT2 of
In one embodiment of the invention, the DNN 1100 shown in
OptVt=ADNN,2×(R(ADNN,1×R(ADNN,0×[Vt,CS,AR]+bDNN,0)+bDNN,1))+bDNN,N2.
The equation above for three (3) previous reads may be generalized for N previous reads as below:
OptVt=ADNN,N×R(ADNN,N-1×R(ADNN,N-2×R( . . . R(ADNN,0×[Vt,CS,AR]+bDNN,0) . . . )+bDNN,N-2)+bDNN,N-1)+bDNN,N.
In the equations above, R represents a set activation function. In various embodiments of the present invention, the set activation function may be a rectified linear unit (ReLU) activation function. Other activation functions may be applicable. As noted above, the DNN's weights may be in the form of multiple matrices ADNN,0, ADNN,1, . . . ADNN,N and bias vectors bDNN,0, bDNN,1, . . . , bDNN,N.
Referring to
At operation 1320, the optimal read threshold determiner 930 may obtain the read threshold set, a checksum value and an asymmetric ratio of ones count and zeros count which are associated with decoding of the selected page according to each read operation.
At operation 1330, the optimal read threshold determiner 930 may provide the obtained read threshold set, checksum value and asymmetric ratio as input information of the neural network.
At operation 1340, the optimal read threshold determiner 930 may estimate, by the neural network 1100, an optimal read threshold voltage based on the input information and weights for the input information, the weights including a combination of multiple matrices and bias vectors.
In various embodiments of the present invention, the neural network 1100 may be trained to obtain the weights for the input information such that a set loss function is minimized. In this embodiment, the set loss function may include a differentiable loss function to measure an error between outputs of the neural network and the optimal read threshold voltage.
In other embodiments, the neural network 1100 may estimate the optimal read threshold voltage based on a combination of the input information, the multiple matrices, the bias vectors and a particular activation function. In this embodiment, the optimal read threshold voltage may be estimated based on Equation above: OptVt=ADNN,N×R(ADNN,N-1×R (ADNN,N-2×R( . . . R(ADNN,0×[Vt, CS, AR]+bDNN,0) . . . )+bDNN,N-2)+bDNN,N-1)+bDNN,N, where OptVt represents the optimal read threshold voltage, Vt represents the read threshold set, CS represents the checksum values, AR represents the asymmetric ratio of ones count to zeros count, ADNN,N represents a (N+1)th matrix, bDNN,N represents a (N+1)th bias vector, and R represents the activation function. In this embodiment, the activation function can include a rectified linear unit (ReLU) activation function.
In various embodiments of the present invention, the selected page may include a least significant bit (LSB) page of a triple level cell (TLC). In this embodiment, the read threshold voltage set may include a first read threshold voltage for distinguishing an erased state and a first program state and a second read threshold voltage for distinguishing a second program state and a third program state.
As described above, embodiments of the present invention can provide a scheme for estimating an optimal read threshold voltage in a memory system (e.g., SSD). Embodiments of the present invention can estimate an optimal read threshold voltage using information obtained from previous read attempts through a neural network without any additional sensing operations and any intermediated parameter estimations. Thus, these embodiments can improve quality of service (QoS) and reliability of the memory system.
Although the foregoing embodiments have been illustrated and described in some detail for purposes of clarity and understanding, the present invention is not limited to the details provided. There are many alternative ways of implementing the invention, as one skilled in the art will appreciate in light of the foregoing disclosure. The disclosed embodiments are thus illustrative, not restrictive. The present invention is intended to embrace all modifications and alternatives of the disclosed embodiment. Furthermore, the disclosed embodiments may be combined to form additional embodiments.
