STORAGE DEVICE AND METHOD FOR OPERATING STORAGE DEVICE - Patent application
By adjusting the number of states and program parameters based on word line strength, the storage device mitigates aging effects, enhancing the lifespan and reliability of non-volatile memory.
Patent Information
- Application Number
- JP2020195711
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2019-11-28
- Filing Date
- 2020-11-26
- Publication Date
- 2024-12-04
- Estimated Expiration
- 2040-11-26
AI Technical Summary
Flash memory word lines deteriorate due to aging from high voltage programming and erasing, leading to reduced lifespan and reliability of memory blocks.
A storage device adjusts the number of states programmed into each word line based on its strength, grouping word lines, and dynamically adjusting program parameters to reduce programming time deviation and extend the lifespan of memory blocks.
This approach prevents individual word lines from becoming unreadable or unwritable due to aging, thereby extending the lifespan and improving the reliability of non-volatile memory devices.
Smart Images

Figure 0007594415000003 
Figure 0007594415000004 
Figure 0007594415000005
Abstract
Description
[Technical field]
[0001] The present invention relates to a storage device, and more particularly to a storage device and method for adjusting the number of states programmed into non-volatile memory cells for each word line. [Background technology]
[0002] Flash memory, which is a non-volatile memory, can maintain stored data even when the power is cut off. When programming and erasing are performed on a word line of a flash memory, a high voltage is applied to a memory cell connected to the word line, causing the word line to age and the strength of the word line to decrease. When the life of one word line included in a memory block ends, the life of the memory block also ends. Therefore, wear leveling is required for each word line. [Prior art documents] [Patent documents]
[0003] [Patent Document 1] JP 2019-79520 A Summary of the Invention [Problem to be solved by the invention]
[0004] The present invention has been made in consideration of the above-mentioned conventional problems, and an object of the present invention is to provide a storage device and an operating method of the storage device that adjusts the number of states indicating the number of levels to which memory cells are programmed for each of a plurality of word lines based on the strength of each of the plurality of word lines. [Means for solving the problem]
[0005] In order to achieve the above-mentioned object, a storage device according to one aspect of the present invention includes a non-volatile memory including a plurality of memory blocks each including a plurality of word lines, and a memory controller that determines a word line strength for each of the plurality of word lines, adjusts a number of states for each of the plurality of word lines based on the word line strength, and adjusts program parameters for each of the plurality of word lines so that program time deviation between the plurality of word lines is reduced.
[0006] In order to achieve the above object, one aspect of the present invention provides an operating method for a storage device including a non-volatile memory, comprising the steps of: calculating a word line strength for each of a plurality of word lines of the non-volatile memory; adjusting a number of states for each of the plurality of word lines based on the word line strength of each of the plurality of word lines; and adjusting program parameters for each of the word lines so that program time deviation between the word lines is reduced.
[0007] In order to achieve the above object, according to another aspect of the present invention, a method for operating a storage device including a non-volatile memory includes the steps of: calculating a word line strength for each of a plurality of word lines included in the non-volatile memory; grouping the plurality of word lines into a plurality of word line groups based on the word line strength; adjusting the number of states of each of the plurality of word line groups for each word line group; and adjusting program parameters of each of the word line groups so that a program time deviation between the word line groups for each word line group is reduced. Effect of the Invention
[0008] According to the present invention, it is possible to prevent a particular word line from becoming unable to be written to and / or read from due to aging, thereby extending the lifespan of a memory block and improving the reliability of non-volatile memories and storage devices. [Brief description of the drawings]
[0009] [Figure 1] 1 is a block diagram showing a storage device according to an embodiment of the present invention; [Figure 2A] 11 illustrates an example in which the number of states of a word line is adjusted in a storage device according to an embodiment of the present invention; [Figure 2B] 11 illustrates an example in which the number of states of a word line is adjusted in a storage device according to an embodiment of the present invention; [Figure 3A] 1 is a block diagram illustrating a memory controller according to an embodiment of the present invention; [Figure 3B] 3B is a flowchart illustrating a method of operation of the memory controller of FIG. 3A. [Figure 4A] 1 is a flowchart illustrating a method for determining whether to adjust the number of states in a memory controller according to an embodiment of the present invention. [Figure 4B] 1 is a flowchart illustrating a method for determining whether to adjust the number of states in a memory controller according to an embodiment of the present invention. [Figure 4C] 1 is a flowchart illustrating a method for determining whether to adjust the number of states in a memory controller according to an embodiment of the present invention. [Diagram 5] FIG. 2 is a system diagram of a memory controller according to one embodiment of the present invention. [Figure 6] 1 is a block diagram illustrating a non-volatile memory according to one embodiment of the present invention. [Figure 7A] FIG. 7 is a block diagram showing an example of a memory cell array of FIG. 6; [Figure 7B] FIG. 7 is a block diagram showing an example of a memory cell array of FIG. 6; [Figure 8A]2 is a circuit diagram illustrating an example of a memory block included in a memory controller according to an embodiment of the present invention; [Figure 8B] FIG. 8B is a perspective view of the memory block of FIG. 8A. [Figure 9A] 1 is a graph showing an example of a threshold voltage distribution of a word line according to an embodiment of the present invention; [Figure 9B] 9B is a table illustrating an example of program parameters and status indicators for each of the word lines of FIG. 9A. [Figure 10] 4 is a timing diagram illustrating a program operation of a non-volatile memory according to an embodiment of the present invention. [Figure 11A] 11 is a graph showing an example of a distribution of threshold voltages of word lines when the number of states of a word line is adjusted according to an embodiment of the present invention; [Figure 11B] 11B is a table illustrating an example of program parameters and status indicators for each of the word lines of FIG. 11A. [Figure 12A] 11 is a graph showing an example of a distribution of threshold voltages of word lines when the number of states of a word line is adjusted according to an embodiment of the present invention; [Figure 12B] 12B is a table illustrating an example of program parameters and status indicators for each of the word lines of FIG. 12A. [Figure 13] 4 is a table illustrating an example of how fractional bits per memory cell may be stored according to an embodiment of the present invention. [Figure 14A] FIG. 10 illustrates an example of how fractional bits per memory cell are stored in accordance with an embodiment of the present invention. [Figure 14B] FIG. 10 illustrates an example of how fractional bits per memory cell are stored in accordance with an embodiment of the present invention. [Figure 15A] 1 is a block diagram illustrating a memory controller according to an embodiment of the present invention; [Figure 15B] 15B is a flowchart illustrating a method of operation of the memory controller of FIG. 15A. [Figure 16]FIG. 2 illustrates a method of operation of a reinforcement learning module according to one embodiment of the present invention. [Figure 17A] 1 is a block diagram illustrating a memory controller according to an embodiment of the present invention; [Figure 17B] 17B is a flowchart illustrating a method of operation of the memory controller of FIG. 17A. [Figure 18A] FIG. 2 illustrates an example of grouping of word lines according to an embodiment of the present invention. [Figure 18B] FIG. 2 illustrates an example of grouping of word lines according to an embodiment of the present invention. [Figure 19] 1 is a block diagram illustrating an SSD system according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0010] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, specific examples of the embodiments of the present invention will be described in detail with reference to the drawings.
[0011] FIG. 1 is a block diagram showing a storage device according to an embodiment of the present invention.
[0012] Referring to FIG. 1, storage device 100 includes a memory controller 110 and a non-volatile memory (NVM) 120, and storage device 100 communicates with a host via various interfaces to store data received from the host or transmit stored data to the host.
[0013] In one embodiment, storage device 100 is an internal memory device built into an electronic device. For example, storage device 100 is a solid static driver (SSD), an embedded universal flash storage (UFS), or an embedded multi-media card (eMMC). In one embodiment, storage device 100 is an external memory that is removable from an electronic device. For example, storage device 100 is a portable SSD, UFS, compact flash (CF) card, secure digital (SD) card, micro secure digital (Micro-SD) card, mini secure digital (Mini-SD) card, extreme digital (xD) card, or memory stick.
[0014] The electronic device incorporating storage device 100 or connected to storage device 100 may be, for example, a personal computer (PC), a data server, a network-attached storage, an internet of things (IoT) device, or a portable electronic device. The portable electronic device may be a laptop computer, a mobile phone, a smartphone, a tablet PC, a personal digital assistant (PDA), an enterprise digital assistant (EDA), a digital still camera, a digital video camera, an audio device, a portable multimedia player (PMP), a personal navigation device (PND), an MP3 player, a handheld game console, an e-book, a wearable device, or the like.
[0015] The non-volatile memory 120 refers to a memory or memory device that has the property of maintaining stored data even when the power supply is interrupted. Therefore, even if the power supplied to the storage device 100 is interrupted, the data stored in the non-volatile memory 120 is maintained. For example, the data is user data.
[0016] The non-volatile memory 120 operates under the control of the memory controller 110. For example, the non-volatile memory 120 receives a command CMD, an address ADDR, and a control signal CTRL from the memory controller 110, writes data DT corresponding to the address ADDR in response to the received command CMD, or reads data DT corresponding to the address ADDR, and transmits the read data DT to the memory controller 110.
[0017] The non-volatile memory 120 includes a plurality of memory blocks (BLK1 to BLKz), each of which includes a plurality of word lines. Here, z is a positive integer of 2 or more, and may be changed in various ways according to an embodiment. A plurality of memory cells are connected to the word line, and a plurality of memory cells connected to the same word line constitute a physical page. If the memory cell is a multi-level cell that stores data of 2 or more bits, data corresponding to a plurality of logical pages is stored in the word line. For example, a memory block is an erase unit, and a page is a program and read unit. That is, according to an embodiment, erasure is performed in units of memory blocks, and programming and reading are performed in units of pages. In an embodiment, the non-volatile memory 120 includes a plurality of planes, a plurality of dies, or a plurality of chips.
[0018] In one embodiment, the non-volatile memory 120 is a NAND flash memory device. In one embodiment, the non-volatile memory 120 is a planar NAND flash memory device (see FIG. 2A) having a two-dimensional horizontal array structure, or the non-volatile memory 120 is a vertical NAND flash (VNAN) memory device (see FIG. 3A) having a three-dimensional array structure. However, without being limited thereto, the non-volatile memory 120 may be a resistive memory device such as a resistive random access memory (ReRAM), a phase change RAM (PRAM), or a magnetic RAM (MRAM). In the following, for convenience of explanation, the non-volatile memory 120 is described as a NAND flash memory, but it will be understood that the present invention is not limited thereto.
[0019] The memory controller 110 controls the overall operation of the storage device 100. The memory controller 110 also controls write, read, and erase operations performed on the non-volatile memory 120, and sets and adjusts parameters related to the write, read, and erase operations.
[0020] For example, the memory controller 110 writes data received from the host to the non-volatile memory 120 in response to a request from the host, or reads data DT from the non-volatile memory 120 and transmits the read data DT to the host. The memory controller 110 encodes the data DT and transmits it to the non-volatile memory 120, and the non-volatile memory 120 programs the encoded data in page units to each word line. The memory controller 110 decodes the data DT read from the non-volatile memory 120.
[0021] A memory cell is programmed with one of a number of states (i.e., an erased state and one or more programmed states) that are assigned by the bits (either integer bits or non-integer bits, e.g., fractional bits) that are set. For example, if a memory cell stores two bits of data, the memory cell can store four (=2 2 If the memory cell stores 2.5 bits of data, the memory cell is programmed with one of six states, whereby data is written to the non-volatile memory 120.
[0022] The memory controller 110 sets or adjusts the number of states of each of a plurality of word lines included in the memory unit in the non-volatile memory 120. Specifically, the memory controller 110 sets or adjusts the number of states (an erased state and at least one programmed state) to be programmed in memory cells connected to each of a plurality of word lines (i.e., the number of states). At this time, the memory unit may be a block unit, a chip unit, or a wafer unit.
[0023] The memory controller 110 calculates a number of word line strengths (hereinafter, referred to as strengths) corresponding to each of the word lines (i.e., calculates the strengths of each of the word lines), and sets or adjusts the number of states of each of the word lines based on the calculated strengths. The memory controller 110 sets or adjusts the number of states of each of the word lines such that a word line with a relatively high strength, e.g., a strong word line, is assigned a greater number of states than a word line with a relatively low strength, e.g., a weak word line.
[0024] In NAND flash, when a number of program or erase operations are performed on a word line, stress (e.g., high voltage) is repeatedly applied to the word line, which causes the word line to age. The stress applied to each word line varies over time, and therefore the word lines have different strengths (i.e., word line strengths). As the word line ages, the strength of the word line decreases, causing the word line to become weak.
[0025] The memory controller 110 performs wear leveling on a word line basis by adjusting the number of states of each of the word lines based on the strengths corresponding to each of the word lines. The more states a word line is assigned to, the more stress is applied to the word line when the word line is programmed. A smaller number of states is assigned to weak word lines and a relatively larger number of states is assigned to strong word lines, thereby reducing the stress applied to the weak word lines. This prevents a particular word line from aging faster than other word lines.
[0026] In one embodiment, the memory controller 110 groups the word lines into word line groups based on the word line strengths of the word lines and dynamically adjusts the number of states and program parameters for each of the word line groups, i.e., in one embodiment, the word lines in a word line group are adjusted together rather than individually.
[0027] In one embodiment, the memory controller 110 adjusts the number of states of each of the word lines through reinforcement learning. The memory controller 110 performs reinforcement learning based on information that has been profiled in advance.
[0028] In one embodiment, the memory controller 110 periodically or aperiodically adjusts the number of states of each of the plurality of word lines during run time of the storage device 100. However, without being limited thereto, in another embodiment, the memory controller 110 presets the number of states of each of the plurality of word lines (e.g., the preset number of states is stored in memory in advance) and controls program and read operations of the non-volatile memory 120 based on the preset number of states during run time of the storage device 100.
[0029] In one embodiment, the memory controller 110 dynamically adjusts program parameters for each of the word lines by adjusting the number of states of each of the word lines. For example, a program time for a word line with a large number of states is longer than a program time for a word line with a small number of states. Here, the program time means a time from when programming starts to when programming is performed on a word line. The memory controller 110 dynamically adjusts program parameters for each of the word lines to reduce program time variation between the word lines.
[0030] As described above, the storage device 100 according to the present embodiment calculates the strength of each of the plurality of word lines, and adjusts the number of states for each word line based on the strengths corresponding to each of the plurality of word lines. A relatively large number of states are assigned to the strong word lines, and a relatively small number of states are assigned to the weak word lines. Thus, the number of states for the strong word lines is increased, and the number of states for the weak word lines is decreased. After the number of states is adjusted, the storage device 100 dynamically adjusts program parameters for each word line to reduce program time deviation between the word lines.
[0031] When writing and / or reading becomes impossible in one of the word lines constituting a memory block, i.e., when decoding of data read from the word line becomes impossible, further writing and reading to the memory block becomes impossible. The storage device 100 according to the present embodiment performs wear leveling for each word line by adjusting the number of states for each word line so that a relatively large number of states are assigned to strong word lines and a relatively small number of states are assigned to weak word lines. As a result, it is possible to prevent a specific word line from becoming unable to write and / or read due to aging, thereby extending the life of the memory block and improving the reliability of the non-volatile memory 120 and the storage device 100.
[0032] 2A and 2B are diagrams illustrating an example in which the number of states of a word line is adjusted in a storage device according to an embodiment of the present invention.
[0033] For convenience of explanation, two word lines, i.e., a first word line WL1 and a second word line WL2, are taken as an example in Fig. 2A, and three word lines, i.e., a first word line WL1, a second word line WL2, and a third word line WL3, are taken as an example in Fig. 2B, but the present invention is not limited thereto, and the adjustment of the number of states is performed for a plurality of word lines included in a memory unit, for example, a memory block, a memory chip, or a memory lot.
[0034] 2A, eight states (i.e., an erase state E and a first program state P1 to a seventh program state P7) are assigned to the first word line WL1 and the second word line WL2, respectively, and in a program step, the first word line WL1 and the second word line WL2 are programmed to one of the eight states, so that three bits of data are stored in the memory cells of the first word line WL1 and the second word line WL2, respectively.
[0035] As the program, read, and erase operations for the first and second word lines WL1 and WL2 proceed, the first and second word lines WL1 and WL2 age, and the strength of the first word line WL1 becomes relatively high and the strength of the second word line WL2 becomes relatively low. For example, if the number of program and / or read operations for the first word line WL1 is greater than the number of program and / or read operations for the second word line WL2, the strength of the first word line WL1 becomes higher than the strength of the second word line WL2.
[0036] The number of states of each of the first word line WL1 and the second word line WL2 is adjusted so that more states are assigned to the strong word line and fewer states are assigned to the weak word line. The number of states of the first word line WL1 is increased and the number of states of the second word line WL2 is decreased. For example, the number of states of the first word line WL1 is increased from 8 to 12, and the number of states of the second word line WL2 is decreased from 8 to 4. As shown in the figure, 16 states (e.g., erase state E and first program state P1 to 15th program state P15) are assigned to the first word line WL1, and 4 states (e.g., erase state E and first program state P1 to 3rd program state P3) are assigned to the second word line WL2. Thus, 4 bits of data are stored in the memory cell of the first word line WL1, and 2 bits of data are stored in the memory cell of the second word line WL2.
[0037] 2B, the first word line WL1, the second word line WL2, and the third word line WL3 are each assigned four states (i.e., an erase state E, and a first program state P1 to a third program state P3), and in a program step, the first word line WL1, the second word line WL2, and the third word line WL3 are programmed with one of eight states, so that two bits of data are stored in each of the memory cells of the first word line WL1, the second word line WL2, and the third word line WL3.
[0038] As the program, read, and erase operations proceed for each of the word lines, the first word line WL1, the second word line WL2, and the third word line WL3 age. The strengths of the first word line WL1, the second word line WL2, and the third word line WL3 are -1, -2, and -3, respectively. The first word line WL1 is the strongest and the third word line WL3 is the weakest.
[0039] The number of states of each of the first word line WL1, the second word line WL2, and the third word line WL3 is adjusted so that a greater number of states are assigned to the strong word line and a smaller number of states are assigned to the weak word line. For example, the number of states of the first word line WL1 is increased from 4 to 6, the number of states of the second word line WL2 is maintained at 4, and the number of states of the third word line WL3 is decreased from 4 to 3. As a result, 2.5 bits of data are stored in the memory cell of the first word line WL1, 2 bits of data are stored in the memory cell of the second word line WL2, and 1.5 bits of data are stored in the memory cell of the third word line WL3.
[0040] As described with reference to Figures 2A and 2B, the state number adjustment module (included in the memory controller 110 of Figure 1) adjusts the state number of each of the plurality of word lines based on the strength of each of the plurality of word lines. In one embodiment, as shown in Figure 2B, the state number of the word line is 2, which represents an integer number of bits. s (where s is a natural number), and fractional bits of data are stored in the memory cells of the word line.
[0041] Fig. 3A is a block diagram illustrating a memory controller according to an embodiment of the present invention, and Fig. 3B is a flowchart illustrating an operation method of the memory controller of Fig. 3A. The operation method of Fig. 3B may be performed at run-time (e.g., a stage where operations such as writing, reading, and erasing are performed on the non-volatile memory 120 based on a request from a host) of the storage device 100 (Fig. 1) or before run-time (e.g., a manufacturing stage or a startup stage of the storage device).
[0042] Referring to FIG. 3A, the memory controller 110 includes a strength calculation module SCM, a state number adjustment module CAM, and a program parameter control module PCM.
[0043] In the embodiments of the present invention, the term "module" is used to refer to a component according to an embodiment. The "module" may be realized by various hardware devices such as an integrated circuit (IC), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or a complex programmable logic device (CPLD), firmware or software operated by the hardware device, or a combination of a hardware device and software. The "module" may also include a circuit or intellectual property (IP) configured with semiconductor elements.
[0044] The strength calculation module SCM calculates a strength for each of the plurality of word lines. The strength calculation module SCM calculates the strength of the word line based on program parameters and status metrics (e.g., status metrics for program, read, and erase operations).
[0045] The state number adjustment module CAM adjusts the state number of each of the plurality of word lines based on the calculated strength for each of the plurality of word lines.
[0046] The program parameter control module PCM adjusts program parameters of each of the word lines, for example, the program parameters include at least one of a program start voltage, a program voltage difference, and a program verify voltage, such that a difference in time at which programming is performed between the word lines (hereinafter, referred to as program time deviation) is reduced.
[0047] 3B, the memory controller 110 acquires a status indicator for each of a plurality of word lines (step S110). For example, the strength calculation module SCM acquires the status indicator based on a signal received from the non-volatile memory 120. The status indicator includes, for example, an actual program time, an erase time, a read time, a read error bit, a read level shift value, and a number of states. For example, every time a program is performed on a word line of the non-volatile memory 120, the memory controller 110 acquires and stores the actual program time. In one embodiment, the memory controller 110 acquires and stores the status indicator every time a program operation is performed, and in another embodiment, the memory controller 110 acquires and stores the status indicator periodically.
[0048] The strength calculation module SCM calculates a plurality of word line strengths corresponding to the plurality of word lines (operation S120). That is, the strength calculation module SCM calculates the strength of each of the plurality of word lines. The strength calculation module SCM calculates the strength of the word line based on the program parameters and the state indicators collected in operation S110.
[0049] In one embodiment, the strength calculation module SCM calculates the strength based on the expected program time and the actual program time for the word line. For example, the strength calculation module SCM calculates the expected program time for the word line based on the preset program parameters and the preset number of states, and calculates the strength based on the difference between the expected program time and the actual program time. In one embodiment, the strength calculation module SCM includes a strength calculation function that uses the program parameters and at least one state indicator as variables. However, the present invention is not limited thereto, and for example, the strength calculation module SCM calculates the strength by further considering at least one of other state indicators, such as erase time, read error bit, and read level shift value, when calculating the strength.
[0050] Thereafter, the state number adjustment module CAM adjusts the number of states of each of the word lines based on the strengths of the word lines (step S130). The state number adjustment module CAM adjusts the number of states of the word lines so that a large number of states are assigned to strong word lines and a small number of states are assigned to weak word lines. As a result, wear leveling is performed for each word line, and the strength difference between the word lines is reduced.
[0051] Meanwhile, steps S120 and S130 are performed based on reinforcement learning. For example, the strength calculation module SCM and the state number adjustment module CAM are embodied as a reinforcement learning agent, and the agent calculates the state number for each word line based on the program time, the state number, program parameters, and state index stored for each word line. The calculation of the state number for each word line through reinforcement learning will be described in detail below with reference to Figures 15A to 15C.
[0052] The program parameter control module PCM adjusts the program parameters of each of the word lines by adjusting the number of states (step S140). The program parameter control module PCM adjusts the program parameters of each of the word lines such that a difference in time at which programming is performed between the word lines (program time deviation) is reduced.
[0053] For example, in FIG. 2A, the number of states of the first word line WL1 is adjusted to 16, and the number of states of the second word line WL2 is adjusted to 4, so that when programming is performed based on the same program parameters for the first word line WL1 and the second word line WL2, the time for programming the first word line WL1, i.e., the program time of the first word line WL1, is longer than the program time of the second word line WL2. That is, if each of the 15 program states (P1 to P15) of the first word line WL1 after the number of states is adjusted has the same program time as each of the seven program states (P1 to P7) of the first word line WL1 before the number of states is adjusted, and each of the three program states (P1 to P3) of the second word line WL2 after the number of states is adjusted has the same program time as each of the seven program states (P1 to P7) of the second word line WL2 before the number of states is adjusted, then the total program time of the first word line WL1 after the number of states is adjusted will be significantly longer than the total program time of the second word line WL2.
[0054] The program parameter control module PCM adjusts the program parameters of the first word line WL1 and / or the second word line WL2 so that the actual program time of the first word line WL1 is similar to the actual program time of the second word line WL2. That is, the program time of each state of the first word line WL1 is shortened, and the program time of each state of the second word line WL2 is extended. The total program time of the first word line WL1 becomes similar to the total program time of the second word line WL2, and the deviation of the program time is reduced. For example, the program parameter control module PCM increases the program start voltage of the first word line WL1 or decreases the program start voltage of the second word line WL2 to reduce the program time deviation between the first word line WL1 and the second word line WL2. The program parameter control module PCM adjusts at least one of the program start voltage, the program voltage difference, and the program verify voltage for each of the plurality of word lines.
[0055] The memory controller 110 transmits the number of states and program parameters of each of the word lines to the non-volatile memory 120 (step S150). In one embodiment, the memory controller 110 transmits the number of states and program parameters of the word line together with an address indicating the word line on which the program is to be executed to the non-volatile memory 120, and the non-volatile memory 120 performs a program operation on the word line indicated by the address based on the received number of states and program parameters. In another embodiment, the memory controller 110 transmits the number of states and program parameters of each of the word lines to the non-volatile memory 120, and the non-volatile memory 120 stores the received number of states and program parameters of each of the word lines. Thereafter, the memory controller 110 transmits an address on which the program is to be executed to the non-volatile memory 120, and the non-volatile memory 120 performs a program operation on the word line corresponding to the received address based on the number of states and program parameters.
[0056] 4A, 4B, and 4C are flow charts illustrating a method for determining whether to adjust the number of states in a memory controller according to an embodiment of the present invention.
[0057] Referring to FIG. 4A, the memory controller 110 determines whether to adjust the number of states based on the program time difference ΔtPGM of a plurality of word lines.
[0058] A program operation is performed on each of the word lines (step S11), and the memory controller 110 monitors the program time of each of the word lines (step S12). After transmitting a program command CMD to the word lines, the memory controller 110 obtains the program time of the word lines based on response signals, such as a ready signal and a busy signal, received from the non-volatile memory 120. For example, the memory controller 110 obtains the program time based on the time from when the busy signal is received until when the ready signal is received.
[0059] The memory controller 110 determines whether the program time difference ΔtPGM between the plurality of word lines is equal to or greater than the threshold time Tth (step S13). For example, the memory controller 110 determines whether the maximum value of the program time difference ΔtPGM between the plurality of word lines is equal to or greater than the threshold time Tth. Alternatively, the memory controller 110 adjusts the number of states in units of at least two word lines and determines whether the program time difference ΔtPGM between the at least two word lines is equal to or greater than the threshold time Tth.
[0060] If the program time difference ΔtPGM of the word lines is equal to or greater than the threshold time Tth, the memory controller 110 adjusts the number of states of each of the word lines (step S14). The memory controller 110 adjusts the number of states by the method described with reference to FIG. 3B.
[0061] On the other hand, if the program time difference ΔtPGM of the word lines is less than the threshold time Tth, the memory controller 110 maintains the number of states of each of the word lines (step S15). The preset number of states for each of the word lines is maintained.
[0062] Referring to FIG. 4B, the memory controller 110 determines whether to adjust the number of states based on program and erase cycles (P / E cycles) of a number of word lines.
[0063] A program operation, a read operation, and an erase operation are performed on each of the plurality of word lines (step S21), and at this time, the memory controller 110 stores a P / E cycle for each word line and updates the P / E cycle every time a program operation or a read operation is performed.
[0064] The memory controller 110 determines whether the P / E cycles are equal to or greater than a threshold value VUth (step S22). For example, the memory controller 110 determines whether the P / E cycles of one or more word lines among the plurality of word lines are equal to or greater than a threshold value VUth. As another example, the memory controller 110 determines whether the average of the P / E cycles of each of the plurality of word lines, i.e., the average P / E cycles, is equal to or greater than a threshold value VUth.
[0065] The memory controller 110 adjusts the number of states of each of the word lines when the P / E cycle is equal to or greater than the threshold value VUth (step S23).The memory controller 110 maintains the number of states of each of the word lines when the P / E cycle is less than the threshold value VUth (step S24).
[0066] Referring to FIG. 4C, when the storage device 100 (FIG. 1) is powered on, the memory controller 110 determines whether to adjust the number of states.
[0067] The storage device 100 is powered on (step S31). The storage device 100 receives power from an external source and is powered on based on the received power. The memory controller 110 determines whether the number of states needs to be adjusted (step S32). For example, the memory controller 110 determines whether a condition for adjusting the number of states is satisfied during the startup process. For example, as described with reference to FIGS. 4A and 4B, the memory controller 110 determines whether a condition for adjusting the number of states is satisfied based on the program time difference ΔtPGM of a plurality of word lines or program and erase cycles (P / E cycles) of a plurality of word lines. However, the present invention is not limited thereto, and the condition for adjusting the number of states may be set in various manners.
[0068] If it is determined that the number of states needs to be adjusted, i.e., if the conditions for performing the adjustment of the number of states are satisfied, the memory controller 110 adjusts the number of states of each of the plurality of word lines (step S33), and if it is determined that the adjustment of the number of states is not necessary, the memory controller 110 maintains the preset number of states for each of the plurality of word lines (step S34).
[0069] The memory controller 110 periodically or aperiodically adjusts the number of states of the word lines, and determines whether to adjust the number of states according to the methods described with reference to FIGS. 4A, 4B, and 4C, for example.
[0070] FIG. 5 is a system diagram of a memory controller according to one embodiment of the present invention.
[0071] 5, the memory controller 110 includes a processor 111, a random access memory (RAM) 112, an encoding and decoding module 113, a read only memory (ROM) 114, a host interface 115, and a memory interface 116. The components of the memory controller 110, such as the processor 111, the RAM 112, the encoding and decoding module 113, the ROM 114, the host interface 115, and the memory interface 116, transmit and receive data to and from each other via a system bus 117. The memory controller 110 may further include other components.
[0072] The processor 111 may include a central processing unit (CPU) or a microprocessor, and controls the overall operation of the memory controller 110. The processor 111 may include one or more processor cores that execute program code or a set of instructions of the program code configured to perform a specific operation. The processor 111 executes instruction codes of firmware stored in the ROM 114. In one embodiment, the processor 111 may be implemented as a multi-core processor, for example, a dual-core processor or a quad-core processor.
[0073] The RAM 112 operates as at least one of a buffer memory, a cache memory, an operating memory, and a main memory of the memory controller 111. The RAM 112 is implemented by a dynamic random access memory (DRAM), a static random access memory (SRAM), etc., and firmware is loaded into the RAM 112. The firmware includes program codes that implement the operating algorithm of the memory controller 110 described with reference to FIGS. 1 to 4C. For example, the firmware includes a strength calculation module SCM, a state number adjustment module CAM, and a program parameter control module PCM.
[0074] The encoding and decoding module 113 performs encoding and decoding on data and performs error correction on the data. The encoding and decoding module 113 performs encoding on data stored in the non-volatile memory 120 and generates an error correction code. The encoding and decoding module 113 decodes data read from the non-volatile memory 120, detects an error in the data based on the error correction code, and corrects the detected error. In one embodiment, the encoding and decoding module 113 is also embodied by hardware. In one embodiment, the encoding and decoding module 113 is also embodied by software or firmware and is loaded into the RAM 112. If the data read from the non-volatile memory 120 cannot be decoded, i.e., if the error in the data is corrected by the error correction code, the word line from which the data was read and the memory block including the word line are determined to have reached the end of their life.
[0075] The ROM 114 stores various information required to operate the memory controller 110. Exemplarily, the ROM 114 stores various information in the form of firmware. When the storage device 100 (FIG. 1) is powered on, the firmware stored in the ROM 114 is loaded into the RAM 112.
[0076] The memory controller 110 communicates with an external device (e.g., a host) via a host interface 115. Exemplarily, the host interface 115 includes a variety of interfaces such as universal serial bus (USB), multimedia card (MMC), peripheral component interconnection (PCI), PCI-E (PCI-express), advanced technology attachment (ATA), serial-ATA (SATA), parallel-ATA (PATA), small computer small interface (SCSI), enhanced small disk interface (ESDI), integrated drive electronics (IDE), mobile industry processor interface (MIPI), and nonvolatile memory-express (NVM-e).
[0077] The memory interface 116 provides an interface between the memory controller 110 and the non-volatile memory 200. For example, commands, addresses, control signals (e.g., state numbers and program parameters), program data, read data, and response signals are transmitted and received between the memory controller 110 and the non-volatile memory 200 via the memory interface 116.
[0078] FIG. 6 is a block diagram illustrating a non-volatile memory according to one embodiment of the present invention.
[0079] 6, the non-volatile memory 120 includes a memory cell array 121, a control logic 122, a voltage generator 123, an address decoder 124, a page buffer circuit 125, and a data input / output circuit 126.
[0080] The memory cell array 121 is connected to a plurality of word lines WL and a plurality of bit lines BL. The memory blocks are connected to one or more string select lines (SSL) and one or more ground select lines (GSL). The memory cell array 121 includes a plurality of memory blocks, each of which includes a plurality of memory cells (MC1 to MC8) (FIG. 8A) arranged in an area where a plurality of word lines WLs and a plurality of bit lines BLs intersect.
[0081] In one embodiment, each of the plurality of memory cells stores one bit of data or multiple bits of data, hi one embodiment, each of the plurality of memory cells stores a fractional bit of data.
[0082] When an erase voltage is applied to the memory cell array 121, the memory cells are in an erased state, and when a program voltage is applied to a selected word line of the memory cell array 121, the memory cells connected to the selected word line are programmed to a set state. As shown in Figures 2A and 2B, each of the memory cells has an erased state E and at least one program state distinguished by a threshold voltage.
[0083] Based on the command CMD, address ADDR, and control signal CTRL received from the memory controller 110, the control logic 122 outputs various control signals to program data into the memory cell array 121, read data from the memory cell array 121, or erase the memory cell array 121.
[0084] Various control signals output from the control logic 122 are provided to the voltage generator 123, the address decoder 124, the page buffer circuit 125, and the data input / output circuit 126. Specifically, the control logic 122 provides a voltage control signal CON_vol to the voltage generator 123, a row address R-ADDR to the address decoder 123, a program control signal CON_P to the page buffer circuit 125, and a data input / output control signal CON_I / O to the data input / output circuit 126. However, the present invention is not limited thereto, and the control logic 122 may further provide other control signals to the voltage generator 123, the address decoder 124, the page buffer circuit 125, and the data input / output circuit 126.
[0085] The voltage generator 123 generates various voltages for performing a program operation, a read operation, and an erase operation on the memory cell array 121 based on the voltage control signal CON_vol. Specifically, the voltage generator 123 generates word line driving voltages VWLs for driving a plurality of word lines WLs. The word line driving voltages VWLs are, for example, a program voltage (or a write voltage), a read voltage, an erase voltage, an inhibit voltage, or a program verify voltage. The voltage generator 123 further generates a string selection line driving voltage for driving a plurality of string selection lines SSL, and a ground selection line driving voltage for driving a plurality of ground selection lines GSL.
[0086] The address decoder 114 is connected to the memory cell array 121 via a plurality of word lines WLs, and selects some of the plurality of word lines WLs in response to a row address R-ADDR received from the control logic 121. During a program operation, the row decoder 124 applies a program voltage to the selected word lines and an inhibit voltage to the unselected word lines.
[0087] The page buffer circuit 125 is connected to the memory cell array 121 via a plurality of bit lines BLs. In particular, during a program operation, the page buffer circuit 125 operates as a write driver and programs a state corresponding to data DT to be stored in the memory cell array 121. During a read operation, the page buffer 125 operates as a sense amplifier and outputs the data DT stored in the memory cell array 121.
[0088] The data input / output circuit 126 transmits data DT input from the outside to the page buffer circuit 125, or transmits data DT output from the page buffer circuit 125 to the memory controller 110 (FIG. 1) via a number of input / output pins or data buses.
[0089] According to the present embodiment, the control logic 122 generates the voltage control signal CON_vol and the program control signal CON_P based on the number of states of each of the word lines WLs and program parameters received from the memory controller 110 (FIG. 1). The program parameters include at least one of a program start voltage, a program voltage difference, and a program verify voltage, and the control logic 122 generates the voltage control signal CON_vol based on the program parameters.
[0090] In addition, the control logic 122 generates a program control signal CON_P that controls each of the memory cells in the word line to be programmed to one of the states assigned to the word line based on the number of states received from the memory controller 110 when programming is performed on a specific word line, and provides the program control signal CON_P to the page buffer circuit 125.
[0091] 7A and 7B are block diagrams illustrating examples of the memory cell array of FIG.
[0092] Referring to Figures 7A and 7B, the memory cell array (121a and 121b) includes a plurality of memory blocks, such as a first memory block BLK1, a second memory block BLK2, and a third memory block BLK3, and each of the first memory block BLK1, the second memory block BLK2, and the third memory block BLK3 has a three-dimensional structure (or a vertical structure).
[0093] Each of the first memory block BLK1, the second memory block BLK2, and the third memory block BLK3 includes structures extending along a plurality of directions (x, y, z) corresponding to three dimensions. Each of the first memory block BLK1, the second memory block BLK2, and the third memory block BLK3 includes a plurality of NAND cell strings extending along a direction perpendicular to the substrate SUB (e.g., the z direction).
[0094] As shown in FIG. 7A, the first memory block BLK1, the second memory block BLK2, and the third memory block BLK3 are stacked in a direction perpendicular to the substrate SUB, or as shown in FIG. 7B, the first memory block BLK1, the second memory block BLK2, and the third memory block BLK3 are arranged side by side in a direction parallel to the substrate SUB (e.g., the x direction or y direction).
[0095] FIG. 8A is a circuit diagram illustrating an example of a memory block included in a memory controller according to an embodiment of the present invention.
[0096] 8A, the memory block BLK is a vertical structure NAND flash memory, and at least one of the memory blocks shown in FIGS. 7A and 7B is implemented by the memory block BLK shown in FIG. 8A.
[0097] The memory block BLK includes a plurality of NAND strings (NS11 to NS33), a plurality of word lines (WL1 to WL8), a plurality of bit lines (BL1 to BL3), a ground selection line GSL, a plurality of string selection lines (SSL1 to SSL3), and a common source line CSL, where the numbers of NAND strings, word lines, bit lines, ground selection lines, and string selection lines may be variously changed according to the embodiment.
[0098] NAND strings NS11 to NS33 are connected between a plurality of bit lines BL1 to BL3 and a common source line CSL. Each NAND string (e.g., NAND string NS11) includes a string selection transistor SST, a plurality of memory cells MC1 to MC8, and a ground selection transistor GST, which are connected in series. In one embodiment, at least one dummy cell is disposed between the string selection transistor SST and the first memory cell MC1 and / or between the string selection transistor SST and the eighth memory cell MC8.
[0099] The string selection transistors SST are connected to string selection lines SSL1 to SSL3. The memory cells MC1 to MC8 are connected to corresponding word lines WL1 to WL8. The ground selection transistors GST are connected to a ground selection line GSL. The string selection transistors SST are connected to corresponding bit lines BL, and the ground selection transistors GST are connected to a common source line CSL.
[0100] In the plurality of NAND strings NS11 to NS33, word lines of the same height (e.g., the first word line WL1 connected to each of the plurality of NAND strings NS11 to NS33) are commonly connected, and the string selection lines SSL1 to SSL3 are separated. When the first string selection line SSL1 is selected and a program voltage is applied to the first word line WL1, memory cells included in the NAND strings NS11, NS12, NS13 of the first row and connected to the first word line WL1 are programmed. Thus, programming is performed in page units.
[0101] FIG. 8B is a perspective view of the memory block of FIG. 8A.
[0102] 8B, the memory block BLK is formed in a direction perpendicular to a substrate SUB. A common source line CSL is disposed on the substrate SUB, and a gate electrode GE and an insulating film IL are stacked on the substrate SUB. A charge storage layer CS is formed between the gate electrode GE and the insulating film IL.
[0103] A plurality of gate electrodes GE and insulating films IL stacked on top of each other are vertically patterned to form a pillar P. The pillar P has a V-shape. The pillar P is connected to the substrate SUB through the gate electrodes GE and insulating films IL. An outer portion O of the pillar P is made of a semiconductor material and functions as a channel region, and an inner portion I of the pillar P is made of an insulating material such as silicon oxide.
[0104] The gate electrodes GE of the memory block BLK are respectively connected to a ground selection line GSL, a plurality of word lines (WL1 to WL8), and a string selection line SSL. And, the pillars P of the memory block BLK are connected to a plurality of bit lines (BL1 to BL3). In FIG. 8B, the memory block BLK is illustrated as including two selection lines (GSL, SSL), eight word lines (WL1 to WL8), and three bit lines (BL1 to BL3), but in reality, there may be more or less than these.
[0105] As shown in Figures 8A and 8B, in a memory block BLK having a three-dimensional vertical structure, the position of a word line affects the strength of the word line. For example, a word line located at the top or bottom of the memory block BLK, for example, the first word line WL1 adjacent to the ground selection line GSL or the eighth word line WL8 adjacent to the string selection line SSL, is subjected to more stress than other word lines during program, verify, and erase due to structural characteristics of the memory block BLK, and thus has a relatively lower strength than other word lines. Meanwhile, a word line located at the center of the memory block BLK has a relatively higher strength than other word lines. However, the present invention is not limited thereto, and the tendency of the strengths corresponding to each of the word lines may vary depending on the manufacturing process of the memory block BLK, the level of the voltage applied to the word line, etc.
[0106] The memory controller 110 (FIG. 1) according to an embodiment of the present invention reflects the structural characteristics of the memory block BLK having a three-dimensional vertical structure when calculating the strength of the multiple word lines, and takes into account the position of each of the multiple word lines within the memory block BLK (e.g., a center portion or an edge portion, a position adjacent to a GSL or SSL, etc.).
[0107] Figure 9A is a graph showing an example of threshold voltage distribution of word lines according to an embodiment of the present invention, and Figure 9B is a table showing example program parameters and status indicators for each of the word lines of Figure 9A. Figure 9B shows an example of the change in program parameters and status indicators due to aging of each of the word lines.
[0108] 9A and 9B, for convenience of explanation, the word line threshold voltage distribution and table will be described with reference to the fourth word line WL4 and the twenty-third word line WL23. The fourth word line WL4 and the twenty-third word line WL23 may be disposed at different positions or adjacent to each other within the memory block.
[0109] 9A, initially, eight states are assigned to the fourth word line WL4 and the twenty-third word line WL23. For example, memory cells of the fourth word line WL4 and the twenty-third word line WL23 are programmed to an erase state E and one of the first program state P1 through the seventh program state P7. Three bits of data are stored per memory cell.
[0110] 9B, initially, the program parameters and state indicators of the fourth word line WL4 and the twenty-third word line WL23 are the same. For example, the program start voltage V ST , Program voltage difference ΔV ISPP , and the program verify voltage V VF are set to be equal. Referring to FIG. 10, the program start voltage V ST , Program voltage difference ΔV ISPP , and the program verify voltage V VF This article explains:
[0111] 10 is a timing diagram showing a program operation of a non-volatile memory according to an embodiment of the present invention, in which the horizontal axis indicates time and the vertical axis indicates a voltage applied to a word line WL during a program operation.
[0112] 10, the program operation includes a number of program loops (LOOP1 to LOOPj) (j is a natural number equal to or greater than 2). In each of the program loops (LOOP1 to LOOPj), program voltages (Vpgm1, Vpgm2, Vpgm3, ..., Vpgmj) (also called program pulses) are applied to the word lines WL. After the program voltages are applied, a program verify voltage V VF By using the ISPP (incremental step pulse programming) method, the voltage levels of the program voltages (Vpgm1, Vpgm2, Vpgm3, . . . , Vpgmj) applied in a plurality of program loops (LOOP1 to LOOPj) are increased.
[0113] Meanwhile, in the first program loop LOOP1, the first program voltage Vpgm1 applied to the word line WL is equal to the program start voltage V ST The voltage difference between the program voltages, for example, the voltage difference between the first program voltage Vpgm1 and the second program voltage Vpgm2, is called the program voltage difference ΔV ISPP In one embodiment, the voltage difference between the program voltages in successive program loops is the same. For example, the voltage difference between the first program voltage Vpgm1 and the second program voltage Vpgm2 is the same as the voltage difference between the second program voltage Vpgm2 and the third program voltage Vpgm3. In another embodiment, the voltage difference between the program voltages in successive program loops is different. For example, the voltage difference between the second program voltage Vpgm2 and the third program voltage Vpgm3 is greater than the voltage difference between the first program voltage Vpgm1 and the second program voltage Vpgm2.
[0114] Program verify voltage V VF is a voltage for verifying whether a plurality of memory cells connected to a word line WL are programmed to a target state, and the threshold voltage of the memory cell is set to a program verify voltage V VFIf so, the memory cell is determined to have been programmed to the target state, and the threshold voltage of the memory cell is equal to or higher than the program verify voltage V. VF If it is lower, it is determined that the memory cell was not programmed to the target state.
[0115] Referring again to FIG. 9B, the expected program time tPGM for each of the fourth word line WL4 and the twenty-third word line WL23 is E and the actual program time tPGM R is equal to 5ms (milliseconds). Expected program time tPGM E and the actual program time tPGM R The difference between the word line strength STR (or aging value) and the expected program time tPGM E and the actual program time tPGM R are equal, but the word line strength STR is "0".
[0116] Program, read, and erase operations are performed on the fourth word line WL4 and the twenty-third word line WL23. The fourth word line WL4 and the twenty-third word line WL23 are programmed based on the same program parameters. The fourth word line WL4 and the twenty-third word line WL23 age as the program, read, and erase operations are performed. Thus, the actual program time tPGM of the fourth word line WL4 R is reduced to 4 ms, and the actual program time tPGM R The expected program time tPGM is reduced to 3 ms. E and the actual program time tPGM R Due to the difference between these, the strength STR of the fourth word line WL4 is calculated as "-1", and the strength STR of the twenty-third word line WL23 is calculated as "-2".
[0117] Next, adjustment of the number of states and adjustment of program parameters of the fourth word line WL4 and the twenty-third word line WL23 will be described with reference to FIGS. 11A and 11B.
[0118] Figure 11A is a graph showing an example of a threshold voltage distribution of a word line when the number of states of the word line is adjusted according to an embodiment of the present invention, and Figure 11B is a table showing an example of program parameters and state indexes of each word line of Figure 11A. Figure 11B shows an example of the change in state index and adjustment of program parameters as the number of states of the word line is adjusted.
[0119] 11A, the number of states of the fourth word line WL4 and the 23rd word line WL23 are adjusted. In FIG. 9B, since the strength STR of the fourth word line WL4 is relatively higher than the strength of the 23rd word line WL23, the number of states of the fourth word line WL4 is increased from 8 to 16, and the number of states of the 23rd word line WL23 is decreased from 8 to 4, as shown. The memory cells of the fourth word line WL4 are programmed to the erase state E and one of the first program state P1 through the 15th program state P15, and the memory cells of the 23rd word line WL23 are programmed to the erase state E and one of the first program state P1 through the third program state P3.
[0120] As described with reference to FIG. 9A, before the number of states is adjusted, 3 bits of data are stored in each memory cell of the fourth word line WL4 and the 23rd word line WL4, and 6 bits of data are stored in two memory cells of the fourth word line WL4 and the 23rd word line WL23. After the number of states is adjusted, 4 bits of data are stored in the memory cell of the fourth word line WL4, and 2 bits of data are stored in the memory cell of the 23rd word line WL23. Therefore, 6 bits of data are stored in two memory cells of the fourth word line WL4 and the 23rd word line WL23. The amount of data stored in the fourth word line WL4 and the 23rd word line WL4 does not change before and after the number of states is adjusted.
[0121] Referring to FIG. 11B, the expected program time tPGM of the fourth word line WL4 is E is extended to 7 ms, and the expected program time tPGM E The actual program time tPGM of the fourth word line WL4 is reduced to 4 ms because the strength STR is the same before and after the state number adjustment. R is 6 ms, and the actual program time tPGM R The actual program time tPGM between the fourth word line WL4 and the twenty-third word line WL23 is 2 ms. R In order to reduce the difference, the program parameters of the fourth word line WL4 and / or the twenty-third word line WL23 are adjusted.
[0122] For example, as shown in FIG. 11B, the program start voltage V ST is increased from 14.5V to 15V, and the program voltage difference ΔV ISPP is increased from 0.5V to 0.6V, and the program verify voltage V VF The expected program time tPGM of the fourth word line WL4 is reduced by 0.5V from the default voltage DF. E The actual program time tPGM is reduced to 5 ms. RThe programming start voltage V ST is reduced from 14.5V to 14V, and the program voltage difference ΔV ISPP is reduced from 0.5V to 0.4V, and the program verify voltage VVF is increased by 0.5V from the default voltage DF. As a result, the expected program time tPGM E The actual program time tPGM is extended to 6 ms. R The actual program time tPGM of the fourth word line WL4 and the 23rd word line WL23 is extended to 4 ms. R will be identical or similar.
[0123] In FIG. 11B, the actual program time tPGM of the fourth word line WL4 R is shortened, and the actual program time tPGM R In the above embodiment, the program parameters are adjusted to extend the actual program time tPGM of the fourth word line WL4. R In another embodiment, the program parameters of the fourth word line WL4 are adjusted to reduce the actual program time tPGM R The program parameters are adjusted so that the
[0124] Meanwhile, referring to FIGS. 11A and 11B, when the number of states of the fourth word line WL4 and the twenty-third word line WL23 is two, sIn the above embodiment, the number of states of the fourth word line WL4 and the 23rd word line WL23 is adjusted to s (where s is a natural number) so that integer bit data is stored in the memory cells of the fourth word line WL4 and the 23rd word line WL23. However, the present invention is not limited thereto, and the number of states of the fourth word line WL4 and the 23rd word line WL23 is adjusted so that fractional bit data is stored in the memory cells of the fourth word line WL4 and the 23rd word line WL23. This will be described with reference to FIGS. 12A to 14B.
[0125] Figure 12A is a graph showing an example of a threshold voltage distribution of a word line when the number of states of the word line is adjusted according to an embodiment of the present invention, and Figure 12B is a table showing an example of program parameters and state indexes of each word line of Figure 12A. Figure 12B shows an example of a change in state index and program parameter adjustment due to the adjustment of the number of states of the word line.
[0126] Referring to FIG. 12A, the number of states of the fourth word line WL4 and the 23rd word line WL23 is adjusted. In FIG. 9B, since the strength STR of the fourth word line WL4 is relatively higher than the strength of the 23rd word line WL23, the number of states of the fourth word line WL4 is increased from 8 to 12, and the number of states of the 23rd word line WL23 is decreased from 8 to 6, as shown. The memory cell of the fourth word line WL4 is programmed to the erase state E and one of the first program state P1 to the 11th program state P11, and the memory cell of the 23rd word line WL23 is programmed to the erase state E and one of the first program state P1 to the 5th program state P5. Since 3.5 bits of data are stored in the memory cell of the fourth word line WL4 and 2.5 bits of data are stored in the 23rd word line WL23, a total of 6 bits of data are stored in the two memory cells of the fourth word line WL4 and the 23rd word line WL23.
[0127] As described with reference to FIG 9A, before the number of states is adjusted, 6-bit data is stored in two memory cells of the fourth word line WL4 and the twenty-third word line WL23. Therefore, the amount of data stored in the fourth word line WL4 and the twenty-third word line WL4 does not change before and after the number of states is adjusted. The method of storing fractional bits in memory cells of word lines will be described below with reference to FIGS. 13 to 14B.
[0128] Referring to FIG. 12B, the expected program time tPGM of the fourth word line WL4 is E is extended to 6 ms, and the expected program time tPGM E The actual program time tPGM of the fourth word line WL4 is reduced to 4.5 ms because the strength STR is the same before and after the state number adjustment. R is 5 ms, and the actual program time tPGM R The actual program time tPGM between the fourth word line WL4 and the twenty-third word line WL23 is 2.5 ms. R In order to reduce the difference, the program parameters of the fourth word line WL4 and / or the twenty-third word line WL23 are adjusted.
[0129] For example, as shown in FIG. 12B, the program start voltage VST of the fourth word line WL4 is increased from 14.5V to 14.75V, and the program voltage difference ΔV ISPP is increased from 0.5V to 0.55V, and the program verify voltage V VF is lowered by 0.25V from the default voltage DF. As a result, the expected program time tPGM of the fourth word line WL4 is E The actual program time tPGM is reduced to 5 ms. R In addition, the program start voltage VST of the 23rd word line WL23 is reduced from 14.5V to 14.25V, and the program voltage difference ΔV ISPP is reduced from 0.5V to 0.45V, and the program verify voltage V VFis increased by 0.25V from the default voltage DF. As a result, the expected program time tPGM of the 23rd word line WL23 is E The actual program time tPGM is extended to 6 ms. R The actual program time tPGM of the fourth word line WL4 and the 23rd word line WL23 is extended to 4 ms. R will be identical or similar.
[0130] FIG. 13 is a table illustrating one example of how fractional bits per memory cell are stored according to one embodiment of the present invention.
[0131] FIG. 13 illustrates an embodiment in which a total of 5 bits of data are stored in the first memory cell MC1 and the second memory cell MC2, with 2.5 bits of data being stored per memory cell.
[0132] Referring to FIG. 13, a program operation and a read operation are performed on a first memory cell MC1 and a second memory cell MC2 as a pair (or as a group). The first memory cell MC1 and the second memory cell MC2 are assigned six states, for example, an erase state E and a first program state P1 to a fifth program state P5, to each of the first memory cell MC1 and the second memory cell MC2. The combination of these states corresponds to 32 states (2 5 A total of 36 states are derived by multiplying the 6 states of the first memory cell MC1 and the 6 states of the second memory cell MC2, and the 32 states excluding 4 of these, for example, the first state S0 to the 32nd state S31, correspond to 32 states corresponding to 5 bits of data, for example, the erase state E and the first program state P1 to the 31st program state P31.
[0133] Therefore, the encoding and decoding module 113 of the memory controller 110 (FIG. 5) treats the first memory cell MC1 and the second memory cell MC2 as a pair and performs encoding and decoding. In the non-volatile memory 120, the first memory cell MC1 and the second memory cell MC2 are simultaneously programmed and read, so that 2.5 bits of data are stored and read in each of the first memory cell MC1 and the second memory cell MC2.
[0134] 13 shows an embodiment in which 5 bits of data are stored in two memory cells, thereby storing fractional bit data of 2.5 bits per memory cell. However, the present invention is not limited thereto, and fractional bit data is stored per memory cell by performing program and read operations on groups of three or more memory cells. In this case, fractional bit data is stored, but in order for fractional bit data to be stored per memory cell, Equation 1 must be satisfied.
[0135]
number
[0136] Here, n is a natural number equal to or greater than 2 and indicates the number of memory cells to be simultaneously programmed and read, Sk indicates the number of states assigned to the kth memory cell (k is a natural number from 1 to n), and B indicates the total number of bits stored in the memory cells. For example, if a total of 5 bits of data are stored in two memory cells, the product of the numbers of states (S1, S2) assigned to each of the two memory cells must be 25, i.e., 32 or more. As shown in FIG. 13, the numbers of states (S1 and S2) assigned to the first memory cell MC1 and the second memory cell MC2 are both 6, so according to Equation 1, TIFF0007594415000002.tif16128 is 36, which is greater than or equal to 32. Therefore, a total of 5 bits of data are stored in the first memory cell MC1 and the second memory cell MC2. For example, even if 5 states and 8 states, or 6 states and 7 states are assigned to the first memory cell MC1 and the second memory cell MC2, respectively, Equation 1 is satisfied, and therefore a total of 5 bits of data are stored in the first memory cell MC1 and the second memory cell MC2.
[0137] 14A and 14B are diagrams illustrating an example of how fractional bits per memory cell may be stored according to one embodiment of the present invention.
[0138] As described with reference to FIG. 13, in order to store fractional bits of data per memory cell, multiple memory cells must operate as pairs or groups and program and read operations must be performed on multiple memory cells simultaneously.
[0139] FIG. 14A illustrates an embodiment in which two memory cells on one word line operate as a pair, and FIG. 14B illustrates an embodiment in which two memory cells on two word lines operate as a pair.
[0140] Referring to FIG. 14A, the fourth word line WL4 includes a plurality of memory cells (C10 to C1m) (m is a positive even number equal to or greater than 4), and the twenty-third word line WL23 includes a plurality of memory cells (C20 to C2m). In the plurality of memory cells (C10 to C1m) of the fourth word line WL4, two memory cells (e.g., C10 and C11, C12 and C13, etc.) operate as a pair, and 7 bits of data are stored in each pair of two memory cells. That is, 7 bits of data are stored in the pair of C10 and C11, and 7 bits of data are stored in the pair of C12 and C13. For example, when 12 states are assigned to the fourth word line WL4, the product of the number of states assigned to C10 and C11 is 144. 144 is multiplied by 128 (=2 7 ) and 256 (=2 8) Therefore, 7 bits of data are stored in the pair of C10 and C11, resulting in 3.5 bits of data being stored in each of C10 and C11.
[0141] Among the memory cells (C20 to C2m) of the 23rd word line WL23, two memory cells (e.g., C20 and C21, C22 and C23, etc.) operate as a pair, and 5 bits of data are stored per pair of two memory cells. For example, if 6 states are assigned to the 23rd word line WL23, the product of the number of states assigned to C20 and C21 is 36. 36 is 32 (=2 5 ) and 64 (=2 6 ) Therefore, 5 bits of data are stored in the pair of C20 and C21, resulting in 2.5 bits of data being stored in each of C20 and C21.
[0142] In this way, 3.5 bits of data are stored in the memory cell (e.g., C10) of the fourth word line WL4, and 2.5 bits of data are stored in the memory cell (e.g., C20) of the 23rd word line WL23, so that a total of 6 bits of data are stored in the two memory cells (e.g., C10 and C20) of the fourth word line WL4 and the 23rd word line WL23. Therefore, even if the number of states of the fourth word line WL4 is adjusted from 8 to 12 and the number of states of the 23rd word line WL23 is adjusted from 8 to 6, the amount of data stored in the fourth word line WL4 and the 23rd word line WL23 is the same before and after the adjustment of the number of states.
[0143] FIG. 14A illustrates an embodiment in which two memory cells on one word line operate as a pair, and FIG. 14B illustrates an embodiment in which two memory cells on two word lines operate as a pair.
[0144] Referring to FIG. 14B, one memory cell each in the fourth word line WL4 and the 23rd word line WL23, i.e., two memory cells, operate as a pair. For example, C10 in the fourth word line WL4 and C20 in the 23rd word line WL23 operate as a pair. That is, C10 in the fourth word line WL4 and C20 in the 23rd word line WL23 are programmed and read simultaneously. For example, if 12 state numbers are assigned to the fourth word line WL4 and 6 state numbers are assigned to the 23rd word line WL23, the product of the state numbers assigned to C10 and C20 is 72 (=12×6). 72 is multiplied by 64 (=2 6 ) and 128 (=2 7 ), a total of 6 bits of data are stored in C10 and C20, resulting in 3.5 bits and 2.5 bits being stored in C10 and C20, respectively. As another example, if 11 state numbers are assigned to the fourth word line WL4 and 6 state numbers are assigned to the 23rd word line WL23, the product of the state numbers assigned to C10 and C20 is 66 (=11×6), which can be multiplied by 64 (=2 6 ) and 128 (=2 7 ), so a total of 6 bits of data are stored in C10 and C20.
[0145] Therefore, even if the number of states of the fourth word line WL4 is adjusted from 8 to 12 or 11, and the number of states of the 23rd word line WL23 is adjusted from 8 to 6, the amount of data stored in the fourth word line WL4 and the 23rd word line WL23 is the same before and after the adjustment of the number of states.
[0146] As described with reference to FIGS. 13 to 14B, by adjusting the number of states, two s A number of states other than 2 are assigned to the word line. s In this case, the number of states can be adjusted more precisely than when the number of states is allocated, and wear leveling can be more easily performed for each word line.
[0147] FIG. 15A is a block diagram that generally illustrates a memory controller in accordance with one embodiment of the present invention, and FIG. 15B is a flow chart illustrating a method of operation of the memory controller of FIG. 15A.
[0148] Referring to FIG. 15A, the memory controller 110a includes a reinforcement learning module RLM, a profiling information storage unit PIS, and a program parameter control module PCM.
[0149] The reinforcement learning module RLM adjusts the number of states of each of a plurality of word lines through reinforcement learning. The reinforcement learning module RLM adjusts the number of states by using the word line strength as a feature. As described with reference to FIG. 3B, the word line strength is calculated by calculating an expected program time based on at least one of program parameters, such as a program start voltage, a program voltage difference, and a program verify voltage, and calculating the strength of the corresponding word line based on the expected program time and the actual program time. In addition, in the process of calculating the word line strength, related features, such as a read level shifting value and a read error bit, are considered in consideration of the situation in which the read is performed.
[0150] The reinforcement learning module RLM is embodied, for example, by firmware, stored in the ROM 114 (FIG. 5), loaded into the RAM 112 at runtime of the storage device 100 (FIG. 1), and executed by the processor 111.
[0151] The profiling information storage unit PIS stores pre-profiled characteristic information (e.g., configuration information, initial value information for the number of states for each of a plurality of word lines, threshold value information for the number of states to be assigned to a word line, etc.). For example, the RAM 112 of the memory controller 110 (FIG. 5) operates as the profiling information storage unit PIS. For example, the profiled characteristic information is stored in the non-volatile memory 120 (FIG. 1) or the ROM 114 (FIG. 5) and loaded into the RAM 112 when the storage device 100 (FIG. 1) is started up.
[0152] When the program time deviation between the word lines increases due to the adjustment of the number of states, the program parameter control module PCM adjusts the program parameters so that the program time deviation is reduced.
[0153] 15B, the memory controller 110a acquires state indicators for each of a plurality of word lines (step S210). The memory controller 110a stores the actual program time, erase time, read time, read error bit, read level shift value, number of states, and the like.
[0154] The reinforcement learning module RLM adjusts the number of states of each of the word lines based on the state indicators and program parameters acquired in step S210 (step S220). At this time, the reinforcement learning module RLM performs reinforcement learning based on the profiling information PI to prevent reinforcement learning from learning in an unpredictable direction, for example, in a direction in which the intensity difference between the word lines increases. For example, the reinforcement learning module RLM performs reinforcement learning based on the initial number of states for each of the word lines. As another example, the reinforcement learning module RLM performs reinforcement learning on threshold value information on the number of states assigned to a word line, for example, the maximum number of states assigned to one word line or a threshold value of the difference in the number of states of the word lines. This prevents reinforcement learning from learning under the worst conditions.
[0155] The program parameter control module PCM adjusts the program parameters of each of the word lines by adjusting the number of states (step S230), and then the memory controller 110 transmits the number of states and the program parameters of each of the word lines to the non-volatile memory 120 (step S240).
[0156] FIG. 16 illustrates a method of operation of a reinforcement learning module according to one embodiment of the present invention.
[0157] 16, the reinforcement learning module RLM sets a reinforcement learning state based on the status information of each word line of the non-volatile memory 120. For example, the program parameter PP, the actual program time tPGM E The features such as the number of states SC, and the erase time tBERS, the read level shift value RLS, and the read error bit REB are set in the reinforcement learning state. E Based on this, the strength STR (or degree of aging) of the word line is determined.
[0158] The reinforcement learning module RLM adjusts (increases or decreases) the number of states for each word line, i.e., adjusts the number of states for each of the plurality of word lines, as an action of reinforcement learning. The reinforcement learning module RLM takes an action to maximize a reward based on the state of reinforcement learning. As a result, the number of states for each of the plurality of word lines increases or decreases. As described with reference to FIG. 4A and FIG. 4B, the program time difference ΔtPGM or P / E cycles of the plurality of word lines is set as a trigger condition for performing the action.
[0159] A reward for reinforcement learning is set based on the difference dSTR in strength between multiple word lines calculated based on the features used when setting the state of reinforcement learning. For example, a function value F(dSTR) calculated based on the difference dSTR in strength is set as the reward for reinforcement learning. The reinforcement learning module RLM takes an action to maximize the function value F(dSTR), that is, to minimize the difference dSTR in strength between multiple word lines.
[0160] In one embodiment, when the program time difference ΔtPGM is set as a trigger condition for performing an action, the time difference dT between actions, i.e., the time until the number of states is further adjusted after the number of states is adjusted, is set as the reward of reinforcement learning. If the number of states is adjusted appropriately, the time difference dT between actions is increased.
[0161] FIG. 17A is a block diagram that generally illustrates a memory controller in accordance with one embodiment of the present invention, and FIG. 17B is a flow chart illustrating a method of operation of the memory controller of FIG. 17A.
[0162] Figure 17A is a modified embodiment of the method of Figure 3A, and the method of Figure 17B is a modified embodiment of the method of Figure 3B, so differences will be mainly described.
[0163] Referring to FIG. 17A, the memory controller 110b includes a strength calculation module SCM, a clustering module CLM, a state number adjustment module CAM, and a program parameter control module PCM.
[0164] The intensity calculation module SCM, the state number adjustment module CAM, and the program parameter control module PCM have been described with reference to FIG. 3A, so a duplicated description will be omitted.
[0165] The clustering module CLM groups the word lines into a number of word line groups by clustering the word lines based on a characteristic, such as a word line strength, the number of states adjustment module CAM adjusts the number of states for each of the word line groups, and the program parameter control module PCM adjusts program parameters for each of the word line groups.
[0166] 17B, the memory controller 110 acquires a status indicator for each of the word lines (step S310). The status indicator may include, for example, an actual program time, an erase time, a read time, a read error bit, a read level shift value, a number of states, etc.
[0167] The strength calculation module SCM calculates a plurality of word line strengths respectively corresponding to the plurality of word lines (operation S320). The strength calculation module SCM calculates the strengths of the word lines based on the program parameters and the status indicators collected in operation S310.
[0168] The clustering module CLM determines whether grouping of the word lines is necessary (step S330), and if grouping is necessary, performs clustering on the word lines based on a characteristic, for example, a word line strength (step S340). The word lines are grouped into a plurality of word line groups.
[0169] For example, if the number of word lines managed for adjusting the number of states is greater than a predetermined value, the load on the state number adjustment module CAM increases, and the amount of information to be managed, for example, the amount of data for acquiring state indicators, increases. If the number of word lines is greater than a predetermined value, the clustering module CLM determines that grouping is necessary and performs clustering according to the characteristics of the word lines, i.e., grouping based on a clustering technique.
[0170] The clustering is performed using at least one of a K-means clustering algorithm, a K-medoids clustering algorithm, a hierarchical clustering algorithm, a density-based clustering algorithm, and a neural network algorithm, for example.
[0171] The state number adjustment module CAM adjusts the number of states (step S350). The state number adjustment module CAM adjusts the number of states for each of the word line groups based on a plurality of word line strengths corresponding to each of the word line groups. If the clustering of step S340 is not performed, the state number adjustment module CAM adjusts the number of states for each of the word lines.
[0172] The program parameter control module PCM adjusts program parameters for each of the word line groups or each of the word lines (step S360).
[0173] The memory controller 110 transmits the number of states and program parameters for each of the word lines or each of the word line groups to the non-volatile memory 120 (step S370).
[0174] 18A and 18B are diagrams illustrating an example of grouping of word lines according to an embodiment of the present invention.
[0175] Referring to FIG. 18A, state adjustment is performed for each memory block BLK, and a plurality of word lines included in the memory block BLK, for example, the first word line WL1 to the twelfth word line WL12, are grouped into a plurality of word line groups.
[0176] A plurality of word lines, for example, the first word line WL1 to the twelfth word line WL12, each have a different strength STR, and a clustering module CLM (FIG. 17A) of the memory controller 110b (FIG. 17A) performs clustering based on the strength STR to group the first word line WL1 to the twelfth word line WL12 into a plurality of word line groups, for example, the first group G1, the second group G2, and the third group G3. For example, the clustering module CLM sets a clustering criterion based on the strength STR distribution of the first word line WL1 to the twelfth word line WL12. For example, the clustering module CLM groups word lines whose strength STR is 0 or less and greater than -2 into the first group G1, groups word lines whose strength STR is -2 or less and greater than -3 into the second group G2, and groups word lines whose strength STR is -3 or less into the third group G2. As a result, the third word line WL3, the fourth word line WL4, the fifth word line WL5, and the twelfth word line WL12 are grouped into a first group G1, the second word line WL2, the sixth word line WL6, the tenth word line WL10, and the eleventh word line WL11 are grouped into a second group G2, and the first word line WL1, the seventh word line WL7, the eighth word line WL8, and the ninth word line WL9 are grouped into a third group G3.
[0177] Referring to FIG. 18B, the number of states is adjusted for each of a plurality of memory blocks (BLK1, BLK2), and a plurality of word lines provided in the plurality of memory blocks (BLK1, BLK2), for example, the first word line WL1 to the eighth word line WL8 of the first memory block BLK1 and the first word line WL1 to the eighth word line WL8 of the second memory block BLK2, are grouped into a plurality of word line groups.
[0178] The first word line WL1 to the eighth word line WL8 in each of the first memory block BLK1 and the second memory block BLK2 each have a different strength STR, and the clustering module CLM (FIG. 17A) of the memory controller 110b (FIG. 17A) performs clustering based on the strength STR to group the first word line WL1 to the eighth word line WL8 of the first memory block BLK1 and the second memory block BLK2 into a plurality of word line groups, for example, a first group G1, a second group G2, and a third group G3.
[0179] For example, the clustering module CLM sets a clustering criterion based on the strength STR distribution of the first word line WL1 to the eighth word line WL8 of the first memory block BLK1 and the second memory block BLK2. For example, the clustering module CLM groups word lines whose strength STR is equal to or less than 0 and greater than -2 into a first group G1, groups word lines whose strength STR is equal to or less than -2 and greater than -3 into a second group G2, and groups word lines whose strength STR is equal to or less than -3 into a third group G2.
[0180] As a result, the fifth word line WL5 of the first memory block BLK1 and the fourth word line WL4 to the sixth word line WL6 of the second memory block BLK2 are grouped into a first group G1, the third word line WL3 and the fourth word line WL4 of the first memory block BLK1 and the second word line WL2, the third word line WL3, and the seventh word line WL7 of the second memory block BLK2 are grouped into a second group G2, and the first word line WL1, the second word line WL2, the seventh word line WL7, and the eighth word line WL8 of the first memory block BLK1 and the first word line WL1 and the eighth word line WL8 of the second memory block BLK2 are grouped into a third group G3.
[0181] For the first word line group G1 to the third word line group G3 in FIGS. 18A and 18B, the number of states is adjusted based on the intensity representative of each word line.
[0182] For example, an average strength for each word line group may be calculated as a strength representative of the word line group, and the number of states may be adjusted based on the average strength. As another example, a minimum strength for each word line group may be calculated as a strength representative of the word line group, and the number of states may be adjusted based on the average strength. However, this is merely one embodiment, and the number of states for multiple word line groups may be adjusted in various ways.
[0183] Meanwhile, Fig. 18A shows an embodiment in which the number of states is adjusted in units of memory blocks, and Fig. 18B shows an embodiment in which the number of states is adjusted in units of multiple memory blocks. However, the present invention is not limited thereto, and the number of states may be adjusted in units of chips or wafers. For example, when the number of states is adjusted in units of wafers, the number of states is set for each word line or each word line group during the manufacturing stage of the storage device, and the set number of states is stored in a non-volatile storage location, for example, a ROM, of the non-volatile memory device 120 (Fig. 1) or the memory controller 110 (Fig. 1) and is used at runtime of the storage device 100.
[0184] FIG. 19 is a block diagram illustrating an SSD system according to one embodiment of the present invention.
[0185] 19, the SD system 1000 includes a host 1100 and an SSD 1200. The SSD 1200 communicates with the host 1100 to exchange signals, such as commands, addresses, and data, with the host 1100. The SSD 1200 receives power from the host 1100 to operate.
[0186] The SSD 1200 includes an SSD controller 1210, a plurality of non-volatile memory devices (1220, 1230, 1240), and a buffer memory 1300. The plurality of non-volatile memory devices (1220, 1230, 1240) are realized by a plurality of semiconductor chips. The SSD controller 1210 communicates with the memory devices (1220, 1230, 1240) through a plurality of channels (CH1, CH2, CHn). The buffer memory 1300 is realized by a volatile memory or a resistive memory, for example, the buffer memory 1300 is a DRAM. The buffer memory 1300 temporarily stores data received from the host 1100 and to be stored in the plurality of non-volatile memory devices (1220, 1230, 1240), or temporarily stores data read from the plurality of non-volatile memory devices (1220, 1230, 1240) and to be transmitted to the host 1100.
[0187] The storage device 100 and the memory controller 110 described with reference to Figs. 1 to 18B are applied to the SSD 1200 and the SSD controller 1210, respectively. The SSD controller 1210 adjusts the number of states for each of a plurality of word lines or a plurality of word line groups for a plurality of non-volatile memory devices (1220, 1230, 1240) in units of memory blocks, a plurality of memory blocks, or a semiconductor chip, and adjusts program parameters for each of the plurality of word lines or the plurality of word line groups. This improves the life span and data reliability of the SSD 1200.
[0188] The storage system according to the above-described embodiment of the present invention may be installed or applied not only to the SSD system 1000 but also to a memory card system, a universal flash storage (UFS), an embedded storage, and the like.
[0189] According to one embodiment, the operations described herein as being performed by the memory controller 110 are performed by at least one processor executing program code including instructions corresponding to the operations. The instructions are stored in a memory. As used herein, the term "processor" refers to a hardware-implemented data processing device including circuitry physically structured to perform certain operations, including, for example, operations expressed by instructions and / or code included in a program. In at least one embodiment, hardware-implemented data processing devices include, but are not limited to, microprocessors, central processing units (CPUs), processor cores, multiprocessor cores, multiprocessors, application specific integrated circuits (ASICs), and field programmable gate arrays (FPGAs).
[0190] Although the embodiments of the present invention have been described in detail above with reference to the drawings, the present invention is not limited to the above-described embodiments and can be modified in various ways without departing from the technical spirit of the present invention. [Explanation of symbols]
[0191] 100 Storage 110, 110a, 110b memory controller 111 Processor 112 RAM 113 Encoding / Decoding Module 114 ROM 115 Host Interface 116 Memory Interface 117 Bus 120 Non-Volatile Memory (NVM) 121, 121a, 121b memory cell array 122 Control Logic 123 Voltage Generation Unit 124 Address Decoder 125 Page Buffer Circuit 126 Data Input / Output Circuit 1000 SD System 1100 Host 1200 SSD 1210 SSD Controller 1220, 1230, 1240 FLASH1~FLASHn 1300 Buffer Memory BL1~BL3 Bit lines BLK Memory block BLK1~BLK3 Memory blocks 1~3 CAM Status Quantity Adjustment Module CLM Clustering Module CS charge storage layer CSL Common Source Line GE gate electrode GSL Ground Selection Line GST Ground Select Transistor I inside IL insulating film MC1~MC8 memory cells NS11~NS13, NS21~NS23, NS31~NS33 NAND strings O Outer part P Pira PCM Program Parameter Control Module PIS Profiling Information Storage Unit RLM Reinforcement Learning Module SCM strength calculation module SSL, SSL1~SSL3 string selection line SST String Select Transistor SUB board WL1~WL8 Word Lines
Claims
1. a non-volatile memory including a plurality of memory blocks each including a plurality of word lines; a memory controller that determines a word line strength for each of the plurality of word lines, adjusts a number of states for each of the plurality of word lines based on the word line strength, and adjusts program parameters for each of the plurality of word lines so that program time deviation between the plurality of word lines is reduced.
2. 2. The storage device of claim 1, wherein the memory controller adjusts the number of states for each of the plurality of word lines so that a number of states assigned to a first word line having a first word line strength is greater than a number of states assigned to a second word line having a second word line strength weaker than the first word line strength.
3. 2. The storage device of claim 1, wherein the memory controller calculates the word line strength based on an actual program time and an expected program time when a program is executed on a word line.
4. 2. The storage device of claim 1, wherein the memory controller adjusts the number of states of each of the word lines when a difference between actual program times of at least two of the plurality of word lines is equal to or greater than a reference time.
5. The memory controller determines the number of states for each of the word lines by dividing the number of states by integer bits of data. s 2. The storage device according to claim 1, wherein the number of bits is adjusted to one (s is a natural number).
6. 2. The storage device of claim 1, wherein the memory controller adjusts the number of states for the plurality of word lines in units of N word lines, and adjusts the number of states for each of the N word lines to a number of states according to fractional bit data.
7. The memory controller includes: a memory controller for storing two memory cells for every two memory cells included in one of the N word lines; s 7. The storage device according to claim 6, wherein the number of states is adjusted so that the number of states is set.
8. The memory controller includes two memory cells for each of the N memory cells corresponding to the N word lines. s 7. The storage device according to claim 6, wherein the number of states is adjusted so that the number of states is set.
9. 2. The storage device of claim 1, wherein the memory controller dynamically groups the word lines into a plurality of word line groups based on the word line strength, and collectively adjusts the number of states of each of the word lines included in each of the plurality of word line groups by word line group.
10. 2. The storage device of claim 1, wherein the memory controller acquires program parameters and state indicators for each of the plurality of word lines, and adjusts the number of states through reinforcement learning based on the program parameters and the state indicators.
11. The storage device according to claim 10 , wherein the memory controller performs the reinforcement learning based on pre-stored profiling information.
12. 2. The storage device of claim 1, wherein the memory controller stores a number of states and a number of program parameters corresponding to each of the plurality of word lines, and controls the execution of a program on the non-volatile memory based on the number of states and the program parameters adjusted for each of the plurality of word lines at runtime of the storage device.
13. 2. The storage device of claim 1, wherein the memory controller dynamically adjusts the number of states and the program parameters for each of the word lines at run-time of the storage device.
14. 2. The storage device of claim 1, wherein the non-volatile memory comprises an array in which a plurality of non-volatile memory cells are vertically stacked on a substrate.
15. A method of operating a storage device including a non-volatile memory, comprising: calculating a word line strength for each of a plurality of word lines of the non-volatile memory; adjusting a number of states for each of the plurality of word lines based on the word line strength of each of the plurality of word lines; and adjusting program parameters of each of the word lines such that program time deviations between the word lines are reduced.
16. The step of calculating the word line strength comprises: calculating an expected program time for the word line based on preset program parameters and a number of states; 16. The method of claim 15, further comprising: calculating the word line strength based on an expected program time and a measured actual program time of the word line.
17. 16. The method of claim 15, wherein the adjusting of the number of states is performed when a maximum value of a program time difference between the word lines is equal to or greater than a threshold time.
18. 16. The method of claim 15, wherein the adjusting step comprises determining the number of first states and the number of second states for two of the plurality of word lines such that a product of a first state number and a second state number set for each of the two word lines is equal to or greater than a product of a first previous state number and a second previous state number previously set for each of the two word lines.
19. 16. The method of claim 15, wherein the steps of calculating the word line strength and adjusting the number of states are performed via reinforcement learning.
20. A method of operating a storage device including a non-volatile memory, comprising: calculating a word line strength for each of a plurality of word lines included in the non-volatile memory; grouping the word lines into a plurality of word line groups based on the word line strengths; adjusting the number of states of each of the plurality of word line groups for each of the word line groups; adjusting program parameters of each of the word line groups such that a program time deviation between the word line groups is reduced for each of the word line groups.
Citation Information
Patent Citations
Non-volatile memory with dynamic multi-mode operation
JP2010518546A
Memory system
JP2012203692A
Error correction coding device and method of flash memory
JP2015138498A
Storage device and operating method thereof
JP2019079520A
