Method for equalizing bit error rate of memory device

By implementing ECC interleaving in non-volatile storage devices, the bit error rate is balanced, the problem of excessive burden on error correction circuits is solved, and the reliability and stability of data storage are improved.

CN110580926BActive Publication Date: 2024-11-19SAMSUNG ELECTRONICS CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN201910423205.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2019-03-20
Filing Date
2019-05-21
Publication Date
2024-11-19
Estimated Expiration
2039-05-21

AI Technical Summary

Technical Problem

In existing non-volatile storage devices, error correction circuits are overburdened when faced with unbalanced bit error rates, resulting in limited bit error correction capabilities. Furthermore, they cannot effectively correct excessive bit errors, affecting the reliability of data storage.

Method used

By implementing ECC interleaving in storage devices, information data is divided into storage areas with different bit error rates. Through interleaving at the levels of word lines, bit lines, stacking layers, and chips, the bit error rate is balanced, reducing the burden on error correction circuits.

Benefits of technology

It effectively reduces the imbalance of bit error rate, improves the correction capability of error correction circuit, and enhances the reliability and stability of data storage.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN110580926B_ABST
    Figure CN110580926B_ABST
Patent Text Reader

Abstract

A method for equalizing bit error rates of a storage device is provided. When a codeword including information data and parity bits of the information data is written into a memory cell array, the storage device selectively performs an error correction code (ECC) interleaving operation according to the resistance distribution characteristics of the memory cells. In an ECC interleaving operation according to an example, an ECC sector including information data is divided into a first ECC sub-sector and a second ECC sub-sector, the first ECC sub-sector is written into a memory cell of a first storage area having a high bit error rate (BER), and the second ECC sub-sector is written into a memory cell of a second storage area having a low BER.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of Korean Patent Application No. 10-2018-0065658, filed in the Korean Intellectual Property Office on June 7, 2018, and U.S. Patent Application No. 16 / 358,884, filed in the U.S. Patent and Trademark Office on March 20, 2019, the disclosures of which are incorporated herein by reference in their entireties. Background Art

[0003] The inventive concept relates to a nonvolatile memory device, and more particularly, to a memory device capable of reducing a burden on an error correction circuit through bit error rate (BER) equalization, an operation method thereof, and a memory system including the memory device.

[0004] Non-volatile memory devices may include resistive memory such as phase change RAM (PRAM), resistive RAM (RRAM), and magnetoresistive RAM (MRAM). Resistive memory uses a variable resistor represented by a resistance state based on a variable resistor for storing data as a memory cell. The material used to form the resistive memory generally has a resistance value that changes according to the magnitude and / or direction of a current or voltage, and generally has a non-volatile property such that the resistance value remains constant even when the current or voltage is cut off.

[0005] However, the resistance value of the variable resistor varies depending on its position in the PRAM memory cell array, so the bit error rate (BER) varies depending on the physical position of the PRAM memory cell storing the data. During the read operation, the error correction code (ECC) circuit can detect the bit errors of the data read from the PRAM, and can correct the detected bit errors to obtain error-free data. However, the number of error bits that can be corrected by the ECC circuit is limited. Therefore, when the number of detected bit errors is within the correction capability, the resistive memory will rewrite the data with bit errors after the ECC circuit corrects the bit errors (and thus store data without bit errors) to avoid further bit errors in the data (which will exceed the correction capability of the ECC circuit). On the other hand, when the number of detected error bits exceeds the correction capability, the ECC circuit cannot correct the detected bit errors. Because the data with a large number of bit errors will be rewritten in the memory after the ECC circuit performs such bit error correction, the BER imbalance in the memory may cause a large burden on the data retained in the memory. Summary of the invention

[0006] The present invention is conceived to provide a memory device capable of reducing the burden of an error correction circuit by bit error rate equalization, an operation method thereof, and a memory system including the memory device. In some examples, when a codeword including information data and a parity bit of the information data is written into a memory cell array, the memory device selectively performs an error correction code (ECC) interleaving operation according to the resistance distribution characteristics of the memory cell. In an ECC interleaving operation according to an example, an ECC sector including information data is divided into a first ECC sub-sector and a second ECC sub-sector, the first ECC sub-sector is written into a memory cell of a first storage area with a high bit error rate (BER), and the second ECC sub-sector is written into a memory cell of a second storage area with a low BER.

[0007] In some examples, a memory device includes a memory cell array including a plurality of memory tiles, each of which includes a plurality of word lines and a plurality of bit lines; and a control circuit configured to control a codeword including information data to be written into the memory cell array and a parity bit of the information data. The control circuit is also configured to divide an error correction code (ECC) sector including the information data into at least two ECC sub-sectors, and control one of the at least two ECC sub-sectors to be written into a first storage area with a high bit error rate, and another of the at least two ECC sub-sectors to be written into a second storage area with a low bit error rate, and the bit error rates of the information data written into the first storage area and the information data written into the second storage area are balanced.

[0008] In some examples, a memory system includes a plurality of memory chips; and a memory controller configured to generate parity bits by encoding information data to be written to the plurality of memory chips, generate a codeword including the information data and the parity bits, and provide the generated codeword to the plurality of memory chips. The memory controller is also configured to divide an error correction code (ECC) sector including the information data into at least two ECC sub-sectors, and control one of the at least two ECC sub-sectors to be written to a first storage area with a high bit error rate, and another of the at least two ECC sub-sectors to be written to a second storage area with a low bit error rate, and the bit error rates of the information data written to the first storage area and the information data written to the second storage area are balanced. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Embodiments of the present invention will be more clearly understood through the following detailed description in conjunction with the accompanying drawings, in which:

[0010] Figure 1 is a schematic diagram of a memory cell included in a resistive memory device according to an exemplary embodiment of the inventive concept;

[0011] Figure 2is used to describe the Figure 1 A graph of a write pulse of a memory cell;

[0012] Figure 3 Is used to describe Figure 1 A graph of characteristics of a memory cell;

[0013] Figure 4 is a block diagram of a memory system according to an embodiment of the inventive concept;

[0014] Figure 5A is shown in more detail including Figure 4 A block diagram of a memory device in a memory system, Figure 5B It is shown Figure 5A A block diagram of exemplary details of a memory device, and Figure 5C yes Figure 5B A top view of an exemplary layout of a memory device;

[0015] Figure 6 It is shown that the Figure 5A and / or Figure 5B A circuit diagram of exemplary details of a memory cell array in a memory device;

[0016] 7A to 7C It is shown Figure 6 A circuit diagram of a modified example of a memory cell of;

[0017] Figure 8 is shown in more detail including Figure 4 A block diagram of a memory controller in a memory system;

[0018] Fig.9A It is shown that it can be included in Figure 5A and Figure 5B A block diagram of exemplary details of a memory cell array in a memory device, and Fig. 9B and Fig. 9C shows an exemplary resistance distribution degradation of a tile group;

[0019] FIG. 10A to FIG. 10D It is used to explain Fig.9A A diagram of an exemplary resistance degradation distribution characteristic of memory cells in a memory slice group of FIG.

[0020] FIG. 11A to FIG. 11C is a view for explaining an error correction code (ECC) interleaving operation according to an embodiment of the inventive concept;

[0021] Fig.12 is a diagram for explaining an ECC interleaving operation according to an embodiment of the inventive concept;

[0022] Fig.13is a diagram for explaining an ECC interleaving operation according to an embodiment of the inventive concept;

[0023] FIG. 14A to FIG. 14C It is used to explain Fig.9A A view of a second example of a distribution characteristic of memory cells in a memory slice group of FIG.

[0024] FIG. 15A to FIG. 15D is a view for explaining an ECC interleaving operation according to an embodiment of the inventive concept;

[0025] Fig.16 is a diagram for explaining an ECC interleaving operation according to an embodiment of the inventive concept;

[0026] Fig.17A and Fig. 17B is a view for explaining the distribution characteristics of memory cells in a three-dimensional (3D) structure, wherein Figure 6 The memory chips are stacked vertically;

[0027] Fig.18 is a diagram for explaining an ECC interleaving operation according to an embodiment of the inventive concept;

[0028] Fig.19 is a schematic diagram of a memory module including a memory device according to an embodiment of the inventive concept;

[0029] Fig. 20 is a flowchart of an ECC interleaving operation of a memory device according to an embodiment of the inventive concept; and

[0030] Fig.21 is a block diagram of a system including a memory device according to an embodiment of the inventive concept. DETAILED DESCRIPTION

[0031] Figure 1 is a schematic diagram of a memory cell MC included in a resistive memory device according to an exemplary embodiment of the inventive concept.

[0032] Reference Figure 1 , the memory cell MC may include a variable resistor R and a switch SW. The switch SW may be implemented using various devices. For example, the switch SW may be a transistor, or the switch SW may be a diode. Figure 1 , the variable resistor R may include a phase change layer 1, an upper electrode 2 disposed on the phase change layer 1, and a lower electrode 3 disposed below the phase change layer 1. In one embodiment, the phase change layer 1 includes a germanium (Ge)-antimony (Sb)-tellurium (Te) (GST) material.

[0033] The GST material can be programmed between an amorphous state having a relatively high resistivity (which may be referred to herein as "high resistivity") and a crystalline state having a relatively low resistivity (which may be referred to herein as "low resistivity"). The GST material can be programmed by heating the GST material. The heating level and time may determine whether the GST material remains in the amorphous state or the crystalline state after the GST material cools. The high resistivity and the low resistivity may represent the programming values ​​of the logic "0" and the logic "1" of the memory cell MC, respectively, and may be sensed by measuring the resistivity of the GST material. Alternatively, the high resistivity and the low resistivity may represent the programming values ​​of the logic "1" and the logic "0" of the memory cell MC, respectively.

[0034] The pulse current I may be applied to the memory cell MC through the lower electrode 3. When the pulse current I flows through the memory cell MC for a very short time, the applied pulse current I may generate heat above a certain threshold only at a portion of the phase change layer adjacent to the lower electrode 3. In this case, a portion of the phase change layer 1 (e.g., Figure 1 The shaded area in the figure can be changed to a crystalline state (or set state) or an amorphous state (or reset state). Whether the crystalline state or the amorphous state is obtained may be due to the difference in the heating profile.

[0035] Figure 2 is to show that when different write pulses are applied to Figure 1 Graph of an exemplary heating profile obtained when the memory cell MC of FIG. 1 is a graph ... Figure 2 TP1 indicates the crystallization temperature of the phase change layer 1 , and TP2 indicates the melting point of the phase change layer 1 .

[0036] Figure 3 Is used to describe Figure 1 The graph of the characteristics of the memory cell MC shows the ideal distribution of a single-level cell programmed with one bit of data. Figure 3 , the horizontal axis indicates resistance, and the vertical axis indicates the number of memory cells.

[0037] As described above, the variable resistor R of the memory cell MC can be programmed to have a low resistance state LRS or a high resistance state HRS. The programming operation (e.g., as described with respect to Figure 2 The programming operation (described in detail) of switching the variable resistor R from the low resistance state LRS to the high resistance state HRS by applying a write pulse to the memory cell MC may be referred to as a set operation or a set write operation. Figure 2 The operation described above may be referred to as a reset operation or a reset write operation.

[0038] Any resistance between the distribution based on the low resistance state LRS and the distribution based on the high resistance state HRS can be set as the reference resistance RR. In a read operation from a specific memory cell MC, when the read result is equal to or greater than the reference resistance RR, reset data (logic "0") corresponding to the high resistance state HRS can be determined, and when the read result is less than the reference resistance RR, set data (logic "1") corresponding to the low resistance state LRS can be determined.

[0039] Figure 4 1 is a block diagram of a memory system 10 according to an embodiment of the inventive concept. Figure 4 , the memory system 10 may include a memory device 100 and a memory controller 200. Each of the memory device 100 and the memory controller 200 may be formed as a semiconductor chip (e.g., an integrated circuit formed as a semiconductor chip). In response to a write / read request from a host HOST, the memory controller 200 may control the memory device 100 so that data is read from the memory device 100 or data is written to the memory device 100. In detail, the memory controller 200 may control programming (or writing) and reading operations on the memory device 100 by providing an address ADDR, a command CMD, and a control signal CTRL to the memory device 100 via corresponding signal lines connecting the memory 100 and the memory controller 200 (e.g., forming a bus). Data to be written and read may be transmitted between the memory controller 200 and the memory device 100 (e.g., a serial communication line). Figure 4 DATA in the display).

[0040] The memory device 100 may include a memory cell array 110 and a control circuit 130. The memory cell array 110 may be composed of a plurality of memory cells MC (such as Figure 1Those described) are formed, and a plurality of memory cells MC may be arranged at the intersection between a plurality of first signal lines and a plurality of second signal lines. According to an exemplary embodiment, the plurality of first signal lines may be a plurality of word lines, and the plurality of second signal lines may be a plurality of bit lines. According to another exemplary embodiment, the plurality of first signal lines may be a plurality of bit lines, and the plurality of second signal lines may be a plurality of word lines. The plurality of word lines may be connected to a row decoder and activated by a row decoder (the row decoder may decode a row address to activate the corresponding word lines). The plurality of bit lines may be connected to a data buffer (e.g., via an I / O line) to send data to and from the memory cells MC. The memory device 100 including the memory cell array 110 may constitute a cross-point memory device.

[0041] According to an example of the present embodiment, a plurality of memory cells MC may be resistive memory cells, each memory cell including a variable resistor R. For example, each variable resistor may be formed of a phase change material (e.g., Ge-Sb-Te (GST)) having a resistance that changes according to a temperature applied thereto, and the memory device 100 may be a phase change RAM (PRAM). As another example, each variable resistor may include an upper electrode, a lower electrode, and a transition metal oxide (or a composite metal oxide) between the upper electrode and the lower electrode, and the memory device 100 may be a resistive RAM (RRAM). As another example, each variable resistor may include a magnetic upper electrode, a magnetic lower electrode, and a dielectric between the magnetic upper electrode and the lower electrode, and the memory device 100 may be a magnetoresistive RAM (MRAM). Therefore, the memory device 100 may be a resistive memory device (whether PRAM, RRAM, or MRAM), and the memory system 10 may be referred to as a resistive memory system.

[0042] According to one embodiment, each memory cell may be a single-level cell (SLC) for storing only one bit of data. In this embodiment, the memory cell may have a plurality of storage units, such as a storage unit, a storage unit, and a storage unit. Figure 3According to another embodiment, each memory cell may be a multi-level cell (MLC) for storing two or more data bits. In this embodiment, the memory cell may have four or more resistance distributions depending on the state of the stored data. According to another embodiment, each MLC memory cell may be a three-level cell (TLC) for storing three bits of data. In this embodiment, the memory cell may have eight resistance distributions depending on the state of the stored data. However, the inventive concept is not limited thereto. For example, in another embodiment, the memory cell array 110 may include memory cells each capable of storing four or more bits of data. Alternatively, the memory cell array 110 may include a combination of SLC and / or two-bit MLC and / or TLC.

[0043] The control circuit 130 may control the overall operation of the memory device 100. According to the present embodiment, the control circuit 130 may control the voltage level and the timing of applying the voltage applied to the first signal line and the second signal line connected to the memory cell array 110. The voltage may include an operating voltage applied to the selected first signal line and the second signal line, and prohibit the voltage applied to the unselected first signal line and the second signal line.

[0044] The memory controller 200 may include an error correction code (ECC) circuit 210. The ECC circuit 210 may perform ECC encoding on write data and perform ECC decoding on read data. In more detail, the ECC circuit 210 may perform ECC encoding on information data received from the host HOST during a write operation, thereby generating a codeword and providing the codeword to the memory device 100. The ECC circuit 210 may perform ECC decoding on read data (i.e., read codewords) received from the memory device 100 during a read operation, thereby generating information data and providing the information data to the host HOST. Although the ECC circuit 210 is provided with the memory controller 200, the ECC circuit 210 may be formed as a part of the memory device 100 instead. Therefore, via its ECC circuit, the memory device 100 may perform ECC encoding on information data received from the memory controller 200 (which may be the same as the memory controller provided by the host HOST) to generate a codeword that is subsequently written to the memory cell array 110, and may perform ECC decoding on the codeword stored and read from the memory cell array 110 to obtain information data that is subsequently sent to the memory controller 200. Alternatively, providing the ECC circuit as part of the memory device 100 (which may have the same structure and operation as the ECC circuit 210) is applicable to all embodiments described herein. However, it should also be understood that providing the ECC circuit as part of the memory device 100 will not exclude further ECC encoding / decoding of the ECC circuit in the controller 200.

[0045] According to the present embodiment, the ECC circuit 210 may perform ECC encoding and ECC decoding by using a low density parity check (LDPC) code. According to another embodiment, the ECC circuit 210 may perform ECC encoding and ECC decoding by using algorithms such as Reed-Solomon (RS) codes, Hamming codes, cyclic redundancy codes (CRC), etc. The ECC circuit 210 may perform ECC encoding on information data of a predetermined size to generate a check code (e.g., an LDPC code or other error correction code), which together with the information data forms a codeword (also referred to herein as an ECC sector).

[0046] The resistance distribution may degrade depending on the position of the memory cells MC of the memory cell array 110. Therefore, the probability of error occurrence of the data stored in the memory cells MC may increase, and the frequency of error occurrence may vary depending on the position of the memory cells MC. This BER imbalance may be an important burden in determining the correction capability of the ECC circuit 210. In order to reduce the BER imbalance, the control circuit 130 may control the execution of the ECC interleaving operation when the information data included in the codeword provided by the ECC circuit is written to the memory cell array 110. The ECC interleaving operation may include a word line interleaving operation, a bit line interleaving operation, a convolution interleaving operation and / or a stacked layer interleaving operation. The interleaving operation described herein may refer to a memory access operation to different relative positions in corresponding parts of the memory.

[0047] Figure 5A is shown in more detail including Figure 4 1 is a block diagram of a memory device 100 in a memory system 10 of the present invention.

[0048] Reference Figure 5A , the memory device 100 may include a memory cell array 110, a write / read circuit 120, a control circuit 130, a reference signal generating circuit 140, a voltage generating circuit 150, a row decoder 160, and a column decoder 170. The write / read circuit 120 may include a sense amplifier array (SA) 121 including one or more sense amplifiers and a write driver circuit (WD) 122 including one or more write drivers.

[0049] The memory cells included in the memory cell array 110 may be connected to a plurality of first signal lines and a plurality of second signal lines. According to an exemplary embodiment, the plurality of first signal lines may be a plurality of word lines WL, and the plurality of second signal lines may be a plurality of bit lines BL. In response to various voltage signals or current signals provided through the bit lines BL and the word lines WL, data may be written to or read from the selected memory cells (e.g., those connected to the word lines selected by the row decoder 160), and data may be prevented from being written to or read from the non-selected memory cells (e.g., those not connected to the word lines selected by the row decoder 160).

[0050] The memory device 100 may receive an address ADDR for identifying a memory cell to be accessed and a command CMD. The address ADDR may include a row address RA for identifying and selecting one of the word lines WL of the memory cell array 110 and a column address CA for identifying and selecting a subset of the bit lines BL of the memory cell array 110. The row decoder 160 may perform a word line selection operation (e.g., activate the selected word line) in response to the row address RA. The column decoder 170 may perform a bit line selection operation (e.g., select a subset of the bit lines BL to access the memory cells connected to the selected subset of the bit lines BL and connected to the activated selected word lines) in response to the column address CA. Such an access operation may be a read operation or a write operation (and it should be understood that the access operation mentioned herein generally refers to such a read and write operation unless the context indicates otherwise).

[0051] The write / read circuit 120 may be connected to the bit lines BL of the memory cell array 110 to write data to or read data from the memory cells. The voltage generating circuit 150 may generate a write voltage Vwrite used in a write operation and a read voltage Vread used in a read operation. The write voltage Vwrite may include a set voltage and a reset voltage (e.g., those described herein) related to the write operation. The write voltage Vwrite and the read voltage Vread may be provided to those bit lines BL selected by the column decoder 170 through the column decoder 170.

[0052] The reference signal generating circuit 140 may generate a reference voltage Vref and / or a reference current Iref, one or both of which may be used in a data reading operation. For example, the sense amplifier of the SA 121 may be connected to a node (e.g., a sense node) of a selected bit line BL to determine the data of the memory cell MC to which the bit line BL is connected. The data value of the memory cell MC may be determined by an operation of comparing the voltage of the sense node with the reference voltage Vref. Alternatively, when a current sensing method is implemented, the reference signal generating circuit 140 may generate a reference current Iref and provide it to the memory cell array 110. The data value may be determined by an operation of comparing the voltage of the sense node caused by the reference current Iref with the reference voltage Vref. A plurality of bit lines BL may be selected and data may be read simultaneously from a plurality of memory cells to which they are connected (e.g., by one of the sensing methods described herein), and a plurality of sense amplifiers of the SA 121 may each be connected to a corresponding sense node of a corresponding one of the bit lines BL thus selected. For example, different column addresses may identify different subsets of the bit lines BL and be decoded by the column decoder 170 to select them.

[0053] The write / read circuit 120 may provide the result of determining the read data to the control circuit 130. The control circuit 130 may control the write and read operations of the memory cell array 110 by referring to the result of the determination.

[0054] The control circuit 130 may output various control signals CTRL_RW for writing data to or reading data from the memory cell array 110 based on a command CMD, an address ADDR, and a control signal CTRL received from the outside (in this example, from the memory controller 200). Therefore, the control circuit 130 may control the internal operation of the memory device 100.

[0055] The control circuit 130 may determine the ECC interleaving method so that when the Figure 4 When the information data in the codeword provided by the ECC circuit 210 is written to the memory cell array 110, the BER can be equalized. Once the ECC interleaving method is determined, when writing the information data to the memory cell array 110, the control circuit 130 can selectively perform ECC interleaving operations, including word line interleaving operations, bit line interleaving operations, convolution interleaving operations, and stacked layer interleaving operations.

[0056] When the selected ECC interleaving operation is performed, the control circuit 130 may change the addressing of the memory cell array 110. The control circuit 130 may control the row decoder 160 and the column decoder 170 so that the addressing of the word lines and / or bit lines of the memory cell array 110 is changed according to the ECC interleaving operation (which may be selected in some examples).

[0057] The control circuit 130 may perform a word line interleaving operation on different memory slices included in the memory cell array 110. When the control circuit 130 performs a word line interleaving operation on the first and second memory slices, the control circuit 130 may change word line addressing of one or both of the first and second memory slice groups.

[0058] The control circuit 130 may perform a word line interleaving operation with respect to different banks included in the memory cell array 110. When the control circuit 130 performs the word line interleaving operation on the first memory bank and the second memory bank, the control circuit 130 may change word line addressing of one or both of the first memory bank and the second memory bank.

[0059] The control circuit 130 may perform a bit line interleaving operation on different memory slices included in the memory cell array 110. When the control circuit 130 performs a bit line interleaving operation on the first and second memory slices, the control circuit 130 may change the bit line addressing of one or both of the first and second memory slice groups.

[0060] The control circuit 130 may perform a convolution interleave operation with respect to different memory slices included in the memory cell array 110. When the control circuit 130 performs a convolution interleave operation on the first memory slice and the second memory slice, the control circuit 130 may change word line addressing and bit line addressing of one or both of the first memory slice group and the second memory slice group.

[0061] The control circuit 130 may perform a stacked layer interleaving operation on the stacked memory slices included in the memory cell array 110. When the control circuit 130 performs the stacked layer interleaving operation on the stacked memory slices, the control circuit 130 may change the word line addressing and the bit line addressing of one or both of the stacked memory slice groups.

[0062] Therefore, when the control circuit 130 selectively performs the ECC interleave operation, BER imbalance according to the locations of memory cells included in the memory cell array 110 may be reduced.

[0063] Figure 5B Shows Figure 5A An exemplary embodiment of a memory device. Figure 5BThe memory device is organized into a plurality of memory banks, including a bank memory array 110a to 110d (forming the memory array 110), a bank row decoder 160a to 160d (forming the row decoder 160), a bank column decoder 170a to 170d (forming the column decoder 170), and a plurality of sense amplifier / write driver arrays (SA / WD) 120a to 120d (forming the write driver circuit 120). Each bank is independently and individually operable, so that access operations of the banks can be performed independently of each other (e.g., in response to different externally received commands CMD, such as from the controller 200), and such access operations can be performed simultaneously. Figure 5B As shown, each bank is provided with its own (and independently operable) row decoder and column decoder. Bank control logic 180 may control the operation of the memory array banks by providing timing and enable signals (e.g., in response to receiving and decoding bank addresses and other control signals from control circuitry 130).

[0064] Figure 5B Further exemplary details are also shown, including an I / O gating circuit 190 configured to connect the bit lines BL selected by the column decoder 170 to the data I / O buffer (i.e., those bit lines BL identified by the column address CA decoded by the column decoder 170). In addition, the memory device 100 may include a command register 132 and an address register 134 to latch externally received commands CMD and addresses ADDR (e.g., provided by the memory controller 200).

[0065] Figure 5B A control circuit 130 is also shown, including a command decoder 130a, an address modification circuit 130b, and a mode register 130c. The command decoder 130a can receive an externally received command CMD from a command register 132, decode it, and provide a corresponding internal command (e.g., a control signal) to implement the operation of the received command CMD. The address modification circuit 130b can receive an externally received address ADDR from an address register 134, and modify one or more bits of the received address ADDR.

[0066] Although the address modification circuit 130b is shown as being continuous and located away from the memory bank, the address modification circuit 130b may include a distributed circuit portion, such as adjacent to the bank row decoders 160a to 160d and / or adjacent to the bank column decoders 170a to 170d. The address modification circuit 130b may invert one or more address bits of a subset of all addresses ADDR. For example, the address modification circuit 130b may invert one or more address bits of the addresses ADDR used to identify some memory banks, but not invert address bits of the addresses ADDR used to identify other memory banks. For example, when an access operation is to be performed on banks 110b and 110d (e.g., which may correspond to odd bank addresses), the address modification circuit 130b may (i) invert a selected one (e.g., a predetermined one) of the address bits of the row address RA, (ii) invert a selected one (e.g., a predetermined one) of the address bits of the column address CA, or (iii) invert a selected one (e.g., a predetermined one) of the address bits of the row address RA and the column address CA. When an access operation is to be performed on banks 110a and 110c, address modification circuit 130b cannot perform modification.

[0067] The address modification circuit 130b may be configurable. For example, the mode register 130c may include a register that can be programmed (e.g., by the memory controller 200) via a mode register access command to set which address bits should be inverted. Alternatively, the address modification circuit 130b may be programmable as part of the manufacturing process and include a fuse set that can be set (e.g., via laser or other one-time programming) to identify the (multiple) address bits to be inverted. Alternatively, the address modification circuit 130b may be hardwired, such as by an inverter inserted between the address output by the control circuit 130 and some of the body row decoders 160a to 160d and / or some of the body column decoders 170a to 170d.

[0068] Figure 6 It is shown Figure 5A and Figure 5B 1 is a circuit diagram of a portion of the memory cell array 110 in further detail. The memory cell array 110 may include a plurality of memory chips, and Figure 6 An example of a memory tile TILE is shown.

[0069] Reference Figure 6 The memory tile TILE may be formed of a plurality of word lines WL0 to WLm, a plurality of bit lines BL0 to BLm, and a plurality of memory cells MC. The number of word lines WL, the number of bit lines BL, and the number of memory cells MC may be selected based on various design criteria and may be variously changed according to some embodiments.

[0070] According to the present embodiment, each of the plurality of memory cells MC may include a variable resistor R and a selector D. Herein, the variable resistor R may be a variable resistor material, and the selector D may be referred to as (and constitute) a switch.

[0071] Each memory cell MC may include a variable resistor R and a selector D connected in series between a corresponding one of a plurality of bit lines BL0 to BLm and a corresponding one of a plurality of word lines WL0 to WLm. Figure 6 In the example shown, the series connection of each memory cell MC is WL-RD-BL (word line-variable resistor-selector bit line), but the order of series connection can also be implemented as WL-DR-BL (for example, the selector D can be connected between one of the multiple bit lines BL0 to BLm and the variable resistor R, and the variable resistor R can be connected between the selector D and one of the multiple word lines WL0 to WLm).

[0072] The selector D may be connected between one of the plurality of word lines WL0 to WLm and the variable resistor R to control the current supply to the variable resistor R based on the voltage applied to the connected word line and the bit line. In some examples, the selector D may be a PN or PIN junction diode. In this case, the anode of the diode may be connected to the variable resistor R, and the cathode thereof may be connected to one of the plurality of bit lines BL0 to BLm. When the voltage difference between the anode and cathode of the diode is greater than the threshold voltage of the diode, the diode may be turned on, and thus current may be supplied to the variable resistor R.

[0073] 7A to 7C is a diagram showing a method for forming a memory cell array 110 (and forming Figure 6 A circuit diagram of an alternative example of a memory cell MC of a TILE. Fig. 7A , the memory cell MCa may include a variable resistor Ra, and the variable resistor Ra may be connected between the bit line BL and the word line WL. The memory cell Mca may store data due to voltages respectively applied to the bit line BL and the word line WL.

[0074] Reference Figure 7B , the memory cell MCb may include a variable resistor Rb and a bidirectional diode Db. The variable resistor Rb may include a resistance material for storing data. The bidirectional diode Db may be connected between the variable resistor Rb and the bit line BL, and the variable resistor Rb may be connected between the word line WL and the bidirectional diode Db. The positions of the bidirectional diode Db and the variable resistor Rb may be switched. The bidirectional diode Db may block leakage current flowing through an unselected memory cell.

[0075] Reference Figure 7C, the memory cell MCc may include a variable resistor Rc and a transistor TR. The transistor TR may be used as a selector (ie, a switch) for supplying current to the variable resistor Rc or blocking current from flowing to the variable resistor Rc based on the voltage of the word line WL. Figure 7C In an embodiment of the present invention, in addition to the word line WL, a source line SL for controlling the voltage level of both ends of the variable resistor Rc may be provided. The transistor TR may be connected between the variable resistor Rc and the bit line BL, and the variable resistor Rc may be connected between the source line SL and the transistor TR. The positions of the transistor TR and the variable resistor Rc may be switched. The memory cell MCc may be selected or not selected based on whether the transistor TR driven by the word line WL is turned on or off.

[0076] Figure 8 is shown in more detail including Figure 4 1 is a block diagram of a memory controller 200 in a memory system 10.

[0077] Reference Figure 8 , the memory controller 200 may include an ECC circuit 210, a processor 220, a RAM 230, a host interface (I / F) 240, and a memory I / F 250. The ECC circuit 210 may generate a parity bit (check code) by performing ECC encoding on the information data received from the host HOST during a write operation, generate a codeword, which is a unit for performing error correction, and provide the codeword to the memory device 100. The codeword may include information data and a parity bit (e.g., a check code or an error correction code generated by the ECC circuit 210). The information data may have the same format or may be reformatted compared to the state when it is received from the host HOST. Similarly, the codeword may also be reformatted in some examples, for example, to provide a codeword with a more balanced number of logic high ("1") bits and logic low ("0") bits (e.g., by processing the codeword output by the ECC circuit 210 using a bit balancing circuit (not shown) using a conventional bit balancing method before storing the codeword in the ECC sector of the memory device). The reformatting may also be in the form of data randomization and / or data encryption, and may be implemented using a randomization circuit and / or an encryption circuit (not shown). In some examples, such reformatting may change (e.g., increase) the number of bits of the resulting reformatted codeword (and / or information data). The ECC circuit 210 may perform ECC decoding on the read data (i.e., codeword) received from the memory device 100 during a read operation, thereby generating information data and providing the information data to the host HOST.

[0078] When the number of error bits in the codeword is within the correction capability, the ECC circuit 210 may correct the error. On the other hand, when the number of error bits in the codeword exceeds the correction capability, the ECC circuit 210 may not correct the error. Thus, BER imbalance may be a great burden in determining the correction capability of the ECC circuit 210. If the BER imbalance can be reduced, the correction capability of the ECC circuit 210 can be reduced, so that the ECC circuit 210 can easily implement the error correction algorithm.

[0079] The processor 220 may include a central processing unit (e.g., or a microprocessor), and may control the overall operation of the memory controller 200. In more detail, the processor 220 may be configured by a code stored in the RAM 230 to control the overall operation of the memory controller, such as controlling the timing of the operation of the firmware of the memory controller 200. The RAM 230 may be used as an operating memory, a cache memory, or a buffer memory of the processor 220.

[0080] The host I / F 240 interfaces with the host HOST to receive requests for memory operations from the host HOST. For example, the host I / F 240 receives various requests for various operations from the host, such as data read and write requests, and the memory controller 200 uses these requests to generate various internal signals for memory operations on the memory device 100.

[0081] The memory I / F 250 may provide an interface between the memory controller 200 and the memory device 100. For example, write data and read data may be transmitted to and received from the memory device 100 through the memory I / F 250. The memory I / F 250 may provide commands and addresses to the memory device 100, receive various information from the memory device 100, and provide various information to internal components of the memory controller 200.

[0082] Fig.9A It is shown Figure 5A 1 is a block diagram of exemplary details of the memory cell array 110. Figure 5B Each of the bank memory arrays 110a-110d in the Fig.9A The memory cell array 110 can be implemented using a memory chip group, wherein Figure 6 The memory tile group TILE is arranged two-dimensionally in the column direction and the row direction. The memory tile group may include a plurality of memory racks BAY, and each of the memory racks BAY may include a plurality of memory tile TILEs (such as Figure 6 Multiple memory tiles TILE) in the memory rack BAY can be arranged in the column direction. Fig.9A128 memory racks BAY are shown, each of which includes 4 memory tiles TILE1 to TILE4. The inventive concept is not limited thereto, and the number of memory racks BAY and the number of memory tiles TILE may vary according to embodiments.

[0083] Reference Fig.9A , the memory tile group TILE_GR can be set to a storage area to which data is written or read in units of pages. A page represents a group of memory cells selected from memory cells connected to a single word line that can perform write and read operations at the same time. For example, assume that the data bits corresponding to each page are 128 bits (or 16 bytes). In this case, 1 bit of data can be written to each of the 128 memory racks BAY that are commonly connected to a word line in the memory tile group TILE_GR, or read from it.

[0084] In the case of forming certain access circuits, the tiles of group TILE_GR can be spaced apart from each other. For example, the access circuits located in the area between the tiles may include one or more of a sub-wordline driver, a bitline sense amplifier, a selection circuit for connecting the bitline to an I / O line, a wordline and a wordline strapping connection between a plurality of sub-wordlines. Within each tile, the rows of memory cells (and corresponding wordlines) may be spaced apart at a first pitch, wherein adjacent rows of memory cells (and adjacent wordlines) located in different tiles are spaced apart from each other at a distance greater than the first pitch, providing space for the access circuits. Similarly, within each tile, the columns of memory cells (and corresponding bitlines) may be spaced apart at a second pitch, wherein adjacent columns of memory cells (and adjacent bitlines) located in different tiles are spaced apart from each other at a distance greater than the second pitch, providing space for the access circuits. See, for example, exemplary layouts described in US 2015 / 0364178 (with subarray blocks SCB) and US 2009 / 0141567, both of which are incorporated herein by reference.

[0085] FIG. 10A to FIG. 10C It is used to explain Fig.9A FIG. 1 is a diagram showing a first example of a distribution characteristic of memory cells in a memory tile group TILE_GR. Fig. 10B and Fig. 10C , the horizontal axis indicates resistance, and the vertical axis indicates the number of memory cells.

[0086] Reference Fig. 10A The memory tile group TILE_GR may have a plurality of word lines WL arranged sequentially from the bottom to the top of the memory tile group TILE_GR. 0 To WL n-1 (For example, word line WL having a bottom edge defining the memory tile group TILE_GR0 and WL which defines the upper edge n-1 ). It should be understood that Fig. 10A Reference mark WL 0 To WL n-1 (and similar notations elsewhere in this disclosure) corresponds to a word line address (e.g., a row address RA). For example, Fig. 10A WL 0 To WL n-1 The labels indicate the word lines corresponding to the ones addressed therein (eg, respectively, by the row address RA 0 To RA n-1 Using this addressing nomenclature (WL 0 To WL n-1 ) may also be used to refer to a specific (ie, physical) word line (or group of word lines). It should be understood that modification of the word line addressing does not modify the physical word line, only the identification of the physical word line. Similar identification of the bit line BL is also implemented in the present disclosure. It is expected that Figure 3 As shown, connected to the word line WL 0 To WL n-1 Each of the memory cells has a low resistance state LRS or a high resistance state HRS (relative to a reference resistance RR).

[0087] However, the word line WL connected to the center of the memory tile group TILE_GR n / 2 The memory cell shows Figure 3 The normal resistance distribution shown in FIG. 1 is connected to the word line WL on the edge of the memory tile group TILE_GR. 0 and WL n-1 The memory cell shows Fig. 10B and Fig. 10C The resistance distribution shown is widened and / or possibly shifted. The resistance distribution of the word lines may be widened for word lines at or near the edge of the memory tile group TILE_GR (e.g., compared to those word lines at the center of the memory tile group TILE_GR (e.g., in this example, WL n / 2 ) memory cell, the word line WL 0 and WL n-1 The memory cell, word line WL 0 and WL n-1 Can have a wider resistance distribution).

[0088] Reference Fig. 10B , along the word line WL located at the center of the memory tile group TILE_GR n / 2 To the word line WL at the lower edge of the memory tile group TILE_GR 0 direction, the resistance of the memory cell decreases and its distribution widens and shifts. Fig. 10B Also shown is the word line WL n / 4 and WL n / 8 The exemplary resistance distribution widens / shifts the word line WL n / 4 and WL n / 8 are located at the far lower edge of the memory tile group TILE_GR (in this example, corresponding to WL 0 ) at 1 / 4 and 1 / 8 of the vertical size (in the column direction).

[0089] Reference Fig. 10C , along the word line WL from the center of the memory tile group TILE_GR n / 2 To the word line WL at the upper edge of the memory tile group TILE_GR n-1 direction, the resistance of the memory cell decreases and its distribution widens and shifts. Fig. 10C Also shown is the word line WL 3n / 4 and WL7 n / 8 The exemplary resistance distribution widens / shifts the word line WL 3n / 4 and WL 7n / 8 are located at the far upper edge of the memory tile group TILE_GR (in this example, corresponding to WL n-1 ) at 1 / 4 and 1 / 8 of the vertical size (in the column direction).

[0090] exist Fig. 10B and Fig. 10C In the resistance distribution characteristics, the resistivity of the reset data of the memory cell can be reduced in the direction toward the lower edge and upper edge of the memory slice group TILE_GR. The lower edge and upper edge of the memory slice group TILE_GR belong to the resistance distribution degradation area. For the memory cells in the resistance distribution degradation area, the sensing margin between the set data of logic "1" and the reset data of logic "0" is reduced. Therefore, the sensing of the reset data of logic "0" becomes more uncertain, slower or generally less reliable. Therefore, the BER of the word line increases as its position is closer to the edge of the memory slice group. In contrast, the resistivity of the reset data of the memory cell located in the center of the memory slice group TILE_GR is relatively maintained, so the BER may be reduced. When the memory cells on the lower and upper edges with large BER and the memory cells at the center with small BER are mixed with each other and then store data, the BER imbalance between these memory cell groups can be reduced. The ECC interleaving operation in which the word line interleaving operation is performed to achieve BER balance will now be described.

[0091] FIG. 11A to FIG. 11C is a view for explaining an ECC interleaving operation according to an embodiment of the inventive concept.

[0092] Reference Fig.11A, you can use Fig.9A The two memory tile groups TILE_GR are used to perform word line interleaving operations. Figure 8 The ECC circuit 210 can generate a codeword by generating parity bits based on the information data received from the host. The information data received from the host can be divided into a plurality of parts that are subsequently used to form an ECC sector. For example, if the ECC circuit 210 is configured to generate parity data to form a 256-bit codeword, the information data received from the host HOST can be divided into a plurality of blocks of 146 bits each. For example, the host HOST can send a storage command to the memory controller 200 to store a large block of data (e.g., a 4MB file), and send the 4MB information data to the memory controller 200 together with the storage command. The data block can be sent to the memory controller 200 by the host HOST as a stream of information data. The information data block sent by the host HOST to the memory controller 200 can be divided into a plurality of smaller data blocks (e.g., 146 bits). Each data block may be sent to the ECC circuit 210 (e.g., sequentially for sequential processing), and the ECC circuit 210 may generate parity data (e.g., 110 bits) for each data block, wherein the data block and the corresponding parity data form a codeword (e.g., 256 bits). Thus, a plurality of codewords are generated, each codeword comprising an information data block from the host HOST (comprising a portion of a larger information data block sent by the host HOST) and corresponding parity data.

[0093] Multiple codewords sequentially generated by the ECC circuit 210 can be sequentially stored in a single storage device, or spread out on multiple storage devices, as described elsewhere herein. For example, the memory controller 200 can allocate a portion of the available memory for storing multiple codewords, and send one or more write commands and multiple codewords to the storage device. The memory controller can send at least one memory address to the storage device, which identifies the location of the memory in which the storage device stores the codeword. For example, the memory controller 200 can send a memory address (e.g., including a row address and a column address) to the storage device together with each write command, wherein each write command is associated with a codeword (e.g., a single write command is valid for storing a single codeword). In this case, the memory controller 200 can send each subsequent write command together with the corresponding incremental memory address and the subsequently sent codeword (e.g., for each subsequently sent codeword, sequentially increase each row address or sequentially increase each column address). Alternatively, the memory controller 200 can provide a starting memory address with a burst type write command to the storage device for identifying the initial position for storing the first codeword and sending a codeword sequence associated with the write command. In a burst write command, the memory device may initially store a first codeword at a memory portion identified by a starting memory address, a second codeword at a memory portion identified by a starting address+1, and a third codeword at a memory portion identified by a starting address+2. For example, the memory may include a row address counter and a column address counter that initially store a starting address (e.g., storing a starting row address and a starting column address, respectively), which are sent from the memory together with the write command. The row address of the row address counter and the column address of the column address counter may be provided to a row decoder and a column decoder, respectively, which are decoded by the row decoder and the column decoder to select the corresponding word line and column storing the first codeword. After the first codeword is stored in a memory location identified by the starting row address and the starting column address, one of the column address counter and the row address may be incremented, and the next codeword may be stored at a memory address identified by the incremented address. In addition, the other of the column address counter and the row address counter may be incremented after a certain number of codewords are written. It should be noted that a single memory address (e.g., sent with a write command) may be used to store a single codeword at multiple memory locations identified by the same single memory address (e.g., according to any of the examples described herein regarding ECC sub-sectors SEC1 and SEC2), or multiple memory addresses may be used (e.g., multiple addresses sent with a write command or generated internally by the storage device).In many implementations, the same row address and the same column address are used to store multiple ECC sub-sectors (e.g., SEC1 and SEC2) of a single ECC sector (i.e., multiple portions of a codeword are stored at different memory locations) (which may be in relatively different memory locations according to any embodiment described herein).

[0094] Other circuits and methods may be used. In general, the result of storing a block of information data may include storing a series of codewords using a set of word lines identified by a series of row addresses (e.g., in order in a range of row addresses). Similarly, a block of information data may include storing a series of codewords at memory locations identified by a series of column addresses and / or in an order of column addresses.

[0095] When storing an ECC sector, the ECC sector is divided into a plurality of parts (ECC sub-sectors). In this example, the ECC sector includes two ECC sub-sectors, namely a first ECC sub-sector SEC1 and a second ECC sub-sector SEC2. For example, the ECC sector may be formed of 256 bits (or 32 bytes), and each of the first ECC sub-sector SEC1 and the second ECC sub-sector SEC2 may be formed of 128 bits (or 16 bytes).

[0096] The first ECC sub-sector SEC1 may be set to be written to the first memory tile group TILE_GR1, and the second ECC sub-sector SEC2 may be set to be written to the second memory tile group TILE_GR2. In this case, when the word line interleaving operation is performed, the first ECC sub-sector SEC1 may be written to the word line WL connected to the lower edge of the first memory tile group TILE_GR1. 0 The second ECC sub-sector SEC2 may be written to the word line WL connected to the center of the second memory tile group TILE_GR2. 0 A memory cell is selected from among the memory cells.

[0097] Word lines WL in the first memory tile group TILE_GR1 0 and the word line WL in the second memory tile group TILE_GR2 0 The word line WL0 in the first memory tile group TILE_GR1 can be set to be a word line close to the lower edge of the first memory tile group TILE_GR1, and the word line WL0 in the second memory tile group TILE_GR2 can be set to be a word line far from the lower edge of the second memory tile group TILE_GR2.

[0098] In some operations, regarding the word line interleaving operation of different memory tiles, the word line WL located at the lower edge of the first memory tile group TILE_GR1 may be interleaved. 0 to the word line WL located at its upper edge n-1 A write operation (e.g., writing a sequence of ECC sectors sequentially processed by the ECC circuit 210, as described elsewhere herein) is sequentially performed on the first memory tile group TILE_GR1 in the direction of . The word line WL located at the center of the second memory tile group TILE_GR2 may be 0 to the word line WL located at its upper edge n / 2-1 direction, and then from the word line WL located at its lower edge n / 2 to the word line WL located at its center n- Of course, other operations may be performed to randomly access the word lines of the tiles TILE_GR1 and TILE_GR2, and other sequences of access operations (including other sequences of write operations) may be implemented.

[0099] In order to perform word line interleaving operations with respect to different memory blocks, such as Fig. 11B As shown, Figure 5A and Figure 5B The control circuit 130 can change the word line addressing of one or both of the first memory tile group TILE_GR1 and the second memory tile group TILE_GR1. The row address RA[i:0] of the first memory tile group TILE_GR1 can be sequentially increased in the word line WL located at the lower edge of the first memory tile group TILE_GR1. 0 To the word line WL located at the upper edge of the first memory tile group TILE_GR1 n-1 The word lines of the first memory tile group TILE_GR1 are respectively identified (and sequentially selected during the access operation) in the direction of . The row address RA[i:0] of the second memory tile group TILE_GR2 can be sequentially increased from the word line WL located at the center of the second memory tile group TILE_GR2. 0 To the word line WL located at the upper edge of the second memory tile group TILE_GR2 n / 2-1 , and then from the word line WL located at the lower edge of the second memory tile group TILE_GR2 n / 2 To the word line WL located at the center of the second memory tile group TILE_GR2 n-1 (and is adjacent to WL which is also located at the center of the second memory tile group TILE_GR2 0 ) direction respectively identifies (and sequentially selects during the access operation) the word lines of the second memory tile group TILE_GR2.

[0100] The order of sequential write operations performed in word line interleaving operations on different memory chips can be changed. For example, the control circuit 130 can be set in the opposite manner. Fig. 11B The word line addressing sequence is such that the word line WL located at the lower edge of the second memory tile group TILE_GR2 is 0 To the word line WL located above it n-1 The write operation is sequentially performed on the second memory tile group TILE_GR2 in the direction of 0 to the word line WL located at its upper edge n / 2-1 , then, from the word line WL located at its lower edge n / 2 to the word line WL located at its center n-1 In the direction of , write operations are sequentially performed on the first memory tile group TILE_GR1.

[0101] Assuming that bit errors occur at a relatively high bit error rate in memory cells on the lower and upper edges of the first memory tile TILE_GR1 and the second memory tile group TILE_GR2 (compared to memory cells in the middle of these tile groups TILE_GR1 and TILE_GR2), after performing the word line interleaving operation, the bit errors of the ECC sector can be equalized by mixing the memory cells on the lower or upper edge with the memory cells at the center. Therefore, the BER imbalance between word lines can be reduced by applying the word line interleaving operation with respect to different memory tiles.

[0102] The present embodiment shows an ECC sector, which includes a first ECC sub-sector SEC1 and a second ECC sub-sector SEC2 written to different relative positions in two different slice groups. However, the ECC sector can be divided into other numbers of sub-sectors, and the sub-sectors can be written to the memory slice groups at different relative positions in these memory slice groups respectively. In this case, the sub-sectors can be mixed in the memory cells of different memory slice groups and written therein via word line interleaving operations, so that the bit errors of the ECC sector can be balanced. The ECC sector can be divided into n sub-sectors, where n is an integer equal to 2 or greater, and each sub-sector can be written to a corresponding one of the n memory slice groups at m different relative positions in each slice group, where m is an integer in the range of 2 to n.

[0103] Fig. 11C shows word line addressing, which differs from the Fig.11A Word line addressing is performed in a word line interleaving operation.

[0104] Reference Fig. 11C In the first memory tile group TILE_GR1, the word line WL 0It may be one of the word lines belonging to the resistance distribution degradation area. In the first memory tile group TILE_GR1, the word line WL 0 These are word lines other than the word lines located at the lower edge of the first memory tile group TILE_GR1.

[0105] In the second memory tile group TILE_GR2, the word line WL 0 It can be one of the word lines belonging to the normal resistance distribution area. In the second memory tile group TILE_GR2, the word line WL 0 It may be a word line other than the word line located at the center of the second memory tile group TILE_GR2.

[0106] When performing a word line interleaving operation, the first ECC sub-sector SEC1 may be written to the word line WL connected to the first memory tile group TILE_GR1. 0 The second ECC sub-sector SEC2 can be written to the word line WL connected to the second memory slice group TILE_GR2. 0 A memory cell is selected from among the memory cells.

[0107] Fig. 10D An example of ECC interleaving using word line interleaving is shown, which can be implemented when resistance distribution degradation occurs at the edge of each tile TILE of the tile group TILE_GR of the memory array. Fig. 10D As shown, resistance distribution degradation may occur for memory cells connected to word lines at the top and bottom edges of each tile. Fig. 9B It shows that when each tile TILE of the tile group TILE_GR includes Fig. 10D When the resistance distribution shown in FIG. 1 is degraded, the resistance distribution of the entire tile group TILE_GR is degraded. When the word line is connected to the memory cell row of the tile group TILE_GR extending horizontally across the entire tile group TILE_GR, the resistance distribution of the entire tile group TILE_GR is degraded compared to those farther from the upper and lower edges of the tile TILE (e.g. Fig. 9B WLi), for word lines located near the upper and lower edges of the TILE (e.g. Fig. 9B If WLj in the resistor distribution is too low, resistance distribution degradation may occur.

[0108] In order to balance the BER, the ECC sector can be divided into sub-sectors SEC1 and SEC2, and stored in different tile groups TILE_GR1 and TILE_GR2, as shown above. Figures 11A to 11CAs described, one or both of the row addressing of the word lines of TILE_GR1 and TILE_GR2 are modified so that whenever one of the sub-sectors SEC1 and SEC2 is stored at a position adjacent to the upper edge or lower edge of the tile TILE, the other of the sub-sectors SEC1 and SEC2 is stored on a memory cell connected to the word line WL away from the upper edge and the lower edge of the tile TILE. For example, assume that the row address RA decoded by the decoder 160 to select one of the word lines includes i bits. The row address RA can be formed by bits RA0, RA1, RA2, ... RAi-3, RAi-2, RAi-1, RAi in the order of the least significant bit (RA0) to the most significant bit (RAi). One of the more significant bits (for example, one of RAi-3, RAi-2, RAi-1) can be inverted for all row addresses used to access the tile group TILE_GR1 to store the sub-sector SEC1, while the row addressing of TILE GR2 can be kept unchanged. In Fig. 9B In the example of FIG. 1 , row address bits RAi-1, RAi may be decoded to identify and select one of four horizontal blocks of a slice containing a word line to be accessed, while row address bits RA0, RA1, RA2, ... RAi-3, RAi-2 may be decoded to identify and select a word line in a horizontal block (e.g., row address bits RA0, RA1, RA2, ... RAi-3, RAi-2 may be decoded to identify and select a word line in each slice). Fig. 10D The word line WL shown in 0 To W1 m-1 ). In this case, the row address RAi-2 can be inverted for the tile group TILE_GR1 and remain unchanged for the tile group TILE_GR2. For the tile group TILE_GR1, the word lines accessed sequentially in the sequential order corresponding to the row address can be in the order from the center of each tile to the upper edge of the tile, and then in the order from the bottom of each tile to the center of the tile. The corresponding word lines accessed in the tile group TILE_GR2 (i.e., for the same row address in the same order) can be in the order from the bottom of the tile to the top of the tile. It should be understood that the modification of the more significant bits (such as RAi-2 as described above) can be completed in combination with other address modifications described herein, such as inverting the most significant bit RAi of one of the tile groups TILE_GR1 and TILE_GR2, so that other patterns of resistance distribution degradation areas can be extended between portions of the ECC sector to achieve BER balance.

[0109] Fig.12 is a diagram for explaining an ECC interleaving operation according to an embodiment of the inventive concept.

[0110] Reference Fig.12, performing word line interleaving operations on different memory banks. The first memory tile group TILE_GR1 may belong to the first memory bank BANK1, and the second memory tile group TILE_GR2 may belong to the second memory bank BANK2. The first ECC sub-sector SEC1 of the ECC sector may be provided to be written to the first memory tile group TILE_GR1 of the first memory bank BANK1, and the second ECC sector SEC2 may be provided to be written to the second memory tile group TILE_GR2 of the second memory bank BANK2.

[0111] In the word line interleaving operation with respect to different memory banks, the word line WL located at the lower edge of the first memory tile group TILE_GR1 may be 0 to the word line WL located at its upper edge n-1 The write operation is sequentially performed on the first memory tile group TILE_GR1 of the first memory bank BANK1 in the direction of 0 to the word line WL located at its upper edge n / 2-1 , and then from the word line WL located at its lower edge n / 2 to the word line WL located at its center n-1 A write operation is sequentially performed on the second memory tile group TILE_GR2 of the second memory bank BANK2 in the direction of.

[0112] To perform word line interleaving operations with respect to different memory banks, as discussed herein, Figure 5A and Figure 5B The control circuit 13 may change the word line addressing of at least one of the first memory bank BANK1 and the second memory bank BANK2.

[0113] According to one embodiment, Figure 5A and Figure 5B The memory cell array 110 may include a plurality of memory banks. Figure 5A and Figure 5B , the memory cell array 110 can be arranged adjacent to a peripheral circuit area in which circuit blocks different from the memory cell array 110 (e.g., a write / read circuit 120, a control circuit 130, a reference signal generating circuit 140, a set voltage generating circuit 150, a row decoder 160, and a column decoder 170) are arranged. Figure 5C1 is a top view showing the layout of an exemplary memory chip having four memory cell arrays 110a to 110d arranged adjacent to various peripheral circuits. The BER of the memory cells of the memory cells close to the peripheral circuit area among the multiple memory cells of the memory cell array 110 may be higher than the BER of the memory cells of the memory cells far from the peripheral circuit area. In this case, the first memory cell BANK1 may be set as a memory cell close to the peripheral circuit area, and the second memory cell BANK2 may be set as a memory cell far from the peripheral circuit area. Moreover, the two memory cells may be arranged adjacent to the peripheral circuit area, and a word line interleaving operation may be performed so that each row address identifies a first word line in the first memory cell and a second word line in the second memory cell, wherein the distances from the first word line and the second word line from the peripheral circuit are different to balance the BER. The Vth distribution degradation zone may be formed within the body memory array at a position adjacent to the peripheral circuit area.

[0114] When a word line interleaving operation is performed with respect to different memory banks, the bit errors of the ECC sector can be mixed in the memory cells on the lower edge or the upper edge and in the memory cells at the center, so that the mixed bit error rate (BER) can be balanced between different word lines. Therefore, by applying the word line interleaving operation with respect to different memory banks, the BER imbalance between word lines can be reduced.

[0115] Fig.13 is a diagram for explaining an ECC interleaving operation according to an embodiment of the inventive concept.

[0116] Reference Fig.13 , performing word line interleaving operations on different memory chips. The first memory chip group TILE_GR1 may belong to the first memory chip CHIP1, and the second memory chip group TILE_GR2 may belong to the second memory chip CHIP2. The first ECC sub-sector SEC1 of the ECC sector may be provided to be written to the first memory chip group TILE_GR1 of the first memory chip CHIP1, and the second ECC sub-sector SEC2 may be provided to be written to the second memory chip group TILE_GR2 of the second memory chip CHIP2.

[0117] A word line interleaving operation may be applied with respect to different memory chips, so that a word line WL located at a lower edge of the first memory tile group TILE_GR1 may be interleaved. 0 to the word line WL located at its upper edge n-1 A write operation is sequentially performed on the first memory chip CHIP1 in the direction of the first memory chip group TILE_GR1, and a word line WL located at the center of the second memory chip group TILE_GR2 may be written. 0 to the word line WL located at its upper edge n / 2-1, and then from the word line WL located at its lower edge n / 2 to the word line WL located at its center n-1 In the direction of , a write operation is sequentially performed on the second memory tile group TILE_GR2 of the second memory chip CHIP2.

[0118] When a word line interleaving operation is performed on different memory chips, the bit errors of the ECC sector can be mixed in the memory cells on the lower edge or the upper edge and in the memory cells at the center, so that the bit error rate (BER) can be balanced between different word lines. Specifically, the Vth distribution degradation area can be located at the position of the memory array of the memory chip adjacent to the edge of the memory chip. Therefore, the BER imbalance between word lines can be reduced by applying a word line interleaving operation on different memory chips.

[0119] FIG. 14A to FIG. 14C It is used to explain Fig.9A FIG. 1 is a diagram showing a second example of a distribution characteristic of memory cells in a memory tile group TILE_GR. Fig. 14B and Fig. 14C , the horizontal axis indicates resistance, and the vertical axis indicates the number of memory cells.

[0120] Reference Fig.14A , the memory tile group TILE_GR can have multiple bit lines BL 0 To BL n-1 . It is expected that Figure 3 As shown, connected to the bit line BL 0 To BL n-1 Each of the memory cells has a low resistance state LRS or a high resistance state HRS (relative to a reference resistance RR).

[0121] However, the bit line BL connected to the center of the memory tile group TILE_GR n / 2 The memory cell shows normal resistance distribution characteristics, such as Figure 3 As shown, Fig. 14B and Fig. 14C As shown, the memory cells connected to the bit lines BL0 and BLn-1 located on the edge of the memory tile group TILE_GR show a widened resistance distribution characteristic.

[0122] Reference Fig. 14B , from the bit line BL located at the center of the memory tile group TILE_GR n / 2 In the direction of the bit line BL0 located at the left edge of the memory tile group TILE_GR, the resistance of the memory cells decreases and the distribution thereof widens.

[0123] Reference Fig. 14C, from the bit line BL located at the center of the memory tile group TILE_GR n / 2 To the bit line BL located at the right edge of the memory tile group TILE_GR n-1 In the direction of , the resistance of the memory cell decreases and its distribution widens.

[0124] exist Fig. 14B and Fig. 14C In the resistance distribution characteristics, the resistivity of the reset data of the memory cell can be reduced in the direction toward the lower edge and the upper edge of the memory slice group TILE_GR. The left edge and the right edge of the memory slice group TILE_GR belong to the resistance distribution degradation area. For the memory cells in the resistance distribution degradation area, the sensing margin between the set data of logic "1" and the reset data of logic "0" is reduced. Therefore, the sensing of the reset data of logic "0" becomes more uncertain, slower or generally less reliable. Therefore, the BER increases. On the other hand, the resistivity of the reset data of the memory cell located in the center of the memory slice group TILE_GR is maintained, so the BER may be reduced. When the memory cells on the left and right edges with large BER and the memory cells at the center with small BER are mixed with each other and then store data, the BER imbalance between the bit lines can be reduced. The ECC interleaving operation of performing a bit line interleaving operation to achieve BER equalization will now be described.

[0125] FIG. 15A to FIG. 15D is a view for explaining an ECC interleaving operation according to an embodiment of the inventive concept.

[0126] Reference Fig.15A , you can use Fig.9A The two memory slice groups TILE_GR of the ECC sector may perform a bit line interleaving operation. The ECC sector may include a first ECC sub-sector SEC1 and a second ECC sub-sector SEC2. The data structure of the ECC sector may include, for example, 256 bits (or 32 bytes), and each of the first ECC sub-sector SEC1 and the second ECC sub-sector SEC2 may include 128 bits (or 16 bytes).

[0127] The first ECC sub-sector SEC1 may be provided to be written to the first memory tile group TILE_GR1, and the second ECC sub-sector SEC2 may be provided to be written to the second memory tile group TILE_GR2. In this case, when the bit line interleaving operation is performed, the first ECC sub-sector SEC1 may be written to the word line WL connected to the first memory tile group TILE_GR1. i The bit line BL connected to the left edge of the first memory tile group TILE_GR1 is 0 To BL 127The memory cell of the second ECC sub-sector SEC2 can be written to the word line WL connected to the second memory tile group TILE_GR2. i The bit line BL connected to the center of the second memory tile group TILE_GR2 is 0 To BL 127 The memory cell. Word line WL i Indicates word lines on the same level in the first memory tile TILE_GR1 and the second memory tile group TILE_GR2.

[0128] The bit line BL in the first memory tile group TILE_GR1 0 and the bit line BL in the second memory tile group TILE_GR2 0 The bit line BL in the first memory tile group TILE_GR1 and the second memory tile group TILE_GR2 may be located at different distances from the respective left edges. 0 can be set to be close to the bit line on the left edge of the first memory tile group TILE_GR1, and the bit line BL in the second memory tile group TILE_GR2 0 The bit line may be set to be away from the left edge of the second memory tile group TILE_GR2.

[0129] In the bit line interleaving operation with respect to different memory tiles, the bit line BL located at the left edge of the first memory tile group TILE_GR1 may be interleaved. 0 to the bit line BL located at its right edge n-1 The write operation is sequentially performed on the first memory tile group TILE_GR1 in the direction of 0 to the bit line BL located at its right edge n / 2-1 , and then from the bit line BL located at its left edge n / 2 to the bit line BL located at its center n-1 (near bit line BL 0 ) direction, write operations are sequentially performed on the second memory tile group TILE_GR2.

[0130] In order to perform bit line interleaving operations on different memory slices, such as Fig. 15B As shown, Figure 5A and Figure 5BThe control circuit 130 can change the bit line addressing of one or both of the first memory tile group TILE_GR1 and the second tile group TILE_GR2. Sequentially increasing the column address CA[j:0] of the first tile group TILE_GR1 can respectively identify (and sequentially select during access operations) the bit lines BL located at the left edge of the first tile group TILE_GR1. 0 To the bit line BL located at the right edge of the first memory tile group TILE_GR1 n-1 The bit lines of the first memory tile group TILE_GR1 in the direction of the first memory tile group TILE_GR2. Sequentially increasing the column address CA[j:0] of the second tile group TILE_GR2 can respectively identify (and sequentially select during the access operation) the bit lines BL located at the center of the second memory tile group TILE_GR2. 0 To the bit line BL located at the right edge of the second memory tile group TILE_GR2 n / 2-1 , and then from the bit line BL located at the left edge of the second memory tile group TILE_GR2 n / 2 To the bit line BL located at the center of the second memory tile group TILE_GR2 n-1 The bit lines of the second tile group TILE_GR2 in the direction of.

[0131] The order of write operations performed in the bit line interleaving operation on different memory slices can be changed. For example, the control circuit 130 can set Fig. 15B The bit line addressing sequence is such that the bit line BL located at the left edge of the second memory tile group TILE_GR2 is 0 To the bit line BL on the right n-1 The write operation is sequentially performed on the second memory tile group TILE_GR2 in the direction of 0 to the bit line BL located at its right edge n / 2-1 , and then from the bit line BL located at its left edge n / 2 to the bit line BL located at its center n-1 In the direction of , write operations are sequentially performed on the first memory tile group TILE_GR1. These write operations can each be associated with a different column address, which causes the selection of the associated one or more bit lines BL by the column decoder. In some examples, multiple bit lines BL can be identified by the same column address, and a single column address CAi can be decoded by the column decoder and cause multiple bit lines to be selected and accessed simultaneously (for example, to provide or obtain corresponding data bits from multiple bit lines at the same time). Some or all of these multiple bit lines (identified by a single column address CAi) can access the same memory tile group TILE.

[0132] Assuming that a relatively high bit error rate occurs in the memory cells on the left and right edges of the first memory tile TILE_GR1 and the second memory tile group TILE_GR2 (compared to the memory cells in the middle), after performing the bit line interleaving operation, the bit errors of the ECC sector can be equalized by mixing the memory cells on the left or right edge with the memory cells at the center. Therefore, the BER imbalance between bit lines can be reduced by applying the bit line interleaving operation to different memory tiles.

[0133] Fig. 15C Shows the same Fig.15A The bit line addressing performed in the bit line interleaving operation is different from the bit line addressing.

[0134] Reference Fig. 15C In the first memory tile group TILE_GR1, the bit line BL 0 It can be one of the bit lines belonging to the resistance distribution degradation area. In the first memory tile group TILE_GR1, the bit line BL 0 It may be a bit line other than the bit line located at the left edge among the bit lines belonging to the resistance distribution deterioration region.

[0135] In the second memory tile group TILE_GR2, the bit line BL 0 It can be one of the bit lines belonging to the normal resistance distribution area. In the second memory tile group TILE_GR2, the bit line BL 0 It may be a bit line other than the bit line located at the center among the bit lines belonging to the normal resistance distribution area.

[0136] When the bit line interleaving operation is performed, the first ECC sub-sector SEC1 may be written to the word line WL connected to the first memory tile group TILE_GR1. i The memory cell connected to the bit line BL 0 To BL 127 The memory cell of the second ECC sub-sector SEC2 can be written to the word line WL connected to the second memory tile group TILE_GR2. i The memory cells are connected to the bit line BL 0 To BL 127 of memory cells.

[0137] Above about FIG. 15A to FIG. 15C The described bit line interleaving operation and address modification can also be used to address resistance distribution degradation that occurs at the vertical edges of each tile. Fig. 9CAn example is shown in which a tile group TILE_GR formed by a two-dimensional array of tiles has a resistance distribution degradation at the vertical edge of each tile TILE (and, as a result, at the vertical edge of each rack BAY). When accessing the memory cells of TILE_GR, after selecting and activating a word line by the row decoder 160 (via decoding the row address RA), the column decoder 170 can select a subset of memory cells connected to the selected word line, such a subset of memory cells being identified and selected by the column address CA. When decoding the column address and selecting the subset of memory cells of the selected word line, one bit line in each rack BAY can be selected and connected to the corresponding I / O line (e.g., using Fig. 15B I / O gating 190) such that in response to each column address (e.g., Fig. 9C 128 bits of TILE_GR), providing one bit per rack BAY. However, when the order of bit line selection is the same (e.g., with respect to the less significant bits of the column address), the column addresses may select bit lines (and therefore memory cells) having the same relative position within each of their corresponding racks. For example, Fig. 9C As shown, column address CAi can select bit lines all located at the left edge of its corresponding rack, while column address CAj can select bit lines all located in the middle of its rack. Therefore, the resulting data bits provided by such access operations can have different resistance degradation rates depending on their column addressing.

[0138] about FIG. 15A to FIG. 15C The column addressing modifications described can be implemented to account for the resulting BER imbalance. Fig.15D As represented, by modifying one of the column addresses CA of one or both of the tile groups TILE_GR1 and TILE_GR2 as described, when a group of bits (e.g., 128 bits of sub-sector SEC1) provided by one of the tile groups (e.g., TILE_GR1) provides bits from memory cells that are all in a corresponding resistance distribution degradation region (e.g., adjacent to a vertical edge of the corresponding tile), the other of the tile groups (e.g., TILE_GR2) will provide a group of bits (e.g., 128 bits of sub-sector SEC2) that are located outside such a resistance distribution degradation region (e.g., in the middle of the corresponding tile). Therefore, the resistance distribution degradation of each codeword (formed by the combined sub-sectors SEC1 and SEC2) will be balanced, resulting in BER balance.

[0139] Fig.16 is a diagram for explaining an ECC interleaving operation according to an embodiment of the inventive concept.

[0140] Reference Fig.16 , you can use Fig.9AThe convolution interleaving operation can be performed by synthesizing the two memory slice groups TILE_GR. Fig.11A The bit line interleaving operation and Fig.15A The ECC sector may include a first ECC sub-sector SEC1 and a second ECC sub-sector SEC2.

[0141] The first ECC sub-sector SEC1 may be provided to write to the first memory tile group TILE_GR1, and the second ECC sub-sector SEC2 may be provided to write to the second memory tile group TILE_GR2. In this case, when performing a convolution interleaving operation, the first ECC sub-sector SEC1 may be written to the word line WL connected to the lower edge of the first memory tile group TILE_GR1. 0 The memory cells are connected to the bit line BL located on the left edge of the first memory tile group TILE_GR1. 0 To BL 127 The second ECC sub-sector SEC2 can be written to the memory cell connected to the word line WL0 located at the center of the second memory slice group TILE_GR2 and the bit line BL located at the center of the second memory slice group TILE_GR2. 0 To BL 127 of memory cells.

[0142] In the convolution interleaving operation with respect to different memory tiles, the word line WL located at the lower edge of the first memory tile group TILE_GR1 may be 0 to the word line WL located at its upper edge n-1 The described write operation is sequentially performed on the first memory tile group TILE_GR1 in the direction of 0 to the bit line BL located at its right edge n-1 The described write operations sequentially performed on the first memory tile group TILE_GR1 in the direction of TILE_GR2 can be combined. In addition, in the word line WL located at the center of the second memory tile group TILE_GR2 0 to the word line WL located at its upper edge n / 2-1 , and then from the word line WL located at its lower edge n / 2 to the word line WL located at its center n-1 The write operation described above is sequentially performed on the second memory slice group TILE_GR2 in the direction of 0 to the bit line BL located at its right edge n / 2-1 , and then from the bit line BL located at its left edge n / 2to the bit line BL located at its center n-1 The described write operations sequentially performed on the second memory tile group TILE_GR2 in the direction of can be combined.

[0143] To perform convolution interleaving operations on different memory slices, Figure 5A and Figure 5B The control circuit 130 may change the word line addressing and the bit line addressing of one or both of the first memory tile group TILE_GR1 and the second memory tile group TILE_GR2.

[0144] When the convolution interleave operation is performed on different memory slices, the bit errors of the ECC sector can be mixed in the memory cells at the lower edge or upper edge and the memory cells at the center, and can be balanced among the ECC sectors. Therefore, by applying the convolution interleave operation, the BER imbalance between word lines and bit lines can be reduced.

[0145] Fig.17A and Fig. 17B is a view for explaining the distribution characteristics of memory cells in a three-dimensional (3D) structure, wherein Figure 6 The memory tiles are stacked vertically. Fig. 17B , the horizontal axis indicates resistance, and the vertical axis indicates the number of memory cells. Figure 6 The memory tiles TILE may be stacked in a height direction. Each of the memory tiles TILEa to TILEd may have a 3D structure in which each memory tile extends in a column direction, a row direction, and a height direction.

[0146] Reference Fig.17A , the stacked memory tiles TILEa to TILEd may be commonly connected to each other via a plurality of global bit lines. The word lines belonging to each of the memory tiles TILEa to TILEd may be independent of the other memory tiles. Figure 3 As shown, the memory cells of each of the memory tiles TILEa to TILEd are expected to have a low resistance state LRS or a high resistance state HRS (relative to the reference resistance RR).

[0147] However, the memory cells of the memory tile TILEd located at the upper end among the stacked memory tiles TILEa to TILEd may show a normal resistance distribution, such as Figure 3 As shown, the memory cell of the memory tile TILEa located at the lower end of the stacked memory tiles TILEa to TILEd may show a widened resistance distribution, as shown in FIG. Fig. 17B shown.

[0148] Reference Fig. 17B, in the direction from the memory tile TILEd located at the upper end to the memory tile TILEa located at the lower end, the resistance of the memory cell decreases and its distribution widens.

[0149] exist Fig. 17B In the resistance distribution of the stacked memory chip TILEa to TILEd, the BER in the direction of the memory chips TILEc, TILEb and TILEa in the lower part can be increased, and the BER of the memory chip TILEd located at the upper end can be reduced. When the memory cells of the memory chip TILEa located at the lower end and having a large BER and the memory cells of the memory chip TILEd located at the upper end and having a small BER are mixed with each other and then store data, the BER imbalance can be reduced. Now, an ECC interleaving operation in which a stacked layer interleaving operation is performed to achieve BER equalization will be described.

[0150] Fig.18 1 is a diagram for explaining an ECC interleaving operation according to an embodiment of the inventive concept. Fig.18 In the embodiment, the memory device 100 may be formed by stacking memory chips. More specifically, the memory device 100 may be formed by stacking stacked layers, wherein each stacked layer includes a two-dimensional array of memory chips (e.g., Fig.9A For ease of illustration, only stacking layers Stack0 and Stack1 are shown, and only two blocks are shown for each of these stacking layers. However, it is contemplated that each stacking layer has additional stacking layers and memory chips. Each stacking layer may include a separate semiconductor chip, wherein the stacks are interconnected by TSVs (through substrate vias). Alternatively, all (or some adjacent) stacking layers of the stacking layer may be formed in a single semiconductor chip.

[0151] Reference Fig.18 , the first memory chip TILEa1 and the second memory chip TILEa2 may be arranged on the first stacking layer Stack0 as the lower part. The third memory chip TILEb1 and the fourth memory chip TILEb2 may be arranged on the second stacking layer Stack1 as the upper part. The third memory chip TILEb1 may be stacked on the first memory chip TILEa1 of the first stacking layer Stack0, and the fourth memory chip TILEb2 may be stacked on the second memory chip TILEa2 of the first stacking layer Stack0.

[0152] In the stacked first to fourth memory tiles TILEa1 to TILEb2, by applying Fig.16 The stacked layer interleaving operation is performed by performing a convolution interleaving operation of the ECC sector. The ECC sector may include a first ECC sub-sector SEC1 and a second ECC sub-sector SEC2.

[0153] The first ECC sub-sector SEC1 may be provided to be written to the first memory tile TILEa1 of the first stack layer Stack0, and the second ECC sub-sector SEC2 may be provided to be written to the fourth memory tile TILEb2 of the second stack layer Stack1. In this case, when the stack layer interleaving operation is performed, the first ECC sub-sector SEC1 may be written to the word line WL connected to the lower edge of the first memory tile TILEa1. 0 The bit line BL connected to the left edge of the first memory tile TILEa1 is connected to the bit line BL connected to the left edge of the first memory tile TILEa1. 0 To BL 127 The memory cell of the second ECC sub-sector SEC2 can be written to the word line WL connected to the center of the fourth memory tile TILEb2. 0 The bit line BL connected to the center of the fourth memory tile TILEb2 is 0 To BL 127 of memory cells.

[0154] In the stacking layer interleaving operation with respect to different memory tiles, the word line WL located on the lower edge of the first memory tile TILEa1 may be 0 to the word line WL located at its upper edge n-1 The write operation may be sequentially performed on the first memory tile TILEa1 in the direction of 0 to the bit line BL located on its right edge n-1 The write operations sequentially performed on the first memory tile TILEa1 in the direction of can be combined. In addition, the write operations can be performed on the word line WL located at the center of the fourth memory tile TILEb2. 0 to the word line WL located at its upper edge n / 2-1 , and then from the word line WL located at its lower edge n / 2 to the word line WL located at its center n-1 The write operation may be sequentially performed on the fourth memory tile TILEb2 in the direction of 0 to the bit line BL located on its right edge n / 2-1 , and then from the bit line BL located at its left edge n / 2 to the bit line BL located at its center n-1 Write operations sequentially performed on the fourth memory tile TILEb2 in the direction of can be combined.

[0155] To perform stacked layer interleaving operations on stacked memory slices, Figure 5A and Figure 5BThe control circuit 130 may change one or both of the word line addressing and the bit line addressing of one or more of the stacked first to fourth memory tiles TILEa1 to TILEb2.

[0156] When a stacked layer interleaving operation is performed on stacked memory slices, the bit errors of the ECC sectors can be mixed in the memory cells of the lower memory slices and the memory cells of the higher memory slices in the stacked memory slices and can be balanced. Therefore, the BER imbalance between the stacked memory slices can be reduced by applying the stacked layer interleaving operation.

[0157] Fig.19 is a schematic diagram of a memory module 1900 including a memory device according to an embodiment of the inventive concept.

[0158] References Fig.19 , the memory module 1900 may include a plurality of memory chips 100_1 to 100_5. The plurality of memory chips 100_1 to 100_5 may be connected to Figure 5A and / or the memory device 100 of 5B are the same or similar.

[0159] The memory module 1900 may include a buffer 1910 that receives and buffers signals from the memory module 1900 via a plurality of signal lines. Figure 8 The buffer 1910 may include an ECC circuit 210, which is connected to the memory controller 200 and provides the commands, addresses and data to the memory chips 100_1 to 100_5. Figure 8 The ECC circuit 210 may be the same or similar to the ECC circuit 210 of the memory module 1900. The ECC circuit 210 may perform ECC encoding on the information data received by the memory module 1900 during a write operation, thereby generating a codeword and providing the codeword to a memory chip selected from the memory chips 100_1 to 100_5. The ECC circuit 210 may perform ECC decoding on the read data (i.e., codeword) received from the memory chips 100_1 to 100_5 selected during a read operation, thereby generating information data and providing the information data to the memory controller 200.

[0160] Buffer 1910 can be referenced above Fig.13 The word line interleaving operation is performed on different memory chips described. Fig.13, the first memory slice group TILE_GR1 may belong to the first memory chip 100_1, and the second memory slice group TILE_GR2 may belong to the second memory chip 100_2. The buffer 1910 may provide the first ECC sub-sector SEC1 of the ECC sector to be written to the first memory slice group TILE_GR1 of the first memory chip 100_1, and provide its second ECC sub-sector SEC2 to be written to the second memory slice group TILE_GR2 of the second memory slice 100_2 at different relative positions in the first slice group TILE_GR1 and the second slice group TILE_GR2, as described herein.

[0161] In the word line interleaving operation with respect to different memory chips, the word line WL located at the lower edge of the first memory tile group TILE_GR1 may be interleaved. 0 to the word line WL located at its upper edge n-1 A write operation is sequentially performed on the first memory chip 100_1 in the direction of the first memory tile group TILE_GR1, and a write operation may be performed on the word line WL located at the center of the second memory tile group TILE_GR2. 0 to the word line WL located at its upper edge n / 2-1 , and then from the word line WL located at its lower edge n / 2 to the word line WL located at its center n-1 Instead of or in addition to the word line interleaving operation, an interleaving operation other than the word line interleaving operation may be performed on the first and second tile groups TILE_GR1 and TILE_GR2 in the manner described herein.

[0162] Fig. 20 is a flowchart of an ECC interleaving operation of a memory device according to an embodiment of the inventive concept.

[0163] Reference Fig. 20 In operation S2000, when the conduction Figure 5A When the memory device 100 is connected to the memory device 100, the ECC interleaving operation starts. For example, the ECC interleaving operation may start during a write operation or a read operation of the memory device 100. According to an embodiment, the ECC interleaving operation may start according to the above reference. FIG. 10A to FIG. 10C , FIG. 14A to FIG. 14C ,as well as Fig.17A and Fig. 17B The ECC interleaving operation begins with the described distribution characteristics of the memory cells in the memory cell array 110.

[0164] When the ECC interleaving operation starts in operation S2000, the memory device 100 may receive a codeword including information data and one or more parity bits of the information data, and may divide the ECC sector including the information data into at least two ECC sub-sectors. When the memory device 100 writes the codeword into a memory cell array including a plurality of memory slices, the memory device 100 may write one of the at least two ECC sub-sectors into a first memory area having a high BER and may write another of the at least two ECC sub-sectors into a second memory area having a low BER. The memory device 100 may perform one or more of a word line interleaving operation (S2010), a body interleaving operation (S2020), a bit line interleaving operation (S2030), a convolution interleaving operation (S2040), and a stacked layer interleaving operation (S2050), so that the BER of the plurality of information data written into the first memory area and the second memory area can be balanced. The ECC interleaving operation may include a chip interleaving operation (S2060).

[0165] In the word line interleaving operation ( S2010 ), information data may be mixed in memory cells of a first word line connected to a first memory area and memory cells of a second word line connected to a second memory area, and may be written therein (eg, in the manner described herein).

[0166] In the body interleaving operation (S2020), information data can be mixed in memory cells connected to a first word line of a first body arranged close to a peripheral circuit block among multiple bodies, and in memory cells connected to a second word line of a second body arranged away from the peripheral circuit block among multiple bodies, and can be written therein (for example, in the manner described herein).

[0167] In the bit line interleaving operation (S2030), information data may be mixed in memory cells connected to the first word line and the first bit line of the first storage area and memory cells connected to the second word line and the second bit line of the second storage area, and may be written therein (e.g., in the manner described herein).

[0168] In the convolution interleaving operation (S2040), information data may be mixed in memory cells connected to a first word line and a first bit line of a first storage region and memory cells connected to a second word line and a second bit line of a second storage region, and may be written therein (e.g., in the manner described herein).

[0169] In the stacked layer interleaving operation (S2050), information data may be mixed in memory cells of a first word line and a first bit line connected to a lower memory slice in the stack, and in memory cells of a second word line and a second bit line connected to an upper memory slice in the stack, and may be written thereto (e.g., in the manner described herein).

[0170] In a chip interleaving operation (S2060), information data may be mixed in memory cells of a first word line connected to a first memory chip among multiple memory chips and in memory cells of a second word line connected to a second memory chip among multiple memory chips, and may be written thereto (e.g., in the manner described herein).

[0171] Fig.21 is a block diagram of a system 2100 according to an embodiment of the inventive concept.

[0172] Reference Fig.21 , the system 2100 may include a processing unit 2110, a volatile memory unit 2120, a resistive memory unit 2130, and a large-capacity memory unit 2140. As described herein, the resistive memory unit 2130 may include one or more of the memory device 100 and / or the memory module 1900. The system 2100 may be a general-purpose or special-purpose computer system, such as a mobile device, a personal computer, a server computer, a programmable home appliance, or a mainframe computer.

[0173] The processing unit 2110 may run an operating system and a plurality of software systems and perform certain calculations or tasks. The processing unit 2110 may be a microprocessor, a central processing unit (CPU), and / or a graphics processing unit (GPU).

[0174] The volatile memory unit 2120 is used as an operation memory or a cache memory of the system 2100 and refers to a medium for storing data in a short-term or temporary manner. The volatile memory unit 2120 may include at least one memory device (eg, DRAM, SRAM).

[0175] The resistance memory unit 2130 may be used as a cache memory of the large-capacity memory unit 2140. The resistance memory unit 2130 may store a portion of data of an application or an operating system that is frequently accessed. The resistance memory unit 2130 may include at least one memory device (e.g., a PRAM). Since data access to the resistance memory unit 2130 is much faster than data access to the large-capacity memory unit 2140 such as a hard disk drive (HDD), the resistance memory unit 2130 may be used as a cache memory. The resistance memory unit 2130 may be implemented as described above with respect to Figures 1 to 20One or more embodiments described.

[0176] When the resistance memory unit 2130 writes the codeword including the information data and the parity bit of the information data into the memory cell array, the resistance memory unit 2130 may divide the ECC sector including the information data into the first ECC sub-sector and the second ECC sub-sector, and may perform an ECC interleaving operation so that the first ECC sub-sector is written into the first storage area with a high BER, and the second ECC sub-sector is written into the second storage area with a low BER. Therefore, the respective BERs of the plurality of pieces of information data written into the first storage area and the second storage area may be equalized.

[0177] The mass storage unit 2140 may be implemented as a HDD, a solid state drive (SDD), a peripheral component interconnect express (PCIe) memory module, a non-volatile memory express (NVMe), etc. Optionally, one or more layers of the mass storage unit 2140 may be implemented as one or more network accessible devices and / or services, such as NVMe-over Fabrics (NVMe-oF) and / or remote direct memory access (RDMA) connected clients, servers, server farms, server clusters, application servers, or message servers. The mass storage unit 2140 refers to a storage medium used when the system 2100 stores user data for a long time. The mass storage unit 2140 may store applications, program data, etc.

[0178] While the inventive concept has been particularly shown and described with reference to embodiments thereof, it will be understood that various changes in form and details may be made therein without departing from the spirit and scope of the appended claims.

Claims

1. A resistive memory device, include: a plurality of banks, each bank comprising a bank row decoder connected to a plurality of word lines of the bank, a bank column decoder connected to a plurality of bit lines of the bank, and a bank array of memory cells having memory cells arranged in a plurality of rows and columns, each row of memory cells connected to a corresponding one of the word lines, and each column of memory cells connected to a corresponding one of the bit lines, the memory cells of each bank being organized into a plurality of slices, each of the memory cells comprising a variable resistor programmable to store at least one bit of data; an address register configured to provide a sequence of memory addresses, each memory address comprising a bank address identifying one of a plurality of banks, a row address identifying a wordline location in the bank identified by the bank address, and a column address identifying a subset of bitlines connected to the wordlines identified by the row address; control circuitry configured to provide each row address and column address of each memory address stored by the address register to a bank row decoder and a column decoder, respectively, of a bank identified by a bank address of the corresponding memory address; wherein, for at least each memory address of the subset of memory addresses, the control circuit is configured to modify at least one of a corresponding row address and a corresponding column address such that at least one of the following is true: a relative position of a first word line identified by a first row address in a first bank of the plurality of banks is different from a relative position of a second word line identified by a first row address in a second bank of the plurality of banks, and a relative position of a first bit line identified by a first column address in a first bank of the plurality of banks being different from a corresponding relative position of a second bit line identified by a first column address in a second bank of the plurality of banks; wherein the first word line is located in a first region of the first body, the second word line is located in a second region of the second body, and The first area is located at the edge of the first body, and the second area is located at the center of the second body.

2. The resistive memory device according to claim 1, in, For a subset of memory addresses, the control circuit is configured to modify corresponding row addresses so that a relative position of a first word line identified by a first row address in a first bank of the plurality of banks is different from a relative position of a second word line identified by a first row address in a second bank of the plurality of banks.

3. The resistive memory device according to claim 2, in, The first region includes memory cells having higher resistance distribution degradation than memory cells of the second region.

4. The resistive memory device according to claim 1, in, The first region is located at an edge of a horizontal block of memory tiles of the first bank through which the first word line extends, and the second region is located at a center of a horizontal block of memory tiles of the second bank through which the second word line extends.

5. The resistive memory device according to claim 2, in, With respect to the first bank, a sequence of row addresses in numerical order from a lowest row address of the sequence to a highest row address of the sequence sequentially identifies word lines from a first edge of the first bank to a second edge of the first bank, and Wherein, with respect to the second body, a sequence of row addresses in numerical order from the lowest row address of the sequence to the highest row address of the sequence sequentially identifies, in positional order, word lines from the center of the second body to the first edge of the second body, and then from the second edge of the first body to the center of the second body.

6. The resistive memory device according to claim 2, in, With respect to the first bank, a sequence of row addresses in numerical order ranging from a lowest row address of the sequence to a highest row address of the sequence sequentially identifies, in positional order, word lines arranged from a first edge of a first slice of the first bank to a second edge of the first slice of the first bank, and Wherein, with respect to the second body, a sequence of row addresses in numerical order of a sequence from a lowest row address of the sequence to a highest row address of the sequence sequentially identifies, in positional order, word lines from a center of a first piece of the second body to a first edge of a first piece of the second body, and then from a second edge of the first piece of the second body to a center of the first piece of the second body.

7. The resistive memory device according to claim 2, in, The control circuitry includes address modification circuitry configured to invert at least one more significant bit in all row addresses of a subset of the memory addresses.

8. A resistive memory device, include: a plurality of banks, each bank comprising a bank row decoder connected to a plurality of word lines of the bank, a bank column decoder connected to a plurality of bit lines of the bank, and a bank array of memory cells having memory cells arranged in a plurality of rows and columns, each row of memory cells connected to a corresponding one of the word lines, each column of memory cells connected to a corresponding one of the bit lines, the memory cells of each bank being organized into a plurality of slices, each of the memory cells comprising a variable resistor programmable to store at least one bit of data; an address register configured to provide a sequence of memory addresses, each memory address comprising a bank address identifying one of the plurality of banks, a row address identifying a wordline location in the bank identified by the bank address, and a column address identifying a subset of bitlines connected to the wordline identified by the row address; control circuitry configured to provide each row address and column address of each memory address stored by the address register to a bank row decoder and a column decoder, respectively, of a bank identified by a bank address of the corresponding memory address; wherein, for at least each memory address of the subset of memory addresses, the control circuit is configured to modify at least one of a corresponding row address and a corresponding column address such that at least one of the following is true: a relative position of a first word line identified by a first row address in a first bank of the plurality of banks is different from a relative position of a second word line identified by a first row address in a second bank of the plurality of banks, and a relative position of a first bit line identified by a first column address in a first bank of the plurality of banks is different from a corresponding relative position of a second bit line identified by a first column address in a second bank of the plurality of banks, wherein, for a subset of the memory addresses, the control circuit is configured to modify corresponding column addresses so that a relative position of a first bit line identified by a first column address in a first bank of the plurality of banks is different from a corresponding relative position of a second bit line identified by a first column address in a second bank of the plurality of banks; and The first bit line is located at an edge of the first body, and the second bit line is located at a center of the second body.

9. The resistive memory device according to claim 8, in, Memory cells connected to the first bit line have higher resistance distribution degradation than memory cells connected to the second bit line.

10. The resistive memory device according to claim 9, in, The sheets of each body are arranged in columns to form a corresponding sheet rack, The first bit lines are arranged at corresponding edges of corresponding frames of the first body, and the second bit lines are arranged at the center of corresponding frames of the second body.

11. A method of storing data, include: forming a first codeword from a first data block; The first codeword is stored in a resistive memory device, so that a first part of the first codeword is stored in a first resistive memory cell of the resistive memory device, and a second part of the first codeword is stored in a second resistive memory cell of the resistive memory device, wherein the storing the first codeword in the resistive memory device comprises: sending a write command to the resistive memory device to store the first codeword in the resistive memory device, sending a first memory address associated with the write command to the resistive memory device, the memory address comprising a first row address and a first column address, and sending the first codeword to a resistive memory device, wherein the first resistive memory cell is identified by a first row address and a first column address in a first body of the resistive memory device, wherein the second resistive memory cell is identified by a first row address and a first column address in a second body of the resistive memory device, and wherein the first resistive memory cell has a first relative position with respect to the first body, wherein the second resistive memory cell has a second relative position with respect to the second body, wherein the first relative position is different from the second relative position; wherein each of the first body and the second body comprises a plurality of resistive memory cells organized in a plurality of slices, the plurality of slices of each body being organized in a plurality of slice racks, wherein a first word line of the first bank identified by a first row address in the first bank is located in a first sheet of the first bank at a first relative position with respect to the first sheet, wherein a second word line of the second bank identified by a first row address in the second bank is located in the second sheet of the second bank at a second relative position with respect to the second sheet, and The first relative position is located near an edge of the first sheet, and the second relative position is located at a center of the second sheet.

12. The method according to claim 11, in, The first resistance type memory cell is connected to the first word line, the first word line is formed at the first relative position with respect to the first body, and The second resistive memory cell is connected to the second word line, and the second word line is formed at the second relative position with respect to the second body.

13. The method according to claim 11, in, The first resistive memory cell is connected to a first set of bit lines of the first bank identified by a first column address in the first bank, and The second resistive memory cell is connected to a second group of bit lines of the second bank identified by a first column address in the second bank.

14. The method according to claim 13, in, The first group of bit lines are respectively located in corresponding racks at edges of corresponding racks of the first body, and Wherein, the second group of bit lines is located at the center of the corresponding rack of the first body.

15. The method of claim 11, further comprising: include: Receive information data from the host; Dividing the information data into data blocks, including the first data block; performing an error correction encoding operation on each data block to generate a plurality of corresponding codewords, the plurality of codewords including a first codeword, each codeword including a corresponding one of the data blocks and a corresponding parity bit generated from a corresponding error correction encoding operation; as well as storing the plurality of codewords in a resistive memory device such that a first portion of each of the plurality of codewords is stored in a first body of the resistive memory device and a second portion of each of the plurality of codewords is stored in a second body, For each of the plurality of codewords, a relative position of a resistive memory cell in the first body storing a first portion of the codeword is different from a relative position of a resistive memory cell in the second body storing a second portion of the codeword.

16. The method of claim 15, in, The first portion and the second portion of each of the plurality of codewords are identified by the same corresponding row address and the same corresponding column address.