An optimized read reference voltage sensing method based on symmetry of threshold voltage distribution of 3D NAND flash memory

Abstract

The present invention proposes an optimized read reference voltage sensing method based on the symmetry of the threshold voltage distribution of 3D NAND flash memory. First, a pair of adjacent states V OPT perception, assuming the scanning range is [V L , V R ], read the pair of adjacent states at V OPT The OCD curve near the center is set as a discriminant group with three sampling points, and the OCD curve is searched by sliding. For each discriminant group, after obtaining the OCD values ​​of the three sampling points, the discriminant γ of the discriminant group is calculated, which is used to measure the degree to which the left and right sampling points are centrally symmetric about the center point. The discriminant group sliding scan with double-level precision is used to first perform a coarse precision scan to determine V OPT The offset range is then scanned with fine precision to obtain V OPT The present invention can effectively overcome the influence of noise on the threshold voltage distribution of small-granularity storage sample volume on the optimized reference voltage perception; the dual-level precision scanning method is adopted, which only requires a few read sampling operations and introduces less read overhead.

Description

Technical Field

[0001] The present invention belongs to the technical field of solid-state storage read performance, and in particular, relates to an optimized read reference voltage sensing method based on the symmetry of 3D NAND flash memory threshold voltage distribution. Background Art

[0002] The current mainstream 3D NAND flash memory usually supports modifying the read reference voltage to reduce the bit errors in the stored data. Flash memory realizes non-volatile storage by injecting charge into the cell. For the write operation, that is, injecting charge into the cell, the cell will form a threshold voltage after the charge is injected. The internal circuit of the flash memory divides different threshold voltage intervals into different states with the read reference voltage (RRV). The encoding of the state of the cell is the information stored in the cell. For the read operation, the state of the cell is identified, and then the corresponding encoding information is obtained. The erase operation is to clear the charge in the cell. Figure 1 The figure shows a schematic diagram of the state encoding and threshold voltage distribution in a 3-bit / cell flash memory. If the information to be stored in a cell is "101", the internal circuit of the flash memory will inject charge into it and make the threshold voltage in the C state interval divided by the read reference voltage Vc and Vd. When the read operation is performed, the cell will be first identified as the C state, and then the 3-bit information "101" stored in it will be output. When the erase operation is performed, the cell's charge is removed, causing the threshold voltage to drop to the ER state.

[0003] In the actual data storage process, the amount of charge in the cell will change. For example, the charge in the cell will leak slowly during the dwell process. The change in the amount of charge in the cell may cause the state of its threshold voltage to flip, thus causing errors in the stored information. Figure 2 As shown in the figure, the state on the right leaks charge during data storage, causing the threshold voltage distribution to shift left and exceed the default left boundary read reference voltage (default voltage, V default ), and caused a large number of read errors. Now mainstream flash memory chips usually support modifying the read reference voltage between adjacent states, such as Figure 2 As shown, the read reference voltage is modified to the optimized read reference voltage (opimized voltage, V OPT ), which will effectively reduce the bit error introduced by the threshold voltage distribution drift. Summary of the invention

[0004] Based on the above problems, the present invention proposes an optimized read reference voltage perception method based on the symmetry of the threshold voltage distribution of 3D NAND flash memory. For small-granularity read samples, the proposed strategy effectively utilizes the symmetry of the threshold voltage distribution near the optimized read reference voltage, which can effectively alleviate the impact of noise in the threshold voltage distribution on the optimized read reference voltage perception, and achieve accurate perception of the optimized read reference voltage with less read overhead and complexity.

[0005] The present invention is achieved through the following technical solutions:

[0006] An optimized read reference voltage sensing method based on the symmetry of 3D NAND flash memory threshold voltage distribution:

[0007] The method specifically comprises the following steps:

[0008] Step 1: First, execute a pair of adjacent states V OPT perception, assuming the scanning range is [V L ,V R ], read the pair of adjacent states at V OPT The nearby OCD curves are set as three sampling points as a discrimination group, and the OCD curves are searched by sliding;

[0009] Step 2: For each discriminant group, after obtaining the OCD values ​​of the three sampling points, the discriminant γ of the discriminant group is calculated, which is used to measure the degree to which the left and right sampling points are centrally symmetric about the center point;

[0010] Step 3: Use the double-level precision discrimination group sliding scan to first perform a coarse precision scan to determine V OPT The offset range is then scanned with fine precision to finally obtain V OPT The offset value.

[0011] Furthermore, in step one,

[0012] Each discriminant group is composed of OCD obtained by three samplings, which are respectively called OCD left , OCD right and OCDcenter, assuming the offset difference between the three sampling points is ΔV, which satisfies:

[0013] ΔV=s·ΔV min #(7)

[0014] Among them, s is the minimum number of steps contained in the left and right intervals of each discrimination group.

[0015] Furthermore, in step 2,

[0016] The discriminant γ is defined as follows:

[0017] γ=|OCD left +OCD right -2×OCD center |#(8)

[0018] In [V L ,V R ], identify the group with the minimum value of γ, and consider the offset value of the center point in this group to be V OPT ;

[0019] That is, suppose that during the sliding scan of a certain stored data sample, the kth discriminant group γ k Take the minimum value, then the left and right endpoint offset values ​​of the discrimination group are V Lk and V Rk , then the V of the stored data sample OPT =V OPT =V Lk +ΔV=V Rk -ΔV#(9).

[0020] Furthermore, in step three, the following steps are specifically included:

[0021] S1. Coarse precision scanning and optimizing the 1s count distribution near the reference voltage in the form of central symmetry discrimination groups, each discrimination group consists of three large-step equally spaced sampling points;

[0022] S2, evaluate the central symmetry of each discrimination group, and obtain the group number with the best central symmetry and the read reference voltage offset of its left endpoint;

[0023] S3, obtaining a fine-precision scanning interval by using the read reference voltage offset determined by S2;

[0024] S4, again performing a fine-precision sliding scan of the interval determined in S3 in the form of a central symmetry discrimination group;

[0025] S5, evaluating the central symmetry degree of each discrimination group, and obtaining the serial number of the group with the best central symmetry degree;

[0026] S6, taking the center point of the discrimination group determined in S5 as the reference voltage offset as the sensed V OPT .

[0027] Furthermore, in step three,

[0028] First, set the left boundary points of the judgment group to V L 、V L +ΔV, ..., V R -2ΔV to determine V with coarse precision OPTIf the offset range is within, the number of discrimination groups N1 that needs to be obtained is:

[0029]

[0030] Furthermore, in step three,

[0031] After completing the coarse-precision scan, a total of N1 γ values ​​are obtained. Next, a fine-precision scan is performed. Let the discriminant γ of the i-th group be i Take the minimum value and the offset value of its left boundary point yes:

[0032]

[0033] It can be seen that if the offset value of the center point of a certain discrimination group is V in a single scan of the fine Jing OPT , then the range of the left boundary point offset value of the group is The minimum step ΔV of the discriminant group shifted by RRV min To perform a scan, the number of times a fine-precision scan needs to be performed is:

[0034]

[0035] After completing the fine-precision scan, a total of N2 γ values ​​are obtained, plus the offset obtained from the coarse-precision scan. and The two γ values ​​at the position constitute an N2+2 bit sequence, numbered from 0 to N2-1. Let the discriminant γ of the jth one be j Take the minimum value, then the V obtained after fine-grained scanning is OPT The offset value is:

[0036]

[0037] Furthermore, in step three,

[0038] The total number of read samples performed during the dual-level precision scan is:

[0039]

[0040] Right now:

[0041]

[0042] To obtain V OPT True value, need to use the minimum step ΔV min Yes [V L ,V R ]To scan the entire range, the number of reads required is:

[0043]

[0044] An optimized read reference voltage sensing device based on the symmetry of the threshold voltage distribution of 3D NAND flash memory:

[0045] The device comprises: a perception module, a discrimination module and a scanning module;

[0046] The perception module is used to first perform a pair of adjacent states V OPT perception, assuming the scanning range is [V L ,V R ], read the pair of adjacent states at V OPT The nearby OCD curves are set as three sampling points as a discrimination group, and the OCD curves are searched by sliding;

[0047] A discriminant module is used to calculate the discriminant γ of each discriminant group after obtaining the OCD values ​​of the three sampling points, which is used to measure the degree to which the left and right sampling points are centrally symmetric about the center point;

[0048] The scanning module is used to use a dual-level precision discrimination group sliding scan, first performing a coarse precision scan to determine V OPT The offset range is then scanned with fine precision to finally obtain V OPT The offset value.

[0049] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0050] A computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the steps of the above method are implemented.

[0051] Beneficial effects of the present invention

[0052] The present invention obtains the optimized reference voltage by sensing the central symmetric point of the 1s count distribution on the target page, which can effectively overcome the influence of noise on the threshold voltage distribution of small-granularity storage sample quantity on the perception of the optimized reference voltage;

[0053] The present invention adopts a dual-level precision scanning method, which requires only a few reading sampling operations and introduces less reading overhead;

[0054] The present invention only includes simple algebraic and logical judgment operations for processing read samples, uses three large-step equally spaced sampling points as a discrimination group, evaluates the degree of central symmetry of each discrimination group, and uses the central point of the discrimination group with the strongest central symmetry as the perceived optimized read reference voltage, thereby achieving low implementation complexity. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 A schematic diagram of threshold voltage distribution of an existing 3D TLC NAND flash memory;

[0056] Figure 2 A schematic diagram of reducing the error rate of read data by modifying the read reference voltage between adjacent states;

[0057] Figure 3 Schematic diagram of obtaining TVD by retrieving the change of 1s count on the target page;

[0058] Figure 4 The following are example graphs of OCD and TVD on real chips, where (a) is an OCD curve graph between states B and C obtained on a real chip, and (b) is a TVD curve graph between states B and C obtained on a real chip;

[0059] Figure 5 Comparison of the VOPT perceived by the minimum value and the true VOPT in the TVD of block granularity and layer granularity of the present invention, where (a) is the block granularity and (b) is the layer granularity;

[0060] Figure 6 The present invention is based on TVD / OCD symmetry perception V OPT strategy, where (a) is TVD symmetry perception V OPT strategy, (b) is OCD symmetry perception V OPT strategy;

[0061] Figure 7 It is the perception strategy of the central symmetric point on the OCD curve of the present invention;

[0062] Figure 8 The central symmetry evaluation sequence obtained during the sliding scan for the discriminant group;

[0063] Fig. 9 VOPT perception strategy for dual-level precision scanning;

[0064] Fig.10 A schematic diagram of the optimized read reference voltage sensing method proposed by the present invention;

[0065] Fig.11 The experimental results of the proposed method and the traditional method for layer-by-layer perception of VOPT between B and C states are shown in FIG. DETAILED DESCRIPTION

[0066] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0067] Combination Figures 1 to 11 .

[0068] Reasonable setting of the read reference voltage can effectively reduce the hidden dangers of threshold voltage distribution drift on storage reliability. The following describes how to configure V OPT The principle is that if the threshold voltage distribution of a pair of adjacent states satisfies f(v) and g(v) respectively, then the optimized reference voltage V OPT satisfy:

[0069] f(V OPT )=g(V OPT )#(1)

[0070] In order to find this point, it is necessary to obtain the distribution of the two adjacent states at the overlap, which is not easy to obtain without the original written data file during the data storage process. OPT Nearby, f(v) and g(v) are roughly symmetrically distributed, so V' OPT Replace V OPT , V' OPT satisfy:

[0071] f′(V′ OPT )=-g′(V′ OPT )#(2)

[0072] That is, the points where the two distribution derivatives are opposite to each other also satisfy:

[0073] (f(V′ OPT )+g(V′ OPT ))′=0#(3)

[0074] Assume h(v) = f(v) + g(v), that is, h(v) is the distribution after the adjacent states are superimposed, so V' OPT satisfy:

[0075] h′(V′ OPT )=0#(4)

[0076] That is V' OPT is the point where the derivative of h(v) is 0, and also the point where h(v) is in V OPTThe minimum value point near h(v) is easy to obtain during data storage, and it is also called threshold voltage distribution (TVD). The following describes how to obtain it.

[0077] Target page: The current mainstream 3D NAND flash memory chip uses Gray code to encode each state, that is, the encoding of each pair of adjacent states differs by only one bit. The target page is the page type where the different encoding bits are located, such as Figure 3 As shown in the figure, when the RRV between a pair of adjacent states is modified and two read operations are performed, a certain number of state flips will occur between the two read data, and the type is Y→X. Specifically, the cells in the RRV offset interval will be identified as the Y state in the first read and as the X state in the second read. In theory, it can be considered that for the page type where the bits with the same encoding of these two states are located, the 1s count on it will not change, and the same number of bit flips as the state flips will occur on the target page. For example, assuming that the adjacent states are B(001) and C(101), after the first read, the RRV between B and C is shifted to the right and a second read is performed. Then, by comparing the two data, a certain amount of C→B state flips can be retrieved. For the low page and the middle page, the 1s counts thereon will theoretically not change, because the bit flip types caused by C→B on the low page and the middle page are 1→1 and 0→0, respectively, while a decrease in the number of 1s will be retrieved on the high page, because the bit flip type caused by C→B on the high page is 1→0. In summary, in theory, it can be considered that as long as the change in the 1s count on the target page after the RRV modification is retrieved, the number of superimposed cells of the adjacent states in the modification interval can be obtained.

[0078] To obtain the TVD of a pair of adjacent states at the overlap, the operation to be performed is to continuously modify the RRV in between and count the 1s count changes on the target page. After the scanning operation is completed, the 1s count distribution (OCD) within the scanning range can be obtained. According to formula (5) or formula (6), the OCD is subtracted item by item to obtain the TVD after the two states are superimposed. In formulas (5) and (6), v is the RRV offset value (offset) used for scanning, and ΔV min is the minimum step value of the internal circuit offset RRV of the flash memory. The specific formula to be selected depends on the change of the 1s count on the target page during the scanning process. If the TVD between the A and B states is scanned from left to right (1s increasing), then formula (5) is selected. If the TVD between the B and C states is scanned from left to right (1s decreasing), then formula (6) is selected.

[0079] TVD(v)=OCD(v+ΔV min )-OCD(v)#(5)

[0080] TVD(v)=OCD(v)-OCD(v+ΔV min )#(6)

[0081] Figure 4 (a) and (b) are OCD and TVD curves between states B and C obtained on a real chip, respectively. The storage sample has undergone 5k wear operations and 1 year of retention loss at room temperature in advance, and the written data is a random pattern.

[0082] For block-sized read samples (block is the smallest unit of flash memory read, write and erase), the noise on the TVD is weak, which is reflected as a relatively smooth curve, such as Figure 5 (a) shows that, therefore, at this time, the minimum point on the curve can be directly obtained to perceive V OPT ,exist Figure 5 In (a), the real V OPT It is 22 ΔV to the left. min By searching for the TVD minimum value, V OPT It is 21 ΔV to the left. min At the same time, the two are relatively close, but for the layer-granularity read samples (a layer is the basic physical unit of stacking in the 3D flash memory process, and a block is composed of several layers stacked), the TVD will show strong noise, such as Figure 5 (b) shows the TVD scanning result of a layer. At this time, if we sense V by obtaining the minimum point on the curve, OPT , local noise may affect V OPT The perception of Figure 5 In (b), the real V OPT It is 23 ΔV to the left. min By searching for the TVD minimum value, V OPT It is 15 ΔV to the left. min There is a large error between the two.

[0083] From the above, we can see that for small sample readings, due to the large noise on the TVD, if the sampling point only contains a ΔV min If the TVD is within the range, the noise may affect V OPT The solution proposed by this method is that in the processing of reading samples, the sampling point should contain multiple ΔV min In addition, in the process of reading the flash memory, the original information obtained is OCD, and TVD needs to be obtained by the difference between two consecutive sample points of OCD, which will introduce additional reading overhead and increase the complexity of sample processing. Therefore, the present invention proposes a V based on OCD.OPT Perception method.

[0084] like Figure 5 As shown, the traditional V OPT The perception method is based on obtaining the minimum value point of TVD. The perception method proposed in the present invention is based on the TVD in V OPT Symmetry near Figure 6 As shown in (a), TVD is at V OPT The interval near V is roughly mirror-symmetric. From equations (5) and (6), we can see that OCD is OPT The interval near is roughly symmetrical to the center, so V OPT The perception problem is transformed into the perception problem of the central symmetric point of the OCD curve.

[0085] An optimized read reference voltage sensing method based on the symmetry of 3D NAND flash memory threshold voltage distribution:

[0086] The method specifically comprises the following steps:

[0087] Step 1: First, execute a pair of adjacent states V OPT perception, assuming the scanning range is [V L ,V R ], read the pair of adjacent states at V OPT The nearby OCD curves are set as three sampling points as a discrimination group, and the OCD curves are searched by sliding;

[0088] Step 2: For each discriminant group, after obtaining the OCD values ​​of the three sampling points, the discriminant γ of the discriminant group is calculated, which is used to measure the degree to which the left and right sampling points are centrally symmetric about the center point;

[0089] Step 3: Use the double-level precision discrimination group sliding scan to first perform a coarse precision scan to determine V OPT The offset range is then scanned with fine precision to finally obtain V OPT The offset value.

[0090] First, V can be preliminarily determined by using large steps OPT The red dotted line is in the range, and then the range is scanned point by point in small steps to determine V OPT In the interval delimited by the blue dotted line, the reading overhead caused by scanning with a small step in the interval delimited by the dotted line can be avoided, and at the same time, the V OPT perception accuracy.

[0091] In step one,

[0092] Each discriminant group is composed of OCD obtained by three samplings, which are respectively called OCDleft , OCD right and OCDcenter, assuming the offset difference between the three sampling points is ΔV, which satisfies:

[0093] ΔV=s·ΔV min #(7)

[0094] Wherein, s is the minimum number of steps included in the left and right intervals of each discrimination group. In this embodiment, the value of s is 16.

[0095] In step 2,

[0096] The discriminant γ is defined as follows:

[0097] γ=|OCD left +OCD right -2×OCD center |#(8)

[0098] In [V L ,V R ], identify the group with the minimum value of γ, and consider the offset value of the center point in this group to be V OPT ;

[0099] That is, suppose that during the sliding scan of a certain stored data sample, the kth discriminant group γ k Take the minimum value, then the left and right endpoint offset values ​​of the discrimination group are V Lk and V Rk , then the V of the stored data sample OPT for:

[0100] V OPT =V Lk +ΔV=V Rk -ΔV#(9).

[0101] In addition, γ in V OPT The vicinity shows good smoothness, such as Figure 8 As shown, there is no obvious noise on the distribution curve of γ, only, this is due to the fact that each γ is 32 ΔV min The comprehensive treatment results of TVD in the interval effectively alleviate the ΔV min The local noise of TVD in the interval has an influence on V OPT The impact of perception.

[0102] In summary, the proposed perception method can effectively overcome the problem of large noise on TVD in small-granularity storage samples. However, if V OPT Although accurate V OPT , but also introduces larger read latency;

[0103] In order to further reduce the reading overhead of the perception method, the present invention adopts a dual-level precision discrimination group sliding scanning design, the general principle of which is as follows: Fig. 9 shown.

[0104] In step three, the following steps are specifically included:

[0105] S1. Coarse precision scanning and optimizing the 1s count distribution near the reference voltage in the form of central symmetry discrimination groups, each discrimination group consists of three large-step equally spaced sampling points;

[0106] S2, evaluate the central symmetry of each discrimination group, and obtain the group number with the best central symmetry and the read reference voltage offset of its left endpoint;

[0107] S3, obtaining a fine-precision scanning interval by using the read reference voltage offset determined by S2;

[0108] S4, again performing a fine-precision sliding scan of the interval determined in S3 in the form of a central symmetry discrimination group;

[0109] S5, evaluating the central symmetry degree of each discrimination group, and obtaining the serial number of the group with the best central symmetry degree;

[0110] S6, taking the center point of the discrimination group determined in S5 as the reference voltage offset as the sensed V OPT Optimize read reference voltage.

[0111] In step three,

[0112] The dual-level precision perception method is explained in detail. First, the left boundary points of the judgment group are set to V L 、V L +ΔV, ..., V R -2ΔV to determine V with coarse precision OPT If the offset range is within, the number of discrimination groups N1 that needs to be obtained is:

[0113]

[0114] After completing the coarse-precision scan, a total of N1 γ values ​​are obtained. Next, a fine-precision scan is performed. Let the discriminant γ of the i-th group be i Take the minimum value and the offset value of its left boundary point yes:

[0115]

[0116] It can be seen that if the offset value of the center point of a certain discrimination group is V in a single scan of the fine Jing OPT , then the range of the left boundary point offset value of the group is The minimum step ΔV of the discriminant group shifted by RRV min To perform a scan, the number of times a fine-precision scan needs to be performed is:

[0117]

[0118] After completing the fine-precision scan, a total of N2 γ values ​​are obtained, plus the offset obtained from the coarse-precision scan. and The two γ values ​​at the position constitute an N2+2 bit sequence, numbered from 0 to N2-1. Let the discriminant γ of the jth one be j Take the minimum value, then the V obtained after fine-grained scanning is OPT The offset value is:

[0119]

[0120] The total number of read samples performed during the dual-level precision scan is:

[0121]

[0122] Right now:

[0123]

[0124] To obtain V OPT True value, need to use the minimum step ΔV min Yes [V L ,V R ]To scan the entire range, the number of reads required is:

[0125]

[0126] In this test, L takes 96 minimum steps ΔV min , the value of s is 16, and it is substituted into equation (15) and equation (16) respectively, then it can be seen that the number of readings of the proposed perception method and the standard perception method are 38 and 97 times respectively, that is, the reading overhead is reduced by about 61%. In addition, it can be seen from the above description that the processing process of the sample by the proposed method only includes simple algebraic and logical judgment operations, so it is more convenient to use in actual scenarios.

[0127] In summary, Fig.10 The schematic diagram of the proposed dual-level precision optimized read reference voltage sensing method based on threshold voltage distribution symmetry is shown:

[0128] An optimized read reference voltage sensing device based on the symmetry of the threshold voltage distribution of a 3D NAND flash memory, characterized in that:

[0129] The device comprises: a perception module, a discrimination module and a scanning module;

[0130] The perception module is used to first perform a pair of adjacent states V OPT perception, assuming the scanning range is [V L ,V R ], read the pair of adjacent states at V OPT The nearby OCD curves are set as three sampling points as a discrimination group, and the OCD curves are searched by sliding;

[0131] A discriminant module is used to calculate the discriminant γ of each discriminant group after obtaining the OCD values ​​of the three sampling points, which is used to measure the degree to which the left and right sampling points are centrally symmetric about the center point;

[0132] The scanning module is used to use a dual-level precision discrimination group sliding scan, first performing a coarse precision scan to determine V OPT The offset range is then scanned with fine precision to finally obtain V OPT The offset value.

[0133] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0134] A computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the steps of the above method are implemented.

[0135] Proposed optimized read reference voltage perception effect verification:

[0136] Fig.11 The VOPT between states B and C is retrieved layer by layer according to the three methods. The black curve is the true value of VOPT, and the blue curve is Figure 5 The red curve is the VOPT perceived based on the TVD minimum point in the traditional method. Figure 6 As shown in the figure, the VOPT perceived based on the central symmetry of OCD in the present invention, it can be seen that the error between the method of the present invention and the true value is within an acceptable range, and its root mean square error (RMSE) is 0.9158. There is a large error between the traditional method and the true value, and its root mean square error (RMSE) is 3.1367. In addition, as mentioned above, the method of the present invention only requires a smaller reading overhead and has lower complexity in sample processing than the traditional method. In summary, the experimental results based on the real chip verify the effectiveness of the proposed VOPT perception strategy.

[0137] The above is a detailed introduction to the optimized read reference voltage sensing method based on the symmetry of the threshold voltage distribution of 3D NAND flash memory proposed in the present invention, and the principles and implementation methods of the present invention are explained. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for optimizing read reference voltage sensing based on the symmetry of threshold voltage distribution of 3D NAND flash memory, characterized in that: The method specifically comprises the following steps: Step 1: First, execute a pair of adjacent states V OPT perception, assuming the scanning range is [V L , V R ], read the pair of adjacent states at V OPT The nearby OCD curves are set as three sampling points as a discrimination group, and the OCD curves are searched by sliding; Each discriminant group is composed of OCD obtained by three samplings, which are respectively called OCD left , OCD right and OCDcenter, assuming the offset difference between the three sampling points is ΔV, which satisfies: ΔV=s·ΔV min #(7) Among them, s is the minimum number of steps contained in the left and right intervals of each discrimination group, V min Adjust the minimum step value of the read reference voltage for flash memory; Step 2: For each discriminant group, after obtaining the OCD values ​​of the three sampling points, the discriminant γ of the discriminant group is calculated, which is used to measure the degree to which the left and right sampling points are centrally symmetric about the center point; The discriminant γ is defined as follows: γ=|OCD left +OCD right -2×OCD center |#(8) In [V L , V R ], identify the group with the minimum value of γ, and consider the offset value of the center point in this group to be V OPT ; That is, suppose that during the sliding scan of a certain stored data sample, the kth discriminant group γ k Take the minimum value, then the left and right endpoint offset values ​​of the discrimination group are V Lk and V Rk , then the V of the stored data sample OPT for: V OPT =V Lk +ΔV=V Rk -ΔV#(9) Step 3: Use the double-level precision discrimination group sliding scan to first perform a coarse precision scan to determine V OPT The offset range is then scanned with fine precision to finally obtain V OPT The offset value.

2. The method according to claim 1, characterized in that: In step three, the following steps are specifically included: S1. Coarse precision scanning and optimizing the 1s count distribution near the reference voltage in the form of central symmetry discrimination groups, each discrimination group consists of three large-step equally spaced sampling points; S2, evaluate the central symmetry of each discrimination group, and obtain the group number with the best central symmetry and the read reference voltage offset of its left endpoint; S3, obtaining a fine-precision scanning interval by using the read reference voltage offset determined by S2; S4, again performing a fine-precision sliding scan of the interval determined in S3 in the form of a central symmetry discrimination group; S5, evaluating the central symmetry degree of each discrimination group, and obtaining the serial number of the group with the best central symmetry degree; S6. Read the reference voltage offset at the center point of the discrimination group with the best central symmetry as the final value of V OPT .

3. The method according to claim 2, characterized in that: In step three, First, set the left boundary points of the judgment group to V L 、V L +ΔV, ..., V R -2ΔV to determine V with coarse precision OPT If the offset range is within, the number of discrimination groups N1 that needs to be obtained is:

4. The method according to claim 3, characterized in that: In step three, After completing the coarse-precision scan, a total of N1 γ values ​​are obtained. Next, a fine-precision scan is performed. Let the discriminant γ of the i-th group be i Take the minimum value and the offset value of its left boundary point yes: It can be seen that if the offset value of the center point of a discriminant group is V in the fine-precision scan, OPT , then the range of the left boundary point offset value of the group is The minimum step ΔV of the discriminant group shifted by RRV min To perform a scan, the number of times a fine-precision scan needs to be performed is: After completing the fine-precision scan, a total of N2 γ values ​​are obtained, plus the offset obtained from the coarse-precision scan. and The two γ values ​​at the position constitute an N2+2 bit sequence, numbered from 0 to N2-1. Let the discriminant γ of the jth one be j Take the minimum value, then the V obtained after fine-grained scanning is OPT The offset value is:

5. The method according to claim 4, characterized in that: In step three, The total number of read samples performed during the dual-level precision scan is: Right now: To obtain V OPT True value, need to use the minimum step ΔV min Yes [V L , V R ]To scan the entire range, the number of reads required is:

6. An optimized read reference voltage sensing device based on the symmetry of the threshold voltage distribution of a 3D NAND flash memory, characterized in that: The device comprises: a perception module, a discrimination module and a scanning module; The perception module is used to first perform a pair of adjacent states V OPT perception, assuming the scanning range is [V L , V R ], read the pair of adjacent states at V OPT The nearby OCD curves are set as three sampling points as a discrimination group, and the OCD curves are searched by sliding; Each discriminant group is composed of OCD obtained by three samplings, which are respectively called OCD left , OCD right and OCDcenter, assuming the offset difference between the three sampling points is ΔV, which satisfies: ΔV=s·ΔV min #(7) Among them, s is the minimum number of steps contained in the left and right intervals of each discrimination group, V min Adjust the minimum step value of the read reference voltage for flash memory; A discriminant module is used to calculate the discriminant γ of each discriminant group after obtaining the OCD values ​​of the three sampling points, which is used to measure the degree to which the left and right sampling points are centrally symmetric about the center point; The discriminant γ is defined as follows: γ=|OCD left +OCD right -2×OCD center |#(8) In [V L , V R ], identify the group with the minimum value of γ, and consider the offset value of the center point in this group to be V OPT ; That is, suppose that during the sliding scan of a certain stored data sample, the kth discriminant group γ k Take the minimum value, then the left and right endpoint offset values ​​of the discrimination group are V Lk and V Rk , then the V of the stored data sample OPT for: V OPT =V Lk +ΔV=V Rk -ΔV#(9) The scanning module is used to use a dual-level precision discrimination group sliding scan, first performing a coarse precision scan to determine V OPT The offset range is then scanned with fine precision to finally obtain V OPT The offset value.

7. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium for storing computer instructions, characterized in that: When the computer instructions are executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images