Method and apparatus for reading data from non-volatile memory

This application is a continuation of U.S. patent application Ser. No. 13/929,319, filed Jun. 27, 2013, titled “METHOD AND APPARATUS FOR READING DATA FROM NON-VOLATILE MEMORY,” which is a continuation of U.S. patent application Ser. No. 11/944,168, filed Nov. 21, 2007, titled “METHOD AND APPARATUS FOR READING DATA FROM FLASH MEMORY,” the disclosures of which is hereby incorporated by reference in their entireties herein.

1. Field of the Invention

Embodiments of the invention relate to memory devices, and more particularly, in one or more embodiments, to flash memory devices.

2. Description of the Related Art

Flash memory devices are non-volatile memory devices which store information on a semiconductor in a way that needs no power to maintain the information stored therein. Flash memory devices typically include an array of memory cells in a matrix form having columns and rows. Each memory cell includes a charge storage node, such as a floating gate transistor formed on a semiconductor substrate. The cells are arranged to form strings such that the source of the transistor of each cell is coupled to the drain of the transistor of a neighboring cell in each string. The memory cell array includes sense lines (often referred to as bit lines), each of which connects to a column of cells in the memory cell array. The memory cell array also includes select lines (often referred to as word lines), extending perpendicular to the bit lines and parallel to one another. Each of the word lines connects to the control gates of the transistors in a row of cells in the memory cell array.

As the geometry of a flash memory is reduced, distances between cells in the memory cell array are also reduced. These reduced distances may incur inter-signal interference (ISI) between neighboring memory cells. The term “inter-signal interference” refers to electromagnetic effect of electrons trapped in the charge storage node of one memory cell on a neighboring memory cell. The inter-signal interference may affect the operation of a flash memory device. Therefore, there is a need for an error-correcting logic or algorithm for flash memory devices with a reduced size.

The embodiments will be better understood from the Detailed Description of Embodiments and from the appended drawings, which are meant to illustrate and not to limit the embodiments, and wherein:

FIG. 1 is a schematic diagram of a memory cell array of a NAND flash memory device according to one embodiment;

FIG. 2 is a schematic cross-section of the NAND flash memory device of FIG. 1 according to one embodiment;

FIG. 3A is a diagram illustrating written data bits and corresponding numbers of trapped electrons in an ideal NAND flash memory array without inter-signal interference;

FIG. 3B is a diagram illustrating data bits read from a NAND flash memory array with inter-signal interference;

FIG. 4 is a schematic block diagram illustrating a NAND flash memory array having inter-signal interference;

FIG. 5A is a diagram illustrating an example of a write operation on a NAND flash memory;

FIG. 5B is a diagram illustrating an example of a read operation on a NAND flash memory;

FIGS. 6A-6J illustrate a method of correcting data read from a NAND flash memory array using a Viterbi algorithm according to one embodiment; and

FIG. 7 illustrates a method of correcting data read from a NAND flash memory array using a Viterbi algorithm according to another embodiment.

As described above, as the geometry of a flash memory device is reduced, distances between cells in the memory device are also reduced. In such a flash memory device, inter-signal interference may affect the operation (particularly, data read operation) of the flash memory device. Therefore, there is a need for an error-correcting logic or algorithm for flash memory devices.

In one embodiment, a flash memory device is provided with an error-correcting logic or algorithm for data read operations, based on a Viterbi algorithm or its variant. Raw data read from a flash memory device is processed using the Viterbi algorithm. A graph called a trellis may be constructed to include all possible combinations of data states that can be stored in a row of cells in the flash memory. A data path with a minimum error is determined using a Viterbi algorithm, thereby providing correct data.

FIG. 1 illustrates a NAND flash memory array according to one embodiment. The illustrated flash memory array includes first to M-th bit lines BL 0 -BLM and first to N-th word lines WL 0 -WLN. The bit lines BL 0 -BLM extend parallel to one another in a column direction. The word lines WL 0 -WLN extend parallel to one another in a row direction. The NAND flash memory array also includes select transistors 120 a , 120 b used for selecting a bit line.

Each bit line includes a string of floating gate transistors coupled in series source to drain. For example, the second bit line BL 1 includes floating gate transistors 110 connected in series. The control gates of floating gate transistors 110 of cells in the same row are coupled to the same word line. Each of the floating gate transistors 110 forms a memory cell that stores a charge (or a lack of charge), wherein the amount of stored charge can be used to represent, for example, one or more states, and wherein the one or more states can represent one or more digits (e.g., bits) of data. The memory cell can be either a single-level cell (SLC) or a multi-level cell (MLC). In one embodiment, the amounts of charge stored in the floating gate transistors 110 may be detected by sensing currents flowing through the floating gate transistors 110 . In another embodiment, the amounts of charge stored in the floating gate transistors 110 may be detected by sensing the threshold voltage values of the floating gate transistors 110 .

FIG. 2 illustrates a cross-section of the floating gate transistors 110 in the second bit line BL 1 . The floating gate transistors 110 are formed on a substrate 201 . Each of the floating gate transistors 110 includes a source region 210 (which is a drain region for a neighboring transistor), a drain region 212 (which is a source region for a neighboring transistor), a doped channel region 214 , a first dielectric (e.g., a tunnel oxide) 216 , a floating gate 218 , a second dielectric (e.g., a gate oxide, wherein the tunnel and gate oxide can be formed of the same or different material) 220 , and a control gate 222 . The tunnel oxide 216 is formed on the channel region 214 to insulate the floating gate 218 from the channel region 214 . The gate dielectric 220 physically and electrically separates the floating gate 218 from the control gate 222 . The control gate 222 is coupled to an appropriate word line, e.g., word line WL 1 . Electrons can be trapped on the floating gate 218 and be used to store data.

Referring to FIG. 3A , memory cells include a certain number (including zero) of electrons trapped therein, depending on the stored data. For the sake of illustration, suppose that a floating gate transistor has no trapped electrons in its floating gate to store “1” (can be vice versa). Also suppose that a floating gate transistor forming a memory cell can trap 8 electrons in its floating gate to store “0.” The numbers of electrons herein are arbitrary numbers for the sake of explanation, and a skilled artisan will appreciate that the numbers of electrons vary widely depending on the design of the memory cell array.

Ideally, when a stored data bit is “1,” a current sensed to detect the state of the memory cell would indicate that the cell has no trapped electrons (i.e., the presence of a current flow). On the other hand, when a stored data bit is “0,” a current sensed to detect the state of the memory cell would indicate that the cell has 8 trapped electrons (i.e., the absence of a current flow).

A current sensed to detect the state of the memory cell may indicate that the number of trapped electrons is in a continuous range rather than either 0 or 8. Therefore, a threshold value for a sensed current is set to determine whether the memory cell has data of either “1” or “0” (or more states for multi-level cells). In the illustrated example, the threshold value can be 4, i.e., if there are 4 or more trapped electrons, the stored data is “0”; if there are less than 4 trapped electrons, the stored data is “1.” For example, in FIG. 3B , the trapped electrons are 0, 11, 5, and 0 in a row of memory cells. The trapped electrons indicate that the stored data b 0 , b 1 , b 2 , b 3 are 1, 0, 0, 1.

As the geometry of the NAND flash memory is reduced, distances between cells in the memory cell array are also reduced. These reduced distances may incur inter-signal interference (ISI) between neighboring memory cells. Referring to FIG. 4 , a NAND flash memory includes a memory cell array 400 . In the illustrated embodiment, a memory cell 410 may experience inter-signal interference from neighboring memory cells 421 - 424 in the same row and/or in the same column.

This inter-signal interference may affect read operations of the NAND flash memory. For example, if one or more of the adjacent memory cells 421 - 424 has 8 electrons trapped therein, the memory cell 410 may appear when sensed as if it has 2 more electrons than it actually has. For example, the memory cell 410 may appear when sensed as if it has 2 electrons even if it actually has no electrons, or 10 electrons even if it actually has 8 electrons.

Because of the inter-signal interference, when a number of electrons is close to a threshold value, the data can be read inaccurately. For example, when the number of actually trapped electrons is 3, the number sensed may correspond to 5 when there is inter-signal interference. For example, referring back to FIG. 3B , the third bit b 2 may be in fact 0 because the read number of electrons (“5”) may be a result of inter-signal interference. In the illustrated example, the threshold is “4.” The cell storing the third bit b 2 may appear as if it has 5 electrons even though it actually has 3 electrons.

Referring to FIGS. 5A-5B and 6 A- 6 J, a method of confirming/correcting data read from a NAND flash memory array according to one embodiment will now be described in detail. In the illustrated embodiment, a Viterbi algorithm is used for confirming/correcting data read from a NAND flash memory. A Viterbi algorithm is an algorithm for finding the most likely sequence of hidden states, which is called the Viterbi path. Given observed patterns of trapped electrons in a row of memory cells, the Viterbi algorithm is used to determine the most likely states of the memory cells in the row in consideration of inter-signal interference.

In the illustrated embodiment, it is assumed that one cell in a row exerts inter-signal interference only on a cell immediately next to the one cell in the same row (e.g., in the left-to-right direction). However, a skilled artisan will appreciate that the Viterbi algorithm can be extended to other neighboring cells around any given cell.

Suppose that a cell storing logic “1” has 0 electrons in its floating gate and that a cell storing logic “0” has 8 electrons in its floating gate. A particular cell is not affected by a neighboring cell immediately next to the particular cell in the same row on the left side, if the neighboring cell stores “1” (e.g., no trapped electrons). If, however, the neighboring cell stores “0” (e.g., 8 trapped electrons), the particular cell may experience inter-signal interference from the neighboring cell as if it has two more electrons in addition to its own trapped electrons.

Referring to FIG. 5A , for the purpose of illustration, suppose that data written on four cells b 1 -b 4 in row 1 are 1, 0, 1, 1. Referring to FIG. 5B , now suppose that sensed information from the cells b 1 -b 4 indicate that the cells b 1 -b 4 store 0, 11, 5, and 0 electrons, respectively. These numbers of electrons for cells b 1 -b 4 will be reused in the example of FIGS. 6A-6J as the read number. If the threshold value is 4 electrons (if there are 4 or more electrons, the stored data is 0; if there are less than 4 electrons, the stored data is 1), the read data can be translated to 1, 0, 0, 1 which is not identical to the written data, 1, 0, 1, 1.

Referring to FIGS. 6A-6J , a Viterbi algorithm is used to confirm/correct the data read from the row of cells b 1 -b 4 . In the Viterbi algorithm, a path of minimum error along a trellis is determined to indicate the correct data. Errors associated with data states along the trellis are determined by a deviation D which is represented by Equation 1:

D =(RE−PE) 2   (1)

In Equation 1, RE is a read number of electrons, and PE is a possible number of electrons in a next state along the trellis. The RE data used will be 0, 11, 5, and 0. As a path is taken along the trellis, the error is accumulated. An accumulated error Ei is represented by Equation 2:

In Equation 2, Ei is an accumulated error at each destination state, and Di is a deviation at the destination state. The alternative paths (from 0 or 1) leading to each state (1 or 0) are compared to one another. Only the path with a minimum error is selected at each destination state. This process is repeated until the paths to all the given states are determined. The selected paths serve to indicate the correct data stored in the memory cells.

In FIGS. 6A-6J , the upper row represents states where a data bit stored in a memory cell is 1 whereas the lower row represents states where a data bit stored in a memory cell is 0. Each number adjacent to an arrow in FIGS. 6A-6J is a number of electrons that can be expected to be sensed in the next cell. The numbers of electrons herein are arbitrary for the sake of explanation, and a skilled artisan will appreciate that the numbers of electrons vary widely depending on the design of the memory cell array and sensitivity of cells and/or sense circuits.

The numbers had prior states all “1,” i.e., the initial condition is that the accumulated error is 0. In FIG. 6A , the state of an initial cell b 0 can be either at 1 or at 0. For the path where the initial cell b 0 is at 1 and a first cell b 1 is at 1, the possible number of electrons (PE) stored in the first cell b 1 is 0 because there is no inter-signal interference by the initial cell b 0 on the first cell b 1 . In the example, the read number of electrons (RE) is 0 for b 1 , and an accumulated error E 1 a associated with this path (1 to 1) is (RE−PE) 2 =(0−0) 2 =0.

For the path where the initial cell b 0 is at 1 and the first cell b 1 is at 0 (path 1 to 0), the possible number of electrons in the first cell b 1 is 8 because the first cell b 1 would have 8 electrons without inter-signal interference by the initial cell b 0 . Thus, an accumulated error E 1 c associated with this path (1 to 0) is (0-8) 2 =64.

On the other hand, for the path where the initial cell b 0 is at 0 and the first cell b 1 is at 1, the possible number of electrons sensed is 2 because the first cell b 1 would store no electrons therein but experience inter-signal interference (2 electrons) from the initial cell b 0 . Thus, an accumulated error E 1 b associated with this path (0 to 1) is (0−2) 2 =4.

For the path where the initial cell b 0 is at 0 and the first cell b 1 is at 0, the first cell b 1 would store 8 electrons with inter-signal interference (2 electrons) from the initial cell b 0 , thus acting as if it has 10 electrons. Thus, an accumulated error E 1 d associated with this path (0 to 0) is (0−10) 2 =100.

Based on the amounts of the accumulated errors, one of the paths leading to each of the destination states (0 or 1) is eliminated, as indicated by being crossed out. In FIG. 6B , for destination state 1, the arrow from 0 to 1 has been crossed out because the accumulated error E 1 b associated with this path from 0 to 1 is greater than the accumulated error E 1 a associated with the other path from 1 to 1. Similarly, the arrow from 0 to 0 has been crossed out because the accumulated error E 1 d is greater than the accumulated error E 1 c . The accumulated errors are drawn inside the circles indicating the destination states.

In FIG. 6C , the read number of electrons (RE) for the second cell b 2 in the illustrated example is 11. The first cell b 1 can be either at 1 or at 0. For the computation of accumulated error, the first cell b 1 has 0 for state 1 and 64 for state 0. For the path where the first cell b 1 is at 1 and a second cell b 2 is at 1, the possible number of electrons sensed for the second cell b 2 would be 0 electrons because the second cell b 2 has no stored electron therein with no inter-signal interference from the first cell b 1 . Thus, a deviation D is (11−0) 2 =121. For the path where the first cell b 1 is at 1 and the second cell b 2 is at 0, a deviation D is (11−8) 2 =9. The accumulated errors E 2 a , E 2 c are 121 and 9, respectively.

On the other hand, for the path where the first cell b 1 is at 0, the accumulated error starts at 64. For the path to destination state 1 of the second cell b 2 , a deviation D is (11-2) 2 =81, and an accumulated error E 2 b is 145 because the accumulated error of the preceding path adds to the deviation of the current path. For the path where the first cell b 1 is at 0 and the second cell b 2 is at 0, a deviation D is (11−10) 2 =1, and an accumulated error E 2 d is 65.

Again, based on the amounts of the accumulated errors, one of the paths leading to each of the destination states (0 or 1) is eliminated, as indicated by being crossed out. In FIG. 6D , the arrow from 0 to 1 has been crossed out because the accumulated error E 2 b of 145 associated with the path from 0 to 1 is greater than the accumulated error E 2 a associated with the other path from 1 to 1. Similarly, the arrow from 0 to 0 has been crossed out because the accumulated error E 2 d associated with the path from 0 to 0 is greater than the accumulated error E 2 c associated with the other path from 1 to 0.

In FIG. 6E , the read number of electrons (RE) for the third cell b 3 is 5 for the illustrated example. The state of the second cell b 2 can be either at 1 or at 0. For the path where the second cell b 2 is at 1 and a third cell b 3 is at 1, a deviation D is (5−0) 2 =25, and an accumulated error E 3 a is 146. For the path where the second cell b 2 is at 1 and the third cell b 3 is at 0, a deviation D is (5−8) 2 =9, and an accumulated error E 3 c is 130.

On the other hand, for the path where the second cell b 2 is at 0, and the third cell b 3 is at 1, a deviation D is (5−2) 2 =9, and an accumulated error E 3 b is 18. For the path where the second cell b 2 is at 0 and the third cell b 3 is at 0, a deviation D is (5−10) 2 =25, and an accumulated error E 3 d is 34.

Again, based on the amounts of the accumulated errors, one of the paths leading to each state (0 or 1) is eliminated, e.g., crossed out. In FIG. 6F , the arrow from 1 to 1 has been crossed out because the accumulated error E 3 a is greater than the accumulated error E 3 b for the path from 0 to 1. Similarly, the arrow from 1 to 0 has been eliminated because the accumulated error E 3 c for the path from 1 to 0 is greater than the accumulated error E 3 d for the path from 0 to 0.

In FIG. 6G , the read number of electrons (RE) for the fourth cell b 4 is 0 for the illustrated example. The state of the third cell b 3 can be either at 1 or at 0. For the path where the third cell b 3 is at 1 and a fourth cell b 4 is at 1, a deviation D is (0−0) 2 =0, and an accumulated error E 4 a is 18. For the path where the third cell b 3 is at 1 and the fourth cell b 4 is at 0, a deviation D is (0−8) 2 =64, and an accumulated error E 4 c is 82.

On the other hand, for the path where the third cell b 3 is at 0 and the fourth cell b 4 is at 1, a deviation D is (0−2) 2 =4, and an accumulated error E 4 b is 38. For the path where the third cell b 3 is at 0 and the fourth cell b 4 is at 0, a deviation D is (0−10) 2 =100, and an accumulated error E 4 d is 134.

Again, based on the amounts of the accumulated errors, one of the paths leading to each of the states (0 or 1) is eliminated. In FIG. 6H , the arrow from 0 to 1 has been crossed out because the accumulated error E 4 b for the path from 0 to 1 is greater than the accumulated error E 4 a for the path from 1 to 1. Similarly, the arrow from 0 to 0 has been crossed out because the accumulated error E 4 d for the path from 0 to 0 is greater than the accumulated error E 4 c for the path from 1 to 0.

FIG. 6I illustrates a resulting path including the arrows remaining after the completion of the process described above. Now, errors at possible states for the fourth cell b 4 are compared to each other. In the illustrated example, for the destination state “1,” the accumulated error is 18. For the destination state “0,” the accumulated error is 82. The smaller the error is, the more likely the cell has the state. Thus, it is more likely that the fourth cell b 4 stores “1.” Then, a path is taken backward from state “1” of the fourth cell b 4 . Thus, the third cell b 3 likely stores 1. The second cell b 2 likely stores 0. In addition, the first cell b 1 likely stores 1. Thus, the correct data is 1, 0, 1, 1, which is identical to the written data, as shown in FIG. 6J .

In another embodiment, inter-signal interference on a particular cell by two or more neighboring cells may be taken into account in determining correct data using a Viterbi algorithm. For example, inter-signal interference on a particular cell by a neighboring cell in a row immediately above the particular cell can be taken into account in processing raw data using the Viterbi algorithm. Referring to FIG. 7 , a delta (A) is added to a possible number of electrons in each state. The delta (A) represents the inter-signal interference from a neighboring cell in a row immediately above a particular cell. In one embodiment in which inter-signal interference between cells in the same row amounts to 2 electrons, the delta may be in a range between about 0 and about 2. The detailed process of the Viterbi algorithm can be as described earlier with reference to FIG. 6A-6J except that the delta is added to the possible number of electrons in each state.

In the embodiments described above, the cells are configured to store one of two states, i.e., single level cells. In other embodiments, multi-level cells can be used to store multi-levels, e.g., more than two states. In such embodiments, a Viterbi algorithm can also be adapted to confirm or correct data read from the cells.

As described above, a Viterbi algorithm may be performed on every row of data bits read during read operations. In certain embodiments, a Viterbi algorithm may be performed only on a selected block of data which is suspected to have at least one error. In other embodiments, a Viterbi algorithm may be performed only on a selected block of data which includes a value close to a threshold value (e.g., 5 electrons where the threshold is 4 electrons) or in response to an uncorrectable error to supplement error correction codes (ECC). A skilled artisan will appreciate that various alternative ways of applying a Viterbi algorithm to NAND flash read operations are also possible. In addition, a skilled artisan will appreciate that any modified Viterbi algorithm or a similar algorithm can also be adapted for flash read operations.

In the embodiments described above, a Viterbi algorithm can be performed by any suitable processor or circuit within the NAND flash memory device. In other embodiments, an external processor or circuit may be provided to perform the Viterbi algorithm. A skilled artisan will appreciate that any suitable configuration of processors circuits can be used for performing the Viterbi algorithm as described above.

A flash memory device according to the embodiments described above can be incorporated in various electronic devices. Examples of the electronic devices can include, but are not limited to, consumer electronic products, electronic circuits, electronic circuit components, parts of the consumer electronic products, electronic test equipments, etc. Examples of the consumer electronic products include, but are not limited to, a mobile phone, a telephone, a television, a computer monitor, a computer, a hand-held computer, a personal digital assistant (PDA), a microwave, a refrigerator, a stereo system, a cassette recorder or player, a DVD player, a CD player, a VCR, an MP3 player, a radio, a camcorder, an optical camera, a digital camera, a washer, a dryer, a washer/dryer, a copier, a facsimile machine, a scanner, a multi functional peripheral device, a wrist watch, a clock, etc. Further, the electronic device can include unfinished products.

One embodiment is a method of retrieving data from a memory cell. The method includes determining values associated with data stored by selected memory cells in an array of memory cells; and processing the determined values in accordance with a Viterbi algorithm so as to determine the data stored in the selected memory cells.

Another embodiment is a method of retrieving data from a memory cell. The method includes sensing an electrical condition of a first memory cell within an array of memory cells; and determining a logical state for the first memory cell based at least in part on the electrical condition and an electrical condition and/or logical state associated with at least one memory cell of the array adjacent to the first memory cell.

Yet another embodiment is an apparatus including an array of memory cells. Each of the memory cells is configured to store charges indicative of a data digit. The apparatus also includes a sense circuit configured to detect values of the charges stored in selected ones of the memory cells. The apparatus is configured to process the detected values in accordance with a Viterbi algorithm.

Although this invention has been described in terms of certain embodiments, other embodiments that are apparent to those of ordinary skill in the art, including embodiments that do not provide all of the features and advantages set forth herein, are also within the scope of this invention. Moreover, the various embodiments described above can be combined to provide further embodiments. In addition, certain features shown in the context of one embodiment can be incorporated into other embodiments as well. Accordingly, the scope of the present invention is defined only by reference to the appended claims.