Apparatus and method for two-step read of resistive random access memory

By adopting a two-step reading technology in magnetoresistive random access memory, using global and separate reading reference values ​​for codeword reading, the problem of memory unit reduction and stability in the prior art is solved, and the problem of memory unit reduction is improved Read reliability and durability.

CN119068951APending Publication Date: 2024-12-03SANDISK TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410039237.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2023-07-19
Filing Date
2024-01-09
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing magnetoresistive random access memory (MRAM) faces bit interference and magnetization switching stability problems when reducing the size of memory cells, and the self-reference reading technology requires multiple write operations, which affects read reliability and durability.

Method used

Using two-step reading technology, the first reading is first used to read the reference value. If it fails, the search table will search for individual reading reference value for the second reading, which reduces the write operation and improves the read reliability. .

Benefits of technology

Through the two-step reading technology, codewords can be read quickly and reliably, reducing read delay and write errors, and improving memory durability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119068951A_ABST
    Figure CN119068951A_ABST
Patent Text Reader

Abstract

The invention relates to an apparatus and method for two-step read of a resistive random access memory. An apparatus is provided that includes a memory array and a control circuit. The memory array includes non-volatile memory cells, each non-volatile memory cell including a resistive random access memory element. A control circuit configured to receive a read command specifying addresses of the first set of non-volatile memory cells; performing a first read of the first set of non-volatile memory cells using a first predetermined read reference value to provide first read data; retrieving a second predetermined read reference value corresponding to the specified address from the memory when the first read is performed; and in response to a condition on the first read data being satisfied, performing a second read of the first set of non-volatile memory cells using the second predetermined read reference value to provide second read data.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] Priority declaration

[0002] This application claims priority to U.S. Provisional Patent Application No. 63 / 504,665, filed on May 26, 2023, entitled “APPARATUS AND METHOD FOR TWO-STEP READ OF RESISTIVE RANDOMACCESS MEMORY,” which is incorporated herein by reference in its entirety. Background Art

[0003] Memory is widely used in various electronic devices, such as cellular phones, digital cameras, personal digital assistants, medical electronics, mobile computing devices, non-mobile computing devices, and data servers. Memory can be non-volatile memory or volatile memory. Non-volatile memory allows information to be stored and retained even when it is not connected to a power source (e.g., a battery).

[0004] An example of non-volatile memory is magnetoresistive random access memory (MRAM), which uses magnetization to represent stored data, as opposed to some other memory technologies that use charge to store data. Generally speaking, MRAM includes a large number of magnetic memory cells formed on a semiconductor substrate, where each memory cell represents a data bit. Data bits are written to the memory cells by changing the magnetization direction of the magnetic elements within the memory cells, and the bits are read by measuring the resistance of the memory cells (low resistance typically represents a "0" bit and high resistance typically represents a "1" bit). As used herein, the magnetization direction is the orientation direction of the magnetic moment.

[0005] Although MRAM is a promising technology, many challenges remain. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] Like numbered elements refer to common parts in different figures.

[0007] Figure 1A is a block diagram of an MRAM memory cell.

[0008] Figure 1B is a block diagram of an MRAM memory cell.

[0009] Figure 1C is a block diagram of an MRAM memory cell.

[0010] Figure 1D is a block diagram of an MRAM memory cell.

[0011] Figure 1E is a block diagram of an MRAM memory cell.

[0012] Figure 2A Example current versus voltage characteristics of an MRAM memory cell are depicted.

[0013] Figure 2B is a graphical representation of example antiparallel and parallel resistance values ​​in a set of memory cells.

[0014] Figure 2C A graph of bit error rate versus read reference is depicted for a large number of memory cells in a memory array.

[0015] Figure 3A is a flow chart of one embodiment of a process for creating a lookup table of individual read reference values ​​for a set of codewords.

[0016] Figure 3B is a flow chart of one embodiment of an alternative process for creating a lookup table of separate read reference values ​​for a set of codewords.

[0017] Figure 4A is a table showing fail bit counts for example codewords and example candidate read reference values.

[0018] Figure 4B Depicted Figure 4A A table showing example digital fail bit count values.

[0019] Figure 4C Describes the use of Figure 3A process and use Figure 4B An example lookup table of entries created based on the sample data and the expected maximum fail bit count of two bits.

[0020] Figure 4D Describes the use of Figure 3A process and use Figure 4B An example lookup table of entries created based on the example data and the expected maximum fail bit count of five bits.

[0021] Figure 5 is a flow chart of one embodiment of a process for reading a codeword.

[0022] Figure 6 is a flow chart of one embodiment of another process for reading a codeword.

[0023] Figure 7 is a block diagram of a memory system that uses the read techniques described herein. DETAILED DESCRIPTION

[0024] A technique for fast two-step read of resistive random access memory cells (e.g., MRAM cells) is described. In one embodiment, a codeword is read from a group of memory cells using a two-step read process. As used herein, a "codeword" includes a predetermined number of data bits (e.g., 16 bits, 64 bits, 1024 bits, etc.). In one embodiment, a codeword includes a plurality of bits that can be encoded and decoded by an error correction coding (ECC) engine (typically on-die). In one embodiment, a memory array includes a plurality of codewords, and each codeword has an associated address.

[0025] In one embodiment, a first read of a codeword is performed using a first predetermined read reference value that is used to distinguish a first state (e.g., 0) from a second state (e.g., 1) for each bit in the codeword. As used herein, the first predetermined read reference value is also referred to as a "global read reference value."

[0026] In one embodiment, the global read reference value is selected to read all codewords of the memory array. In one embodiment, the global read reference value may be determined during wafer sort and has a value that minimizes the number of bit errors for all codewords of the memory array.

[0027] In one embodiment, a compact fast memory (e.g., a content addressable memory) stores a table (e.g., a lookup table) of codeword addresses and corresponding second read reference values. As used herein, the second predetermined read reference value is also referred to as a "separate read reference value."

[0028] In one embodiment, each individual read reference value is selected to read a corresponding codeword of the memory array. In one embodiment, each individual read reference value may be determined during wafer sort and has a value that minimizes the number of bit errors for the corresponding codeword.

[0029] In one embodiment, when the first read of a codeword is performed using a global read reference value, the lookup table is searched to determine if there is an entry for the codeword address. If there is such an entry, the individual read reference value corresponding to the codeword address is retrieved from the lookup table.

[0030] In one embodiment, if the first read fails (e.g., the ECC engine cannot successfully decode the read bits), and if the lookup table includes an entry for the codeword address, a second read of the codeword is performed using the corresponding separate read reference value. In one embodiment, if the lookup table does not include an entry for the codeword address, the memory controller may perform further read failure processing (e.g., using a self-referenced read technique to read the specified codeword).

[0031] Without wishing to be bound by any particular theory, it is believed that the two-step read technique described below may be able to read more bits without having to perform a self-referenced read, which requires multiple write operations and may increase write errors and reduce endurance. In this regard, without being bound by any particular theory, it is believed that the two-step read technique described below can improve read reliability.

[0032] Without wishing to be bound by any particular theory, it is believed that the two-step read technique described below can minimize latency during the first read operation by using a fast lookup memory and searching for a separate read reference value. In this regard, without being bound by any particular theory, it is believed that if the first read fails, the second read can be started quickly and therefore can be fast.

[0033] Figure 1A is a schematic perspective view of a conventional MRAM memory cell 10 using field induced switching. In general, the MRAM memory cell 10 includes a magnetic tunnel junction (MTJ) 12, which includes an upper ferromagnetic layer 14, a lower ferromagnetic layer 16, and a tunnel barrier (TB) 18, which is an insulating layer between the two ferromagnetic layers. In this example, the upper ferromagnetic layer 14 is a free layer (FL), the magnetization direction of which can be switched. The lower ferromagnetic layer 16 is a pinned (or fixed) layer (PL), the magnetization direction of which does not change.

[0034] When the magnetization direction in the free layer 14 is parallel to the magnetization direction of the pinned layer 16, the resistance (RP) across the MRAM memory cell 10 (referred to herein as "parallel resistance RP") is low due at least in part to spin-dependent scattering of minority electrons. When the magnetization direction in the free layer 14 is antiparallel to the magnetization direction of the pinned layer 16, the resistance (R) across the MRAM memory cell 10 is low. AP ) (referred to herein as the “antiparallel resistance R AP ”) is higher due at least in part to spin-dependent scattering of minority and majority electrons. The memory state (“0” or “1”) of the MRAM memory cell 10 is read by measuring the resistance of the MRAM memory cell 10. In this regard, the electrical conductors 20 / 22 attached to the MRAM memory cell 10 are used to read the MRAM data.

[0035] The magnetization direction in the free layer 14 changes in response to a current 24 flowing in a digit line 26 and in response to a current 28 flowing in a write line 20 which generates magnetic fields 30 and 32, respectively. Figure 1AThe situation is depicted where current 24 in digit line 26 flows out of the page and current 28 in write line 20 flows from left to right. Therefore, magnetic fields 30 and 32 are orthogonal, which will cause the magnetization direction in free layer 14 to switch from parallel to antiparallel with the magnetization direction of pinned layer 16. By reversing the polarity of current 28 in write line 20 while keeping the polarity of current 24 in digit line 26 constant, the orientation of the bit is switched.

[0036] Above for Figure 1A The field induced switching technique described in MRAM memory cell 10 has some practical limitations, particularly when designs require scaling the MRAM memory cell to smaller sizes. For example, because the technique requires two sets of magnetic field write lines, the MRAM memory cell array is susceptible to bit disturbances (i.e., neighboring cells may be inadvertently altered in response to a write current directed to a given cell).

[0037] Furthermore, reducing the physical size of the MRAM memory cell results in lower magnetic stability against magnetization switching due to thermal fluctuations. The stability of the bit can be enhanced by utilizing magnetic materials with large magnetic anisotropy and therefore large switching fields for the free layer, but the currents required to generate a strong enough magnetic field to switch the bit are impractical in existing applications.

[0038] Spin transfer torque (STT) switching is another technique used to program MRAM memory cells. Figure 1B Schematic representation of the STT switching technology for an MRAM memory cell 50, which includes a magnetic tunnel junction 52, which includes an upper ferromagnetic layer 54, a lower ferromagnetic layer 56, and a tunnel barrier 58, which is an insulating layer between the two ferromagnetic layers. In this example, the lower ferromagnetic layer 56 is a free layer, the magnetization direction of which can be switched. The upper ferromagnetic layer 54 is a pinned (or fixed) layer, the magnetization direction of which is not easily changed.

[0039] When the magnetization direction in the free layer 56 is parallel to the magnetization direction of the pinned layer 54, the parallel resistance RP across the MRAM memory cell 50 is low. When the magnetization direction in the free layer 56 is antiparallel to the magnetization direction of the pinned layer 54, the antiparallel resistance R across the MRAM memory cell 50 is high. AP is higher. The memory state ("0" or "1") of the memory cell 50 is read by measuring the resistance of the MRAM memory cell 50. In this regard, the electrical conductors 60 / 62 attached to the MRAM memory cell 50 are used to read the MRAM data. By design, both the parallel configuration and the anti-parallel configuration remain stable in the quiescent state and / or during a read operation (at a sufficiently low read current).

[0040] In the rest of the text and drawings, the direction of the write current is defined as the direction of electron flow. Therefore, the term write current refers to the electron current.

[0041] To "set" the MRAM memory cell bit value (i.e., select the direction of the free layer magnetization), an electrical write current 64 is applied from conductor 60 to conductor 62. The electrons in the write current 64 become spin-polarized as they pass through the pinned layer 54, because the pinned layer 54 is a ferromagnetic metal. Although the conduction electrons in the ferromagnetic metal will have a spin orientation that is co-linear with the magnetization direction, the vast majority of them will have a particular orientation that is parallel to the magnetization direction, resulting in a net spin-polarized current. (Electron spin refers to angular momentum that is proportional to but antiparallel to the direction of the electron's magnetic moment, but for ease of discussion, this directional distinction will not be used from now on).

[0042] When spin-polarized electrons tunnel through the tunnel barrier 58, the conservation of angular momentum can cause a torque to be exerted on both the free layer 56 and the pinned layer 54, but the torque is not sufficient (by design) to affect the magnetization direction of the pinned layer 54. In contrast, if the initial magnetization direction of the free layer 56 is antiparallel to the pinned layer 54, the torque is sufficient (by design) to switch the magnetization direction of the free layer 56 to be parallel to the magnetization direction of the pinned layer 54. The parallel magnetization will then remain stable before and after this write current is turned off.

[0043] In contrast, if the magnetization of the free layer 56 and the magnetization of the pinned layer 54 are initially parallel, then by applying a write current in the opposite direction to that described above, the magnetization direction STT of the free layer 56 can be switched to be antiparallel to the magnetization direction of the pinned layer 54. Therefore, via the same STT physics, the magnetization direction of the free layer 56 can be deterministically set to either of the two stable orientations by judiciously choosing the write current direction (polarity).

[0044] Figure 1B The MRAM memory cell 50 uses a material in which both the pinned layer magnetization and the free layer magnetization are in the in-plane direction. In contrast, Figure 1C A schematic representation of an STT switching MRAM memory cell 70 is depicted in which both the pinned layer magnetization and the free layer magnetization are in perpendicular directions. The MRAM memory cell 70 includes a magnetic tunnel junction 72, which includes an upper ferromagnetic layer 74, a lower ferromagnetic layer 76, and a tunnel barrier 78, which is an insulating layer between the two ferromagnetic layers. In this example, the lower ferromagnetic layer 76 is a free layer, the magnetization direction of which can be switched. The upper ferromagnetic layer 74 is a pinned (or fixed) layer, the magnetization direction of which is not easily changed.

[0045] When the magnetization direction in the free layer 76 is parallel to the magnetization direction of the pinned layer 74, the resistance RP across the MRAM memory cell 70 is low. When the magnetization direction in the free layer 76 is antiparallel to the magnetization direction of the pinned layer 74, the resistance R across the MRAM memory cell 70 is high. AP is higher. The memory state ("0" or "1") of the MRAM memory cell 70 is read by measuring the resistance of the MRAM memory cell 70. In this regard, electrical conductors 80 / 82 attached to the MRAM memory cell 70 are used to read the MRAM data. By design, both the parallel configuration and the antiparallel configuration remain stable in the quiescent state and / or during a read operation (at a sufficiently low read current). To "set" the MRAM memory cell bit value (i.e., select the direction of the free layer magnetization), an electrical write current 84 is applied from conductor 80 to conductor 82, and the MRAM memory cell 70 is set as described above with respect to Figure 1B Operate as described.

[0046] Compared to the earliest MRAM memory cells that used a magnetic field from a current-carrying conductor close to the MRAM memory cell, STT switching technology requires lower power, which virtually eliminates the problem of adjacent bit interference and has more favorable scaling for higher cell densities (MRAM memory cell size reduction). The latter problem also favors STT-MRAM, where the free layer magnetization and the pinned layer magnetization are oriented perpendicular to the film plane, rather than in-plane. However, in practice, STT switching requires the full write current to flow through the tunnel barrier, which negatively affects the long-term reliability of the STT MRAM memory cell due to the necessary stress of moderate to high write voltages across the tunnel barrier.

[0047] Figure 1D An alternative MRAM memory cell 100 utilizing spin-orbit torque (SOT) for switching the free layer using a spin current is depicted. The spin of an electron is an intrinsic angular momentum, separate from the angular momentum due to the orbital motion of the electron. In a solid, the spins of many electrons can act together to affect the magnetic and electronic properties of the material, such as giving it a permanent magnetic moment as in a ferromagnet.

[0048] In many materials, electron spins have an equal degree of up and down directions, and no transport properties depend on spin. However, various techniques can be used to generate a spin-polarized population of electrons, which creates an excess of spin-up or spin-down electrons in order to change the properties of a material. Such a population of spin-polarized electrons moving through a common material in a common direction is called a spin current. As described herein, spin currents can be used to operate MRAM memory cells.

[0049] Generally speaking, when a charge current is applied in the longitudinal direction (in-plane direction), the spin Hall effect (SHE) can be used to generate a spin current flowing in the lateral direction (perpendicular to the plane direction). The spin polarization direction of the spin current generated by this spin Hall effect is in the in-plane direction orthogonal to the charge current. The MRAM memory cell 100 includes three terminals A, B and C, a magnetic tunnel junction 102 and a SHE material 120.

[0050] In one embodiment, the magnetic tunnel junction 102 may include a free layer, a tunnel barrier, and a pinned layer. In another embodiment, the magnetic tunnel junction 102 includes a pinned layer 106, an interlayer coupling (ILC) layer 108, a reference layer (RL) 110, a tunnel barrier 112, and a free layer 114. The interlayer coupling layer 108 promotes a strong antiferromagnetic (i.e., antiparallel) coupling between the pinned layer 106 and the reference layer 110, so that their net magnetic moments are mostly offset, thereby greatly reducing the unwanted stray field on the free layer. The spin Hall effect layer 104 includes a heavy metal with a strong SHE, such as platinum, tantalum, or tungsten. The magnetization direction of the free layer 114 switches between upward and downward.

[0051] The advantage of the SOT switching design utilizing the spin Hall effect is that the write current 116 only passes through the spin Hall effect layer 104 and does not flow through the tunnel barrier 112. This eliminates the above-mentioned long-term degradation of the tunnel barrier 112 by the switching current in the existing STT switching design for MRAM memory cells.

[0052] Figure 1E A schematic representation of a voltage controlled magnetic anisotropy (VCMA) MRAM memory cell 120 is depicted. The VCMA memory cell 120 is structurally similar to Figure 1C The STT switching MRAM memory cell 70 is similar to the MRAM memory cell 70, but there are some differences. The MRAM memory cell 120 includes a magnetic tunnel junction 122, which includes an upper ferromagnetic layer 124, a lower ferromagnetic layer 126, a spacer layer (SP) 128, a tunnel barrier 138, and a reference layer 140. The upper ferromagnetic layer 124 is also called a bias layer 124, and the lower ferromagnetic layer 126 is also called a free layer 126.

[0053] When the magnetization direction in the free layer 126 is parallel to the magnetization direction of the reference layer 140, the resistance RP across the MRAM memory cell 120 is low. When the magnetization direction in the free layer 126 is antiparallel to the magnetization direction of the reference layer 140, the resistance R across the MRAM memory cell 120 is low. APis higher. The memory state (“0” or “1”) of the MRAM memory cell 120 is read by measuring the resistance of the MRAM memory cell 120. In this regard, the electrical conductors 142 / 144 attached to the MRAM memory cell 120 are used to read the MRAM data. The reading process is similar to Figure 1C The reading process of the MRAM memory cell 70 is the same, except that the read polarity is selected to be opposite to the write polarity.

[0054] In one embodiment, the write process of the MRAM memory cell 120 is as follows: (1) read the MRAM memory cell 120 to determine its state, (2) if the MRAM memory cell 120 is in the desired write state, terminate the write process, (3) otherwise, apply a write voltage to the MRAM memory cell 120 with a polarity that reduces the magnetic anisotropy of the free layer to almost zero (when the write voltage is applied, the free layer will precess around the magnetic field direction defined by the magnetic bias layer 124), (4) remove the write voltage after 1 / 2 precession cycle, (5) repeat steps 1 to 4 until the correct state is written or the write process times out.

[0055] In an embodiment, the materials and interfaces of the free layer 126, the spacer layer 128, and the tunnel barrier 130 are selected to provide a large VCMA coefficient for the FL 126 to maximize the degree to which the free layer magnetic anisotropy changes with applied voltage. This may require the use of materials that increase the resistance of the tunnel barrier 130. In order to achieve reliable switching, the free layer 126 generally requires a small non-zero in-plane bias field. In one embodiment, the bias field is generated by the bias layer 124. Alternatively, the in-plane field can be supplied by a magnet external to the MRAM memory cell 120, or by an Oersted field generated by passing a current through a wire close to the memory cell.

[0056] Figure 2A Depicted is an MRAM memory cell, e.g. Figures 1A to 1D Example current-to-voltage characteristics of MRAM memory cells 10, 50, 70, and 100. For simplicity, Figure 2A The discussion will refer to Figure 1B In one embodiment, the MRAM memory cell 50 can be in a "parallel state" (P) ( Figure 2A ) and the "antiparallel state" (AP) ( Figure 2A In the parallel state P, the magnetization direction in the free layer 56 is parallel to the magnetization direction of the pinned layer 54. In the antiparallel state AP, the magnetization direction in the free layer 56 is antiparallel to the magnetization direction of the pinned layer 54.

[0057] If the MRAM memory cell 50 is initially in the antiparallel state AP, and an electron write current 64 is applied from the conductor 60 to the conductor 62, the MRAM memory cell 50 will switch to the parallel state P. Figure 2A In the embodiment depicted in FIG. 1 , the MRAM memory cell 50 has a write voltage V between the conductor 62 and the conductor 60 of about 1.12V. AP-P Switch from the antiparallel state AP to the parallel state P.

[0058] Conversely, if the MRAM memory cell 50 is initially in the parallel state P, and an electron write current 64 is applied from the conductor 62 to the conductor 60, the MRAM memory cell 50 will switch to the antiparallel state AP. Figure 2A In the embodiment depicted in FIG. 1 , the MRAM memory cell 50 has a write voltage V between the conductor 62 and the conductor 60 of about 0.36V. P-AP The state switches from the parallel state P to the antiparallel state AP.

[0059] By crossing the MRAM memory cell (e.g. Figure 1B A read reference (eg, a read reference voltage V rd ) to determine the state of the MRAM memory cell. The applied read reference V rd A current proportional to the resistance of the MRAM memory cell is generated. Figure 2A In the embodiment depicted in FIG. 1 , at a read reference V of about 0.56 V rd In the anti-parallel state AP, the MRAM memory cell 50 has an anti-parallel resistance R AP , and in the parallel state P has a parallel resistance R P .

[0060] Therefore, by detecting the individual MRAM memory cells at the read reference V rd The resistance under the resistor and then compare the detected resistance with the threshold resistance value R T The comparison is performed to determine the state of the MRAM memory cell. If the resistance of the MRAM memory cell detected is greater than the threshold resistance value R T , it is determined that the MRAM memory cell is in the antiparallel state AP, and if the detected resistance of the MRAM memory cell is less than the threshold resistance value R T , it is determined that the MRAM memory cell is in the parallel state P.

[0061] This reading technique is referred to herein as “single reference reading” and is very effective for reading individual MRAM memory cells because the antiparallel resistances R AP The value of the parallel resistance R and the value of the parallel resistance RP remain relatively constant. However, the antiparallel resistance R of a group of MRAM memory cells (e.g., in a memory array of MRAM memory cells) AP The values ​​of RP and the parallel resistance RP often show large variations between different MRAM memory cells.

[0062] In addition, if Figure 2B As depicted in Figure 1, the antiparallel resistors R AP The values ​​of RP and parallel resistance RP often overlap. Due to this overlap, a single reference read of a group of memory cells can only accurately distinguish a limited number of bits. For simplicity, the remaining description uses the term "codeword" to describe a group of multiple memory cells that store a predetermined number of data bits (e.g., 16 bits, 64 bits, 1024 bits, or some other number of data bits). In one embodiment, a codeword includes a number of bits that can be encoded and decoded by an error correction coding (ECC) engine (usually on-die).

[0063] For example, Figure 2C Plotting a bit error rate (BER) of a plurality of memory cells (eg, all memory cells in a memory array including a plurality of memory cells) versus a read reference Vr d The graph shows that for all memory cells, an “optimal” read reference V of about 0.73 V is used. rd oAchieve a lowest bit error rate of approximately 7%.

[0064] Although a single read reference V rd oSuccessfully read some codewords, but for others, a single read reference V rd o The number of bit errors that occur may exceed the maximum number of bit errors that can be corrected by the ECC engine. In this case, an alternative reading technique (e.g., self-referenced reading) is typically used to read the codeword.

[0065] For example, a self-referenced read technique first reads a memory cell, then writes it to a first memory state, performs a second read, and then compares the results of the first read with the second read. If the two read results are the same, it is determined that the memory cell was initially in the first memory state and has been in the first memory state. However, if the two read results are different, it is determined that the memory cell was initially in the second memory state but is now in the first memory state. Therefore, a write operation is required to restore the memory cell to the second memory state.

[0066] Although such self-referenced read techniques can be used to successfully read codewords that cannot be successfully read using a single reference read, self-referenced reads require at least two read and one write operations, and may require additional write operations. Therefore, such self-referenced read techniques take a long time to complete and are susceptible to write errors, and may also reduce memory cell endurance.

[0067] To overcome these deficiencies, a two-step read technique for reading MRAM memory cells is proposed. Specifically, a technique for fast two-step read of resistive random access memory cells (e.g., MRAM cells) is described. In one embodiment, a codeword having a specified address is read from a group of memory cells using a two-step read process. In one embodiment, a first read of the codeword is performed using a global read reference value.

[0068] In one embodiment, when performing a first read, a lookup table is searched to determine if there is an entry for the specified address. In one embodiment, the lookup table includes a codeword address and a corresponding individual read reference value. In one embodiment, if the lookup table includes an entry for the specified address, a second read of the codeword is performed using the corresponding individual read reference value for the codeword.

[0069] In one embodiment, the global read reference value is the read reference value selected for reading all codewords of the memory array. In one embodiment, the global read reference value may be determined during wafer sort and has a value that minimizes the number of bit errors for all codewords of the memory array.

[0070] In one embodiment, each individual read reference value is selected to read a corresponding codeword of the memory array. In one embodiment, each individual read reference value may be determined during wafer sort and has a value that minimizes the number of bit errors for that codeword.

[0071] Figure 3A is a flow chart of one embodiment of a process 300a for creating a lookup table of individual read reference values ​​for a set of codewords. In one embodiment, process 300a may be performed by system control circuitry, as described below.

[0072] In one embodiment, the lookup table is addressed by a codeword address. In one embodiment, the set of codewords includes all codewords of the memory array. In another embodiment, the set of codewords includes less than all codewords of the memory array.

[0073] For simplicity, the remaining description assumes that the set of codewords includes m codewords CW i, i = 0, 1, 2, ..., m-1. In addition, for simplicity, the remaining description uses the term "all codewords" to refer to all codewords CW in the set of m codewords. i , and use the term "each code word CW i " is used to represent each codeword CW in the group of m codewords i .

[0074] At step 302, data is written into all codewords CW i In one embodiment, the codeword CW i All bits of the codeword CW may be written to the first memory state (e.g., 0). i All bits of are written to the second memory state (eg, 1).

[0075] At step 304, each codeword CW is sequentially read using a plurality of candidate read reference values. i In one embodiment, n candidate read reference values ​​v are used. rdj Read each codeword CW in turn i , j = 0, 1, 2, ..., n-1. For example, the candidate reading reference value v may be selected based on simulation results, empirical evidence, or a combination of both, or by some other method. rdj .

[0076] At step 306, based on the reading result of step 304, for each codeword CW i and each candidate reads a reference value v rdj Determine the fail bit count FB ij Alternatively, for each codeword CW i and each candidate reads a reference value v rdj Determine the bit error rate BER ij .

[0077] Figure 4A An example where m=8 and n=4 is depicted in table form. In the example shown, codeword CW0 is read using the candidate read reference value v rd0 Read with fail bit count FB 00 , codeword CW1 is read using the candidate reference value v rd2 Read with fail bit count FB 12 , codeword CW5 uses the candidate read reference value v rd1 Read with fail bit count FB 51 ,etc.

[0078] Figure 4B Depicted Figure 4A The table below will give an example of a digital failure bit count value. For example, codeword CW2 uses the candidate read reference value vrd2 Read with fail bit count FB 22 =3, codeword CW4 uses the candidate read reference value v rd0 Read with fail bit count FB 40 =4, codeword CW6 uses the candidate read reference value V rd1 Read with fail bit count FB 61 =2, and so on.

[0079] Reference again Figure 3A At step 308, for each codeword CW i Determine the corresponding single reading reference value v rdi In one embodiment, the codeword CW i The single reading reference value v rdi The one with the lowest fail bit count FB ij , j = 0, 1, 2, ..., n-1. Therefore, use Figure 4B The example value of V corresponds to the single reading reference value rdi as follows:

[0080] Codeword Minimum FBC Corresponding single read reference <![CDATA[CW0]]> <![CDATA[FB 01 =0]]> <![CDATA[v rd0 =in rd1 ]]> <![CDATA[CW1]]> <![CDATA[FB 11 =0]]> <![CDATA[v rd1 =in rd1 ]]> <![CDATA[CW2]]> <![CDATA[FB 20 =0]]> <![CDATA[v rd2 =in rd0 ]]> <![CDATA[CW3]]> <![CDATA[FB 31 =0]]> <![CDATA[v rd3 =in rd1 ]]> <![CDATA[CW4]]> <![CDATA[FB 41 =0]]> <![CDATA[v rd4 =in rd2 ]]> <![CDATA[CW5]]> <![CDATA[FB 53 =0]]> <![CDATA[v rd5 =in rd3 ]]> <![CDATA[CW6]]> <![CDATA[FB 60 =1]]> <![CDATA[v rd6 =in rd0 ]]> <![CDATA[CW7]]> <![CDATA[FB 72 =2]]> <![CDATA[v rd7 =in rd2 ]]>

[0081] Reference again Figure 3A At step 310, determine all code words CW i The global read reference value v rdg In one embodiment, all codewords CW i The global read reference value v rdg The fail bit count FB ij The lowest reading reference value V rdj , i = 0, 1, 2, ..., m-1. Therefore, use Figure 4B Example value, read the reference value v rd1 Make the sum the lowest (20 failed bits), and therefore in this example, make all codewords CW i The global read reference value is v rdg =v rd1 .

[0082] Reference again Figure 3A At step 312, each codeword CW is assigned i The expected maximum fail bit count FB M For example, the expected maximum fail bit count FB M It can be 2, 3, etc.

[0083] At step 314, when using the global read reference value v rdg When reading a codeword, identify the failed bit count FBi jGreater than the expected maximum fail bit count FB M Any codeword CWi (i.e., FB ij >FB M ).

[0084] For example, refer again to Figure 4B Example value if the desired maximum fail bit count FB M = 2 bits, then when using the global read reference value V rdg =V rd1 When reading codewords, the failure bit count FBi of codewords CW2, CW5, and CW7 is j Both are greater than the expected maximum failure bit count FB M Alternatively, if the desired maximum fail bit count FB M = 5 bits, then when using the global read reference value v rdg =v rd1 When reading the codeword, the failure bit count FB of codeword CW5 ij Greater than the expected maximum fail bit count FB M .

[0085] Reference again Figure 3A At step 316, for each codeword CW identified in step 314 i Create entries in the lookup table. In one embodiment, each lookup table entry includes each codeword CW identified in step 314. i The address and the identified codeword CW i The corresponding single reading reference value V rdi .

[0086] Figure 4C The process 300a includes step 316 and uses Figure 4B Sample data and expected FB M =2 bits of the maximum fail bit count. Specifically, the lookup table 400c includes: a first entry, which includes the address of the codeword CW2 and its corresponding individual read reference value V rd0 ; A second entry, the second entry including the address of the code word CW5 and its corresponding separate read reference value v rd3 ; and a third entry, the third entry including the address of the code word CW7 and its corresponding separate read reference value v rd2 .

[0087] Figure 4D The process 300a is depicted as including step 316 and using Figure 4B Sample data and expected FB M==5 bits of maximum fail bit count. Specifically, the lookup table 400d includes a single entry, which includes the address of the codeword CW5 and its corresponding single read reference value v rd3 .

[0088] Thus, without wishing to be bound by any particular theory, it is believed that these examples illustrate the effect of the expected maximum fail bit count FB M decreases, the number of lookup table entries usually increases because when using the global read reference value v rdg Read codeword CW i When more codewords have failed bit counts FB ij May be greater than the expected maximum fail bit count FB M .

[0089] In the above example process 300a, at step 302, all codewords CW i The bits are written to a first memory state (e.g., 0) or a second memory state (e.g., 1). However, for some memory devices, when a codeword is written to the first memory state, the codeword may have a first fail bit count, and when the codeword is written to the second memory state, the codeword may have a second fail bit count. In some embodiments, the first fail bit count may be higher or lower than the second fail bit count. That is, depending on the data written to the codeword, the codeword may have a higher fail bit count.

[0090] Figure 3B is a flow chart of one embodiment of an alternative process 300b for creating a lookup table of individual read reference values ​​for a set of codewords. Figure 3A The example process 300a is similar to that of FIG. 300a, but also includes converting each codeword CW i All bits of write are written to both the first memory state and the second memory state. In one embodiment, process 300b may be performed by system control circuitry, as described below.

[0091] At step 318, during the first write operation, the codeword CW i All bits of are written to the first memory state (eg, 0).

[0092] At step 320, during a first read operation, each codeword CW is sequentially read using a plurality of candidate read reference values. i In one embodiment, n candidate read reference values ​​v are used. rdj Read each codeword CW in turn i , j = 0, 1, 2, ..., n-1. For example, the candidate reading reference value V may be selected based on simulation results, empirical evidence, or a combination of both, or by some other method.rdj .

[0093] At step 322, during a second write operation, the codeword CW i All bits of are written to the second memory state (eg, 1).

[0094] At step 324, during a second read operation, each codeword CW is sequentially read using the same plurality of candidate read reference values ​​used at step 320. i .

[0095] At step 326, based on the reading result of step 320 and the reading result of step 324, for each codeword CW i and each candidate reads a reference value v rdj Determine the fail bit count FB ij In one embodiment, for each codeword CW i , the determined fail bit count FB ij is the larger of the failed bit counts from the first read operation and the second read operation. Alternatively, for each codeword CW i and each candidate reads a reference value v rdj Determine the bit error rate BER ii In one embodiment, for each codeword CW i , the bit error rate BER determined ij is the larger of the bit error rates from the first read operation and the second read operation.

[0096] The remaining steps 308 to 316 of process 300b are the same as those of process 300a, except that the fail bit count FB determined at step 326 is used. ij , the failure bit count is based on each codeword CW i Without wishing to be bound by any particular theory, it is believed that process 300b may be able to identify a single read reference value that works for both the first memory state and the second memory state.

[0097] At steps 302, 318 and 322 in the above example processes 300a and 300b, the codeword CW i All bits of the codeword CW are written into the first memory state or the second memory state. i All bits of are written to the first memory state or the second memory state. It will be understood by those skilled in the art that the codeword bits may also be written using a predetermined random bit pattern (e.g., "0101010...", "0100010...", "11110000..." or some other predetermined random bit pattern).

[0098] In one embodiment, the above example processes 300a and 300b may be performed once. For example, the above processes 300a and 300b may be performed at wafer sort time, and the lookup table entries created at step 316 may be stored in a lookup table stored in a memory on a memory die. In one embodiment, the lookup table may be stored in a compact, fast memory structure such as a content addressable memory (CAM) or other type of memory.

[0099] In other embodiments, the above example processes 300a and 300b may be performed more than once. For example, the above example processes 300a and 300b may be initially performed at wafer sort to create initial entries in the lookup table, and may then be performed one or more additional times to create updated or additional entries in the lookup table over the life of the memory die.

[0100] Figure 5 is a flow chart of one embodiment of a process 500 for reading a codeword. In one embodiment, process 500 may be performed by system control circuitry, as described below.

[0101] At step 502, a read command is received. For example, the host device may send a read command to the memory controller. In one embodiment, the read command specifies one or more codewords CW to be read. i For simplicity, the following description assumes that the received read command specifies a single codeword CW i ADD a single address i .

[0102] At step 504, a global read reference value is used, for example, Figure 3A Course 300a or Figure 3B The global read reference value v determined at step 310 of process 300b rdg , to read the specified address ADD i The code word CWi at .

[0103] At step 506, the user Figure 3A Course 300a or Figure 3B The lookup table created by process 300b (e.g., Figure 4C Lookup table 400c or Figure 4D The lookup table 400d) is searched to determine whether the lookup table includes a i Entries.

[0104] At step 508, it is determined whether the lookup table includes a matching address ADD i If it is determined at step 508 that the lookup table does not include a matching address ADDi , then at step 510, the matching flag I m Set equal to 0.

[0105] If it is determined at step 508 that the lookup table includes a matching address ADD i , then at step 512, the address ADD is retrieved from the lookup table i The corresponding single read reference value v of the code word at rdi At step 514, the matching flag I m Set equal to 1.

[0106] In one embodiment, steps 506 to 514 are performed in parallel with the read operation performed in step 504. In one embodiment, steps 506 to 514 are performed while the read operation is performed in step 504. In one embodiment, the time required to complete steps 506 to 514 is less than or equal to the time required to complete the read operation performed in step 504.

[0107] At step 516, the data read from step 504 is decoded (eg, using an ECC engine). The ECC engine can typically successfully correct the maximum number of bit errors. M (For example, ECC M =5 bits, 7 bits, or some other maximum number of bits). If the data read at step 504 includes less than or equal to the ECC M bit error, the ECC engine can successfully correct the error and the ECC operation will pass. However, if the data read in step 504 includes more than the ECC M If a bit error occurs, the ECC engine cannot successfully correct the error and the ECC operation will fail.

[0108] At step 518, it is determined whether the ECC operation performed at step 516 passes. If it is determined at step 518 that the ECC engine successfully decodes the data read at step 504, then at step 520 the decoded data is provided to the host.

[0109] However, if it is determined at step 518 that the ECC engine cannot successfully decode the data read at step 504, then a match flag is determined at step 522. m Is it equal to l?

[0110] If it is determined at step 522 that the match flag I m is not equal to 1, then a failure handling operation may be performed at step 524. In one embodiment, the failure handling operation may include i Performs an overwrite read operation.

[0111] However, if the match flag I is determined at step 522 m is equal to 1, then read the specified address ADD again at step 526 i The codeword CWi at step 512 is obtained, but this time using the corresponding single read reference value v retrieved from the lookup table at step 512. rdi .

[0112] At step 528, the match flag I m is set equal to 0, and process 500 then loops back to step 516 to decode the read data, this example using the data read at step 526 .

[0113] At step 518, it is determined whether the ECC operation performed at step 516 passes. If it is determined at step 518 that the ECC engine successfully decodes the data read at step 504, then at step 520 the decoded data is provided to the host.

[0114] However, if it is determined at step 518 that the ECC engine cannot successfully decode the data read at step 504, then a match flag is determined at step 522. m Is it equal to 1? Because the match flag I m is reset to 0 at step 528 so that a failure handling operation may be performed at step 524, for example, as described above.

[0115] Without wishing to be bound by any particular theory, it is believed that process 500 can be modified to reduce the time required to read a codeword. Specifically, first use the global read reference value v rdg To read the specified codeword CW i Without wishing to be bound by any particular theory, it is believed that in many cases this reading will be successful even if the lookup table includes a codeword CW for a given i Address ADD i Entries.

[0116] However, if the first read is unsuccessful and if the look-up table includes a i Address ADD i entry, the corresponding single read reference value v retrieved from the lookup table is used rdi To read the specified code word CW for the second time i Without wishing to be bound by any particular theory, it is believed that this second read will be successful because the second read uses the designated codeword CW i Specially selected individual reading reference value v rdi .

[0117] However, if the second read is unsuccessful, a failure handling operation may be used, such as iPerform an overwrite read to successfully read the codeword.

[0118] Without wishing to be bound by any particular theory, it is believed that because the lookup table is searched and any matching individual read reference values ​​v are retrieved in parallel with the first read operation rdi , so the second read operation can be performed quickly and without repeated attempts to find a codeword CW that can be used to successfully read the specified codeword i Optional reading of reference value.

[0119] Furthermore, without wishing to be bound by any particular theory, it is believed that the described process can eliminate or substantially reduce the number of rewrite read operations required to read codewords on a memory die.

[0120] Figure 6 is a flow chart of one embodiment of a process 600 for reading a codeword. In one embodiment, process 600 may be performed by system control circuitry, as described below.

[0121] At step 602, a command is received to read a first codeword from a memory array, the memory array comprising a plurality of codewords collectively having a corresponding first predetermined read reference value and each codeword having a corresponding second predetermined read reference value.

[0122] At step 604, the first codeword is read using the first predetermined read reference value to provide first read data.

[0123] At step 606, when the first codeword is read, the second predetermined read reference value corresponding to the first codeword is retrieved from a lookup table.

[0124] At step 608, the first read data is decoded using an error correction code algorithm.

[0125] At step 610, it is determined that the error correction code algorithm cannot correctly decode the first read data.

[0126] At step 612, the first codeword is read using the retrieved second predetermined read reference value to provide second read data.

[0127] Figure 7 700. Memory system 700 includes a memory array 702, which may include any of the memory cells described above. Array terminal lines of memory array 702 include individual word line layers organized into rows, and individual bit line layers organized into columns. However, other orientations may also be implemented.

[0128] The memory system 700 includes a row control circuit 704 whose output 706 is connected to a corresponding word line of the memory array 702. The row control circuit 704 receives a set of M row address signals and one or more various control signals from the system control logic circuit 708, and may generally include circuits such as a row decoder 710, an array terminal driver 712, and a block selector circuit 714 for both read operations and write operations.

[0129] The memory system 700 also includes column control circuitry 716 having input / output 718 connected to corresponding bit lines of the memory array 702. The column control circuitry 718 receives a set of N column address signals and one or more various control signals from the system control logic 708 and may generally include circuitry such as a column decoder 720, array terminal receivers or drivers 722, block selector circuitry 724, and read / write circuits and I / O multiplexers.

[0130] The system control logic 708 receives data and commands from the host and provides output data and status to the host. In other embodiments, the system control logic 708 receives data and commands from a separate controller circuit and provides output data to the controller circuit, wherein the controller circuit communicates with the host. The system control logic 708 may include one or more state machines, registers, and other control logic for controlling the operation of the memory system 700.

[0131] In one embodiment, the system control logic component 708 includes a memory structure (not shown) for storing the lookup tables described above, such as Figure 4C and Figure 4D Lookup tables 400c and 400d. In one embodiment, the memory structure is a compact fast memory structure. In one embodiment, the memory structure is a content addressable memory. Other memory structures may also be used.

[0132] In other embodiments, a portion of the memory array 702 may be used to store the lookup tables described above, such as Figure 4C and Figure 4D Lookup tables 400c and 400d.

[0133] In one embodiment, Figure 7 All components depicted in are arranged on a single integrated circuit. For example, system control logic 708, column control circuitry 716, and row control circuitry 704 are formed on a surface of a substrate, and memory array 702 is formed on or above the substrate.

[0134] In one embodiment, one or more of the memory array 702, the row control circuit 704, the column control circuit 716, and the system control logic 708 constitute a system control circuit that includes a plurality of logic components configured to respectively implement the system control circuit. Figure 3A , Figure 3B , Figure 5 and Figure 6 The hardware and / or software of methods 300a, 300b, 500 and 600.

[0135] Although the techniques described above and Figure 3A , Figure 3B , Figure 5 and Figure 6 Although the exemplary methods 300a, 300b, 500 and 600 are described above, one of ordinary skill in the art will appreciate that the techniques can be used with any resistance switching memory cell, including MRAM memory cells, phase change memory cells, reversible resistance switching random access memory cells (ReRAM), and other resistance switching memory cells. In addition, the techniques described above and Figure 3A , Figure 3B , Figure 5 and Figure 6 The example methods 300a, 300b, 500, and 600 may each be used with a block of multiple resistance-switching memory cells.

[0136] One embodiment includes a device including a memory array and a control circuit. The memory array includes nonvolatile memory cells, each of which includes a resistive random access memory element. The control circuit is configured to receive a read command that specifies an address of a first group of nonvolatile memory cells; perform a first read of the first group of nonvolatile memory cells using a first predetermined read reference value to provide first read data; when performing the first read, retrieve a second predetermined read reference value corresponding to the specified address from a memory; and in response to satisfying a condition regarding the first read data, perform a second read of the first group of nonvolatile memory cells using a second predetermined read reference value to provide second read data.

[0137] One embodiment includes a method comprising: receiving a command to read a first codeword from a memory array, the memory array comprising a plurality of codewords, the plurality of codewords collectively comprising a corresponding first predetermined read reference value, each of the codewords comprising a corresponding second predetermined read reference value; reading the first codeword using the first predetermined read reference value to provide first read data; retrieving the second predetermined read reference value corresponding to the first codeword from a lookup table when reading the first codeword; decoding the first read data using an error correction code algorithm; determining that the error correction code algorithm cannot correctly decode the first read data; and reading the first codeword using the retrieved second predetermined read reference value to provide second read data.

[0138] One embodiment includes a system comprising a control circuit configured to: receive a command from a host device to read a first group of non-volatile memory cells in a memory array including multiple groups of memory cells; perform a first read of the first group of non-volatile memory cells using a first predetermined read reference value; retrieve a second predetermined read reference value from a lookup table; and selectively perform a second read of the first group of non-volatile memory cells using the second predetermined read reference value only if data read during the first read fails to decode correctly. The retrieving step is performed in parallel with the first reading step. The first predetermined read reference value minimizes a fail bit count of memory cells of all groups. The second predetermined read reference minimizes a fail bit count of the first group of memory cells.

[0139] For the purposes of this document, references in the specification to "an embodiment," "one embodiment," "some embodiments," or "another embodiment" may be used to describe different embodiments or the same embodiment.

[0140] For the purposes of this document, a connection may be a direct connection or an indirect connection (e.g., via one or more other components). In some cases, when an element is referred to as being connected or coupled to another element, the element may be directly connected to the other element, or indirectly connected to the other element via an intervening element. When an element is referred to as being directly connected to another element, there are no intervening elements between the element and the other element. Two devices are "in communication" if they are directly connected or indirectly connected, such that they are able to communicate electronic signals between them.

[0141] For the purposes of this document, the term "based on" may be understood as "based, at least in part, on."

[0142] For the purposes of this document, the use of numerical terms such as a "first" object, a "second" object, and a "third" object without additional context may not imply an ordering of the objects, but may be used for identification purposes to identify different objects.

[0143] For the purposes of this document, the term "group" of objects may refer to a "group" of one or more objects.

[0144] The above detailed description has been provided for the purpose of illustration and description. It is not intended to be exhaustive or to limit the precise form disclosed by the invention. Many modifications and variations are possible in light of the above teachings. The described embodiments are selected to best explain the principles of the proposed technology and its practical application, thereby enabling others skilled in the art to best utilize it in various embodiments and in various modifications suitable for the specific use contemplated. The scope of the invention is intended to be defined by the appended claims.

Claims

1. A device, comprising: a memory array comprising a plurality of nonvolatile memory cells, each nonvolatile memory cell comprising a resistive random access memory element; and A control circuit, the control circuit being configured to: receiving a read command specifying an address of a first group of non-volatile memory cells; performing a first read of the first group of nonvolatile memory cells using a first predetermined read reference value to provide first read data; When performing the first read, retrieving a second predetermined read reference value corresponding to the specified address from the memory; as well as In response to the condition regarding the first read data being satisfied, a second read of the first group of non-volatile memory cells is performed using the second predetermined read reference value to provide second read data. 2 . The apparatus of claim 1 , wherein the first predetermined read reference value is selected from a plurality of candidate read reference values.

3. The device according to claim 1, wherein: The memory array includes a plurality of groups of nonvolatile memory cells; and The first predetermined read reference value is selected based on a result of reading all groups of nonvolatile memory cells using a plurality of candidate read reference values.

4. The apparatus according to claim 3, wherein the first predetermined read reference value is selected by: For each candidate read reference value, determining a total number of bit errors resulting from reading each group of nonvolatile memory cells; and The first predetermined read reference value is set as the candidate read reference value having the lowest total number of bit errors. 5 . The apparatus of claim 1 , wherein the second predetermined read reference value is selected from a plurality of candidate read reference values.

6. The device according to claim 1, wherein: The memory array includes a plurality of groups of nonvolatile memory cells; and The second predetermined read reference value is selected based on a result of reading the first group of non-volatile memory cells using a plurality of candidate read reference values.

7. The apparatus of claim 6, wherein the second predetermined read reference value is selected by: for each candidate read reference value, determining a total number of bit errors resulting from reading the first group of nonvolatile memory cells; and The second predetermined read reference value is set as the candidate read reference value having the lowest total number of bit errors.

8. The device of claim 1, wherein the RRAM element comprises a MRAM element.

9. The apparatus of claim 1, wherein the memory array comprises: a plurality of first conductive lines; and A plurality of second conductive lines, each of the plurality of nonvolatile memory cells being connected between a first conductive line among the first conductive lines and a second conductive line among the second conductive lines. 10 . The apparatus of claim 1 , wherein the memory comprises a lookup table configured to store addresses and corresponding second predetermined read reference values.

11. The apparatus of claim 1, wherein the memory comprises a content addressable memory.

12. The device according to claim 1, wherein: The apparatus further comprises a device configured to decode data read from the non-volatile memory cell; and The condition satisfied includes failure to correctly decode the first read data.

13. The apparatus of claim 12, wherein the device comprises an error correction coding engine.

14. A method comprising: receiving a command to read a first codeword from a memory array, the memory array comprising a plurality of codewords, the plurality of codewords collectively comprising a corresponding first predetermined read reference value, each of the codewords comprising a corresponding second predetermined read reference value; reading the first codeword using the first predetermined read reference value to provide first read data; When reading the first codeword, retrieving the second predetermined reading reference value corresponding to the first codeword from a lookup table; Decoding the first read data using an error correction code algorithm; determining that the error correction code algorithm cannot correctly decode the first read data; as well as The first codeword is read using the retrieved second predetermined read reference value to provide second read data.

15. The method according to claim 14, further comprising: decoding the second read data using the error correction code algorithm; as well as It is determined that the error correction code algorithm can correctly decode the second read data.

16. The method of claim 14, wherein the lookup table comprises a content addressable memory. 17 . The method of claim 14 , wherein a time required to retrieve the second predetermined read reference value from the lookup table is less than or equal to a time required to complete reading of the first codeword.

18. The method of claim 14, wherein the memory array comprises a plurality of nonvolatile memory cells, each nonvolatile memory cell comprising a magnetoresistive random access memory element.

19. A system, comprising: A control circuit, the control circuit being configured to: receiving a command from a host device to read a first group of non-volatile memory cells in a memory array including a plurality of groups of memory cells; performing a first read of the first group of nonvolatile memory cells using a first predetermined read reference value; retrieving a second predetermined read reference value from the lookup table; as well as selectively performing a second read of the first group of nonvolatile memory cells using the second predetermined read reference value only when the data read during the first read fails to be correctly decoded, in: The retrieving step is performed in parallel with the first reading step; The first predetermined read reference value minimizes a fail bit count of all groups of memory cells; and The second predetermined read reference minimizes a fail bit count of the first group of memory cells.

20. The system of claim 19, wherein each of the nonvolatile memory cells comprises a magnetoresistive random access memory element.