Read threshold calibration for long-term sequential reading at cross temperature
Read level tracking (RLT) operations adjust read thresholds based on bit error rate to maintain QoS in memory devices, addressing process non-uniformity and temperature variations, enhancing read performance and reducing latency.
Patent Information
- Application Number
- JP2024531088
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2022-04-06
- Filing Date
- 2023-04-04
- Publication Date
- 2024-12-09
- Estimated Expiration
- 2043-04-04
Smart Images

Figure 0007596604000013 
Figure 0007596604000014 
Figure 0007596604000015
Abstract
Description
[Technical field]
[0001] (CROSS REFERENCE TO RELATED APPLICATIONS) This application claims the benefit of the entire contents of U.S. Non-provisional Application No. 17 / 714,379, filed April 6, 2022, and entitled “Read Threshold Calibration for Extended Sequential Reads at Cross Temperature,” which is incorporated by reference herein for all purposes. [Background technology]
[0002] Apparatus and methods consistent with example embodiments relate to techniques for improving read performance in memory devices.
[0003] In recent years, Quality of Service (QOS) requirements have increased, dictating the latency / probability profile that storage devices target. To improve QOS, system solutions strive to improve performance not just over the average scenario, but over a wide range of different scenarios and command combinations.
[0004] Difficulties in maintaining process uniformity brought about by NOT-AND (NAND) process shrink and three-dimensional (3D) stacking can pose additional challenges in maintaining QOS standards. Lack of process uniformity can result in increased variability, for example, between memory dies, blocks, and pages, as well as between different endurance, retention, temperature, and disturbance conditions. Operation under such diverse conditions requires the use of adaptive systems, particularly in terms of tracking optimal read thresholds to ensure read throughput and input / output operations per second (IOPS) requirements are met and the desired QOS is ensured.
[0005] Because cross-temperature (writing at one temperature and reading at another) affects the flash read threshold, read threshold calibration is often required to read the target word line (WL) at a low bit error rate (BER). This challenge persists in many embedded storage devices and is especially relevant for Internet of Things (IOT) technologies and automotive devices.
[0006] During long sequential reads where the temperature shifts during the read, the system needs to adapt to the temperature shift. Read threshold calibration can be performed through a process of applying a large number of senses and then sequentially calculating syndrome weights (SW) as estimates of the bit error rate (BER) according to different hypotheses related to the number of senses. This process can be called a BER estimation scan (BES).
[0007] An exemplary BES process uses a low density parity check (LDPC) engine to calibrate read threshold levels to reduce the BER. Data programmed into the memory must satisfy a set of parity check constraints. The syndrome weight (SW) is the number of parity check constraints that are not satisfied, and the SW of a read codeword provides an estimate of its BER. For example, in a memory device programmed according to a QLC (4 bits per memory cell) scheme with 16 possible states, reading the middle page requires sensing at four different transition points. Around each of these transition points, several reads are sampled and the combination of read thresholds that minimizes the BER is identified.
[0008] Alternatively, the valley of the cell voltage distribution (CVD) can be determined by measuring the difference in cell conduction between different voltage points in a process called valley search (VS).
[0009] Solutions such as BES and VS are high-latency operations aimed at optimizing the read threshold for a particular WL, and are suitable for read recovery flows, but may not be very effective in maintaining a rapidly shifting read threshold. It is not practical for a system to run BES or VS at high frequencies, as this would significantly degrade performance. Summary of the Invention
[0010] Exemplary embodiments may address any problems and / or shortcomings not described above, as well as other shortcomings, and exemplary embodiments need not overcome the above shortcomings, and may not overcome any of the above problems.
[0011] In accordance with one or more exemplary embodiments, a read level tracking (RLT) operation is used to provide low accuracy, low latency operations that can be performed only if the decode is successful.
[0012] According to one aspect of an exemplary embodiment, a method for calibrating a read threshold voltage includes performing a plurality of read operations, determining to perform a read level tracking (RLT) method, and performing the RLT method, the RLT method including determining a bit error rate (BER) of an instructional word line (WL), determining an adjusted read threshold level based on the BER, and adjusting the read threshold level according to the adjusted read threshold level.
[0013] According to an aspect of another exemplary embodiment, a system for calibrating a read threshold voltage includes a controller in operative communication with at least one memory die, the controller including an interface configured to receive a number of page bits corresponding to a plurality of read operations from the at least one memory die, and a processor configured to determine to perform a read level tracking (RLT) method and to perform the RLT method, the RLT method including determining a bit error rate (BER) of an instructional word line (WL), determining an adjusted read threshold based on the BER, and adjusting a read threshold level according to the adjusted read threshold level.
[0014] According to an aspect of another exemplary embodiment, a system for calibrating a read threshold includes a control means in operative communication with at least one memory die, the control means including an interface means configured to receive a number of page bits corresponding to a number of read operations from the at least one memory die, and a processing means configured to determine to perform a read level tracking (RLT) method and to perform the RLT method, the RLT method including the steps of: determining a bit error rate (BER) of an instructional word line (WL); determining an adjusted read threshold based on the RLT method; and adjusting a read threshold level according to the adjusted read threshold level. [Brief description of the drawings]
[0015] The above and / or other aspects will become apparent and more readily understood from the following description of exemplary embodiments taken in conjunction with the accompanying drawings.
[0016] [Figure 1A] FIG. 1 is a block diagram of an exemplary memory system.
[0017] [Figure 1B] FIG. 2 is a block diagram of an example storage module including multiple memory systems.
[0018] [Figure 1C] FIG. 1 is a block diagram of an exemplary hierarchical storage system.
[0019] [Figure 2A] 1B is a block diagram of an exemplary configuration of components of a controller of the memory system of FIG. 1A.
[0020] [Figure 2B] 1B is a block diagram of an exemplary configuration of components of a memory die of the memory system of FIG. 1A.
[0021] [Diagram 3] 1 is a flow chart of a read threshold calibration method according to the prior art.
[0022] [Figure 4A] 1 is a plot of threshold voltage distribution curves for different storage schemes. [Figure 4B] Same as above. [Figure 4C] Same as above. [Figure 4D] Same as above.
[0023] [Diagram 5] FIG. 1 illustrates multiple sense operations performed to read a lower page of data stored in a 3-bit per cell MLC cell.
[0024] [Figure 6A] The decrease in read threshold level for SLC is shown. [Figure 6B] 1 shows an increase in read threshold levels for SLC.
[0025] [Figure 7] 1 is a flowchart of a read threshold calibration method according to an example embodiment.
[0026] [Figure 8] 11 is a flowchart of a read threshold calibration method according to another exemplary embodiment.
[0027] [Figure 9] FIG. 13 illustrates how the read threshold change can be determined for a single read threshold voltage.
[0028] [Figure 10] 11 is a flowchart of a read threshold calibration method according to another exemplary embodiment.
[0029] [Figure 11] 11 is a flowchart of a read threshold calibration method according to another exemplary embodiment.
[0030] [Figure 12] 11 is a flowchart of a read threshold calibration method according to another exemplary embodiment.
[0031] [Figure 13] 1 illustrates an example result of a read threshold calibration method in accordance with one or more example embodiments. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0032] Reference will now be made in detail to the exemplary embodiments, as illustrated in the accompanying drawings, in which like reference numerals refer to like elements throughout, and in this regard, the exemplary embodiments may have different forms and should not be construed as limited to the description set forth herein.
[0033] As used herein, the terms "INCLUDE," "INCLUDING," "COMPRISE," and / or "COMPRISING" specify the presence of stated features, integers, steps, operations, elements, and / or components, but will be understood as not excluding the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.
[0034] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items. Placing a phrase such as "at least one of" after a list of elements modifies the entire list of elements and does not modify the individual elements of the list. Also, terms such as "unit," "English-ER (-OR)," and "module" used herein refer to an element for performing at least one function or operation, and may be implemented in hardware, software, or a combination of hardware and software.
[0035] 1 illustrates a block diagram of a memory system 100 according to an example embodiment that may perform the methods described herein. The memory system 100 includes a controller 102 and a memory including one or more memory dies 104. As used herein, the term die refers to a set of memory cells and associated circuitry for managing the physical operations of those memory cells formed on a single semiconductor substrate. The controller 102 may interface with a host system and send command sequences to the memory dies 104 for read, program, and erase operations.
[0036] The controller 102, which is a flash memory controller, may take the form of, for example, any one of a processing circuit, a microprocessor or processor and a computer readable medium storing computer readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, an application specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller. The controller 102 may be configured with hardware and / or firmware to perform various functions described below and illustrated in the flow diagrams of the figures described herein. For example, the controller may control a read threshold level in a non-volatile memory. Additionally or alternatively, some of the components shown as being internal to the controller may be stored external to the controller and other components may be used. Additionally or alternatively, the phrase "operably in communication with" may mean to communicate directly or indirectly (wired or wirelessly) through one or more components that may or may not be shown or described herein.
[0037] As described herein, the controller 102 is a device that manages data stored in the memory dies and communicates with a host, such as a computer or electronic device. The controller 102 may have any of a variety of additional functions in addition to the specific functions described herein. For example, the controller 102 may format the memory dies 104 to ensure that they operate properly, map out bad flash memory cells, and allocate spare cells to replace future failed cells. Some portions of the spare cells may be used to hold firmware to operate the controller 102 and implement other features. In operation, when the host needs to read data from or write data to the memory die 104, the host will communicate with the controller 102. If the host provides a logical address where data is to be read / written, the controller 102 can translate the logical address received from the host into a physical address within the memory die 104. Alternatively, the host may provide a physical address. Controller 102 may also perform various memory management functions such as, but not limited to, wear leveling (distributing writes to avoid wearing out certain blocks of memory that are otherwise repeatedly written to) and garbage collection (the process of moving only valid pages of data from a full block to a new block so that the full block can be erased and reused). Controller 102 may also control the methods and operations described below.
[0038] The interface between the controller 102 and the non-volatile memory die 104 may be any suitable interface, such as a flash interface. The memory system 100 may be a card-based system, such as a Secure Digital (SD) or Micro Secure Digital (microSD) card. Alternatively, the memory system 100 may be part of an embedded memory system.
[0039] 1A, memory system 100 is shown to include a single channel between controller 102 and non-volatile memory die 104. However, this is not intended to be limiting and the memory system may include more than one channel, such as a memory system embodying a NAND architecture and including two, four, eight, or more channels between controller 402 and memory die 104. In any one or more of the example embodiments described herein, even though a single channel is illustrated, multiple channels may exist between controller 102 and memory die 104.
[0040] 1B is a block diagram of a storage module 200 including multiple memory systems, according to an example embodiment. Thus, the storage module 200 includes a storage controller 202 that interfaces with a host and a storage system 204 including multiple non-volatile memory systems 200. The interface between the storage controller 202 and the non-volatile memory system 100 can be a bus interface, such as, for example, a Serial Advanced Technology Attachment (SATA), a Peripheral Component Interface Express (PCIE) interface, an Embedded Multimedia Card (EMMC) interface, an SD interface, or a Universal Serial Bus (USB) interface. The storage module 200 can be a solid-state drive (SSD) such as those found in portable computing devices such as laptop computers, tablet computers, and mobile phones.
[0041] 1C is a block diagram of an exemplary hierarchical storage system 210. The hierarchical storage system 210 includes multiple storage controllers 202, each of which controls a respective storage system 204. The host systems 212 may access memory in the hierarchical storage system 210 through a bus interface. Exemplary bus interfaces may include, for example, a Non-Volatile Memory Express (NVME), a Fibre Channel over Ethernet (FCOE) interface, an SD interface, a USB interface, a SATA interface, a PCIE interface, or an eMMC interface. The storage system 210 may be a rack-mounted mass storage system accessible by multiple host computers, such as those found in data centers or other locations where mass storage is needed.
[0042] 2A is a block diagram of an arrangement of components of the controller 102 of the memory system 100 of FIG. 1A. The controller 102 may include a front-end module 108 that interfaces with a host, a back-end module 110 that interfaces with the non-volatile memory die 104, and various other modules that perform various functions of the non-volatile memory system 100. As described herein, the modules may be hardware or a combination of hardware and software. For example, each module may include an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a circuit, a digital logic circuit, an analog circuit, a combination of discrete circuits, gates, or any other type of hardware, or any combination thereof. Additionally or alternatively, each module may include memory hardware that includes instructions executable by a processor or processor circuitry to implement one or more features of the module. If any one of the modules includes a portion of memory that includes instructions executable by a processor, the module may or may not include a processor. In some examples, each module may simply be a portion of memory that includes instructions executable by a processor to implement the features of the corresponding module without the module including any other hardware. Each module may be referred to as a hardware module because each module includes at least some hardware, even if the included hardware includes software.
[0043] The controller 102 may include a buffer manager / bus controller module 114 that manages buffers in a random access memory (RAM) 116 and controls internal bus arbitration for communications on an internal communication bus 117 of the controller 102. A read only memory (ROM) 118 may store and / or access system boot code. Although shown as being located separately from the controller 102, one or both of the RAM 116 and the ROM 118 may be located within the controller 102. According to one or more exemplary embodiments, portions of the RAM 116 and the ROM 118 may be located both internal to the controller 102 and external to the controller 102. According to one or more exemplary embodiments, the controller 102, the RAM 116, and the ROM 118 may be located on separate semiconductor dies.
[0044] The front-end module 108 may include a host interface 120 and a physical layer interface (PHY) 122 that provide an electrical interface with a host or next level storage controller. The type of host interface 120 may depend on the type of memory being used. Exemplary types of host interface 120 may include, but are not limited to, SATA, SATA EXPRESS, SAS, Fibre Channel, USB, PCIE, and NVME. The host interface 120 may facilitate the transfer of data, control signals, and timing signals.
[0045] The back-end module 110 may include an error correcting code (ECC) engine or module 124 that encodes data bytes received from the host and performs decoding and error correction on data bytes read from the non-volatile memory die 104. The ECC engine may be a low density parity check (LDPC) engine that uses a linear error correcting code. The back-end module 110 may also include a command sequencer 126 that generates command sequences, such as program, read, and erase command sequences, that are sent to the non-volatile memory die 104. The back-end module 110 may include a RAID (REDUNDANT ARRAY OF INDEPENDENT DRIVES) module 128 that manages the generation of RAID parity and recovery of failed data. RAID parity may be used as an additional level of integrity protection for data being written to the non-volatile memory system 100. The RAID module 128 may be part of the ECC engine 124. The memory interface 130 provides command sequences to the non-volatile memory die 104 and receives status information from the non-volatile memory die 104. Data to be programmed into and read from the non-volatile memory die 104, along with command sequences and status information, may be communicated through memory interface 130. According to one or more exemplary embodiments, memory interface 130 may be a double data rate (DDR) interface and / or a toggle mode 200, 400, 800, or higher interface. Control layer 132 may control the overall operation of backend module 110.
[0046] Additional modules of non-volatile memory system 100 may include a media management layer 138 that performs certain memory management functions, such as wear leveling, address management, and facilitating folding operations of memory cells of memory die 104. Non-volatile memory system 100 may also include other discrete components 140, such as an external electrical interface, external RAM, resistors, capacitors, or other components that may interface with controller 102. According to one or more exemplary embodiments, one or more of RAID module 128, media management layer 138, and buffer management / bus controller 114 are optional components that may be omitted.
[0047] 2B is a block diagram of an arrangement of components of the memory die 104 of the memory system of FIG. 1A. The memory die 104 includes a memory cell structure 142 that includes a plurality of memory cells, interchangeably referred to as memory elements. A memory cell is an element or component that stores a unit of data having an N-bit data value, where N is 1 or greater. Any suitable type of memory may be used for the memory cells of the memory cell structure 142. By way of example, the memory may be a dynamic random access memory ("DRAM") or static random access memory ("SRAM"), a resistive random access memory ("RERAM"), an electrically erasable programmable read only memory ("EEPROM"), a flash memory (which may be considered a subset of EEPROM), a non-volatile memory such as a ferroelectric random access memory ("FRAM"), a magnetoresistive random access memory ("MRAM"), a phase change memory ("PCM"), or any other element that includes a semiconductor material or other material capable of storing information. Each type of memory may have a different arrangement. For example, a flash memory may be arranged in a NAND arrangement or a NOR arrangement.
[0048] The memory may be formed from passive and / or active elements in any one or more combinations, as will be appreciated by those of skill in the art.
[0049] The memory cells may be configured such that they are connected in series or each element is individually accessible. As a non-limiting example, flash memory devices in a NAND configuration (NAND memory) typically contain memory elements connected in series. NAND memory arrays may be configured such that the array consists of multiple cell groups, in which memory cells that are part of the same cell group share a single bias line, such as a single word line or a single bit line, and are accessed or biased as a group.
[0050] The memory cells forming the memory cell structure 142 of a memory die may be disposed in and / or on a substrate, which may be a wafer over which layers of memory cells are formed, or may be a carrier substrate to which the memory cells are attached after they are formed.
[0051] The memory cells forming memory cell structure 142, or at least a portion of memory cell structure 142, may be arranged two-dimensionally or three-dimensionally in the two dimensions (2D) of a 3D memory cell structure.
[0052] In the prior art method, BES or VS is performed on the representative WL every time WL indicates BER rise. As the indicative WL, the WL at the edge of the block may be selected so that the BER rise is quickly captured. In the scenario of temperature shift during long-term sequential reading, these events will be triggered frequently, and the frequent operation of BES will cause latency, resulting in poor system performance.
[0053] Such a prior art method is shown, for example, in the flow chart of FIG. 3. As described above, this prior art method is triggered by an indication (301) that there is a rise in BER at an indicative WL, which may be a WL at an edge of a block. For example, a rise above a certain threshold, or alternatively, a rise of a certain amount from a previous BER, may trigger this series of read threshold calibration operations. A BES is then performed (302) for the representative WL. The read threshold obtained by the BES is then used for the remainder of the time tag (303) and may be adjusted for WLs other than the representative WL based on the WL index (304).
[0054] In accordance with one or more exemplary embodiments, the read threshold is rate determined and used adjustable for long sequential reads in scenarios where the temperature may shift rapidly during the read. Small shifts in the read threshold (also referred to herein as "tracking") are suitable for such scenarios.
[0055] In comparison to BES and VS, as mentioned above, an approach called Read Level Tracking (RLT) involves performing calibration based on using the host read itself, rather than based on taking multiple readings, such as with VS and BES.
[0056] In implementation, a set of memory cells programmed to the same memory state may have an associated set of threshold voltages as a result of programming, which may be represented as a probability distribution of threshold voltages or simply a threshold distribution of the number of memory cells in the set as a function of threshold voltage.
[0057] A given storage scheme may have a model, reference, ideal, or target threshold voltage distribution, which may be in the form of a continuous probability distribution, such as a Gaussian distribution, for each of the memory states and associated threshold voltage ranges defined by the given storage scheme. The model threshold voltage distribution may represent a model of an ideal threshold voltage distribution for a collection of memory cells programmed to the same memory state. The lower and upper tails of the model threshold voltage distribution may align with or correspond to the upper and lower threshold voltage levels of the ranges with which each of the model threshold voltage distributions is associated.
[0058] 4A-4D are plots of threshold voltage distribution curves of different storage schemes for storing different numbers of bits of data in memory cells. The threshold voltage distribution curves are plotted versus number of memory cells as a function of threshold voltage VTH, with each curve representing a unique data state associated with one or more bits of data. FIG. 4A shows a threshold voltage distribution curve for an SLC storage scheme (1 bit of data per memory cell). FIG. 4B shows a threshold voltage distribution curve for an MLC storage scheme (2 bits of data per memory cell). FIG. 4C shows a threshold voltage distribution curve for a TLC storage scheme (3 bits of data per memory cell). FIG. 4D shows a threshold voltage distribution curve for a QLC storage scheme (4 bits of data per memory cell). Similar threshold voltage distribution curves may be defined or generated for other storage schemes used to store numbers of bits other than (e.g., greater than) 2, 3, or 4 bits per memory cell.
[0059] With reference to FIG. 4A, for memory cells programmed according to an SLC storage scheme, the memory cells can be in either an erased state ER or a programmed state A. With reference to FIG. 4B, for memory cells programmed according to an MLC storage scheme, each memory cell can be in an erased state ER or one of three programmed states A, B, C. With reference to FIG. 4C, for memory cells programmed according to a TLC storage scheme, each memory cell can be in an erased state ER or one of seven programmed states A, B, C, D, E, F, G. With reference to FIG. 4D, for memory cells programmed according to a QLC storage scheme, each memory cell can be in an erased state ER or one of fifteen programmed states A, B, C, D, E, F, G, H, I, J, K, L, M, N, O. It should be noted that the terms "memory state" and "programmed state" are used interchangeably herein. For example, the terms "memory state A" and "programmed state A" are used interchangeably to refer to the same memory state.
[0060] 4A-4D, each program state is associated with a respective verify threshold voltage level (or verify threshold) VV and read voltage threshold level (or read threshold) VR. For example, with respect to the storage scheme of FIGs. 4A-4D, for example, verify threshold VVA and read threshold VRA are associated with memory state A, verify threshold VVB and read threshold VRB are associated with memory state B, and verify threshold VVC and read threshold VRC are associated with memory state C.
[0061] 4A-4D, the verify threshold VV and read threshold VR specify or are positioned relative to the lower limit or tail of the threshold voltage distribution curve. The verify threshold VV is used during programming to define when programming of a memory cell is complete, i.e., further programming of the memory cell is inhibited in response to the threshold voltage of that memory cell exceeding the verify threshold VV of the data state to which the memory cell is being programmed. The read threshold VR is used during a read operation to determine which data state the memory cell has been programmed to.
[0062] Also in Figures 4A-4C, exemplary associated data values are included within or below the threshold voltage distribution curves for each memory state. For example, Figure 4A illustrates an exemplary embodiment in which the threshold voltage distribution curve on the left is associated with the erased state ER and a logic "1" value, and the threshold voltage distribution curve on the right is associated with memory state A and a logic "0" value. In the MLC and TLC storage schemes shown in Figures 4B and 4C, respectively, the data values change sequentially with increasing memory states and threshold voltage ranges. In various exemplary embodiments, the particular data values associated with each data state may change according to a predetermined scheme, such as a Gray code scheme. With reference to Figure 4B, memory cells programmed according to the MLC storage scheme store two bits per cell, with the erased state ER being associated with a data value of "11", memory cells programmed to memory state A being associated with a data value of "01", memory cells programmed to memory state B being associated with a data value of "00", and memory cells programmed to memory state C being associated with a data value of "10". Similarly, with reference to Figure 4C, memory cells programmed according to a TLC storage scheme store three bits per cell, with the erased state ER being associated with a data value of "111", memory state A being associated with a data value of "011", memory state B being associated with a data value of "001", memory state C being associated with a data value of "000", memory state D being associated with a data value of "010", memory state E being associated with a data value of "110", memory state F being associated with a data value of "100", and memory state G being associated with a data value of "101". A similar relationship or correspondence between data values, memory states, and threshold voltage ranges can be made for memory cells programmed according to a QLC storage scheme (four bits per memory cell), as shown in Figure 4D.
[0063] 2B, memory die 104 may further include read / write circuitry 144 that includes multiple or P sense blocks (also referred to as sense modules or sense systems) 146. Sense blocks 146 are configured to participate in reading data from already programmed memory cells, which verifies the memory cells during programming operations.
[0064] Power control circuitry 160 may include any of a variety of circuit topologies or configurations (such as driver circuits, current sources, charge pumps, reference voltage generators, regulators, and pulse generation circuits) or any of various combinations thereof for generating and / or providing voltages at appropriate levels to perform memory operations (including read, program / write, sense, and erase operations). Other types of circuits for generating voltages may be possible. Power control circuitry 160 may also be in communication with and / or controlled by control logic circuitry 154, read / write circuitry 144, and / or sense block 146 to provide voltages at appropriate levels and at appropriate times to perform memory operations.
[0065] During a read operation, the power control circuit 160 may bias the word lines at a read threshold voltage level VR to read the data value of the data stored by the memory cell. The threshold voltage level VR for a given memory state may have different optimal values for different operating conditions, including different process conditions, different program / erase cycles, different retention times, different temperatures, different disturbance conditions, or combinations thereof. For a given storage scheme, the memory system may maintain a data structure, such as a table, at the controller side and / or memory die side that identifies one or more sets of read threshold levels for the power control circuit 160 to use to bias the word lines during a read operation. The data structure may include multiple sets of read threshold levels, each corresponding to a different region of the memory die 104.
[0066] 5, when a first sense operation is performed, the sense circuitry outputs a logic 1 for the TLC cells in erase state ER and a logic 0 for the TLC cells in memory states A-G. When a second sense operation is performed, the sense circuitry outputs a logic 1 for the TLC cells in erase state ER and memory states A-D, and a logic 0 for the TLC cells in memory states E-G. In various exemplary embodiments, the read / write circuitry 144 (or other circuitry on the memory die 104, such as, for example, control logic 154) may perform a NOT XOR (NXOR) operation between the logic bit values of the first sense operation and the logic bit values of the second sense operation to generate read bit values for the lower page bits.
[0067] The ECC module 124 then decodes the read page to generate a decoded page. After the ECC module 124 decodes the read page, the controller 102 may be configured to calculate an error page by comparing (e.g., XORing) the read page with the decoded page. Then, based on the error page and the respective sense result pages, the controller 102 may calculate a number of errors EJ or a bit error rate BERJ for each voltage bin J by counting each error in the error page (e.g., each bit in the error page that is equal to 1) in a corresponding voltage bin counter (e.g., as determined by a corresponding value in the respective sense result page).
[0068] For example, if the individual sensed result for a particular memory cell is 11→read value=NXOR(1,1)=1 and the decoded value is 0 (e.g., indicating an error), the controller 102 increments counter E0→(BER0) corresponding to voltage bin 0. If the individual sensed result for a particular memory cell is 01→read value=NXOR(0,1)=0 and the decoded value is 1 (indicating an error), the controller 102 increments counter E1→(BER1) corresponding to voltage bin 1. If the individual sensed result for a particular memory cell is 00→read value=NXOR(0,0)=1 and the decoded value is 0 (indicating an error), the controller 102 increments counter E2→(BER2) corresponding to voltage bin 2.
[0069] The read / write circuit may perform sense operations to read the middle page and the upper page in a similar manner to complete the execution of the read operation. For each of the middle page and the upper page, the number of sense operations performed may be different and / or the read threshold voltage levels applied to the bit lines may be different. In one exemplary TLC (3 bits per cell) storage scheme, the sense circuit performs three sense operations to read the middle page and two sense operations to read the upper page. However, the decode operation is generally the same, i.e., for a given middle stage or upper stage, the read / write circuit 144 may perform an NXOR operation on the bit values determined from the multiple sense operations to generate the read middle page and the read upper page.
[0070] As described above, the controller 102 determines how to adjust or correct the read threshold voltage level based on the BER of the data read page. For example, as shown in FIGS. 6A and 6B, if BER1→0 and BER0→1 are known, assuming a Gaussian state distribution, the position where the read threshold should be placed can be determined. As shown in FIG. 6A, if BER1→0 > BER0→1, the read threshold moves to the left, and as shown in FIG. 6B, if BER1→0 < BER0→1, the read threshold moves to the right.
[0071] Adjusting the read threshold level based on BER for an MLC storage scheme involving multiple sense operations to read a single page (e.g., lower, middle, upper, etc.) becomes more complex because a single comparison between two counts cannot be performed. For example, referring back to FIG. 5, when reading a lower page in a TLC (3 bits per cell) storage scheme, when the decoder inverts the bit value from logic 0 to logic 1, without more information, the decoder does not know whether the inversion corresponds to the A read threshold level VRA or the E read threshold level.
[0072] To obtain more information, the controller 102 divides the threshold voltage range into voltage bins defined by the read threshold levels and determines the BER for different voltage bins. Additionally, or alternatively, the controller 102 receives the lower page bits from both the first and second sense operations before the NXOR operation is performed, whereby the controller 102 can determine the voltage bin to which each bit belongs. By knowing the voltage bin to which each bit belongs, the controller 102 can then update the accurate count to adjust the BER and, consequently, adjust the threshold voltage level accordingly.
[0073] 5, the A and E threshold voltage levels used for the first and second sense operations divide the threshold voltage distribution curve into three voltage bins. Voltage bin 0 contains the voltage distribution curve for the erased state ER. Voltage bin 1 contains the voltage distribution curves for memory states A through D. Voltage bin 2 contains the voltage distribution curves for memory states E through G.
[0074] To determine the voltage bin, the controller 102 receives the lower page bits from both the first and second sense operations instead of (or in addition to) reading the lower page bits based on a NXOR operation to distinguish bits in voltage bin 0 from voltage bin 2. For example, if the combined bit value indicated from the first and second sense operations for a given bit is "11," the controller 102 knows that the given bit is in voltage bin 0. On the other hand, if the combined bit value indicated from the first and second sense operations for a given bit is "00," the controller 102 knows that the given bit is in voltage bin 2.
[0075] After identifying the voltage bin for each of the lower page bits, the controller 102 may then perform an NXOR operation to generate a read lower page. The controller 102 may then perform error correction via the ECC module 124 by inverting the bit values of the bits. Based on the decoded results and the individual sense pages, the controller 102 may calculate three BER counts: a first count for voltage bin 0, a second count for voltage bin 1, and a third count for voltage bin 2. The first count may indicate a BER0 associated with voltage bin 0, the second count may indicate a BER1 associated with voltage bin 1, and the third count may indicate a BER2 associated with voltage bin 2.
[0076] If the ECC module 124 determines to change a bit having a logic 0 to a logic 1, the controller 102 may determine whether the bit is in voltage bin 0 or voltage bin 2. If the bit is in voltage bin 0, the controller 102 may update the first count. If the bit is in voltage bin 2, the controller 102 may update the third count. Additionally or alternatively, if the ECC module 124 determines to change a bit having a logic 1 to a logic 0, the controller 102 may update the second count associated with voltage bin 1.
[0077] FIG. 7 is a flow chart of a continuous RLT method according to an exemplary embodiment. As shown in FIG. 7, for example, a host may be trying to read a page, but the threshold level is off. When a host read request is received (701), data is sensed from memory based on the host request (702). Then, the code is decoded using ECC, and the decoded data is sent appropriately to the host (703). Assuming the code is successfully decoded, the difference between the decoded page and the corresponding read page can be observed (704) to determine where the error is coming from. Then, a determination can be made as to where / how to move the read threshold (705).
[0078] 8 is a flow chart of a read threshold calibration method according to an exemplary embodiment. According to this exemplary embodiment, continuous RLT is used at an adjustable rate in a long sequential read scenario in the presence of temperature shifts that may be rapid. The exemplary method is based on the understanding that small shifts in the read threshold (also referred to herein as "tracking") performed frequently may provide improved read threshold calibration over read threshold calibrations performed less frequently or only when there is an indication of high BER.
[0079] According to this exemplary method, if there is an indication of a BER higher than a predefined threshold (801), then an RLT is performed on the next read WL (802). As with the RLT described above, this RLT does not interfere with host reads. Then, read thresholds are calibrated (803) based on the difference between the decoded page and the corresponding read page. These read thresholds are then used for the next read WL (804).
[0080] 9 shows how Δ can be created when the BER after successful decoding (or error correction) is known. This exemplary method can be used to calibrate the read threshold after successful ECC decoding, where the BER for each voltage bin can be derived from calculating the number of FBCs in each voltage bin. Each of the bits is mapped to its voltage bin according to the result of a single threshold read, the erroneous bits are summed per voltage bin, and the error vector is passed to a module that calculates the read threshold according to the BER per voltage bin.
[0081] FIG. 9 zooms in on a single threshold. In FIG. 9, D is the distance between the centers of two states, and σ0 and σ1 are the standard deviations of the corresponding states. Here, Δ is the distance of the current read threshold (solid line) from the optimal read threshold (dashed line). N is the number of cells in the page, and N / 8 is the total number of cells programmed in each data state, assuming scrambling. Since the decoding was successful, the number of cells in each data state can also be explicitly counted. E0 and E1 are the number of erroneous bits in each bin, and therefore P0 and P1 are the BER in each voltage bin.
[0082] Using the Gaussian tail distribution function Q, it can be derived as follows:
number
number
number
number
number
[0083] In the above formula,
number
number
number
number
number
number
[0084] For a single threshold, the above formula provides a calculation of Δ that can be used to smoothly tune the read threshold of an SLC page given the error rate in a voltage bin. However, for MLC, TLC, or QLC, several thresholds are involved when reading a single page. From the TLC LP example in Figure 5, BER#1 and BER#3 are calculated exactly, but BER#2 is split between the two thresholds that make up the LP read.
[0085] According to this exemplary embodiment, any RLT overhead is the result of an extra transfer from memory and additional processing, both of which are on the order of 1 microsecond. In contrast, a BES operation incurs at least five read operations, which is about three hundred microseconds (300 μS) of overhead in QLC, even before processing is considered. BES also requires stopping the read operation pipeline and inserting or stalling a BES operation, which results in a poor read threshold being used for a longer period of time. In contrast, according to this exemplary embodiment, the RT of a representative WL can be optimized, or alternatively, the current WL can be used as a reference.
[0086] FIG. 10 is a flow chart of a read threshold calibration method according to another exemplary embodiment. A block of flash memory includes many word lines, not all of which are equal. For example, the read threshold of an outlier WL, such as an edge WL, may be significantly different from the read threshold of other WLs, such as a "center" WL. For example, in one type of block with 96 WLs arranged in two "hemispheres," where the center of the block is also considered an "edge," the edge / outlier WLs may be WLs 0, 1, 46, 47, 48, 49, 94, and 95. Each memory type has its own specifications regarding this hard-coded outlier designation. The definition of an outlier WL may come from another source, for example, if a particular WL is recognized to have some physical defect, or if there is another reason that may make the WL an outlier for the block.
[0087] Therefore, according to this exemplary embodiment, these outlier WLs are taken into account to prevent any miscalculation based on the erroneous assumption that the outlier WLs have the same read threshold as the "center" WL. According to this exemplary embodiment, the outlier WLs are treated differently, by either using the BES for them in case of high BER, or by assuming a successful decode and ignoring the indication of high BER.
[0088] According to this exemplary method, if there is an indication of a BER higher than a predetermined threshold (1001), a determination is made as to whether the WL in question is an outlier WL, such as an edge WL (1002). If the WL is an outlier, a BES operation can be performed on the WL or the high BER of the WL can be ignored (1003). For example, if it is first determined that there is a BER higher than a first threshold (1001) and the WL in question is an outlier (1002-YES), there can be an additional determination of whether the BER is higher than a second threshold that is higher than the first threshold. If the BER is also higher than the second threshold, it can be decided to perform a BES operation, and if the BER is not higher than the second threshold, it can be decided to ignore the outlier WL.
[0089] If the WL in question is not an outlier WL (1002-NO), the method may continue as described above with respect to Figure 7, i.e., perform an RLT on the next read (1004), adjust the read threshold based on the RLT output (1005), and use the adjusted read threshold for the next WL to be read (1006).
[0090] 11 is a flow chart of a read threshold calibration method according to another exemplary embodiment. According to this exemplary embodiment, to account for WL variations, the WL of a block can be divided into zones, and a WL zoning table can be used to modify the RLT results and adjust them to apply to other WLs.
[0091] According to this exemplary embodiment, if there is an indication of a BER higher than a predefined threshold (1101), then perform an RLT for the next WL to be read (1102), infer a read threshold for a representative WL using an inverse WL zoning table (1103), adjust the read threshold based on the RLT output and the WL zoning table (1104), and use the adjusted read threshold for the next WL to be read (1105).
[0092] 12 is a flow chart of a read threshold calibration method according to another exemplary embodiment. In contrast to the above exemplary method, according to this exemplary embodiment, the trigger to perform RLT may be based on the temperature change itself, relative to a previous RLT operation, rather than on an indication of high BER. Thus, this exemplary method may allow improved adjustment of the read threshold before any actual increase in BER occurs.
[0093] According to this exemplary embodiment, the trigger for RLT may be a temperature change relative to a previous RLT operation that is higher than a temperature change threshold (1201). Alternatively, the trigger may be one or more of a temperature change above a temperature change threshold and a BER higher than a predefined threshold (1201). The other operations (1202-1205) of this exemplary method are similar to the operations (1102-1105) of the previous exemplary embodiment described in connection with FIG. 11. Alternatively, the trigger based on temperature change may be combined with any one or more of the exemplary embodiments discussed herein.
[0094] 13 illustrates an example result of a read threshold calibration method according to one or more of the example embodiments described herein. The example is based on a BICS5X3 middle page where data is programmed at -25°C and read at increasing temperatures up to 65°C.
[0095] The curve labeled "optimal" shows fail bit count (FBC) readings with an optimal read threshold, while the curve labeled "default" shows FBC readings with a default read threshold. The curves labeled "system" and "system+WLZ1" show FBC readings with read thresholds determined based on a prior art method and a prior art method using a WL zoning function that adds an offset to the representative WL read threshold for each different WL. The curve labeled "exemplary" shows FBC readings according to one of the exemplary embodiments described herein. It is clear that the reading technique described herein produces a lower FBC figure, resulting in improved reading performance.
[0096] Another factor that affects read performance is the number of calibration operations performed. BES operations are relatively long operations, taking 5-7 extra reads from memory, while RLT operations are relatively short, do not require extra reads, and only a limited number of extra transfers. For example, the overhead of a BES operation may be about 300 microseconds, while the overhead of an RLT operation may be about 20 microseconds (20 μS). Table #1 below shows the number of operations for each of the methods performed for the graph of FIG. 10, and shows the number of operations and the number of operations per read page for each of a prior art system without WL zoning, a prior art system with WL zoning, and a system according to one of the exemplary embodiments described herein. [Table 1] Table #1: Number of operations for each method during testing
[0097] As is evident from the table, the exemplary embodiment described herein results in significantly fewer operations, resulting in latency savings due to reduced overhead.
[0098] According to one aspect of the exemplary embodiment described herein, a method of calibrating a read threshold voltage includes performing a plurality of read operations, determining to perform a read level tracking (RLT) method, and performing the RLT method, the RLT method including determining a bit error rate (BER) of an indicative word line (WL), determining an adjusted read threshold level based on the BER, and adjusting the read threshold level according to the adjusted read threshold level. The determining to perform the RLT method may include determining that the BER of the indicative WL exceeds a BER threshold. The determining to perform the RLT method may further include determining that the indicative WL is not an outlier WL. The outlier WL may be an edge WL. The determining to perform the RLT method may include determining that a temperature change relative to a previous RLT method exceeds a temperature threshold. The determining to perform the RLT method may include at least one of determining that a BER of the indicative WL exceeds a BER threshold and that a temperature change relative to a previous RLT method exceeds a temperature threshold. The step of determining the BER may include determining a plurality of voltage bins and, for each of the plurality of voltage bins, determining a BER, and the step of determining the adjusted read threshold level based on the BER may include, for each of the plurality of voltage bins, calculating a Gaussian tail distribution function based on a corresponding BER, and, for each of the plurality of voltage bins, determining a read threshold distance from an optimal read threshold. The step of determining the plurality of voltage bins may include, for each of the memory states from the plurality of read operations, determining a corresponding voltage bin. The step of determining the BER may include estimating the BER using a corresponding syndrome weight. The step of performing the plurality of read operations may include performing the plurality of read operations according to a host request.The step of determining the adjusted read threshold level based on the BER may include applying a WL zoning table.
[0099] Various terms are used to refer to particular system components. Different companies may refer to the components by different names. This document does not intend to distinguish between components that differ in name but not function.
[0100] Matters of these exemplary embodiments that are apparent to those skilled in the art to which these exemplary embodiments pertain may not be described in detail herein.
[0101] The exemplary embodiments described herein may be considered in an illustrative sense only and not for purposes of limitation. Description of features or aspects within each exemplary embodiment may be considered applicable to other similar features or aspects in other exemplary embodiments.
[0102] Although exemplary embodiments have been described with reference to the drawings, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope as defined by the following claims.
Claims
1. A method for calibrating a read threshold voltage during a sequential read operation, comprising: performing a plurality of sequential read operations; determining whether to perform a read level tracking (RLT) operation following a first read operation on a first word line (WL) based on a comparison of a bit error rate (BER) of the first WL to a predefined threshold; responsive to determining that the BER of the first WL is above the predetermined threshold, performing the RLT operation during a second read operation performed on a second WL; determining an adjusted read threshold level based on a result of the RLT operation performed during the second read operation on the second WL; adjusting a read threshold level according to the adjusted read threshold level for a read operation performed subsequent to the second read operation; performing at least one read operation subsequent to the second read operation in accordance with the adjusted read threshold level. method.
2. The method of claim 1 , wherein the step of determining whether to perform the RLT operation further comprises determining whether the first WL is an outlier WL.
3. The method described in claim 2, wherein the outlier WL is an edge WL.
4. The method of claim 1 , wherein the step of determining whether to perform the RLT operation comprises determining whether a temperature change since a previous RLT operation exceeds a temperature threshold.
5. Determining a plurality of voltage bins; determining a corresponding BER for each of the plurality of voltage bins; determining the BER; calculating, for each of the plurality of voltage bins, a Gaussian tail distribution function based on the corresponding BER; determining a distance of a read threshold from an optimal read threshold for each of said plurality of voltage bins; determining the adjusted read threshold level. The method of claim 1.
6. The method of claim 5 , wherein determining the plurality of voltage bins comprises determining a corresponding voltage bin for each of the plurality of sequential read operations memory states.
7. The method of claim 1 , wherein determining the BER comprises estimating the BER using corresponding syndrome weights.
8. The method of claim 1 , wherein the step of performing the plurality of sequential read operations comprises performing the plurality of sequential read operations in accordance with a host request.
9. The method of claim 1 , wherein the step of determining the adjusted read threshold level based on the results of the RLT operation comprises applying a WL zoning table.
10. A system for calibrating read threshold voltages during sequential read operations, comprising: a controller in operative communication with at least one memory die; The controller: an interface configured to receive a number of bits corresponding to a plurality of sequential read operations from the at least one memory die; 1. A processor comprising: determining whether to perform a read level tracking (RLT) operation following a first read operation on a first word line (WL) based on a comparison of a bit error rate (BER) of the first WL to a predefined threshold; responsive to determining that the BER of the first WL is above the predetermined threshold, performing the RLT operation during a second read operation performed on a second WL; determining an adjusted read threshold based on a result of the RLT operation performed during the second read operation on the second WL; adjusting a read threshold level according to the adjusted read threshold level for a read operation performed subsequent to the second read operation; and a processor configured to perform at least one read operation subsequent to the second read operation in accordance with the adjusted read threshold level. system.
11. The system of claim 10 , wherein the processor is configured to determine whether to perform the RLT operation by determining whether the first WL is an outlier WL.
12. The system described in claim 11, wherein the outlier WL is an edge WL.
13. The system of claim 10 , wherein the processor is configured to determine whether to perform the RLT operation by determining whether a temperature change since a previous RLT operation exceeds a temperature threshold.
14. To determine the BER, the processor determining a plurality of voltage bins; for each of the plurality of voltage bins, determining a corresponding BER; To determine the adjusted read threshold level, the processor: calculating, for each of the plurality of voltage bins, a Gaussian tail distribution function based on the corresponding BER; determining a read threshold distance from an optimal read threshold for each of the plurality of voltage bins. The system of claim 10.
15. A system for calibrating read thresholds during sequential read operations, comprising: control means in operative communication with at least one memory die; The control means an interface means configured to receive a number of bits corresponding to a number of sequential read operations from the at least one memory die; A processing means, determining whether to perform a read level tracking (RLT) operation following a first read operation on a first word line (WL) based on a comparison of a bit error rate (BER) of the first WL to a predefined threshold; responsive to determining that the BER of the first WL is above the predetermined threshold, performing the RLT operation during a second read operation performed on a second WL; determining an adjusted read threshold based on a result of the RLT operation performed during the second read operation on the second WL; adjusting a read threshold level according to the adjusted read threshold level for a read operation performed subsequent to the second read operation; and processing means configured to perform at least one read operation subsequent to the second read operation using the adjusted read threshold level. system.
Citation Information
Patent Citations
Nonvolatile semiconductor storage device
JP2012221522A
Adaptive targeting of read levels in storage devices
US20160232054A1
Single page read level tracking by bit error rate analysis
US20200105353A1
Read threshold calibration using multiple decoders
US20210376854A1
Read voltage tracking method, memory storage device and memory control circuit unit
US9892799B1
