Method and device for self-checking and repairing memory

By configuring the memory test algorithm and starting test logic according to the control signal, recording and repairing memory errors, the problem of resource waste in the prior art is solved, efficient self-test and repair of the memory is realized, and reliability and stability are improved.

CN118942522APending Publication Date: 2024-11-12XEL TECH INC
View PDF 0 Cites 7 Cited by

Patent Information

Application Number
CN202410975951.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-19
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing memory self-testing technology has the problem of resource waste, especially when the fixed detection failure algorithm during chip design cannot meet user needs, and the detection logic after memory cropping still tests and repairs all memory, resulting in wasted resources and time.

Method used

Provide a memory self-test and repair method, by configuring a predefined memory test algorithm according to the control signal definition or test mode, start the memory test logic to obtain read data, record error addresses and data, calculate the failure rate, and decide based on the failure rate threshold. Whether to generate repair information to fix the error. The method includes a data processing module and a state machine, which can flexibly select the number of memory and detection schemes, and reduce resource waste.

Benefits of technology

It realizes automatic detection and rapid repair of memory failures, improves the reliability and stability of memory, and has flexibility and scalability, adapts to different test needs, and solves the problem of resource waste.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118942522A_ABST
    Figure CN118942522A_ABST
Patent Text Reader

Abstract

The invention provides a memory self-checking and repairing method and device, and the method comprises the steps: defining a memory testing algorithm according to a control signal or testing mode configuration, and predefining the memory testing algorithm; starting a memory test logic according to the test algorithm to obtain read data of the to-be-tested memory, the test logic comprising a data processing module; if the read data is inconsistent with the comparison data, recording an error address and error data; calculating a fault rate according to the error address and the error data; if the fault rate is smaller than or equal to a fault rate threshold value, the data processing module is called, and the data processing module is used for generating repair information; and repairing the error address and the error data of the memory to be tested based on the repairing information. According to the method, self-checking and repairing of the memories can be completed through a test logic, the number of the memories and a detection scheme can be flexibly selected through pins or registers, the method is more flexible, and the problem of resource waste is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a method and device for memory self-checking and repairing. Background Art

[0002] When the chip is manufactured, memory is set based on internal functional requirements. Since the memory is relatively complex, it is prone to failure, for example, the read and write control circuit failure during the read and write operation of the memory, and the address decoding circuit failure. The core memory cell array of the memory is composed of tens of thousands of memory cells, and the probability of failure is higher.

[0003] Memory Built-In Self-Test (MBIST) is a test logic integrated inside the chip, which is used to automatically detect and repair faults in the memory during the test phase after chip manufacturing and during the actual use of the chip. When executing MBIST, the test and repair are performed according to the algorithm fixed in the chip's internal logic. If the chip is designed to support only one fault detection algorithm, and the user wants to use another fault detection algorithm, the user will not choose this chip.

[0004] In addition, when the chip is packaged, functions may be cut according to customer needs, resulting in a portion of memory not being used, but the detection logic will still test and repair all the memory inside the chip, wasting resources and time. At the same time, the MBIST function is provided by the memory or backend manufacturer, and each group of memory will have a built-in self-test (BIST) function. The more memory, the more resources are required to implement the BIST function, resulting in a waste of resources. Summary of the invention

[0005] The present application provides a method and device for memory self-checking and repairing to solve the problem of resource waste.

[0006] In a first aspect, the present application provides a method for memory self-checking and repairing, comprising:

[0007] Predefine memory test algorithms based on control signal definitions or test mode configurations;

[0008] Starting a memory test logic according to a test algorithm to obtain read data of the memory to be tested, wherein the test logic includes a data processing module;

[0009] If the read data and the comparison data are inconsistent, record the error address and error data;

[0010] Calculating the failure rate according to the error address and the error data;

[0011] If the failure rate is less than or equal to the failure rate threshold, calling the data processing module, the data processing module is used to generate repair information;

[0012] The error addresses and error data of the memory to be tested are repaired based on the repair information.

[0013] In some feasible embodiments, it also includes:

[0014] If the failure rate is greater than an error threshold, error information is output, wherein the error information indicates that the error address and the error data cannot be repaired.

[0015] In some feasible embodiments, it also includes:

[0016] Obtaining the bit width of the memory to be tested;

[0017] Performing a circular shift on the data in the memory to be tested, wherein the number of shifts is the number of bit width shifts;

[0018] If the designated bit of the comparison data is different from the designated bit of the read data, outputting a first value, the comparison data is the same as the bit sequence of the read data;

[0019] If the number of times the first value is output is 1, repairing the error address and error data of the memory to be tested based on the repair information;

[0020] If the number of times the first value is output is greater than 1, an error message is output.

[0021] In some feasible embodiments, after outputting the error information, the following steps are included:

[0022] triggering a retry mechanism according to the error information to obtain a retry result, wherein the retry mechanism is used to retry accessing the memory;

[0023] If the retry result is successful, outputting correct information, wherein the correct information represents the repair of the error address and error data of the memory to be tested based on the repair information;

[0024] If the retry result is failure, an error message is output.

[0025] In some feasible embodiments, if the failure rate is less than or equal to the failure rate threshold, after calling the data processing module, it also includes:

[0026] Acquire a redundant column according to the error address and the error data;

[0027] The redundant columns are used to repair the error addresses and error data of the memory to be tested.

[0028] In some feasible embodiments, starting the memory test logic according to the test algorithm to obtain the read data of the memory to be tested includes:

[0029] In response to an instruction of a predefined memory test algorithm, setting the test logic to a start state, wherein the configuration information includes at least a start address and an end address;

[0030] Generate a read / write control signal and test data according to the test algorithm and configuration information, wherein the read / write control signal includes an enable signal and an address signal;

[0031] The control signal or test data is sent to the memory encapsulator to trigger a memory write operation or receive read data.

[0032] In some feasible embodiments, the test logic further includes a comparison module;

[0033] If the read data and the comparison data are inconsistent, recording the error address and the error data includes:

[0034] The comparison module is called to compare the read data and the comparison data through comparison logic, and to calculate the failure rate through the error address and the error data, and the comparison module includes the comparison logic of the test logic.

[0035] In some feasible embodiments, the test logic further includes a state machine, and the state machine is used to pull up a comparison enable signal;

[0036] If the read data and the comparison data are inconsistent, recording an error address and error data includes:

[0037] In response to the state machine pulling up a comparison enable signal, comparing the read data with the comparison data;

[0038] A bit-by-bit XOR operation is performed on the read data and the comparison data, and if the XOR result is 1, an error address and error data are recorded.

[0039] In some feasible embodiments, it also includes:

[0040] Calculating a first checksum of the memory data and storing the first checksum in a check bit, wherein the check bit is set in an area of ​​the memory;

[0041] In response to the instruction that the data is read, calculating a second checksum of the memory data;

[0042] If the first checksum is different from the second checksum, calling a data processing module, the data processing module being used to generate repair information;

[0043] The error addresses and error data of the memory to be tested are repaired based on the repair information.

[0044] In a second aspect, the present application further provides a device for memory self-checking and repairing, comprising:

[0045] An acquisition unit is used to predefine a memory test algorithm according to a control signal definition or a test mode configuration; and to start a memory test logic according to the test algorithm to obtain read data of the memory, wherein the test logic includes a data processing module; if the read data and the comparison data are inconsistent, the acquisition unit is also used to record an error address and error data; if the read data and the comparison data are inconsistent, the acquisition unit is also used to record an error address and error data;

[0046] A calculation unit, used for calculating the failure rate according to the error address and the error data;

[0047] a processing unit, wherein if the failure rate is less than or equal to the failure rate threshold, the processing unit is used to call a data processing module, wherein the data processing module is used to generate repair information;

[0048] The processing unit is further configured to repair the error address and error data of the memory to be tested based on the repair information.

[0049] It can be seen from the above technical solutions that the present application provides a method and device for memory self-check and repair, the method comprising: pre-defining a memory test algorithm according to a control signal definition or a test mode configuration; starting the memory test logic according to the test algorithm to obtain the read data of the memory to be tested, the test logic comprising a data processing module; if the read data and the comparison data are inconsistent, recording the error address and error data; calculating the failure rate according to the error address and error data; if the failure rate is less than or equal to the failure rate threshold, calling the data processing module, the data processing module is used to generate repair information; repairing the error address and error data of the memory to be tested based on the repair information. The method can complete the self-check and repair of the memory through a test logic, and can also flexibly select the number of memories and the detection scheme through pins or registers, making the method more flexible to solve the problem of resource waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solution of the present application, the drawings required for use in the embodiments are briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0051] Figure 1 A flowchart of a method for memory self-checking and repairing provided in an embodiment of the present application;

[0052] Figure 2 A schematic diagram of the test logic test process provided in the embodiment of the present application;

[0053] Figure 3 A schematic diagram of the state machine operation process is provided for the embodiment of the present application;

[0054] Figure 4 A schematic diagram of a device for memory self-checking and repairing provided in an embodiment of the present application. DETAILED DESCRIPTION

[0055] The following embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following embodiments do not represent all implementations consistent with the present application. They are only examples of systems and methods consistent with some aspects of the present application as detailed in the claims.

[0056] Memory failures are easily caused by memory configuration errors, poor contact, memory damage, etc. Different test algorithms can be used to detect whether the memory is faulty, such as single-bit errors, multi-bit errors, row failures, column failures, address or data bus failures, etc. The purpose of these detection algorithms is to achieve high fault coverage to ensure that potential failure modes have a chance to be detected, thereby improving the reliability of the memory.

[0057] The test algorithms include March algorithm, Checkboard algorithm, etc. Among them, March algorithm is a series of algorithms used for memory testing, which is used to detect and locate various faults in memory, such as short circuit or open circuit of address lines and data lines, and defects of other storage units. Checkboard algorithm (chessboard test method) is a graphical test method that performs fault detection by creating a specific data pattern similar to a chessboard in the memory. Specifically, it divides the memory into two or more alternating data areas, such as writing logic 0 to a group of addresses and writing logic 1 to adjacent addresses, forming a checkerboard-like distribution. By reading and verifying these data, it is checked whether there are address line, data line crossing errors or storage unit failures. These two detection algorithms can efficiently detect address-related errors because faults usually produce obvious anomalies in the chessboard pattern.

[0058] In addition, it also includes a variety of test algorithms. These test algorithms are test or control logic that have been fixed in the chip design stage and cannot be easily changed through external intervention. That is, the test strategy cannot be easily adjusted according to different test requirements or new fault models discovered.

[0059] Each memory cell has a BIST function. BIST is an automated test technology that allows memory cells to perform self-diagnosis and repair without external test equipment. However, there is a situation where the internal memory capacity of the chip is large, but only part of the memory is activated in actual use. BIST will still test and repair all the memories, which not only wastes test resources but also increases test time. As the number of memories increases, the resources required to implement BIST also increase, resulting in overall resource waste.

[0060] The present application provides a method and device for memory self-checking and repairing, which can realize automatic detection and rapid repair of memory faults, improve the reliability and stability of the memory. At the same time, it also has flexibility and scalability, can adapt to different testing requirements and scenarios, and solve the problem of resource waste.

[0061] See also Figure 1 , Figure 2 Some embodiments of the present application provide a method for memory self-checking and repairing, including:

[0062] S100: pre-define a memory test algorithm according to a control signal definition or a test mode configuration.

[0063] The control signal may be a control chip pin, which is a metal contact on a chip package for connecting to an external circuit or device. These pins allow current, signals or data to be transmitted between the inside and outside of the chip.

[0064] Test mode configuration is to configure the corresponding registers, which are hardware storage units used to store configuration information, allowing the chip's operating parameters and behaviors to be adjusted through programming. When determining the detection scheme, the corresponding test mode can be activated or configured by writing predefined values ​​to specific registers.

[0065] Control the chip pins or configure the corresponding registers to determine the test algorithm to be used. The test algorithm can be different test algorithms, for example, MARLR algorithm, MOVI algorithm.

[0066] The MARLR algorithm is a memory address read loop test (Memory Address Read Loop Test), that is, the system reads each address of the memory in a loop to check whether the read data is consistent with the expected data.

[0067] MOVI (Memory Operation Verification Insertion) reads and verifies data immediately after writing it to ensure the correctness of the write operation.

[0068] S200: Start the memory test logic according to the test algorithm to obtain read data of the memory to be tested.

[0069] When the test algorithm is determined, in some embodiments, in response to the instructions of the predefined memory test algorithm, the test logic MBIST is set to the startup state, and MBIST_ctrl (MBIST controller) generates read and write control information, read and write data according to the test algorithm and configuration information, for example, the starting address and end address of the memory, and sends the comparison enable signal, address signal and data to the RAM encapsulator when the read data reaches the RAM encapsulator.

[0070] MBIST controller is a memory built-in self-test controller, which is a built-in test logic that detects and verifies the performance of the memory by generating and executing test vectors for the memory. These test vectors include operations such as read, write and verify, and the purpose is to find potential faults in the memory, such as address decoding errors, cell damage and access time errors. MBIST controller allows memory cells to be tested without the need for external test equipment, which can verify the memory during the production stage or perform health checks regularly during device operation.

[0071] Among them, the read-write control signal indicates when to read or write the memory, and the type of operation, such as sequential read and write, random access, etc. The starting address and ending address of the memory test can make the test cover the entire or part of the specified memory area. Read and write data Based on the test algorithm, the MBIST controller will generate test data for writing into the memory. These data are usually known specific patterns, which are convenient for subsequent comparison and verification.

[0072] The RAM encapsulator is the interface for the memory chip to communicate with the outside world. The RAM encapsulator receives a comparison enable signal from the controller to indicate whether the read data is to be compared with the expected write data after the read operation. It also receives an address signal to specify the exact location in the memory, telling the RAM encapsulator where to read or write data. And receive data signals, which are used to send test data to the specified memory address during a write operation; during a read operation, these signals are not directly provided by the MBIST controller, but by the data read out by the RAM encapsulator after receiving the address signal.

[0073] MBIST_ctrl sends the generated address signal to the RAM encapsulator, indicating the memory location to be accessed. In the write cycle, MBIST_ctrl sends the generated write data to the RAM encapsulator and triggers the write operation through the write enable signal. In the read cycle, MBIST_ctrl issues a read enable signal and receives the read data from the RAM encapsulator.

[0074] When read data comes back from the RAM wrapper, the MBIST controller compares the data read out with the expected data that was previously written in. If the two do not match, there is an error.

[0075] MBIST will perform traversal read and write operations on the memory to be tested according to the requirements of the algorithm to generate read data. In this process, the test logic includes a data processing module, which will collect and process these data in real time to prepare for subsequent fault detection.

[0076] S300: If the read data and the comparison data are inconsistent, the error address and error data are recorded.

[0077] When the read data arrives at the output of the RAM wrapper, MBIST_ctrl will issue a Compare enable signal to start the data comparison logic. Under the control of the Compare enable signal, the MBIST test logic will compare the read data with the expected data written previously.

[0078] If the read data is inconsistent with the comparison data, it is considered that an error is detected. MBIST_ctrl will record the error address and error data or error mode for subsequent analysis and repair.

[0079] S400: Calculate the failure rate according to the error address and the error data.

[0080] In some embodiments, the test logic also includes a state machine and a comparison module, and the comparison module includes the comparison logic of the test logic. The MBIST process is controlled by a state machine, and the state machine advances the test process according to a preset test sequence. When the MBIST state machine determines to start data comparison, it will pull up a comparison enable (compare_en) signal, and the compare_en signal is the input of the comparison (compare) module, which is used to trigger the data comparison operation and compare by the comparison logic. The compare module will perform an XOR operation on the data read from the memory to be tested, that is, the read data and the comparison data. The result of the XOR will indicate which bits do not match between the two.

[0081] See also Figure 3 When the MBIST test is triggered by the enable signal, the MBIST state machine jumps from the MbistIdle (wait) state to the MbistStart (test start) state. The MbistStart state transition indicates that the MBIST test process has started. In the MbistStart state, the state machine jumps to mbistNewIter (new iteration item).

[0082] In the mbistNewIter state, the state machine initializes the iteration and sequence numbering: iter_num (number of iterations) and seq_num (number of sequences per iteration) are assigned initial values. For example, if iter_num is 0, it means only one round of testing is performed. The seq_num value is 5, which means there are 6 test sequences (seq) per round of testing. The corresponding test sequence is generated according to seq_num, which is used to traverse the memory address and detect potential errors.

[0083] After initialization, the state machine enters the mbistRunseq (running test sequence) state and starts to perform memory testing. In this state, MBIST starts from the preset highest address or lowest address and performs sequence detection on the memory one by one. Each sequence traverses all addresses between the specified start address and the end address.

[0084] For each sequence, MBIST will traverse all specified memory addresses. When a sequence has traversed all addresses, the end_addr (end address) signal is pulled high, indicating that the address traversal of the sequence has been completed. In the mbistNextseq (move to the next sequence to be tested) state, MBIST switches to the next sequence according to the end_addr signal and continues to traverse the memory address.

[0085] When all sequences have completed a round of traversal, the end_iter (end of iteration) signal is pulled high to indicate that the current iteration has been completed. The state machine checks the value of iter_num to determine whether to proceed to the next round of iteration. If iter_num is greater than 0, the next round of iteration is performed; otherwise, it indicates that all iterations have been completed and the test is over.

[0086] When all iterations and sequences are executed, the MBIST test is completed. At this point, the state machine may return to the MbistDone (test completed) state, indicating that the test process has ended and may output test results.

[0087] If the result of the XOR operation is non-zero, it indicates an error. At this time, the compare module will record the error data and the address where the error was first detected. When an error is detected, the compare module will pull up the fail signal, indicating that the test failed.

[0088] At the same time, if the error can be repaired by redundant resources, such as redundant columns, the compare module will pull up the repairable signal. In addition to the address and data of the first error, the compare module will continue to test other data rows and record the locations of all error bits.

[0089] Based on the location and number of error bits, the compare module evaluates whether the FRR (Failure Rate Ratio) condition is met. If the error does not meet the FRR condition, that is, there are too many errors and they cannot be repaired by redundant resources, an error message is output and the compare module pulls down the repairable signal, indicating that the error is unrepairable.

[0090] S500: If the failure rate is less than or equal to the failure rate threshold, call a data processing module.

[0091] If the failure rate is less than or equal to the failure rate threshold, indicating that the current error is repairable and all necessary FRR calculations have been completed, the compare module will pull up the frr_calc_done signal to indicate that the FRR calculation has been completed.

[0092] As to whether the error can be repaired, in some embodiments, the bit width of the memory to be tested, that is, the number of data bits that can be processed at one time, is obtained; a circular shift is performed on the data in the memory to be tested, and the number of shifts is the number of bit width times; if the specified bit of the comparison data is different from the specified bit data of the read data, a first value is output, wherein the bit sequence of the comparison data is the same as that of the read data; if the number of times the first value is output is 1, the error address and error data of the memory to be tested are repaired based on the repair information; if the number of times the first value is output is greater than 1, error information is output.

[0093] During this process, each bit in the data will be moved to the next position of its original position until each bit completes a full shift cycle. The data processing module tracks and records the data status after each shift. The circular shift operation can fully detect the read and write capabilities of each bit in the memory, especially those minor faults that may be difficult to expose in conventional testing. This comprehensive testing method can improve the sensitivity and coverage of fault detection.

[0094] After each cyclic shift, the shifted read data is compared with a predefined comparison data with the same bit sequence as the original read data. The designated bit can be a designated bit or bits, or the entire data bit sequence. If an inconsistency is found in the designated bits of the two data, the system will output a specific first value to indicate that a data error has been detected.

[0095] During the test, if the system outputs the first value only once, that is, it indicates that only one data error is detected, it can be considered that this is an isolated error, which may be caused by a single storage unit failure. At this time, a repair operation will be performed on the error address and error data based on the previously generated repair information. The repair operation may include replacing the failed storage unit, rewriting the data, or adjusting the circuit connection.

[0096] If the system outputs the first value greater than 1 time, it indicates that multiple data errors have been detected, and there are serious fault problems in the memory under test, such as multiple storage units failing at the same time or circuit connection problems. In this case, the system will output an error message to prompt the user to pay attention and may need to take further troubleshooting and repair measures.

[0097] In some embodiments, after outputting the error information, a retry mechanism can also be triggered according to the error information to obtain a retry result, and the retry mechanism is used to retry accessing the memory to be tested; if the retry result is successful, correct information is output, and the correct information represents the repair of the error address and error data of the memory to be tested based on the repair information; if the retry result is a failure, an error message is output.

[0098] The retry mechanism is a mechanism that automatically retries to access the memory when a memory access error is detected. The retry mechanism records the current error status, including the error address, error type, and possible error data. When retrying to access the memory under test, the access may change the access mode, such as changing the read speed, adjusting the timing parameters, or using a different access path to eliminate errors caused by temporary factors, such as voltage fluctuations and signal interference.

[0099] You can also set the number of retries and the retry interval based on actual conditions, but too many retries may cause system performance to degrade, while too few retries may not be able to effectively deal with temporary failures. The retry interval can analyze the impact of the time interval between retries on system performance and fault recovery capabilities.

[0100] If the retry result fails, the error message continues to be output. This error message can be defined as a secondary error message and the retry mechanism will not be triggered.

[0101] Perform the same number of circular shift operations on the read data and the comparison data. After each shift, perform an XOR operation on the corresponding bits of the two sets of data to detect whether there is an error. Take the lowest bit of the XOR result or any specified bit for judgment. If an error is detected during the shift process, that is, the specified bit of the XOR result is 1, the position of the error is recorded. If only one error is detected before the end of the circular shift, or the detected errors are isolated, which can be understood as being far apart and exceeding the repair capability range, then these errors are considered repairable. If multiple errors are detected continuously during the circular shift process, or the distance between the errors is too close to be effectively repaired with redundant resources, then these errors are considered unrepairable.

[0102] The data processing module is used to detect errors during the MBIST test and generate corresponding repair information. These repair information may include the location, type and repair strategy of the error, for example, the replacement information of redundant columns and rows. After the detection is completed, the data processing module will store the generated repair information in an internal register or a special storage unit for subsequent use.

[0103] After all the tests of the memory to be tested are completed, the electronic fuse control module will read the repair information stored in the data processing module. The electronic fuse control module will burn the read repair information into the electronic fuse (eFuse). eFuse is a one-time programmable (OTP) storage element used to store key configuration information or repair data of the chip, which cannot be changed during the life cycle of the chip.

[0104] During the subsequent use of the chip, each time the chip is started, the system will automatically load the repair information from the eFuse and store it in the appropriate register. In this way, when the memory address that may cause an error is accessed during the operation of the chip, the repair information will be used to correct the potential error. Since the repair information has been permanently stored in the eFuse and automatically loaded at each startup, the chip does not need to be tested by MBIST during subsequent use. This can improve the startup speed and operating efficiency of the chip.

[0105] S600: Repairing error addresses and error data of the memory to be tested based on the repair information.

[0106] The repair process may include replacing damaged storage cells, adjusting circuit connections, or rewriting correct data, etc. When an error is detected, the erroneous address and data can be repaired based on the repair information, realizing the functions of self-checking and modification.

[0107] By adopting a unified verification algorithm and data processing module, the need to design test logic for each different type of memory can be reduced. In addition, since the error handling and repair logic is shared, there is no need to equip each memory cell or memory block with a repair circuit, which reduces the complexity of the overall logic design and effectively reduces the size of the chip.

[0108] In some embodiments, a first checksum of memory data is calculated and stored in a check bit, which is set in an area of ​​the memory to be tested; in response to an instruction that data is read, a second checksum of the memory data is calculated; if the first checksum is different from the second checksum, a data processing module is called, and the data processing module is used to generate repair information; based on the repair information, the error address and error data of the memory to be tested are repaired.

[0109] When data is written to the memory under test, the checksum of the data is calculated first. The checksum is a value calculated by a specific algorithm to verify the integrity of the data. The calculated checksum is stored in the check bits, which are usually a special area in the memory used to store the check information associated with the data block.

[0110] When data needs to be read from the memory under test, the system will calculate the checksum of these data again and compare the first checksum with the second checksum. If the two checksums are the same, it means that the data has not changed during the storage and reading process, that is, the data is complete. If the two checksums are different, the system believes that the data may be damaged or wrong and needs further processing.

[0111] By calculating a checksum of the data and storing it in an extra check bit, when the data is read, the checksum is recalculated and compared to the stored checksum. If the two do not match, there is an error.

[0112] See also Figure 4 Based on the above method for memory self-checking and repairing, some embodiments of the present application further provide a device for memory self-checking and repairing, including:

[0113] An acquisition unit is used to predefine a memory test algorithm according to a control signal definition or a test mode configuration; and to start a memory test logic according to the test algorithm to obtain read data of the memory to be tested, wherein the test logic includes a data processing module; if the read data and the comparison data are inconsistent, the acquisition unit is also used to record an error address and error data;

[0114] A calculation unit, used for calculating the failure rate according to the error address and the error data;

[0115] a processing unit, wherein if the failure rate is less than or equal to the failure rate threshold, the processing unit is used to call a data processing module, wherein the data processing module is used to generate repair information;

[0116] The processing unit is further configured to repair the error address and error data of the memory to be tested based on the repair information.

[0117] The present application provides a method and device for memory self-check and repair, the method comprising: pre-defining a memory test algorithm according to a control signal definition or a test mode configuration; starting the memory test logic according to the test algorithm to obtain the read data of the memory to be tested, the test logic comprising a data processing module; if the read data and the comparison data are inconsistent, recording the error address and error data; calculating the failure rate according to the error address and error data; if the failure rate is less than or equal to the failure rate threshold, calling the data processing module, the data processing module is used to generate repair information; repairing the error address and error data of the memory to be tested based on the repair information. The method can complete the self-check and repair of the memory through a test logic, and can also flexibly select the number of memories and the detection scheme through pins or registers, making the method more flexible to solve the problem of resource waste.

[0118] Similar parts between the embodiments provided in this application can be referenced to each other. The specific implementation methods provided above are only a few examples under the general concept of this application and do not constitute a limitation on the protection scope of this application. For those skilled in the art, any other implementation methods expanded based on the scheme of this application without creative work belong to the protection scope of this application.

Claims

1. A method for memory self-checking and repairing, characterized in that: include: Predefine memory test algorithms based on control signal definitions or test mode configurations; Starting a memory test logic according to a test algorithm to obtain read data of the memory to be tested, wherein the test logic includes a data processing module; If the read data and the comparison data are inconsistent, record the error address and error data; Calculating the failure rate according to the error address and the error data; If the failure rate is less than or equal to the failure rate threshold, calling the data processing module, the data processing module is used to generate repair information; The error addresses and error data of the memory to be tested are repaired based on the repair information.

2. The method for memory self-checking and repairing according to claim 1, characterized in that: Also includes: If the failure rate is greater than an error threshold, error information is output, wherein the error information indicates that the error address and the error data cannot be repaired.

3. The method for memory self-checking and repairing according to claim 1, characterized in that: Also includes: Obtaining the bit width of the memory to be tested; Performing a circular shift on the data in the memory to be tested, wherein the number of shifts is the number of bit width shifts; If the designated bit of the comparison data is different from the designated bit of the read data, outputting a first value, the comparison data is the same as the bit sequence of the read data; If the number of times the first value is output is 1, repairing the error address and error data of the memory to be tested based on the repair information; If the number of times the first value is output is greater than 1, an error message is output.

4. The method for memory self-checking and repairing according to claim 2 or 3, characterized in that: After the error information is output, it includes: triggering a retry mechanism according to the error information to obtain a retry result, wherein the retry mechanism is used to retry accessing the memory; If the retry result is successful, outputting correct information, wherein the correct information represents the repair of the error address and error data of the memory to be tested based on the repair information; If the retry result is failure, an error message is output.

5. The method for memory self-checking and repairing according to claim 1, characterized in that: If the failure rate is less than or equal to the failure rate threshold, after calling the data processing module, it also includes: Acquire a redundant column according to the error address and the error data; The redundant columns are used to repair the error addresses and error data of the memory to be tested.

6. The method for memory self-checking and repairing according to claim 1, characterized in that: The starting of the memory test logic according to the test algorithm to obtain the read data of the memory to be tested includes: In response to an instruction of a predefined memory test algorithm, setting the test logic to a start state, wherein the configuration information includes at least a start address and an end address; Generate a read / write control signal and test data according to the test algorithm and configuration information, wherein the read / write control signal includes an enable signal and an address signal; The control signal or test data is sent to the memory encapsulator to trigger a memory write operation or receive read data.

7. The method for memory self-checking and repairing according to claim 1, characterized in that: The test logic also includes a comparison module; If the read data and the comparison data are inconsistent, recording the error address and the error data includes: The comparison module is called to compare the read data and the comparison data through comparison logic, and to calculate the failure rate through the error address and the error data, and the comparison module includes the comparison logic of the test logic.

8. The method for memory self-checking and repairing according to claim 1, characterized in that: The test logic also includes a state machine, and the state machine is used to pull high a comparison enable signal; If the read data and the comparison data are inconsistent, recording an error address and error data includes: In response to the state machine pulling up a comparison enable signal, comparing the read data with the comparison data; A bit-by-bit XOR operation is performed on the read data and the comparison data, and if the XOR result is 1, an error address and error data are recorded.

9. The method for memory self-checking and repairing according to claim 1, characterized in that: Also includes: Calculating a first checksum of the memory data and storing the first checksum in a check bit, wherein the check bit is set in an area of ​​the memory; In response to the instruction that the data is read, calculating a second checksum of the memory data; If the first checksum is different from the second checksum, calling a data processing module, the data processing module being used to generate repair information; The error addresses and error data of the memory to be tested are repaired based on the repair information.

10. A device for memory self-checking and repairing, characterized in that: The method for performing memory self-checking and repairing according to any one of claims 1 to 9 comprises: An acquisition unit is used to predefine a memory test algorithm according to a control signal definition or a test mode configuration; and to start a memory test logic according to the test algorithm to obtain read data of the memory to be tested, wherein the test logic includes a data processing module; if the read data and the comparison data are inconsistent, the acquisition unit is also used to record an error address and error data; A calculation unit, used for calculating the failure rate according to the error address and the error data; a processing unit, wherein if the failure rate is less than or equal to the failure rate threshold, the processing unit is used to call a data processing module, wherein the data processing module is used to generate repair information; The processing unit is further configured to repair the error address and error data of the memory to be tested based on the repair information.

Citation Information

Cited By

  • Design method of high-yield magnetic random access memory system under irradiation condition

    CN119559992A

  • Built-in computing macro-cell-oriented built-in self-test architecture, system and method

    CN120673828A

  • Method of handling failure of memory, memory controller and memory system

    CN120766743A

  • Method of handling failure of memory, memory controller and memory system

    CN120766743B

  • Memory and electronic equipment

    CN122531446A