Memory circuit

The memory circuit addresses the need for flexible capacity by using redundant cells and address conversion to switch between large and small capacity modes, enhancing test efficiency and reducing costs.

JP2024156328AActive Publication Date: 2024-11-06SOCIONEXT INC
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
JP2023070700
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-04-24
Publication Date
2024-11-06
Estimated Expiration
2043-04-24

AI Technical Summary

Technical Problem

Existing memory circuits in system LSIs lack the ability to dynamically switch between operating as a single large-capacity memory and multiple small-capacity memories, which is necessary for varying memory requirements in different applications.

Method used

The memory circuit includes multiple memory sections with redundant cells and control units that allow switching between operation modes, converting addresses to access redundant cells when defects are detected, enabling parallel access and repair of defects.

Benefits of technology

This design allows the memory circuit to operate as both a large-capacity and multiple small-capacity memories, reducing test time and manufacturing costs while maintaining usability and efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024156328000001_ABST
    Figure 2024156328000001_ABST
Patent Text Reader

Abstract

To provide a memory circuit capable of operating as a plurality of small-capacity memories or a large-capacity memory by switching an operating mode.SOLUTION: A memory circuit has: a plurality of memory units each including a plurality of first memory cells and a plurality of second memory cells; a plurality of first memory control units for controlling access to each memory unit; and a second memory control unit provided in common to the plurality of memory units. Either of the plurality of memory units has a third memory cell corresponding to the second memory cell of each memory unit. During a first mode, the first memory control unit converts a first address indicating an inaccessible first memory cell into a second address indicating a second memory cell and outputs it to the memory unit. During a second mode, when the second address indicating the second memory cell is included in a second access request, the second memory control unit converts the second address into a third address indicating a third memory cell and outputs it to the memory unit including the third memory cell.SELECTED DRAWING: Figure 1
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a memory circuit. [Background technology]

[0002] In a semiconductor memory device having a plurality of blocks each having memory cells, a method is known in which a redundant column for repairing defects is provided in each block and each redundant block (see, for example, Patent Document 1). In a memory device having a plurality of memory blocks, a method is known in which defects are repaired on a memory block basis (see, for example, Patent Document 2).

[0003] In a nonvolatile memory, when a memory area is defective, a method is known in which a switching circuit switches the output destination of a defective address to a redundant memory area (see, for example, Patent Document 3). Also, in a nonvolatile memory, a method is known in which a redundant memory cell row provided in each of a plurality of memory blocks is used to relieve defective memory cells in other memory blocks (see, for example, Patent Document 4). In a nonvolatile semiconductor memory device, a method is known in which a nonvolatile memory cell with deteriorated write or erase characteristics is automatically replaced with a redundant cell or cell block (see, for example, Patent Document 5). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2003-187591 A [Patent Document 2] JP 2012-509541 A [Patent Document 3] Japanese Patent Application Publication No. 01-128300 [Patent Document 4] Japanese Patent Application Publication No. 05-290598 [Patent Document 5] Japanese Patent Application Publication No. 08-007597 Summary of the Invention [Problem to be solved by the invention]

[0005] However, memory circuits mounted on system LSIs (Large-Scale Integration) and the like use different memory capacities depending on the application, and one memory circuit may be shared by multiple applications. For this reason, it is preferable for this type of memory circuit to have an operation mode in which it operates as one large-capacity memory and an operation mode in which it operates as multiple small-capacity memories, but no such method has been proposed.

[0006] The present invention has been made in view of the above-mentioned points, and has as its object to provide a memory circuit capable of operating as a plurality of small-capacity memories or a large-capacity memory by switching the operation mode. [Means for solving the problem]

[0007] In one aspect of the present invention, a memory circuit includes a plurality of memory units, each including a plurality of first memory cells and a second memory cell that is accessed in place of a defective first memory cell; a plurality of first memory control units provided corresponding to each of the plurality of memory units and controlling access to the corresponding memory unit based on a first access request for each of the plurality of memory units during a first mode; and a second memory control unit provided commonly to the plurality of memory units and controlling access to the plurality of memory units based on a second access request during a second mode, wherein any of the plurality of memory units has a plurality of third memory cells respectively corresponding to the plurality of second memory cells of the plurality of memory units, and each of the plurality of first memory control units, when the first access request includes a first address indicating the first memory cell that cannot be accessed during the first mode, converts the first address into a second address indicating the second memory cell and outputs it to the corresponding memory unit, and when the second access request includes the second address indicating each of the plurality of second memory cells of the plurality of memory units during the second mode, the second memory control unit converts the second address into a third address indicating any of the plurality of third memory cells and outputs it to the memory unit including the third memory cell. Effect of the Invention

[0008] According to the disclosed technique, it is possible to provide a memory circuit that can operate as a plurality of small-capacity memories or a large-capacity memory by switching the operation mode. [Brief description of the drawings]

[0009] [Figure 1] 1 is a block diagram showing an outline of a memory circuit according to a first embodiment; [Diagram 2] 2 is an explanatory diagram showing an example of word redundancy in each memory unit in FIG. 1; [Diagram 3] 2 is a block diagram showing an example of a memory control unit CNT1 in FIG. 1. [Figure 4] 4 is a flow chart showing an example of the operation of the memory control unit CNT1 of FIG. 3. [Diagram 5]1. FIG. 4 is a flow chart showing an example of a first test for detecting defects in each memory unit by each test circuit in FIG. [Figure 6] 10 is a flow chart showing an example of a second test for checking the operation of each memory unit after storing defect information in a nonvolatile memory by each test circuit in FIG. 1. [Figure 7] 11 is a flow chart showing an example of an operation of a memory control unit CNT2 that receives an access request via a system bus. FIG. [Figure 8] FIG. 11 is a flow chart showing an example of the operation of a branch control unit excluding the final stage. [Figure 9] FIG. 11 is a block diagram showing an outline of a memory circuit according to a second embodiment. [Figure 10] 10 is a block diagram showing an example of a memory control unit CNT1 of FIG. 9. [Figure 11] 10 is an explanatory diagram showing an example of word redundancy in the memory block of FIG. 9; [Figure 12] FIG. 13 is a block diagram showing an outline of a memory circuit according to a third embodiment. [Figure 13] FIG. 13 is a block diagram showing an outline of a memory circuit according to a fourth embodiment. [Figure 14] FIG. 13 is a block diagram showing an outline of a memory circuit according to a fifth embodiment. [Figure 15] FIG. 13 is a block diagram showing an outline of a memory circuit according to a sixth embodiment. [Figure 16] 16 is an explanatory diagram showing an example of word redundancy in the memory block of FIG. 15; [Figure 17] 16 is a flow chart showing an example of a first test for detecting defects in each memory unit by each test circuit in FIG. 15. [Figure 18] 16 is a flow chart showing an example of a second test for checking the operation of each memory unit after defect information is stored in the nonvolatile memory by each test circuit of FIG. 15. [Figure 19] FIG. 13 is a block diagram showing an outline of a memory circuit according to a seventh embodiment. [Figure 20] FIG. 13 is a block diagram showing an outline of a memory circuit according to an eighth embodiment. [Figure 21]FIG. 21 is a block diagram illustrating an example of a memory group of FIG. 20. [Figure 22] FIG. 21 is a block diagram illustrating another example of the memory group of FIG. 20. [Figure 23] 2 is a block diagram showing an example of a system in which the memory circuit of FIG. 1 is mounted. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] Hereinafter, an embodiment will be described with reference to the drawings. In the following, the same reference numerals as the names of the signals will be used for signal lines through which information such as signals is transmitted.

[0011] Fig. 1 shows an overview of a memory circuit according to a first embodiment. The memory circuit 100 shown in Fig. 1 is, for example, a static random access memory (SRAM), and is implemented in a semiconductor device such as a system LSI that processes image data. For example, a semiconductor device including the memory circuit 100 is mounted on an imaging device such as a surveillance camera, a head-mounted device such as AR / VR (Augmented Reality / Virtual Reality) glasses, or a digital camera, and generates moving image data to be displayed on a display device.

[0012] The memory circuit 100 has a memory management unit MNG (input side), a memory management unit MNG (output side), a plurality of relay control units RCNT, a plurality of memory blocks MBLK, and a system bus SBUS. The plurality of relay control units RCNT and the plurality of memory blocks MBLK are alternately arranged between the memory management unit MNG (input side) and the memory management unit MNG (output side). In the following, when the memory management unit MNG (input side) and the memory management unit MNG (output side) are described without distinction, they are simply referred to as the memory management unit MNG.

[0013] Each memory block MBLK has a memory unit MEM (MEM0, MEM1, MEM2, or MEM3) including a plurality of memory cells, and a memory control unit CNT1 arranged on the input side of the memory unit MEM. The memory control unit CNT1 is an example of a first memory control unit. Each memory block MBLK also has a test circuit BIST (Built-In Self Test) and a non-volatile memory NVM arranged corresponding to the memory control unit CNT1.

[0014] 1, four memory blocks MBLK are illustrated, but the number of memory blocks MBLK may be any number equal to or greater than 1. When there is one memory block MBLK, the memory block MBLK has a plurality of memory units MEM.

[0015] In the example shown in FIG. 1, each of the memory units MEM0-MEM3 has a plurality of word lines WL and a predetermined number of bubble redundancy word lines BRWL. In addition, in the memory unit MEM3 in the final stage, the predetermined number of word lines WL and the bubble redundancy word lines BRWL also function as sub-word lines SWL. The bubble redundancy word lines BRWL and the sub-word lines SWL are described in FIG. 2. Note that the position of the memory unit MEM3 having the sub-word lines SWL is not limited to the final stage, and may be disposed on the input side of the memory unit MEM0, between the memory units MEM1-MEM2, or between the memory units MEM1-MEM2. The arrangement of the word lines WL and the like of the memory unit MEM is described in FIG. 2.

[0016] In Fig. 1, circles attached to the word line WL, the bubble redundancy word line BRWL, and the sub-word line SWL exemplify memory cells. The memory cell connected to the word line WL is an example of a first memory cell. The memory cell connected to the bubble redundancy word line BRWL is an example of a second memory cell. The memory cell connected to the sub-word line SWL is an example of a third memory cell.

[0017] The memory management unit MNG is connected to the system bus SBUS, and operates during a normal operation mode in which it receives an access request MREQ from the system bus SBUS, but does not operate during a test mode. The normal operation mode is an example of the second mode, and the test mode is an example of the first mode. The access request MREQ received from the system bus SBUS is an example of a second access request. The memory management unit MNG functions as an interface circuit between signals input / output to / from the system bus SBUS and signals input / output to / from the memory block MBLK.

[0018] The memory management unit MNG (input side) receives an access request MREQ including a command CMD, an address AD, and write data WD used to access the memory block MBLK from a higher-level controller such as a CPU (Central Processing Unit) via a system bus SBUS. The access request MREQ is a write request or a read request. The memory management unit MNG (input side) outputs the received access request MREQ to the first-stage relay control unit RCNT.

[0019] The memory management unit MNG (input side) has a memory control unit CNT2 common to a plurality of memory blocks MBLK. The memory control unit CNT2 is an example of a second memory control unit. The memory control unit CNT2 refers to defect information (defective address FAD and redundancy enable REDEN) stored in a plurality of nonvolatile memories NVM or grasps the bubble redundant word lines BRWL of a plurality of memory units MEM as specifications, determines a transfer destination of an access request MREQ, and converts an address AD included in the access request MREQ to the address AD of the determined transfer destination. An example of the operation of the memory control unit CNT2 is shown in FIG. 2 and FIG. 7.

[0020] 1, the memory management unit MNG (input side) and the memory management unit MNG (output side) are arranged separately, but the memory management unit MNG (input side) and the memory management unit MNG (output side) may be arranged as one memory management unit MNG. In this case, a row of a plurality of relay control units RCNT and a plurality of memory blocks MBLK may be arranged in a U-shape.

[0021] For example, the memory management unit MNG (input side) outputs 256-bit write data WD to the first-stage relay control unit RCNT every time it receives 64-bit write data WD included in a write request four times. That is, the memory management unit MNG (input side) converts the serial write data WD received from the system bus SBUS into parallel write data WD and causes one of the memory blocks MBLK to perform a 256-bit write operation.

[0022] The memory management unit MNG (output side) outputs the 256-bit read data RD output from the memory block MBLK to the system bus SBUS in four batches of 64 bits each based on a read request received by the memory management unit MNG (input side). That is, the memory management unit MNG (output side) converts the 256-bit read data RD read from one of the memory blocks MBLK by a read operation into serial read data RD and outputs it to the system bus SBUS.

[0023] 1 may be arranged in parallel between the memory management unit MNG (input side) and the memory management unit MNG (output side). The memory management unit MNG may output 512-bit write data WD in parallel to the two memory blocks MBLK every time it receives 64-bit write data WD included in a write request eight times. The memory management unit MNG may also output 512-bit read data RD, which is read in parallel from the two memory blocks MBLK based on a read operation, to the system bus SBUS eight times as 64-bit read data RD.

[0024] By processing 256-bit or 512-bit data in parallel within the memory circuit 100, the operating frequency of the memory block MBLK can be approximately one-fourth or one-eighth of the operating frequency of the system bus SBUS. Therefore, the power consumption of the memory block MBLK can be reduced compared to the case where 64-bit data is processed sequentially within the memory circuit 100. Since the operating frequency of the memory block MBLK can be lowered, it is possible to provide a sufficient operating margin for the relay control unit RCNT, the memory control unit CNT1, and the memory unit MEM, and it is possible to easily implement the timing design of the circuit, etc.

[0025] The number of bits of data input / output to the memory block BLK may be 2 n times (n is an integer equal to or greater than 1; in this example, n=2 or n=3) the number of bits of data input / output to the system bus SBUS.

[0026] Each test circuit BIST operates during the test mode, outputs an access request to the corresponding memory control unit CNT1, and performs a test on the corresponding memory unit MEM. When each test circuit BIST detects a defect by the test, it stores defect information including a defect address indicating the location of the defect in the non-volatile memory NVM. The access request output by the test circuit BIST is an example of a first access request.

[0027] For example, the nonvolatile memory NVM is a memory in which information can be electrically written, such as eFUSE or a flash memory. The nonvolatile memory NVM may be a magnetoresistive random access memory (MRAM) or a resistive memory. The defect information stored in the nonvolatile memory NVM is retained even when the power supply to the memory circuit 100 is cut off. The defect information stored in the nonvolatile memory NVM is output to the corresponding memory control unit CNT1 and the memory control unit CNT2 of the memory management unit MNG (input side).

[0028] Note that one non-volatile memory NVM commonly connected to the multiple test circuits BISTs and the multiple memory control units CNT1 may be provided in the memory circuit 100 instead of the multiple non-volatile memories NVMs shown in Fig. 1. The functions of the memory control unit CNT1 and the test circuit BISTs are described in Figs. 4 to 6.

[0029] Also, the memory circuit 100 does not need to have a test circuit BIST. In this case, the memory circuit 100 is tested by a test circuit of an LSI tester connected to the memory circuit 100, and the test results (defective addresses) are stored in each nonvolatile memory NVM. However, in this case, since it is necessary to connect each memory control unit CNT1 to an external test circuit during testing, the number of external terminals of the memory circuit 100 increases.

[0030] Each relay control unit RCNT has a selection unit SEL. The relay control unit RCNT operates during normal operation mode and does not operate during test mode. The first-stage relay control unit RCNT outputs the received access request MREQ to the next-stage relay control unit RCNT or the following memory control unit CNT1. The relay control units RCNT except for the first and final stages output the received access request MREQ or read data RD to the next-stage relay control unit RCNT or the following memory control unit CNT1. The final-stage relay control unit RCNT outputs the received read data RD to the memory management unit MNG (output side). In this way, the relay control unit RCNT can sequentially transfer the access request MREQ. Here, the write request includes a command CMD, an address AD, and write data WD, and the read request includes a command CMD and an address AD.

[0031] If the received address AD indicates the address AD assigned to the subsequent memory unit MEM, the selection unit SEL outputs the received access request MREQ to the subsequent memory control unit CNT1. If the received address AD does not indicate the address AD assigned to the subsequent memory unit MEM, or if the selection unit SEL receives read data RD, it outputs the received access request MREQ or read data RD to the next relay control unit RCNT. The selection unit SEL of the final stage relay control unit RCNT outputs the received read data RD to the memory management unit MNG (output side).

[0032] Fig. 2 shows an example of word redundancy in each memory unit MEM in Fig. 1. For example, each of the memory units MEM0-MEM3 has 8190 word lines WL and two bubble redundancy word lines BRWL. In the memory unit MEM3, the six word lines WL and the two bubble redundancy word lines BRWL of the memory management unit MNG (output side) function as sub-word lines SWL.

[0033] As shown in Fig. 2, the bubble redundancy word line BRWL is arranged discretely (so that bubbles are dispersed) between the memory management unit MNG (input side) and the memory management unit MNG (output side). This makes it possible to operate a plurality of memory units MEM independently and in parallel using a plurality of test circuits BIST to perform testing, and to repair word line defects using the bubble redundancy word line BRWL.

[0034] This makes it possible to shorten the test time and improve the test efficiency even when the memory capacity of the memory circuit 100 is large. As a result, it is possible to suppress an increase in the manufacturing cost of the memory circuit 100. Note that a word line defect refers not only to a defect in the word line WL itself, but also to a defect in a memory cell connected to the word line WL, as indicated by an X mark in FIG. 2.

[0035] The eight sub-word lines SWL are assigned to correspond to two bubble redundancy word lines BRWL in each of the four memory units MEM0-MEM3. When an address AD indicated by an access request MREQ for the memory units MEM0-MEM3 indicates a bubble redundancy word line BRWL, the memory control unit CNT2 of the memory management unit MNG (input side) converts the address AD into an address AD indicating the corresponding sub-word line SWL, as indicated by the dashed arrow.

[0036] This allows the addresses of the word lines assigned in the memory circuit 100 to be continuous even when the bubble redundant word lines BRWL are discretely arranged between the memory management unit MNG (input side) and the memory management unit MNG (output side). As a result, it is possible to prevent a decrease in the usability of the memory circuit 100 by a system (e.g., a user or an application) that uses the memory circuit 100.

[0037] 3 shows an example of the memory control unit CNT1 in FIG. 1. The memory control unit CNT1 operates during normal operation mode and test mode. The memory control unit CNT1 has selectors SEL0 and SEL1, a comparator CMP, an AND circuit AND, and a register REG that holds the address BRAD of each bubble redundant word line BRWL. Note that two comparators CMP, two AND circuits AND, two registers REG, and two selectors SEL1 are provided corresponding to each of the two bubble redundant word lines BRWL.

[0038] When the mode MD indicates the test mode (=0), the selector SEL0 selects the access request MREQ received from the test circuit BIST. When the mode MD indicates the normal operation mode (=1), the selector SEL0 selects the access request MREQ received from the relay control unit RCNT. The selector SEL0 outputs the command CMD and write data WD (only during a write operation) included in the selected access request MREQ to the memory unit MEM, and outputs the address AD included in the access request MREQ to the comparator CMP and the selector SEL1.

[0039] Each comparator CMP compares the address AD from the selector SEL0 with the defective address FAD from the nonvolatile memory NVM, and if they match, outputs a high level to the AND circuit AND, and if they do not match, outputs a low level to the AND circuit AND. The nonvolatile memory NVM outputs up to two defective addresses FAD to the two comparators CMP, respectively, corresponding to the two bubble redundant word lines BRWL.

[0040] Each AND circuit AND outputs a high level to the corresponding selector SEL1 when it receives a high level from the corresponding comparator CMP and a high level redundancy enable REDEN indicating that the corresponding defective address FAD is valid. Each AND circuit AND outputs a low level to the corresponding selector SEL1 when it receives a low level from the corresponding comparator CMP or a low level redundancy enable REDEN indicating that the corresponding defective address FAD is invalid.

[0041] When each selector SEL1 receives a high level from the corresponding AND circuit AND, that is, when a defect is relieved using the bubble redundant word line BRWL, it outputs an address BRAD indicating the bubble redundant word line BRWL held in the corresponding register REG to the memory unit MEM.When each selector SEL1 receives a low level from the corresponding AND circuit AND, that is, when a memory cell or the like to be accessed by the access request MREQ operates normally, it outputs an address AD received from the relay control unit RCNT to the memory unit MEM.

[0042] The test circuit BIST operating in the test mode judges a defect when the read data RD read from the memory unit MEM does not match the expected value (i.e., the write data WD). If the defect can be relieved by the bubble redundant word line BRWL, the test circuit BIST stores the defective address FAD, which is the address of the word line WL in which the defect is detected, and the high-level redundancy enable REDEN in the non-volatile memory NVM.

[0043] Fig. 4 shows an example of the operation of the memory control unit CNT1 in Fig. 3. First, in step S10, the memory control unit CNT1 waits until an access request MREQ is received, and when the access request MREQ is received, it carries out step S11.

[0044] In step S11, the memory control unit CNT1 judges whether the access address included in the access request MREQ is a defective address FAD. The defective address FAD is an example of a first address indicating an inaccessible memory cell. If the access address is the defective address FAD, the memory control unit CNT1 performs step S12, and if the access address is not the defective address FAD, the memory control unit CNT1 performs step S13.

[0045] In the first test shown in FIG. 5, since the defective address FAD is not stored in the nonvolatile memory NVM, the memory control unit CNT1 determines in step S11 that there is a mismatch between the access address and the defective address FAD.

[0046] In step S12, the memory control unit CNT1 replaces the access address with an address BRAD indicating any one of the bubble redundant word lines BRWL. The address BRAD indicating the bubble redundant word line BRWL is an example of a second address. In step S13, the memory control unit CNT1 outputs the access address included in the access request MREQ or the address BRAD replaced in step S12 to the memory unit MEM, and performs a read access or a write access to the memory unit MEM. The memory control unit CNT1 repeats the operations of steps S10-S13.

[0047] Fig. 5 shows an example of a first test for detecting defects in each memory unit MEM by each test circuit BIST in Fig. 1. For example, the first test shown in Fig. 5 is performed in parallel by a plurality of test circuits BIST.

[0048] First, in step S20, the test circuit BIST outputs a write request to the memory control unit CNT1. The write request output to the memory control unit CNT1 is an example of a first access request. The memory control unit CNT1 performs write access to the memory unit MEM based on the write request, and writes test data (expected value) to the memory unit MEM.

[0049] Next, in step S21, the test circuit BIST outputs a read request to the memory control unit CNT1. The read request output to the memory control unit CNT1 is an example of a first access request. The memory control unit CNT1 reads and accesses the memory unit MEM based on the read request, and reads data from the memory unit MEM.

[0050] Next, in step S22, the test circuit BIST determines whether the read data matches the expected value (i.e., the write data in step S20). If the read data matches the expected value, there is no defect, so the test circuit BIST performs step S25. If the read data does not match the expected value, the test circuit BIST performs step S23 to repair the defect.

[0051] Next, in step S23, the test circuit BIST judges whether the defect can be repaired by the bubble redundant word line BRWL. If the defect can be repaired, the test circuit BIST performs step S24, and if the defect cannot be repaired, the test circuit BIST performs step S26. For example, an example in which the defect cannot be repaired is when there is no free space in the bubble redundant word line BRWL.

[0052] In step S24, the test circuit BIST stores the defective address FAD and the redundancy enable REDEN (=1) in the nonvolatile memory NVM corresponding to the bubble redundant word line BRWL used to relieve the defect, and executes step S25. In step S25, when the test circuit BIST has tested all areas of the memory unit MEM except for the bubble redundant word line BRWL, it ends the operation shown in Fig. 5, and when there is an area in the memory unit MEM that has not been tested, it returns to the operation of step S20.

[0053] In step S26, the test circuit BIST detects that the tested memory unit MEM has a failure that cannot be relieved by the bubble redundant word line BRWL, and ends the operation shown in FIG.

[0054] Fig. 6 shows an example of a second test for checking the operation of each memory unit MEM after storing failure information in the nonvolatile memory NVM by each test circuit BIST in Fig. 1. For example, the second test shown in Fig. 6 is performed in parallel by a plurality of test circuits BIST and a plurality of memory control units CNT1.

[0055] First, in step S30, the test circuit BIST outputs a write request to the memory control unit CNT1, similarly to step S20 in Fig. 5. When the address included in the write request matches the defective address FAD stored in the non-volatile memory NVM, the memory control unit CNT1 converts the address included in the write request into an address BRAD indicating the bubble redundant word line BRWL, writes access to the memory unit MEM, and writes test data (expected value) to the memory unit MEM. When the access address included in the write request does not match the defective address FAD stored in the non-volatile memory NVM, the memory control unit CNT1 writes access to the memory unit MEM using the address included in the write request, and writes test data (expected value) to the memory unit MEM.

[0056] Next, in step S31, similar to step S21 in Fig. 5, the test circuit BIST outputs a read request to the memory control unit CNT1. Similar to when receiving a write request, if the address included in the read request matches the defective address FAD, the memory control unit CNT1 converts the address included in the read request into the address BRAD indicating the bubble redundant word line BRWL and performs a read access to the memory unit MEM. Also, if the address included in the read request does not match the defective address FAD, the memory control unit CNT1 performs a write access to the memory unit MEM using the address included in the read request.

[0057] Next, in step S32, the test circuit BIST determines whether the read data matches the expected value, similar to step S22 in Fig. 5. If the read data matches the expected value, the test circuit BIST determines that the device operates normally, and performs step S33. If the read data does not match the expected value, the test circuit BIST determines that there is a defect, and performs step S34.

[0058] In step S33, similar to step 25 in FIG. 5, if the test circuit BIST has tested all areas of the memory unit MEM except for the bubble redundant word line BRWL, it ends the operation shown in FIG. 6, and if there is an untested area in the memory unit MEM, it returns to the operation of step S30.

[0059] In step S34, the test circuit BIST detects that there is an unrepairable failure in the tested memory unit MEM, similarly to step S26 in FIG. 5, and ends the operation shown in FIG.

[0060] As shown in Figures 5 and 6, by sequentially performing the first and second tests on the multiple memory units MEM using multiple test circuits BIST, it is possible to detect defects contained in each of the multiple memory units MEM and check the defects after they have been repaired in parallel.

[0061] Fig. 7 shows an example of the operation of the memory control unit CNT2 that receives an access request MREQ via the system bus SBUS. Fig. 7 shows the operation that the memory circuit 100 performs in the normal operation mode.

[0062] First, in step S40, the memory control unit CNT2 waits until the memory management unit MNG receives an access request MREQ, and when the memory management unit MNG receives the access request MREQ, it carries out step S41.

[0063] In step S41, the memory control unit CNT2 judges whether or not the access address included in the access request MREQ is the address BRAD indicating the bubble redundant word line BRWL. If the access address is the address BRAD, the memory control unit CNT2 performs step S42, and if the access address is not the address BRAD, the memory control unit CNT2 performs step S43.

[0064] In step S42, the memory control unit CNT2 converts the access address into an address indicating the sub-word line SWL of the memory unit MEM3, and executes step S43. The address indicating the sub-word line SWL is an example of a third address. In step S43, the memory control unit CNT2 outputs an access request to the first-stage relay control unit RCNT. The memory control unit CNT2 repeats the operations of steps S40-S43.

[0065] Fig. 8 shows an example of the operation of the relay control unit RCNT except for the final stage. Fig. 8 shows the operation that the relay control unit RCNT performs during the normal operation mode. The relay control unit RCNT in the final stage receives read data RD together with a read request, and outputs the received read data RD to the memory management unit MNG (output side).

[0066] First, in step S50, the relay control unit RCNT waits until it receives an access request MREQ or read data RD, and when it receives the access request MREQ or read data RD, it executes step S51. The access request MREQ is supplied from the memory management unit MNG (input side) or the relay control unit RCNT in the previous stage. The read data RD is supplied from the relay control unit RCNT in the previous stage.

[0067] In step S51, if the relay control unit RCNT receives an access request MREQ, it carries out step S52, and if the access request MREQ has not been received, it carries out step S55 since it has received read data RD.

[0068] In step S52, the relay control unit RCNT judges whether the access request MREQ is for the immediately succeeding memory unit MEM. If the access request MREQ is for the succeeding memory unit MEM, the relay control unit RCNT performs step S53, and if the access request MREQ is not for the succeeding memory unit MEM, the relay control unit RCNT performs step S54.

[0069] In step S53, the relay control unit RCNT outputs an access request MREQ to the succeeding memory control unit CNT1, and the operation returns to step S50. In step S54, the relay control unit RCNT outputs an access request MREQ to the next relay control unit RCNT, and the operation returns to step S50.

[0070] In step S55, the repeater control unit RCNT outputs the read data RD to the next repeater control unit RCNT, and the operation returns to step S50.

[0071] As described above, in the first embodiment, in the test mode, the multiple memory units MEM can be accessed in parallel for testing, and in the normal operation mode, the multiple memory units MEM can be accessed as one memory. As a result, it is possible to provide a memory circuit 100 that can operate as multiple small-capacity memories or a large-capacity memory by switching the operation mode.

[0072] When the address AD included in the access request MREQ indicates a bubble redundancy word line BRWL, the memory control unit CNT2 converts the address AD into an address AD indicating a corresponding sub-word line SWL. This makes it possible to make the addresses AD of the word lines assigned in the memory circuit 100 continuous even when the bubble redundancy word lines BRWL are discretely arranged. As a result, it is possible to prevent the usability of the memory circuit 100 from being reduced by the system that uses the memory circuit 100.

[0073] By providing a test circuit BIST and a bubble redundant word line BRWL for each memory unit MEM (memory block MBLK), it is possible to operate and test a plurality of memory units MEM in parallel. In addition, by sequentially performing the first test and the second test on the plurality of memory units MEM using the plurality of test circuits BIST, it is possible to perform detection of defects contained in each of the plurality of memory units MEM and confirmation after the defects are relieved in parallel. This makes it possible to shorten the test time and improve the test efficiency even when the memory capacity of the memory circuit 100 is large. As a result, it is possible to suppress an increase in the manufacturing cost of the memory circuit 100.

[0074] When the access request MREQ does not indicate an access to the subsequent memory block MBLK, the relay control unit RCNT transfers the access request MREQ to the next relay control unit RCNT, bypassing the memory block MBLK. This makes it possible to prevent the command line CMD, the address line AD, the write data line WD, and the read data line DT from being routed through a plurality of memory blocks BLK. As a result, it is possible to suppress an increase in the wiring load, and it is possible to suppress an increase in the access time of the memory circuit 100 while suppressing an increase in power consumption.

[0075] Fig. 9 shows an overview of a memory circuit of the second embodiment. Elements that are the same as or similar to those in Fig. 1 are given the same reference numerals, and detailed description will be omitted. In the memory circuit 101 shown in Fig. 9, data can be read and written in parallel from the outside to each memory unit MEM during normal operation mode. The configuration and function of the memory circuit 101 are the same as those of the memory circuit 100 in Fig. 1, except that the function of the memory control unit CNT1 is added.

[0076] For this reason, each memory control unit CNT1 has a function of receiving an access request MREQ (CMD, AD, WD) from outside the memory circuit 101, and a function of outputting read data RD to outside the memory circuit 101. Each memory control unit CNT1 may receive the access request MREQ via the system bus SBUS. The access request output by each test circuit BIST and the access request MREQ received by each memory control unit CNT1 are an example of a first access request.

[0077] In this embodiment, the normal operation mode has a first normal operation mode in which each memory unit MEM can be accessed in parallel as a small-capacity memory, and a second normal operation mode in which all memory units MEM can be accessed as one large-capacity memory. In the first normal operation mode, an access request MREQ is supplied to each memory control unit CNT1. In the second normal operation mode, an access request MREQ is supplied to the memory management unit MNG (input side). The first normal operation mode is an example of the first mode, and the second normal operation mode is an example of the second mode.

[0078] Fig. 10 shows an example of the memory control unit CNT1 of Fig. 9. Elements that are the same as or similar to those in Fig. 3 are given the same reference numerals, and detailed description will be omitted. The memory control unit CNT1 shown in Fig. 10 has the same configuration as the memory control unit CNT1 of Fig. 3, except that an OR circuit OR is added to the memory control unit CNT1 of Fig. 3.

[0079] The OR circuit OR outputs to the selection unit SEL0 an access request MREQ from the test circuit BIST or an access request MREQ from outside the memory circuit 101. The function of the memory control unit CNT1 shown in Fig. 10 is the same as that of the memory control unit CNT1 in Fig. 3, except that the memory unit MEM can be accessed based on an access request MREQ from outside the memory circuit 101 as well as the relay control unit RCNT during the normal operation mode. Note that read data RD output from the memory unit MEM based on an access request MREQ (read request) received during the normal operation mode is output to the relay control unit RCNT and to the outside of the memory circuit 101.

[0080] Fig. 11 shows an example of word redundancy in the memory block of Fig. 9. Detailed description of operations similar to those in Fig. 2 will be omitted. For example, each of the memory units MEM0-MEM3 has 8190 word lines WL and two bubble redundancy word lines BRWL, similar to Fig. 2. The six word lines WL and two bubble redundancy word lines BRWL of the memory unit MEM3 function as sub-word lines SWL.

[0081] During the test mode and the first normal operation mode, defects in each memory unit MEM are repaired independently for each memory unit MEM. Also, in the first normal operation mode, the process of converting the addresses of the bubble redundancy word lines BRWL of the memory units MEM0-MEM2 into the addresses of the sub-word lines SWL of the memory unit MEM3 of the memory block MBLK1 is not performed. Word repair during the second normal operation mode is performed in the same manner as in FIG. 2. That is, when the address included in the access request MREQ indicates the bubble redundancy word line BRWL, the address included in the access request MREQ is converted into an address indicating the corresponding sub-word line SWL, as indicated by the dashed arrow.

[0082] As described above, the second embodiment can also provide the same effects as the first embodiment. For example, in a test mode, a plurality of memory units MEM are accessed in parallel to be tested, thereby shortening the test time and improving the test efficiency. As a result, an increase in the manufacturing cost of the memory circuit 101 can be suppressed.

[0083] Furthermore, in the second embodiment, the memory circuit 101 has a first normal operation mode in which the multiple memory units MEM can be accessed in parallel as small-capacity memories, and a second normal operation mode in which the multiple memory units MEM can be accessed as one large-capacity memory. This makes it possible to provide a memory circuit 101 that can operate as multiple small-capacity memories or a large-capacity memory by switching the operation mode.

[0084] Fig. 12 shows an overview of a memory circuit of the third embodiment. Elements that are the same as or similar to those in Fig. 1 are given the same reference numerals, and detailed description is omitted. The memory circuit 102 shown in Fig. 12 has three relay control units RCNT arranged between a memory management unit MNG (input side) and a memory management unit MNG (output side), and a memory block MBLK (MBLK0, MBLK1) arranged between two adjacent relay control units RCNT.

[0085] Each memory block MBLK has four memory units MEM (MEM0-MEM3). Each of the memory units MEM0-MEM3 of the memory blocks MBLK0 and MBLK1 has a plurality of word lines WL and a predetermined number of bubble redundant word lines BRWL, similar to each of the memory units MEM in FIG.

[0086] In addition, in the memory unit MEM3 of the memory block BLK1, a predetermined number of word lines WL and bubble redundancy word lines BRWL also function as sub-word lines SWL, similar to the memory unit MEM3 in Fig. 1. The sub-word lines SWL are assigned to correspond to the bubble redundancy word lines BRWL of each memory unit MEM, similar to Fig. 2. When an address AD indicated by an access request MREQ indicates a bubble redundancy word line BRWL, the memory control unit CNT2 of the memory management unit MNG (input side) converts the address AD into an address AD indicating the corresponding sub-word line SWL.

[0087] The memory circuit 102 can be tested independently and in parallel on multiple memory units MEM0-MEM3 of multiple memory blocks MBLK using multiple test circuits BIST as described in Fig. 2. The method of repairing the word lines WL in the memory circuit 102 is similar to the method of repairing the word lines WL shown in Fig. 2, except that the addresses of the bubble redundant word lines BRWL of the memory units MEM0-MEM3 of the memory block MBLK0 and the memory units MEM0-MEM2 of the memory block MBLK1 are converted to the addresses of the sub-word lines SWL of the memory unit MEM3 of the memory block MBLK1.

[0088] In this embodiment, when the access address included in the access request MREQ indicates the bubble redundant word line BRWL, the memory control unit CNT2 converts the access address into an address indicating one of the sub-word lines SWL of the memory unit MEM3 of the memory block MBLK1. The memory control unit CNT2 outputs the converted address to the first-stage relay control unit RCNT. When the access address included in the access request MREQ does not indicate the bubble redundant word line BRWL, the memory control unit CNT2 outputs the access address to the first-stage relay control unit RCNT without converting it.

[0089] The memory circuit 102 may have five relay control units RCNT and four memory blocks BMLK, and each memory block MBLK may be disposed between two adjacent relay control units RCNT. In this case, the method of repairing the word lines WL is the same as the method of repairing the word lines WL shown in Fig. 2, except that each memory block MBLK has four memory units MEM0-MEM3. The number of relay control units RCNT and memory blocks BMLK is not limited to this.

[0090] As described above, the third embodiment can also provide the same effects as the first embodiment. Furthermore, in the third embodiment, each memory block MBLK has a plurality of memory units MEM, a memory control unit CNT1 corresponding to the memory units MEM, a test circuit BIST, and a non-volatile memory NVM. This allows the memory capacity, which is the unit for performing a test, to be smaller than that of the first embodiment. As a result, the test time can be further shortened and the test efficiency can be further improved compared to the first embodiment.

[0091] Fig. 13 shows an overview of a memory circuit according to a fourth embodiment. Elements identical or similar to those in Fig. 1, Fig. 9, and Fig. 12 are given the same reference numerals, and detailed description thereof will be omitted. The memory circuit 103 shown in Fig. 13 has the same configuration and function as the memory circuit 102 shown in Fig. 12, except that, like the memory circuit 101 shown in Fig. 9, data can be read and written in parallel from the outside to each memory unit MEM during normal operation mode.

[0092] In the memory circuit 103 shown in FIG. 13, the normal operation mode includes a first normal operation mode in which each memory unit MEM can be accessed in parallel as a small-capacity memory, and a second normal operation mode in which all memory units MEM can be accessed as one large-capacity memory. In the first normal operation mode, an access request MREQ is supplied to each memory control unit CNT1. In the second normal operation mode, an access request MREQ is supplied to the memory management unit MNG (input side). The method of repairing words in the test mode, the first normal operation mode, and the second normal operation mode is the same as that in FIG. 11, except that each memory block MBLK has multiple memory units MEM.

[0093] As described above, the fourth embodiment can also provide the same effects as the first to third embodiments.

[0094] FIG. 14 shows an overview of the memory circuit of the fifth embodiment. The same or similar elements as those in FIG. 1, FIG. 9, FIG. 12, and FIG. 13 are given the same reference numerals, and detailed description is omitted. The memory circuit 104 shown in FIG. 14 has a plurality of memory blocks MBLK (MBLK0, MBLK1) having the same or similar configuration as the memory block MBLK1 in FIG. 13, and a sub-memory management unit MNGs provided corresponding to each memory block MBLK. The configuration and functions other than the memory block MBLK and the sub-memory management unit MNGs are the same as those of the memory circuit 103 in FIG. 13, except that a function of receiving an access request MREQ from the sub-memory management unit MNGs is added to the relay control unit RCNT.

[0095] The memory circuit 104 has a third normal operation mode in addition to the first and second normal operation modes similar to those of the memory circuit 101 of FIG. 9. In the first normal operation mode, each memory unit MEM can be accessed in parallel as a small-capacity memory. In the second normal operation mode, all memory units MEM can be accessed as one large-capacity memory. In the third normal mode, each memory block MBLK can be accessed in parallel as a medium-capacity memory. This makes it possible to provide a memory circuit 104 that can operate as multiple small-capacity memories, multiple medium-capacity memories, or a large-capacity memory by switching the operation mode.

[0096] The memory units MEM0-MEM2 of each memory block MBLK have a plurality of word lines WL and a predetermined number of bubble redundancy word lines BRWL, similar to each memory unit MEM in Fig. 1. In the memory unit MEM3 of each memory block BLK, the predetermined number of word lines WL and bubble redundancy word lines BRWL also function as sub-word lines SWL, similar to the memory unit MEM3 in Fig. 1. The conversion of the addresses of the bubble redundancy word lines BRWL to the addresses of the sub-word lines SWL of the memory unit MEM3 is performed for each memory block MBLK.

[0097] The sub-memory management unit MNGs is a subset of the memory management unit MNG (input side) and has a different range for converting the address AD. The memory management unit MNG (input side) performs address conversion across multiple memory blocks MBLK, and the sub-memory management unit MNGs performs address conversion within the corresponding memory block MBLK. Each sub-memory management unit MNGs operates during a third normal operation mode in which it receives an access request MREQ from the system bus SBUS, and outputs the received access request MREQ to the corresponding relay control unit RCNT. Each sub-memory management unit MNGs has a sub-memory control unit CNT2s corresponding to the memory control unit CNT2.

[0098] Each sub-memory control unit CNT2s refers to the defect information stored in the non-volatile memory NVM of the corresponding memory block MBLK or grasps the bubble redundant word lines BRWL of multiple memory units MEM as specifications, determines the memory unit MEM to which the access request MREQ is to be transferred, and converts the address AD included in the access request MREQ into the address AD of the determined destination memory unit MEM.

[0099] When the address AD included in the access request MREQ indicates a bubble redundant word line BRWL, each sub-memory control unit CNT2s converts the address AD into an address AD indicating a sub-word line SWL in the corresponding memory block MBLK. This allows the memory units MEM0-MEM3 in each memory block MBLK to be used as an area with continuous addresses AD in the third normal operation mode.

[0100] The word repair method in the test mode, the first normal operation mode, the second normal operation mode, and the third normal operation mode is the same as that in FIG. 11, except that each memory block MBLK has a plurality of memory units MEM. However, in the second normal operation mode, when the address AD indicated by the access request MREQ indicates a bubble redundancy word line BRWL, the memory control unit CNT2 converts the address AD into an address AD indicating a corresponding sub-word line SWL in the memory unit MEM3 of the memory block MBLK1. In the third normal operation mode, when the address AD indicated by the access request MREQ indicates a bubble redundancy word line BRWL, each sub-memory control unit CNT2s converts the address AD into an address AD indicating a corresponding sub-word line SWL in the memory unit MEM3 of the corresponding memory block MBLK. Note that, if the address AD may be discretely assigned by the bubble redundancy word line BRWL, the conversion of the sub-word line SWL into the address AD by the sub-memory control unit CNT2s is not required.

[0101] As described above, the fifth embodiment can also provide the same effects as the first to fourth embodiments. Furthermore, this embodiment can provide a memory circuit 104 that can operate as multiple small capacity memories, multiple medium capacity memories, or a large capacity memory by switching the operation mode.

[0102] FIG. 15 shows an outline of a memory circuit of the sixth embodiment. The same or similar elements as those in FIG. 1 and FIG. 12 are given the same reference numerals, and detailed description is omitted. The memory circuit 105 shown in FIG. 14 has a plurality of relay control units RCNT, as in FIG. 1. The memory circuit 105 also has a memory control unit CNT1, a memory unit MEM (MEM0, MEM1, MEM2 or MEM3), a test circuit BIST, and a non-volatile memory NVM, each of which is disposed between two adjacent relay control units RCNT. In this embodiment, a group GR including two sets of the memory control unit CNT1, the memory unit MEM, the test circuit BIST, and the non-volatile memory NVM is assigned to each memory block MBLK (MBLK0, MBLK1).

[0103] In this embodiment, when one memory unit MEM is tested in each memory block MBLK (i.e., each group GR), the word line WL of the other memory unit MEM is used as a redundant word line. That is, during the test mode, in each group GR, a defect in one memory unit MEM is relieved by the word line WL of the other memory unit MEM.

[0104] The bubble redundancy word lines BRWL are used only in the normal operation mode. In the memory unit MEM3, the word lines WL of the memory management unit MNG (output side) and a predetermined number of the bubble redundancy word lines BRWL function as sub-word lines SWL. The method of repairing defects using the bubble redundancy word lines BRWL in the normal operation mode and the method of using the sub-word lines SWL are the same as those in the first and second embodiments.

[0105] For example, during a test mode, a defective word line WL of one memory unit MEM is replaced with a word line WL having the same number (i.e., the same word line address) as the defective word line WL in the other memory unit MEM. This allows the memory circuit 105 to be tested independently for each group GR, and allows tests of multiple groups GR to be performed in parallel. In the test of each group GR, first, one memory unit MEM is tested, and then the other memory unit MEM is tested.

[0106] The memory cells (circles) connected to each word line WL are an example of the first memory cells. The memory cells (circles) connected to each bubble redundant word line BRWL are an example of the second memory cells. In FIG. 13, the memory circuit 105 has five relay control units RCNT and two memory blocks MBLK, but the numbers of relay control units RCNT and memory blocks BMLK are not limited to this.

[0107] Each test circuit BIST stores a defective address indicating a word line WL in which a defect is detected in a corresponding non-volatile memory NVM in a first test for detecting a defect in each memory unit MEM. Also, each test circuit BIST performs a second test for checking the operation of each memory unit MEM after storing the defect information in the non-volatile memory NVM. That is, each test circuit BIST can perform a second test for checking the operation of each memory unit MEM after storing the defect information in the non-volatile memory NVM based on the result of the first test.

[0108] In the second test, each memory control unit CNT1 can transfer an access request MREQ to the other memory control unit CNT1 in the group GR in accordance with a defective address stored in the nonvolatile memory NVM. Each test circuit BIST can receive read data RD from both of the pair of memory units MEM.

[0109] FIG. 16 shows an example of word redundancy of the memory block MBLK of FIG. 15. For example, each of the memory units MEM0-MEM3 has 8192 word lines. During the normal operation mode, as in FIG. 2, defects in each memory unit MEM are redundancy redundancy word lines BRWL. Also, eight sub-word lines SWL are assigned to correspond to two bubble redundancy word lines BRWL of each of the four memory units MEM0-MEM3. When an address included in an access request MREQ indicates a bubble redundancy word line BRWL, the address included in the access request MREQ is converted to an address indicating the corresponding sub-word line SWL, as shown by the dashed arrow.

[0110] For example, during the test mode, a defect in the memory unit MEM0 indicated by an X mark is repaired using the word line WL of the memory unit MEM1, and a defect in the memory unit MEM1 indicated by an X mark is repaired using the word line WL of the memory unit MEM0. Similarly, during the test mode, a defect in the memory unit MEM2 indicated by an X mark is repaired using the word line WL of the memory unit MEM3, and a defect in the memory unit MEM3 indicated by an X mark is repaired using the word line WL of the memory unit MEM2.

[0111] This makes it possible to shorten the test time by testing a plurality of memory blocks BMLK in parallel in the test mode, while at the same time relieving defects by using the bubble redundant word line BRWL in the normal operation mode.

[0112] Fig. 17 shows an example of a first test for detecting defects in each memory unit MEM by each test circuit BIST in Fig. 15. The same operations as those in Fig. 5 are indicated with the same step numbers, and detailed explanations are omitted. For example, the first test shown in Fig. 17 is performed in parallel by one test circuit BIST of a plurality of groups GR. The test flow shown in Fig. 17 is the same as the test flow in Fig. 5, except that step S23A is performed instead of step S23 in Fig. 5, and step S25A is performed instead of step S25.

[0113] In step S23A, the test circuit BIST judges whether or not the defect can be repaired using the word line WL of the adjacent memory unit MEM. If the defect can be repaired, the test circuit BIST performs step S24, and if the defect cannot be repaired, the test circuit BIST performs step S26. For example, an example in which the defect cannot be repaired is when the word line WL indicated by the defective address FAD in the adjacent memory unit MEM cannot be accessed. After step S24, in step S25A, if the entire area of ​​the memory unit MEM has been tested, the test circuit BIST ends the operation shown in FIG. 17, and if there is an area in the memory unit MEM that has not been tested, returns to the operation of step S20.

[0114] FIG. 18 shows an example of a second test for checking the operation of each memory unit MEM after failure information is stored in the nonvolatile memory NVM by each test circuit BIST of FIG. 15. The same step numbers are indicated for operations similar to those in FIG. 6, and detailed descriptions are omitted. For example, the second test shown in FIG. 18 is performed in parallel by one test circuit BIST of a plurality of groups GR. The test flow shown in FIG. 18 is the same as the test flow in FIG. 5, except that step S33A is performed instead of step S33 in FIG. 6. After step S32, in step S33A, if the test circuit BIST has tested the entire area of ​​the memory unit MEM, the test circuit BIST ends the operation shown in FIG. 18, and if there is an area in the memory unit MEM that has not been tested, the test circuit BIST returns to the operation of step S30.

[0115] When the memory control unit CNT1 receives a write request from the test circuit BIST, if the address included in the write request matches the defective address FAD stored in the nonvolatile memory NVM, the memory control unit CNT1 transfers the write request to the adjacent memory control unit CNT1. When the memory control unit CNT1 receives a write request from the test circuit BIST, if the access address included in the write request does not match the defective address FAD stored in the nonvolatile memory NVM, the memory control unit CNT1 writes access to the memory unit MEM using the address included in the write request and writes test data (expected value) to the memory unit MEM. When the memory control unit CNT1 receives a write request from the adjacent memory control unit CNT1, the memory control unit CNT1 writes access to the memory unit MEM using the address included in the write request and writes test data (expected value) to the memory unit MEM.

[0116] Furthermore, when the memory control unit CNT1 receives a read request from the test circuit BIST, if the address included in the read request matches the defective address FAD stored in the nonvolatile memory NVM, the memory control unit CNT1 transfers the read request to the adjacent memory control unit CNT1. Furthermore, when the memory control unit CNT1 receives a read request from the test circuit BIST, if the access address included in the read request does not match the defective address FAD stored in the nonvolatile memory NVM, the memory control unit CNT1 reads and accesses the memory unit MEM using the address included in the read request. When the memory control unit CNT1 receives a read request from the adjacent memory control unit CNT1, the memory control unit CNT1 reads and accesses the memory unit MEM using the address included in the read request.

[0117] As described above, the sixth embodiment can also provide the same effects as the first embodiment. For example, in a test mode, a plurality of memory units MEM can be accessed in parallel for testing, and in a normal operation mode, a plurality of memory units MEM can be accessed as one memory. As a result, a memory circuit 105 can be provided that can operate as a plurality of small-capacity memories or a large-capacity memory by switching the operation mode. Also, in a test mode, a plurality of memory units MEM can be accessed in parallel for testing, thereby shortening the test time and improving the test efficiency. As a result, an increase in the manufacturing cost of the memory circuit 105 can be suppressed.

[0118] Furthermore, in the sixth embodiment, in each group GR, a defect in one memory unit MEM during the test mode is relieved by the word line WL of the other memory unit MEM. This makes it possible to perform the first test and the second test without allocating the bubble redundancy word line BRWL shown in FIG. 1 to each memory unit MEM, and therefore makes it possible to perform a test of the entire area of ​​each memory unit MEM.

[0119] Fig. 19 shows an overview of a memory circuit according to the seventh embodiment. Elements identical or similar to those in Fig. 15 are given the same reference numerals, and detailed description thereof will be omitted. In the memory circuit 106 shown in Fig. 19, similarly to Fig. 9, data can be read and written in parallel from outside the memory circuit 106 to a plurality of memory units MEM during normal operation mode.

[0120] In this embodiment, similarly to FIG. 9, the normal operation mode includes a first normal operation mode in which a plurality of memory units MEM can be accessed in parallel as small-capacity memories, and a second normal operation mode in which a plurality of memory units MEM can be accessed as one large-capacity memory. In the first normal operation mode, an access request MREQ is supplied to each memory control unit CNT1. In the second normal operation mode, an access request MREQ is supplied to the memory management unit MNG (input side). Note that the test of the memory block MBLK in the test mode is the same as the test of the memory block MBLK shown in FIG. 17 and FIG. 18.

[0121] In this embodiment, in the first normal operation mode, each memory block BLK uses one of a pair of memory units MEM in the group GR as a main memory area, and the other of the pair of memory units MEM as a redundant memory area for repairing defects. That is, the repair method for defects in the first normal operation mode is the same as the repair method in the test mode shown in Fig. 16. On the other hand, the repair method for defects in the second normal operation mode is the same as the repair method in the normal operation mode shown in Fig. 16.

[0122] In the first normal operation mode, it is also possible to independently repair defects for each memory unit MEM, as in the second embodiment (FIGS. 9, 10, and 11). In that case, although not shown in FIG. 19, each of the pair of memory control units CNT1 in each memory block MBLK (i.e., each group GR) is provided with a function of receiving an access request MREQ (CMD, AD, WD) from outside the memory circuit 106, and a function of outputting read data RD to outside the memory circuit 106. Each memory control unit CNT1 may receive an access request MREQ via the system bus SBUS and output read data RD to the system bus SBUS.

[0123] As described above, the seventh embodiment can also provide the same effects as the first to sixth embodiments.

[0124] Fig. 20 shows an outline of a memory circuit of the eighth embodiment. Elements identical or similar to those of the above-mentioned embodiments are given the same reference numerals, and detailed description is omitted. The memory circuit 107 shown in Fig. 20 has five relay control units RCNT arranged between a memory management unit MNG (input side) and a memory management unit MNG (output side), and a memory block MBLK arranged between two adjacent relay control units RCNT.

[0125] The memory block MBLK has two memory groups MEMG each including five relay control units RCNT and four memory blocks MBLK arranged between adjacent relay control units RCNT. In each memory group MEMG, the first-stage relay control unit RCNT is connected to the memory management unit MNG (input side), and the last-stage relay control unit RCNT is connected to the memory management unit MNG (output side). The number and arrangement of memory blocks MBLK mounted on the memory circuit 107 are not limited to the example shown in FIG. 20.

[0126] Each memory block MBLK has 16 memory units MU, 4 vertically and 4 horizontally. Note that the number and arrangement of memory units MU mounted in each memory block MBLK are not limited to the example shown in FIG.

[0127] Although not particularly limited, the size of each memory unit MU is 32k words x 64 bits (=2M bits), and four memory units MU aligned vertically input and output 256 bits of data. The size of a memory block MBLK including 16 memory units MU is 128k words x 256 bits (=32M bits). The size of a memory circuit 107 including eight memory blocks MBLK is 512k words x 512 bits (=256M bits).

[0128] A memory group column MGC is composed of two memory blocks MBLK arranged vertically. In this embodiment, a memory section MEM is composed of four memory units MU arranged vertically. For example, each memory group column MGC is identified by bits [24:23] of a 25-bit address AD. Each memory section MEM is identified by bits [22:21] of a 25-bit address AD.

[0129] The memory circuit 107 operates in response to an access request MREQ received from the system bus SBUS during a second normal operation mode in which all memory units MEM can be accessed as one large-capacity memory. During the second normal operation mode, the relay control unit RCNT on the input side of each memory block MBLK outputs an access request MREQ (a command CMD, an address AD, and write data WD) to the memory block MBLK or the relay control unit RCNT on the output side of the memory block MBLK. Each memory block MBLK outputs read data RD to the relay control unit RCNT on the output side.

[0130] 20, the command line CMD and the address line AD are not routed through multiple memory blocks MBLK. This makes it possible to suppress an increase in the wiring load, and to suppress an increase in the access time of the memory circuit 107 and an increase in power consumption.

[0131] On the other hand, in the first normal operation mode, the memory circuit 107 allows data to be read and written in parallel from outside the memory circuit 107 for each memory unit MEM or for each two memory units MEM.

[0132] Fig. 21 shows an example of the memory blocks MBLK in Fig. 20. For example, Fig. 21 shows four memory blocks MBLK in the central portion away from the memory management unit MNG (input side) and the memory management unit MNG (output side) in Fig. 20.

[0133] Each memory block MBLK has the same or similar configuration as the memory block MBLK in Fig. 12. That is, each memory block MBLK has four memory units MEM, four memory control units CNT1, four test circuits BIST, and four non-volatile memories NVM. As shown in Fig. 15 or 19, each memory block MLK may be configured with two groups GR, and in each group GR, a defect in one memory unit MEM may be relieved by the word line WL of the other memory unit MEM during the test mode.

[0134] Fig. 22 shows another example of the memory blocks MBLK in Fig. 20. Fig. 22 shows four memory blocks MBLK in the central portion away from the memory management unit MNG (input side) and the memory management unit MNG (output side) in Fig. 20.

[0135] Each memory block MBLK has the same configuration as the memory block MBLK in FIG. 21, except that the memory control unit CNT1, the test circuit BIST, and the nonvolatile memory NVM are provided in common to the two memory units MEM. In addition, in each memory block MBLK, the memory control unit CNT1, the test circuit BIST, and the nonvolatile memory NVM may be provided in common to the four memory units MEM. That is, the memory control unit CNT1, the test circuit BIST, and the nonvolatile memory NVM may be provided for each memory block MBLK. Also, as shown in FIG. 15 or FIG. 19, each memory block MLK may be configured with two groups GR, and in each group GR, a defect in one memory unit MEM may be relieved by the word line WL of the other memory unit MEM during the test mode.

[0136] As described above, the eighth embodiment can also provide the same effects as the first to seventh embodiments.

[0137] Fig. 23 shows an example of a system in which the memory circuit 100 in Fig. 1 is mounted. For example, the system 200 shown in Fig. 23 is a head-mounted device such as AR / VR glasses capable of processing moving images, a digital camera, a game machine, or the like. Note that the system 200 may be an image processing system mounted on a vehicle. In addition, the system in which the memory circuit 100 is mounted is not limited to the configuration shown in Fig. 23.

[0138] The system 200 includes a controller 210, an imaging device 220, a display device 230, and an external memory 240. The controller 210 includes a CPU 211, an image processing unit 212, a display processing unit 213, an encoder / decoder 214, an external memory control unit 215, and the memory circuit 100 shown in Fig. 1, which are interconnected via a system bus SBUS. For example, the controller 210 may be designed as a system LSI.

[0139] The controller 210 may have the memory circuit 101 of Fig. 9, the memory circuit 102 of Fig. 12, the memory circuit 103 of Fig. 13, the memory circuit 104 of Fig. 14, the memory circuit 105 of Fig. 15, the memory circuit 106 of Fig. 19, or the memory circuit 107 of Fig. 20. For example, the memory circuits 100, 101, 102, 103, 104, 105, 106, and 107 are SRAMs, but may be other randomly accessible memory circuits such as MRAMs or resistance change memories.

[0140] The CPU 211 controls the entire system 200. The image processing unit 212 processes image data acquired by the imaging device 220, converts the processed image data into frame image data that can be displayed on the display device 230, and stores the frame image data in the memory circuit 100. The display processing unit 213 reads the frame image data from the memory circuit 100 and displays the image on the display device 230. The encoder / decoder 214 encodes the image data before it is stored in the memory circuit 100, and decodes the compressed image data read from the memory circuit 100. The external memory control unit 215 controls access to an external memory 240 such as a DRAM (Dynamic Random Access Memory).

[0141] Although the present invention has been described above based on the embodiments, the present invention is not limited to the requirements shown in the above embodiments. These points can be changed without departing from the spirit of the present invention, and can be appropriately determined according to the application form. [Explanation of symbols]

[0142] 100, 101, 102, 103, 104, 105, 106, 107 Memory circuit 200 Systems 210 Controller 211 CPU 212 Image processing section 213 Display processing unit 214 Encoder / Decoder 215 External memory control unit 220 Imaging device 230 Display device 240 External Memory AD Address BIST test circuit BRWL Bubble Redundant Word Line CMD command CMP comparator CNT1, CNT2, CNT3 memory control section FAD Bad Address GR Group RD Read Data MBLK Memory Block MEM, MEM0, MEM1, MEM2, MEM3 memory section MEMG Memory Group MRG Memory Management Unit MREQ Access Request MU Memory Unit NVM Non-Volatile Memory RCNT Relay control section REDEN Redundancy enable REG Register RWL Redundant Word Line SBUS System bus SEL Selection section SEL0, SEL1 selector SWL Sub-word line WD Write Data WL Word Line

Claims

1. a plurality of memory units each including a plurality of first memory cells and a second memory cell that is accessed in place of a defective first memory cell; a plurality of first memory control units provided corresponding to the plurality of memory units, the first memory control units controlling access to the corresponding memory units based on a first access request to the corresponding memory units during a first mode; a second memory control unit provided in common to the plurality of memory units and controlling access to the plurality of memory units based on a second access request during a second mode; any one of the plurality of memory units has a plurality of third memory cells corresponding to the plurality of second memory cells of the plurality of memory units, each of the plurality of first memory control units converts the first address into a second address indicating the second memory cell when the first access request includes a first address indicating the first memory cell that cannot be accessed during the first mode, and outputs the second address to the corresponding memory unit; When the second address indicating each of the second memory cells of the memory units is included in the second access request during the second mode, the second memory control unit converts the second address into a third address indicating one of the third memory cells, and outputs the third address to the memory unit including the third memory cell. Memory circuit.

2. The first memory control units operate in parallel during the first mode and are capable of accessing the corresponding memory units in parallel.

2. The memory circuit of claim 1.

3. a plurality of relay control units arranged between the second memory control unit and the plurality of memory units which are connected in series and capable of sequentially transferring the second access request from the second memory control unit during the second mode; a memory block including one or more of the memory units is disposed between a pair of the relay control units adjacent to each other; Each of the plurality of relay control units outputs the second access request to the subsequent memory block when the second access request indicates an access to the memory block located immediately after the subsequent memory block, and outputs the second access request to a next relay control unit when the second access request indicates an access to the memory block located after the subsequent memory block.

3. The memory circuit according to claim 1.

4. a non-volatile memory that stores defect information of the memory unit; a plurality of test circuits provided corresponding to the plurality of memory units, the test circuits operating in parallel during the first mode and testing each of the plurality of memory units; each of the plurality of test circuits outputs a write request and a read request as the first access request to the corresponding first memory control unit, and when detecting an inaccessible first memory cell, stores a defective address indicating the inaccessible first memory cell in the nonvolatile memory; Each of the plurality of first memory control units converts the address included in the first access request to the second address and outputs the second address to the corresponding memory unit when an address included in the first access request matches the defective address stored in the non-volatile memory after testing the plurality of memory units by the plurality of test circuits.

3. The memory circuit according to claim 1.

5. Each of the plurality of test circuits stores the defective address in the nonvolatile memory, and then outputs the write request including the defective address stored in the nonvolatile memory and the read request as the first access request to the corresponding first memory control unit, and determines whether read data output from the memory unit based on the read request matches an expected value.

5. The memory circuit of claim 4.

6. a plurality of memory units each including a plurality of first memory cells; a plurality of first memory control units provided corresponding to the plurality of memory units, the first memory control units controlling access to the corresponding memory units based on a first access request to the corresponding memory units during a first mode; a second memory control unit provided in common to the plurality of memory units and controlling access to the plurality of memory units based on a second access request during a second mode; any one of the plurality of memory units has a plurality of second memory cells that are accessed in place of the inaccessible first memory cells of each of the plurality of memory units during the second mode; The plurality of memory units are grouped into pairs of the memory units, one of the pair of first memory control units corresponding to the pair of memory units, during the first mode, outputs the address indicating the first memory cell to the other of the first memory control units when the first access request includes an address indicating the inaccessible first memory cell, outputs the received address to the corresponding memory unit when the first access request includes an address indicating the normal first memory cell, and outputs the address indicating the normal first memory cell to the corresponding memory unit when the first access request includes an address indicating the normal first memory cell; When an address included in the second access request indicates an inaccessible first memory cell during the second mode, the second memory control unit converts the address included in the second access request into an address indicating any one of the plurality of second memory cells and outputs the address to any one of the plurality of memory units. Memory circuit.

7. each of the plurality of memory units is included in one of the plurality of groups; In each of the plurality of groups, the first memory control units corresponding to one of the memory units operate in parallel during the first mode and can access the one of the memory units in parallel. The memory circuit of claim 6.

8. a plurality of relay control units arranged between the second memory control unit and the plurality of memory units which are connected in series and capable of sequentially transferring the second access request from the second memory control unit during the second mode; a memory block including one or more of the memory units is disposed between a pair of the relay control units adjacent to each other; Each of the plurality of relay control units outputs the second access request to the subsequent memory block when the second access request indicates an access to the memory block located immediately after the subsequent memory block, and outputs the second access request to a next relay control unit when the second access request indicates an access to the memory block located after the subsequent memory block.

8. The memory circuit according to claim 6 or 7.

9. a non-volatile memory that stores defect information of the plurality of memory units; a plurality of test circuits provided corresponding to the plurality of memory units, the test circuits operating in parallel during the first mode and testing each of the plurality of memory units; Among the plurality of test circuits, the test circuit corresponding to the one of the first memory control units and the test circuit corresponding to the other of the first memory control units operate alternately; The test circuit in operation outputs a write request and a read request as the first access request to the corresponding first memory control unit, and when an inaccessible first memory cell is detected, stores a defective address indicating the inaccessible first memory cell in the nonvolatile memory; After the defective address is stored in the nonvolatile memory, if an address included in the read request matches the defective address stored in the nonvolatile memory, the one of the first memory control units outputs the address included in the read request to the other of the first memory control units.

8. The memory circuit according to claim 6 or 7.

10. The test circuit in operation stores the defective address in the nonvolatile memory, and then outputs the write request and the read request including the defective address stored in the nonvolatile memory as the first access request to the corresponding first memory control unit, and determines whether read data output from the memory unit not corresponding to the test circuit in operation matches an expected value.

10. The memory circuit of claim 9.

Citation Information

Patent Citations

  • Semiconductor memory device

    JP2005129164A

  • Semiconductor memory device and testing method therefor

    JP2010140528A

  • Method and system for memory test and repair

    US9003252B1

  • Nonvolatile memory

    JP1989128300A

  • Semiconductor memory

    JP1993290598A