Multi-bank semiconductor memory device suitable for integration with logic

[0002] 1. Field of the Invention

[0003] The present invention relates to a semiconductor integrated circuit device, and more particularly relates to a semiconductor memory device fit to mixture with logical processing circuit units such as MPU (Microprocessor), CPU (Central Processing Unit), ASIC (Application Specific Integrated Circuit) and the like, as well as a logic contained semiconductor memory device or a memory contained logical processing circuit unit in which a logical processing circuit unit and a semiconductor memory device are mixed.

[0004] 2. Description of the Background Art

[0005] FIG. 39 schematically shows an entire arrangement of a processor containing DRAM (Dynamic Random Access Memory) in which a conventional DRAM and a processor are formed on the same chip. Referring to FIG. 39, the conventional processor containing DRAM is integrated on a semiconductor chip 900. The microprocessor containing DRAM includes DRAM arrays 902a, 902b, 902cand 902drespectively arranged in four regions #A, #B, #C and #D on semiconductor chip 900. Each of DRAM arrays 902a-902dincludes dynamic memory cells arranged in rows and columns.

[0006] Corresponding to respective DRAM arrays 902a-902d, row decoders 903a, 903b, 903cand 903deach for selecting a row in a corresponding DRAM array, column decoders 904a, 904b, 904cand 904deach for selecting a column in a corresponding memory array, and preamplifier.write driver blocks 905a, 905b, 905cand 905deach for writing/reading of data into/from a corresponding DRAM array are provided. Preamplifier.write driver blocks 905a-905dare respectively coupled with internal read/write data buses 907a, 907b, 907cand 907deach having a width of 32 bits to 256 bits.

[0007] In a region #E between regions #A and #B and regions #C and #D, a processor (CPU) 920 which supplies and receives necessary data to and from DRAM arrays 902a-902dand executes various processings as well as a DRAM control circuit 910 which controls an access to DRAM arrays 902a-902daccording to a command from processor 920 are provided.

[0008] DRAM control circuit 910 generates and sends via an internal bus 912aa control signal for controlling activation/inactivation of row decoders 903aand 903b, column decoders 904aand 904band preamplifier.write driver blocks 905aand 905b, and an address signal for selecting a row and a column. Control circuit 910 also supplies a control signal and an address signal via an internal bus 912bto row decoders 903cand 903d, column decoders 904cand 904d, and preamplifer.write driver blocks 905cand 905d. Internal bus 912bis arranged to extend from DRAM control circuit 910 to regions #C and #D across processor 920.

[0009] Processor 920 is coupled with internal read/write data buses 907a, 907b, 907cand 907dand inputs/outputs necessary data. Internal read/write data buses 907aand 907bis coupled with processor 920 across DRAM control circuit 910.

[0010] In this processor containing DRAM, DRAM arrays 902a-902dare simultaneously driven into an active state, and data are transferred between DRAM arrays 902a-902dand processor 920 via internal read/write data buses 907a-907d. Since internal read/write data buses 907a-907deach has a sufficiently large width of, for example, 32 bits to 256 bits, data of 128 bits to 1024 bits can be transferred in one transfer cycle and a band width for data transfer can be increased. If the DRAM and the processor are discretely provided, the operating speed of the DRAM cannot follow the operating speed of the processor and high speed data transfer becomes impossible accordingly, and the operating speed of the DRAM becomes a bottleneck against processing performance of the processor. The processor containing DRAM described above aims to solve such a problem.

[0011] Particularly by integrating processor 920 and DRAM arrays 902a-902don the same semiconductor chip 900, the bit width of internal read/write data buses 907a-907dcan be increased sufficiently without limitation by external pin terminals. Further, by integrating DRAM arrays 902a-902dand processor 920 on semiconductor chip 900, interconnection line capacitance and resistance of internal read/write data buses 907a-907dcan be reduced relative to those of interconnection lines on a circuit board. As a result, data can be transferred at high speed.

[0012] The processor containing DRAM shown in FIG. 39 has a single bank configuration in which DRAM arrays 902a-902dare all simultaneously activated/inactivated. Considering localization of data to which the processor makes access, a multi-bank configuration containing a plurality of banks is preferable for a high speed operation. In this case, during an access to one bank, another bank is driven from an inactive state to an active state in order to prepare for an access by the processor. Following completion of the access to the one bank, an access is made to this another bank. In the single bank configuration, RAS precharge time is necessary for activating DRAM arrays 902a-902dafter precharging of the DRAM array in order to obtain necessary data again. In the multi-bank configuration, such RAS precharge time is not required and a high speed access is implemented. However, a processor containing the multi-bank DRAM has not been proposed.

[0013] The bus width of internal read/write data buses 907a-907dshould be further increased in order to modify DRAM arrays 902a-902dof FIG. 39 into the multi-bank configuration. In a two-bank configuration, for example, one bank is constituted of two memory arrays. In this case, data can be only transferred simultaneously between processor 920 and two DRAM arrays. In order to implement a data transfer band width which is equal to that of the single bank configuration, the width of internal read/write data buses should be doubled. Since the doubled width of the buses increases the area occupied by interconnection lines, an efficient layout of internal circuits is required for avoiding increase of the chip area. However, such an efficient layout for preventing increase of the chip area in the multi-bank DRAM has not been proposed.

[0014] In addition, in the processor containing DRAM shown in FIG. 39, the length of internal bus 912aand that of internal bus 912bdiffer from each other, resulting in different signal propagation delays. The difference of the length of internal buses 912aand 912bis canceled by the difference of the length of internal read/write data buses 907aand 907band that of buses 907cand 907dso as to obtain the same access time. However, difference of delay time in internal read/write data buses 907a-907dand difference of delay time in internal buses 912aand 912bcannot be made precisely equal to each other due to difference of parasitic capacitance of the interconnection lines or the like. Therefore, some margin is required for data access. Further, since timings of starting/stopping of internal access of the banks and timings of activating/inactivating the arrays are different, collision of data could occur when the banks are accessed in the interleaved manner.

[0015] In this case, if DRAM arrays 902aand 902bconstitute one bank and DRAM arrays 902cand 902dconstitute the other bank in FIG. 39, access times of those banks are different from each other and the access time is determined according to the worst case, so that a high speed access cannot be implemented. If DRAM arrays 902aand 902cconstitute one bank, data access time of DRAM array 902ais different from that of DRAM array 902c. In this case, the access time is also determined according to the worst case, so that a high speed access cannot be implemented.

[0016] In the case of such a processor containing DRAM or a circuit unit in which logic and DRAM are mixed, the processor or unit is often designed on a cell base (because automatic design is used). The processor or unit is hierarchically designed utilizing the DRAM as a macro (functional block). However, a multi-bank DRAM macro having an efficient layout which is fit to mixture with a processor or logic, enables a high speed access and does not increase the occupied area has not been prepared yet.

[0017] One object of the present invention is to provide a semiconductor integrated circuit device including a multi-bank memory having a layout fit to mixture with a logical processing circuit unit.

[0018] Another object of the present invention is to provide a semiconductor integrated circuit device having a multi-bank memory which can implement a high speed access without increasing a chip area.

[0019] Still another object of the present invention is to provide a multi-bank DRAM fit to mixture with a logical processing circuit unit.

[0020] A semiconductor integrated circuit device according to the invention includes a plurality of banks arranged in a rectangular region and to which data access is made alternatively, a bank controller provided in a prescribed region among regions where the plurality of banks are arranged, for controlling at least the data access to the plurality of banks, and an internal access data transfer bus arranged in a region different from a region where the bank controller is formed in the rectangular region, and coupled with each bank for supplying and receiving data.

[0021] By placing the bank controller and the internal access data transfer bus in different regions in the rectangular region, there does not exit a region where the bank controller and the internal access data transfer bus are overlapped viewed in a planar layout, so that even if the bus width increases the bank controller can be arranged without being affected by the increased bus width. Since the bank controller can be arranged without consideration of a crossing to the internal access data transfer bus and the memory, a layout of high density and reduction of an occupied area of a chip are achieved. By placing the bank controller among the plurality of banks, distances between respective banks and the bank controller can be made equal to each other. The difference of access timing due to delay in interconnection lines can be eliminated, resulting in high speed access.

[0022] The foregoing and other objects, features, aspects and advantages of the present invention will become more apparent from the following detailed description of the present invention when taken in conjunction with the accompanying drawings.

[0023] FIG. 1 schematically shows an entire configuration of a semiconductor integrated circuit device according to a first embodiment of the present invention.

[0024] FIGS. 2A and 2B schematically show arrangements of interconnection lines at a crossing of an address signal and control signal bus and a read/write data bus shown in FIG. 1.

[0025] FIG. 3 shows an arrangement of a control bus of the semiconductor integrated circuit device according to the first embodiment of the invention.

[0026] FIG. 4 schematically shows an entire configuration of a semiconductor integrated circuit device according to a second embodiment of the invention.

[0027] FIG. 5 schematically shows an entire configuration of a semiconductor integrated circuit device according to a third embodiment of the invention.

[0028] FIG. 6 schematically shows a configuration of a testing circuit shown in FIG. 5.

[0029] FIG. 7 shows a structure of the testing circuit shown in FIG. 6 in more detail.

[0030] FIG. 8 schematically shows a configuration of a selector shown in FIG. 7.

[0031] FIG. 9 schematically shows a configuration of a compression circuit shown in FIG. 7.

[0032] FIG. 10 schematically shows an entire configuration of a semiconductor integrated circuit device according to a fourth embodiment of the invention.

[0033] FIG. 11 schematically shows an entire configuration of a semiconductor integrated circuit device according to a fifth embodiment of the invention.

[0034] FIG. 12 schematically shows a configuration of a modification of the semiconductor integrated circuit device according to the fifth embodiment of the invention.

[0035] FIG. 13 schematically shows an entire configuration of a semiconductor integrated circuit device according to a sixth embodiment of the invention.

[0036] FIG. 14 schematically shows an entire configuration of a first modification of the semiconductor integrated circuit device according to the sixth embodiment.

[0037] FIG. 15 schematically shows a configuration of a second modification of the semiconductor integrated circuit device according to the sixth embodiment.

[0038] FIG. 16 schematically shows an entire configuration of a semiconductor integrated circuit device according to a seventh embodiment.

[0039] FIG. 17 schematically shows a configuration of a preamplifier.write driver block shown in FIG. 16.

[0040] FIG. 18 schematically shows a configuration of a first modification of the preamplifier.write driver block shown in FIG. 16.

[0041] FIG. 19 schematically shows a configuration of a second modification of the preamplifier.write driver block shown in FIG. 16.

[0042] FIG. 20 schematically shows a configuration of a third modification of the preamplifier.write driver block shown in FIG. 16.

[0043] FIG. 21 schematically shows an entire configuration of a semiconductor integrated circuit device according to an eighth embodiment of the invention.

[0044] FIG. 22 specifically shows a configuration of one subbank of the semiconductor integrated circuit device shown in FIG. 21.

[0045] FIG. 23 shows propagation delay time of read memory cell data signal in the subbank structure shown in FIG. 22.

[0046] FIG. 24 is a timing chart representing data reading operation in the subbank structure shown in FIG. 22.

[0047] FIG. 25A schematically shows an entire configuration of a semiconductor integrated circuit device according to a ninth embodiment, and FIG. 25B schematically shows a configuration of a portion related to one subbank of FIG. 25A.

[0048] FIG. 26 schematically shows a configuration of the power supply of the semiconductor integrated circuit device of FIG. 25A.

[0049] FIG. 27 schematically shows an arrangement of interconnection lines of a memory array section in the subbank of FIG. 25A.

[0050] FIG. 28 schematically shows an arrangement of power supply lines in the ninth embodiment.

[0051] FIG. 29 schematically shows an arrangement of interconnection lines in the semiconductor integrated circuit device according to the ninth embodiment.

[0052] FIG. 30 schematically shows a modification of the semiconductor integrated circuit device of the ninth embodiment.

[0053] FIG. 31 schematically shows an entire configuration of a semiconductor integrated circuit device according to a tenth embodiment of the invention.

[0054] FIG. 32 schematically shows an arrangement of interconnection lines of a memory array unit of the semiconductor integrated circuit device shown in FIG. 31.

[0055] FIG. 33 schematically shows an entire configuration of a semiconductor integrated circuit device according to an eleventh embodiment of the invention.

[0056] FIG. 34 schematically shows an arrangement of banks of the semiconductor integrated circuit device shown in FIG. 33.

[0057] FIG. 35 schematically shows a configuration of the testing circuit section of the eleventh embodiment of the invention.

[0058] FIG. 36 schematically shows a first configuration for switching an output from the testing circuit and an output from the processor.

[0059] FIG. 37 schematically shows a configuration of a switching section of outputs from the testing circuit and from the processor of the semiconductor integrated circuit device according to the eleventh embodiment.

[0060] FIG. 38 schematically shows an entire configuration of a modification of the semiconductor integrated circuit device according to the eleventh embodiment.

[0061] FIG. 39 schematically shows an entire configuration of a semiconductor integrated circuit device according to a twelfth embodiment of the invention.

[0062] FIG. 40 schematically shows configurations of a memory array and a row-related array circuit illustrated in FIG. 39.

[0063] FIG. 41 shows one example of a configuration of a row decoder.driver shown in FIG. 40.

[0064] FIG. 42 shows one example of a configuration of a sense control circuit illustrated in FIG. 40.

[0065] FIG. 43 schematically shows components arranged in an interconnection-dedicated region illustrated in FIG. 39.

[0066] FIG. 44A schematically shows an entire configuration of a semiconductor memory device according to the twelfth embodiment of the invention, and FIG. 44B schematically shows a configuration of the semiconductor integrated circuit device illustrated in FIG. 44A where storage capacity thereof is decreased.

[0067] FIG. 45 shows an application of the twelfth embodiment of the invention.

[0068] FIG. 46 shows another application of the twelfth embodiment of the invention.

[0069] FIG. 47 schematically illustrates a configuration of a modification of the twelfth embodiment of the invention.

[0070] FIG. 48 schematically shows configurations of a memory array and a row-related array circuit illustrated in FIG. 47.

[0071] FIG. 49 schematically shows configurations of respective blocks of a column decoder, a preamplifier, and a write driver corresponding to one column block shown in FIG. 48.

[0072] FIGS. 50A and 50B illustrates a manner in which an array size is changed according to the modification of the twelfth embodiment of the invention.

[0073] FIG. 51 shows an application of the modification of the twelfth embodiment of the invention.

[0074] FIG. 52 schematically shows an entire configuration of a conventional semiconductor integrated circuit device.

[0075] FIG. 1 schematically shows an entire configuration of a semiconductor integrated circuit device according to the first embodiment of the invention. In FIG. 1, a semiconductor integrated circuit device 1 includes a processor (CPU) 2 and a DRAM macro (functional block) 3 which functions as a main memory for processor 2 that are formed to be integrated on a semiconductor chip. Although semiconductor integrated circuit device 1 is shown including processor 2, the processor may be a logical circuit such as a DSP (Digital Signal Processor) which executes a dedicated processing. In order to represent characteristic features of the present invention clearly, semiconductor integrated circuit device 1 is shown including only processor 2 and DRAM macro 3, the semiconductor circuit device may include another macro or cell. However, the semiconductor integrated circuit device is hereinafter described as a processor (or logic) containing DRAM or a DRAM containing processor.

[0076] DRAM macro 3 has a rectangular shape and includes subbanks 4a, 4b, 4cand 4darranged respectively in four regions. Subbanks 4aand 4cconstitute a bank A and sub banks 4band 4dconstitute a bank B. These subbanks 4a-4dhave the same configuration. Specifically, subbank 4aincludes a memory array 4aahaving a plurality of dynamic memory cells arranged in rows and columns, a row decoder 4abfor driving a row in memory array 4aainto a selected state, a column decoder 4acfor selecting a column in memory array 4aa, and a preamplifier.write driver block 4adwhich reads/writes data from/into a memory cell selected by row decoder 4aband column decoder 4ac.

[0077] Similarly, subbank 4bincludes a memory array 4ba, a row decoder 4bb, a column decoder 4bcand a preamplifier.write driver block 4bd. Subbank 4cincludes a memory array 4ca, a row decoder 4cb, a column decoder 4ccand a preamplifier.write driver block 4cd. Subbank 4dincludes a memory array 4da, a row decoder 4db, a column decoder 4dcand a preamplifier.write driver block 4dd.

[0078] The configuration of each of subbanks 4a-4dis identical to that of a normal DRAM. Preamplifier.write driver blocks 4adand 4bdof subbanks 4aand 4bare arranged facing to each other with respect to a central region in the column direction of DRAM macro 3. Similarly, preamplifier.write driver blocks 4cdand 4ddof subbanks 4cand 4dare arranged facing to each other with respect to the central region in the column direction of DRAM macro 3. Subbanks 4aand 4chave layouts that are in mirror symmetry relative to the central region in the column direction of DRAM macro 3, and subbanks 4band 4dhave layouts that are in mirror symmetry relative to the central region in the column direction of DRAM macro 3. Further, subbanks 4aand 4bhave layouts that are in mirror symmetry relative to a central region in the row direction of DRAM macro 3, and subbanks 4cand 4dhave layouts that are in mirror symmetry relative to the central region in the row direction of DRAM macro 3. After completion of the layout of one subbank, the subbank with the completed layout is folded back, and layouts of remaining subbanks are implemented, resulting in easier layout.

[0079] The central region in the column direction of DRAM macro 3 is divided into three regions 10, 11aand 11b. In region 10, a bank A control circuit 7awhich controls selection/non-selection (including data access) of subbank 4avia an internal bus 6aand controls selection/non-selection of subbank 4cvia an internal bus 6c, as well as a bank B control circuit 7bwhich controls selection/non-selection of subbank 4bvia an internal bus 6band controls selection/non-selection of subbank 4dvia an internal bus 6dare arranged. Bank A control circuit 7ais arranged in a region substantially intermediate between subbanks 4aand 4c, and internal buses 6aand 6chave a substantially identical length. Similarly, bank B control circuit 7bis arranged in a substantially intermediate between subbanks 4band 4d, and internal buses 6band 6dhave the same length. Bank A control circuit 7aand bank B control circuit 7bare arranged in alignment with each other in the row direction in region 10. Subbanks 4aand 4bare arranged in alignment with each other along the row direction, and subbanks 4cand 4dare arranged in alignment with each other along the row direction. Accordingly, the length of internal buses 6a-6dare substantially equal, so that the timing of selection/non-selection of subbanks 4a-4dupon selection of bank A can be made the same as that of selection of bank B. It is not necessary to consider a margin for the difference of the timing, and high speed access is possible.

[0080] In region 11a, an internal read/write data bus (internal access data transfer bus) 5ais arranged extending in the row direction and is commonly coupled with preamplifier.write driver blocks 4adand 4bdof subbanks 4aand 4b. In region 11b, an internal read/write data bus (internal access data transfer bus) 5bwhich extends in the row direction and is coupled with preamplifier.write driver blocks 4cdand 4ddof subbanks 4cand 4dis provided.

[0081] The internal read/write data buses 5aand 5bare placed in parallel with bank A control circuit 7aand bank B control circuit 7b. Therefore, there is not a region where bank A control circuit 7aand bank B control circuit 7bcross internal read/write data buses 5aand 5b, and the bank A control circuit 7aand bank B control circuit 7bdo not need to include an empty region for arranging internal read/write data buses 5aand 5b. An arrangement of high density is thus achieved and increase of an area of DRAM macro 3 or increase of chip area can be avoided.

[0082] Internal read/write data buses 5aand 5btransfer data of both bank A and bank B. In order to implement a bank width similar to that of the single bank configuration, the bit width of each of internal read/write data buses 5aand 5bis determined to be 64 bits to 512 bits, for example, which is twice the bit width in the single bank configuration. Although the number of bus lines of internal read/write data buses 5aand 5bincreases, there is no limitation on layout compared with the limitation on layout generated when the crossing of bank A control circuit 7aand bank B control circuit 7band internal read/write data buses 5aand 5bshould be considered. Therefore, more efficient layout of bank A control circuit 7aand bank B control circuit 7bis possible and increase of the area due to increase of the bus lines of internal read/write data buses can be sufficiently compensated for.

[0083] Internal read/write data buses 5aand 5bare electrically coupled with processor 2. In this case, processor 2 is arranged facing to and external to DRAM macro 3 and supplies and receives data to and from internal read/write data buses 5aand 5blocated in regions 11aand 11b. Processor 2 has thus no limitation on layout thereof due to a crossing to internal read/write data buses. Only provision of a data input/output port in the vicinity of regions 11aand 11bof DRAM macro 3 is required. Accordingly, an efficient layout of processor 2 is possible without considering a crossing to arrangement of internal read/write data buses, and a processor of high density can be implemented. Accordingly, semiconductor integrated circuit device 1 which is highly integrated with high density is implemented.

[0084] Internal buses 6a-6dthat transmit a control signal and an address signal are arranged extending in the column direction, while internal read/write data buses 5aand 5bare arranged extending in the row direction. Therefore, there are crossings of internal buses 6a-6dand internal read/write data buses 5aand 5b. In FIG. 1, a crossing of internal bus 6band internal read/write data bus 5ais representatively shown by a block CX indicated with a broken line.

[0085] FIG. 2A schematically shows a configuration of an arrangement of bus lines at crossing CX shown in FIG. 1. FIG. 2A shows a cross sectional structure of one bus line included in internal bus 6b. Internal read/write data bus 5aincludes bus lines 5aa, 5ab, . . . , 5anconstituted of, for example, a second level aluminum interconnection line (2A1). The bus line of internal bus 6bincludes an interconnection line 6bcformed of, for example, a first level aluminum interconnection line (1A1) formed at a layer under internal read/write data bus lines 5aa-5an, as well as interconnection lines 6baand 6beformed in the same layer as internal read/write data bus lines 5aa-5an. Interconnection lines 6baand 6beare connected electrically to interconnection line 6belocated in the lower layer via contacts 6bband 6bdrespectively. Accordingly, even if internal read/write data bus 5aand internal bus 6bare crossed at crossing CX, the internal read/write data bus and the internal bus 6bcan be arranged by using different interconnection layers.

[0086] FIG. 2B shows a structure of crossing CX. In FIG. 2B, an interconnection line 6bhformed of, for example, a third level aluminum interconnection layer (3A1) is provided above the layer of internal read/write data bus lines 5aa-5an, and interconnection lines 6bfand 6bjare formed in the same layer as internal read/write data bus lines 5aa-5an. Interconnection lines 6bfand 6bjare connected to interconnection line 6bhrespectively via contacts 6bgand 6bi. Interconnection lines 6bf, 6bg, 6bh, 6biand 6bjconstitute one internal bus line.

[0087] If the third level aluminum interconnection line is employed, the resistance value thereof is lower than that of the second level aluminum interconnection line. If the aluminum interconnection line is used, an aluminum interconnection line located in an upper layer is subjected to less heat cycles after formation of the upper layer interconnection line for formation of another interconnection. Further, the number of insulating layers and interconnection layers formed in a layer thereabove is small, and stress applied from these layers is small. Therefore, the purity of the upper layer aluminum interconnection line can be made higher than that of the lower layer aluminum interconnection line (Resistance against stress and heat becomes lower as the purity of aluminum increases.). In this case, compared with the case in which the internal bus line is formed using only the second level aluminum interconnection layer, signals (control signal and address signal) can be transmit at a high speed since interconnection line 6bhof low resistance is used.

[0088] The structure of crossing CX is also applied to a crossing of internal read/write data bus 5aand internal bus 6aas well as crossings of internal read/write data bus 5band internal buses 6cand 6d.

[0089] In the structures of crossings shown in FIGS. 2A and 2B, interconnection lines of different layers are used for internal buses 6a-6d. However, internal buses 6a-6dmay be formed of the same interconnection layer, and bus lines of internal read/write data buses 5aand 5bmay be connected to another interconnection layer at the crossing.

[0090] FIG. 3 shows an arrangement of control buses that transmit a control signal and an address signal to processor 2, bank A control circuit 7aand bank B control circuit 7b. In FIG. 3, subbanks 4a-4d, and read/write data buses 5aand 5bincluded in regions 11aand 11bare schematically shown for simplification of the figure. Referring to FIG. 3, a control bus 12awhich transmits a control signal and an address signal from processor 2 is placed in parallel with internal read/write data bus 5ain region 11a, and electrically coupled with bank A control circuit 7aand bank B control circuit 7b. In region 11b, a control bus bwhich transmits a control signal and an address signal from processor 2 is placed in parallel with internal read/write data bus 5b, and electrically coupled with bank A control circuit 7aand bank B control circuit 7b.

[0091] Control buses 12aand 12btransfer a signal for designating a bank, a signal for designating an operation mode and the like.

[0092] In the multi-bank configuration formed of banks A and B including subbanks 4a-4d, there is no crossing of control buses 12aand 12band bank A control circuit 7aand bank B control circuit 7b. Control buses 12aand 12bdo not cross processor 2. Since control buses 12aand 12bare located in regions 11aand 11bwhere internal read/write data buses 5aand 5bare arranged, the interconnection layout of control buses 12aand 12bdoes not adversely affect bank A control circuit 7aand bank B control circuit 7bas well as processor 2. As a result, a semiconductor integrated circuit device of high density can be implemented to reduce the chip area, as is implemented by arrangement of internal read/write data buses 5aand 5b.

[0093] In the arrangement shown in FIG. 3, regions 11aand 11brespectively have control buses 12aand 12barranged therein. However, the control buses may be arranged in only one of regions 11aand 11b.

[0094] According to the first embodiment of the invention, a region where a processor is arranged is separated from a region where a DRAM array and a control circuit are arranged, and different banks are allocated to memory arrays located on the same side relative to the DRAM control circuit (bank control circuit). Therefore, an internal read/write data bus interconnection line arranged commonly for the banks is not required to pass through a processor section and a DRAM control circuit section (bank control circuit section). There is no limitation of layout due to crossing of interconnection lines for the processor and control circuit sections, resulting in an efficient layout and prevention of increase of an occupied area of a circuit. The advantage above is also obtained as to a control bus which transfers a control signal and an address signal to the bank control circuit.

[0095] Since the bank control circuit is arranged in central region 10 of DRAM macro 3 and banks are allocated such that subbanks facing to each other relative to the central region belong to the same bank, the lengths of interconnection lines from the bank control circuit to respective subbanks are made identical to each other to accomplish the same selection/non-selection timing in each subbank, resulting in high speed access.

[0096] Since internal read/write data buses are arranged on both sides of the bank control circuit, the same distances between subbanks belonging to the same bank and the internal read/write data bus, as well as the same delay time in reading/writing of data are accomplished. Accordingly, there is no need to consider a skew resulting from delay in interconnection lines for the timing at which data is defined, resulting in high speed access.

[0097] FIG. 4 schematically shows an entire structure of a semiconductor integrated circuit device according to the second embodiment of the invention. Referring to FIG. 4, semiconductor integrated circuit device 1 includes DRAM macro 3, and processors (CPUs) 2aand 2barranged facing to each other on both sides of DRAM macro 3 in the row direction.

[0098] DRAM macro 3 includes, as in the first embodiment, subbanks 4aand 4cthat are arranged facing to each other with respect to region 10 and constitute bank A, subbanks 4band 4dthat are arranged facing to each other with respect to region 10 and constitute bank B, bank A control circuit 7awhich is arranged between subbanks 4aand 4cin region 10 and controls selection/non-selection (including activation/inactivation and data access) of subbanks 4aand 4cvia internal buses 6aand 6c, and bank B control circuit 7bwhich is arranged between subbanks 4band 4din region 10 and controls selection/non-selection of subbanks 4band 4dvia internal buses 6band 6d.

[0099] In region 11abetween subbanks 4aand 4barranged in alignment with each other in the row direction and region 10, internal read/write data bus 5aarranged extending in the row direction and coupled with subbanks 4aand 4b, as well as control bus 12aarranged extending in the row direction and coupled with bank A control circuit 7aand bank B control circuit 7bare provided. In region 11bbetween subbanks 4cand 4darranged in alignment with each other in the row direction and region 10, internal read/write data bus 5barranged extending in the row direction and coupled with subbanks 4cand 4d, as well as control bus 12barranged extending in the row direction and coupled with bank A control circuit 7aand bank B control circuit 7bare provided. Only one of control buses 12aand 12bmay be provided.

[0100] Internal read/write data buses 5aand 5bas well as control buses 12aand 12bare coupled with processors 2aand 2b.

[0101] Since internal read/write data buses 5aand 5bas well as control buses 12aand 12bare arranged in regions 11aand 11b, in addition to the advantage obtained by the first embodiment, an advantage that processors 2aand 2bcan be easily placed on both sides of DRAM macro 3 without affecting adversely the internal layout of DRAM macro 3 can be obtained. Accordingly, semiconductor integrated circuit device 1 implements a multi-processor system employing DRAM macro 3 as a main memory. A semiconductor integrated circuit device having a superior function and an excellent performance can thus be easily implemented.

[0102] Regarding processors 2aand 2b, one processor may be a main processor and the other processor may be a coprocessor. One of the processors 2aand 2bmay be a processor and the other may be a logical circuit such as a DSP (Digital Signal Processor) which executes a dedicated logical processing. Particularly in the field of ASIC, by employing DRAM macro 3 as shown in FIG. 4 as a memory section, an optimal arrangement of an external logical circuit section is possible without being affected by the layout of DRAM macro 3. As a result, an ASIC where the chip area is efficiently utilized can be easily implemented.

[0103] According to the second embodiment of the invention, as described above, logical circuit units such as processors are arranged on both sides of DRAM macro 3 and are coupled with internal read/write data buses as well as control data buses, a semiconductor integrated circuit device having high function and high performance can be easily implemented. Especially, if DRAM macro 3 is registered as a library, a semiconductor integrated circuit device which efficiently utilizes the area can be easily implemented by using the DRAM macro 3 as the library when the semiconductor integrated circuit device is designed on a cell base.

[0104] FIG. 5 schematically shows an entire structure of a semiconductor integrated circuit device according to the third embodiment of the invention. A difference between the semiconductor integrated circuit device of FIG. 5 and that of the second embodiment shown in FIG. 4 is as described below. Specifically, on both sides of DRAM macro in the row direction, processor 2 and a testing circuit 15 for testing at least the function of DRAM macro 3 are arranged opposing to each other. Other configuration is similar to that of the semiconductor integrated circuit device shown in FIG. 4, corresponding portions have the same reference characters and detailed description thereof is omitted.

[0105] In the configuration of the semiconductor integrated circuit device shown in FIG. 5, internal read/write data buses 5aand 5bas well as control buses 12aand 12bare coupled with processor 2 and testing circuit 15. A function test for DRAM macro 3 which is not directly connected to an external pin terminal of semiconductor integrated circuit device 1 can be easily carried out by providing testing circuit 15 for conducting the function test for DRAM macro 3 in semiconductor integrated circuit device 1.

[0106] If testing circuit 15 is not provided, the function test for DRAM macro 3 has to be carried out via processor 2. In this case, a program for conducting the function test for DRAM macro 3 has to be stored in processor 2, or an instruction for conducting the function test for DRAM macro 3 has to be externally issued to processor 2 to cause processor 2 to output a control signal for conducting the function test, and a result of the test has to be read via processor 2. In the function test for DRAM macro 3 thus carried out via processor 2, a test for the operation timing margin of DRAM macro 3 cannot be correctly carried out (since delay time of processor 2 is added). Further, the capacity of the program limits the number of test patterns generated, the test cannot be carried out satisfactorily, and reliability cannot be fully assured.

[0107] Since testing circuit 15 is provided and DRAM macro 3 can be externally tested via a dedicated testing unit directly in this configuration, a function test necessary for DRAM macro 3 can be easily and correctly carried out.

[0108] Since the testing circuit 15 is provided, a program for testing DRAM macro 3 only is not required to be changed and a test for DRAM macro 3 incorporated in semiconductor integrated circuit device 1 can be carried out (since a direct access can be externally made to DRAM macro 3). Suppose that testing circuit 15 is not provided, a result of a test is stored in processor 2 or a result of a test is determined by processor 2, and the result of the test is output from processor 2. In this case, a logical circuit for conducting a test for DRAM macro 3 is inserted into a path in which an operational processing by processor 2 itself is carried out, so that a signal propagation delay of processor 2 is generated. Provision of testing circuit 15 avoids provision of extra circuits in processor 2, resulting in high speed operation of processor 2.

[0109] FIG. 6 schematically shows a configuration of testing circuit 15 in FIG. 5. The arrangement of pads is emphasized in FIG. 6. Referring to FIG. 6, pads Pa-Pm, Pn-Pq, Pr, Ps-Pu, and pads PDa-PDi and PDj-PDp are arranged surrounding processor 2 along the periphery of semiconductor integrated circuit device 1. Pads Pa-Pm and PDa-PDi are arranged in alignment with each other at one side (upper side) of semiconductor integrated circuit device 1, pads Pr-Pu and PDj-PDp are arranged in alignment with each other at an opposing side of semiconductor integrated circuit device 1, and pads Pa and Pn-Pr are arranged in alignment with each other at the left side of semiconductor integrated circuit device 1. A pad PTa for generating test activation signal TE is placed adjacent to pad PDi.

[0110] In FIG. 6, semiconductor integrated circuit device 1 is shown having pads Pa-Pu and PDa-PDp and PTa are arranged in regions of three sides of processor 2 for simplification of the figure. However, these pads are located along the four sides of semiconductor integrated circuit device 1. These pads Pa-Pu, PDa-PDp and PTa are coupled with external pin terminals (not shown). Pads Pa-Pu and PDa-PDp are coupled with processor 2, and supply and receive signals (including data) to and from processor 2.

[0111] Testing circuit 15 includes an actual testing circuit 15afor conducting a test actually, a select/connect circuit 15bresponsive to activation of test activation signal TE supplied from pad PTa to electrically connect pads PDa-PDi to actual testing circuit 15a, and a select/connect circuit 15cresponsive to activation of test activation signal TE from pad PTa to electrically connect pads PDj-PDp to actual testing circuit 15a. Select/connect circuits 15band 15celectrically connect actual testing circuit 15ato external pin terminals via pads PDa-PDi and PDj-PDp. Testing of DRAM macro 3 is possible via actual testing circuit 15aeven after packaging of semiconductor integrated circuit device 1. Actual testing circuit 15ais coupled with DRAM macro 3 via internal read/write data bus 5 (5aand 5b) and control bus 12 (12aand 12b). Via actual testing circuit 15a, writing/reading of test data is carried out for subbanks contained in DRAM macro 3.

[0112] FIG. 7 specifically shows a configuration of testing circuit 15 of FIG. 6. In FIG. 7, a configuration of a portion of the testing circuit for internal read/write data bus 5aand control bus 12aprovided for subbanks 4aand 4bincluded in DRAM macro 3 is shown. A configuration similar to that shown in FIG. 7 is provided for internal read/write data bus 5band control bus 12bprovided for subbanks 1cand 1dshown in FIG. 5. If only one of control buses 12aand 12bis provided, the actual testing circuit is provided for the one control bus.

[0113] Referring to FIG. 7, select/connect circuit 15bincludes transfer gates TX provided for respective pads PDa-PDi and rendered conductive upon activation of test activation signal TE supplied from pad PTa. Transfer gates TX are provided in the vicinity of pads PDa-PDi as near as possible. In the normal operation, reduction of interconnection line capacitance of pads PDa-PDi as well as reduction of capacitance of a signal input/output portion of processor 2 are achieved, and high speed operation of processor 2 in the normal operation is assured.

[0114] Actual testing circuit 15aincludes: a decoder 15aawhich receives signals supplied to pads PDa-PDb via transfer gates TX of select/connect circuit 15band decodes the received signals; a selector 15abcoupled with internal read/write data bus 5a, selects a set of bus lines 5aa-5anof the internal read/write data bus according to a signal output from decoder 15aa, and electrically couples the selected internal read/write data bus lines to pads PDc-PDd via transfer gates of select/connect circuit 15b; a compression circuit 15acwhich compresses data signals on bus lines 5aa-5anof internal read/write data bus 5ainto one bit data and transmits the compressed data to pad PDe via a corresponding transfer gate TX; and a select gate circuit 15adresponsive to activation of test activation signal TE supplied from pad PTa to connect bus lines 12aa-12amof control bus 12ato corresponding transfer gates TX of select/connect circuit 15b. Select gate circuit 15adis provided for preventing increase of interconnection line capacitance of bus lines 12aa-12amof control bus 12ain the normal operation, and includes transfer gates XT provided corresponding to respective bus lines 12aa-12am.

[0115] In the test operation, processor 2 (see FIG. 6) is in a non-operation state, and set in an output high impedance state. This is accomplished by supplying a control signal indicating testing of DRAM macro 3 to processor 2, or supplying test activation signal TE to processor 2.

[0116] Control bus 12atransmits an address signal and a control signal. The address signal includes a bank designation signal designating a bank, a row address signal designating a row, and a column address signal designating a column. The control signal includes a signal designating writing/reading mode of data, and a signal which provides a timing at which row and column address signals are taken in. If a memory management unit such as a DRAM controller which manages access to DRAM array is contained in the processor, a control signal which is the same as a control signal for a monolithic DRAM device is supplied to DRAM macro 3. Accordingly, a function test for DRAM macro 3 can be carried out following a sequence identical to that for testing of a discrete DRAM device. If a control signal is supplied to pads PDf-PDi and writing/reading of test data is carried out via pads PDc-PDd, testing for DRAM macro 3 can be externally executed correctly after packaging since these pads PDa-PDi are connected to external terminals.

[0117] FIG. 8 shows one example of a configuration of selector 15abof FIG. 7. A configuration of a unit selector circuit provided for one pad PD is illustrated in FIG. 8. A set of eight internal read/write data bus lines 5a0-5a7 and a set of bus lines 5a8-5a15 are provided for the one pad PD. The unit selector circuit includes transfer gates T0-T7 provided respectively corresponding to eight internal read/write data bus lines 5a0-5a7 and selectively rendered conductive in response to decode signals S0-S7 supplied from decoder 15aashown in FIG. 7. Transfer gates T0-T7 are commonly coupled to a node ND. Node ND is electrically coupled to a corresponding pad PD in the test operation mode.

[0118] The decoder decodes the signals supplied from corresponding pads PDa-PDb as an address signal, and drives one of decode signals S0-S7 into a selected state (active state). Accordingly, regarding internal read/write data bus line 5a, one of the eight bus lines in the respective sets of eight bus lines 5a0-5a7 and 5a8-5a15 is electrically coupled to a corresponding pad (in test operation mode).

[0119] Internal read/write data bus 5ahas a width of 64 bits to 512 bits. Since selector 15abis provided, even if the total bus width of internal read/write data buses 5aand 5bis, for example, 1 K bits, 128 pads (pin terminals) are just required when test data is input/output. Even if the bit width of the internal read/write data bus is large, the DRAM macro can be surely tested using external pin terminals. Approximately 200 pin terminals are usually provided in this semiconductor integrated circuit device as in the normal CPU.

[0120] FIG. 9 shows one example of a configuration of compression circuit 15acshown in FIG. 7. Referring to FIG. 9, compression circuit 15acincludes a two-input coincidence detection circuit NEX provided for bus lines on which the same test data is written (bus lines selected by the same selection signal of selection signals S0-S7) of internal read/write data bus lines 5aa-5an. In FIG. 9, a coincidence detection circuit NEXO receiving signals on bus lines 5a0 and 5a8, and a coincidence detection circuit NEX1 receiving signals on bus lines 5a1 and 5a9 are representatively shown.

[0121] Compression circuit 15acfurther includes an AND circuit AN1 receiving signals output from adjacent two coincidence detection circuits NEXi and NEXi+1. In FIG. 9, an AND circuit AN10 receiving signals output from coincidence detection circuits NEX0 and NEX1 is shown representatively. Compression circuit 15acfurther includes an AND circuit AN20 receiving signals output from adjacent AND circuits AN1i and AN1i+1. In FIG. 9, an AND circuit AN20 receiving signal output from AND circuit AN10 at one input is representatively shown. AND gates are successively provided depending on the number of bus lines of internal read/write data bus 5a, and signal P/F indicating whether the logics of the signals on all bus lines 5a0-5ancoincide with each other or not is output from AND circuit ANn at the last stage.

[0122] Coincidence detection circuit NEX0 outputs a signal at an H level if the logics of the signals on corresponding bus lines are identical to each other. The coincidence detection circuit receives data read from memory cells to which data of the same logic is written. If the logics of signals on read/write data bus lines 5a0-5anare the same, signal P/F output from AND circuit ANn is at the H level. If the logics of signals of a pair of bus lines are different, a signal output from a corresponding coincidence detection circuit NEX is at an L level, a signal output from a corresponding AND circuit AN1j is at the L level, and output signal P/F from AND circuit ANn at the last stage is at the L level. Whether the logics of data read from memory cells to which the data is written are identical or not is determined to detect a pass/fail of a memory cell. In this way, memory cells corresponding in number to internal read/write data buses can be simultaneously tested, and time for the testing can be reduced.

[0123] In the configuration of compression circuit 15ac, if the bit width of internal read/write data bus 5ais 1024 bits, the number of gates decreases by half, as 512 gates for the coincidence detection circuit, 256 gates for the AND circuit AN1, 128 gates for the AND circuit AN2, . . . As a result, gate delay of nine stages occurs. However, since high speed access is not required in test operation, there is no problem even if the number of stages of the gates increases. Further, the testing circuit is provided on the outside of the DRAM macro and can occupy an area of a length which is approximately the same as the length of the DRAM macro in the column direction. As a result, the compression circuit can be arranged with a sufficient margin.

[0124] As heretofore described, according to the third embodiment of the invention, a processor (logical processing circuit unit) is placed on one side of the DRAM macro and a testing circuit for executing a function test for the DRAM macro is provided on the other side in the row direction of the DRAM macro. The testing circuit has a function of coupling the internal read/write data bus and the control bus of the DRAM macro to external pin terminals of the semiconductor integrated circuit device, so that the function test for the DRAM macro can be easily conducted even after packaging. The DRAM macro has a configuration identical to that of a normal DRAM (a standard DRAM (if a bank control signal is supplied from a processor separately to bank A and bank B) or a synchronous DRAM), so that a test can be executed following a sequence identical to the test sequence of the normal DRAM, and a test for the DRAM macro can be carried out using a conventional testing unit.

[0125] Further, since the processor is separated from the DRAM macro when the DRAM macro is tested (the processor is set in the output high impedance state or the non-operation state upon activation of test activation signal TE), the processor is not required to manage testing for the DRAM macro and load of the processor is reduced.

[0126] FIG. 10 schematically shows an entire configuration of a semiconductor integrated circuit device according to the fourth embodiment. With reference to FIG. 10, semiconductor integrated circuit device 1 includes processor 2, and DRAM macro 3 serving as a main memory for processor 2. DRAM macro 3 includes four subbanks 4e-4h. Each of subbanks 4e-4hincludes a memory cell array, a row decoder, a column decoder and a preamplifier.write driver block as in the configuration shown in FIG. 1. Subbanks arranged in alignment with each other in the row direction in DRAM macro 3 constitute the same bank. Specifically, subbanks 4eand 4fconstitute bank A, and subbanks 4gand 4hconstitute bank B.

[0127] Subbanks arranged in alignment with each other in the column direction belong to different banks and an internal read/write data bus is provided commonly for them. Specifically, an internal read/write data bus 5cis commonly coupled to subbanks 4eand 4g, an internal read/write data bus 5dis commonly coupled to subbanks 4fand 4h. Internal read/write data buses 5cand 5dare respectively coupled to data buses 5eand 5fin a central region 20 in the column direction. Data buses 5eand 5fextend in the row direction in region 20 and are coupled to processor 2.

[0128] In a central region 21 in the row direction of DRAM macro 3, a bank A control circuit 7cis provided between subbanks 4eand 4f, and a bank B control circuit 7dis provided between subbanks 4gand 4h. Bank A control circuit 7cand bank B control circuit 7dreceive an address signal and a control signal from processor 2 via a control bus 12cprovided in region 20.

[0129] Bank A control circuit 7cand bank B control circuit 7dare placed in a region different from region 20 where data buses are located. Therefore, there is no need to consider a region where data buses are placed when bank A control circuit 7cand bank B control circuit 7dare to be arranged, resulting in implementation of an efficient layout and reduction of an occupied area of the circuit. Processor 2 is merely required to have an interface section for input/output of data and output of a control signal (including an address signal) placed in its central portion in the column direction, and its internal circuits can be appropriately distributed to implement an efficient layout.

[0130] Further, bank A control circuit 7cis located between subbanks 4eand 4fand bank B control circuit 7dis located between subbanks 4gand 4h. Accordingly, propagation delay of the control signal and the address signal is the same and difference of operation timings in subbanks 4e-4hcan be avoided. Therefore, there is no need to consider a margin for the difference of operation timing, and high speed operation is achieved.

[0131] FIG. 11 schematically shows an entire configuration of a semiconductor integrated circuit device according to the fifth embodiment. The semiconductor integrated circuit device shown in FIG. 11 includes processor 2 and DRAM macro 3. Eight subbanks 4a-4dand 4i-4lare provided in DRAM macro 3. Subbanks 4aand 4cconstitute bank A, subbanks 4band 4dconstitute bank B. Subbanks 4iand 4kconstitute bank C, and subbanks 4jand 4lconstitute bank D. Subbanks opposite to each other with respect to central region 10 in the column direction of DRAM macro 3 belong to the same bank. A bank control circuit is placed to be sandwiched between those subbanks belonging to the same bank. Specifically, bank A control circuit 7ais arranged between subbanks 4aand 4c, bank B control circuit 7bis placed between subbanks 4band 4d, bank C control circuit 7eis placed between subbanks 4iand 4k, and bank D control circuit 7f is placed between subbanks 4jand 4l. Internal read/write data bus 5aextending in the row direction in region 11ais connected commonly to subbanks 4a, 4b, 4iand 4j. Similarly, internal read/write data bus 5bextending in the row direction in region 11bis connected commonly to subbanks 4c, 4d, 4kand 4l.

[0132] According to the arrangement shown in FIG. 11, the configuration of subbanks 4a-4d, bank A control circuit 7aand bank B control circuit 7bis identical to that of DRAM macro 3 shown in FIG. 1. By repeatedly duplicating the configuration identical to this two-bank configuration in the row direction, DRAM macro 3 of four-bank configuration is implemented as shown in FIG. 11. The bank can be easily extended by arranging internal read/write data buses 5aand 5bof DRAM macro 3 to extend in the row direction such that the data buses do not cross the bank control circuits. A multi-bank DRAM macro having a small occupied area is thus implemented.

[0133] FIG. 12 schematically shows a configuration of the first modification of the semiconductor integrated circuit device according to the fifth embodiment. DRAM macro 3 shown in FIG. 12 includes eight subbanks 4e-4h, 4m, 4n, 4pand 4q. Those subbanks opposite to each other with respect to central region 20 in the column direction of DRAM macro 3 belong to different banks. Subbanks 4eand 4fconstitute bank A, subbanks 4gand 4hconstitute bank B, subbanks 4mand 4nconstitute bank C, and subbanks 4pand 4qconstitute bank D.

[0134] Between subbanks 4eand 4f, bank A control circuit 7cis located and bank B control circuit 7dis located between subbanks 4gand 4h. Bank C control circuit 7gis arranged between subbanks 4mand 4n, and bank D control circuit 7his arranged between subbanks 4pan 4q. These bank control circuits control the operations of corresponding banks.

[0135] Subbanks opposite to each other about region 20 are connected to each other by the same internal read/write data bus. Specifically, subbanks 4eand 4gare coupled by internal read/write data bus 5c, subbanks 4fand 4hshare internal read/write data bus 5d, subbanks 4mand 4pshare internal read/write data bus 5g, and subbanks 4nand 4qshare internal read/write data bus 5h. Internal read/write data buses 5cand 5gare coupled by internal data bus 5eextending in the row direction in region 20, and internal read/write data buses 5dand 5hare intercoupled by data bus 5fextending in the row direction in region 20. Data buses 5eand 5fare coupled to processor 2.

[0136] In region 20, a control bus which transmits an address signal and a control signal is provided commonly for control circuits 7c, 7d, 7gand 7hwhich control selection/non-selection of a bank. In FIG. 12, the control bus is not shown for simplification of the figure. The control bus is arranged extending in the row direction in region 20.

[0137] In DRAM macro 3 shown in FIG. 12, the configuration of subbanks 4e-4h, bank A control circuit 7cand bank B control circuit 7dis identical to that shown in FIG. 10. The layout of subbanks 4m, 4n, 4pand 4q, bank C control circuit 7gand bank D control circuit 7his the same as the layout of subbanks 4e-4h, bank A control circuit 7cand bank B control circuit 7d. Accordingly, the four-bank configuration is implemented by repeatedly duplicating the layout of the two-bank configuration in the row direction.

[0138] The bank can be easily extended by arranging the internal data bus and the control bus to extend in the row direction such that the buses do not cross the bank control circuit in DRAM macro 3 and the processor.

[0139] If DRAM macro 3 shown in FIG. 1 or 10 is registered as a library, the DRAM macro as the library can be utilized to easily extend the bank (A DRAM macro as a base unit is repeatedly arranged in the row direction).

[0140] FIG. 13 schematically shows an entire configuration of a semiconductor integrated circuit device according to the sixth embodiment. Referring to FIG. 13, semiconductor integrated circuit device 1 includes a processor (CPU) 25 arranged in a central portion of a chip where semiconductor integrated circuit device 1 is formed, subbanks 4e-4harranged corresponding to regions at the four corners of processor 25 respectively, bank A control circuit 7carranged between subbanks 4eand 4f, and bank B control circuit 7darranged between subbanks 4gand 4h. Subbanks 4eand 4fconstitute bank A, and subbanks 4gand 4hconstitute bank B.

[0141] Subbanks 4eand 4gare coupled to each other by internal read/write data bus 5carranged in a region on the outside of processor 25, and internal read/write data bus 5cis coupled to processor 25 by internal data bus 5i. Subbanks 4fand 4hshare internal read/write data bus 5d. Internal read/write data bus 5dis coupled to processor 25 via internal data bus 5j. Processor 25 supplies a control signal and an address signal to bank A control circuit 7cand bank B control circuit 7dvia control buses 12cand 12d.

[0142] In the arrangement shown in FIG. 13, internal read/write data buses 5cand 5ddo not cross bank A control circuit 7cand bank B control circuit 7d.

[0143] Since processor 25 is coupled to internal read/write data buses 5cand 5dvia internal data buses 5iand 5j, bank A control circuit 7c, bank B control circuit 7dand processor 25 can be arranged without considering these data buses so that a DRAM containing processor in which increase of the area is suppressed is implemented. Processor 25 could supply and receive data and a control signal to and from an external unit external to integrated circuit device 1. Communication of data and the control signal between the external unit and the processor 25 is carried out using an interconnection line placed in an appropriate empty region as shown by the alternate long and short dash line in FIG. 13.

[0144] FIG. 14 schematically shows a configuration of the first modification of the sixth embodiment. Semiconductor integrated circuit device 1 shown in FIG. 14 includes a processor 25 arranged in a region which extend in the column direction in central region 21 in the row direction. Subbanks 4aand 4bare arranged to be opposite to each other relative to processor 25, and subbanks 4cand 4dare arranged to be opposite to each other relative to processor 25 on the lower side of processor 25. In central region 10 in the column direction of semiconductor integrated circuit device 1, bank A control circuit 7ais provided between subbanks 4aand 4c, and bank B control circuit 7bis provided between subbanks 4band 4d. Subbanks 4aand 4cconstitute bank A and subbanks 4band 4dconstitute bank B.

[0145] Subbanks 4aand 4bshare internal read/write data bus 5aand subbanks 4cand 4dshare internal read/write data bus 5b. Internal read/write data buses 5aand 5bare arranged extending through processor 25. Processor 25 is provided with empty regions 25aand 25bfor allowing internal read/write data buses 5aand 5bto pass therethrough. Although empty regions 25aand 25bhave no interconnection line in the same layer as that of internal read/write data buses 5aand 5b, a transistor may be formed in these regions.

[0146] Although processor 25 should be provided with empty regions 25aand 25b, processor 25 can be arranged over the entire portion of region 21 extending in the column direction of semiconductor integrated circuit device 1 and internal components of processor 25 can be appropriately distributed. Since bank A control circuit 7aand bank B control circuit 7bare provided in region 10, there is no need to consider a crossing of internal read/write data buses 5aand 5band these control circuits, and an efficient layout is possible to enable the components to be arranged with high density, and an area occupied by those can be reduced.

[0147] According to the arrangement shown in FIG. 14, a logic-mixed DRAM is implemented without increasing an occupied area by the circuits.

[0148] FIG. 15 schematically shows a configuration of the second modification of the sixth embodiment. Referring to FIG. 15, semiconductor integrated circuit device 1 includes, as in the configuration shown in FIG. 14, processor 25 arranged to extend in the column direction in central region 21 with respect to the row direction, and subbanks 4e, 4f, 4gand 4hrespectively arranged in four regions divided by processor 25 and region 10. Subbanks 4eand 4fopposite to each other relative to processor 25 constitute bank A and subbanks 4gand 4hopposite to each other relative to processor 25 constitute bank B. In region 10, bank control circuit 27ais placed between subbanks 4eand 4gand bank control circuit 27bis placed between subbanks 4fand 4h. Bank control circuits 27aand 27brespectively control access operation to banks A and B according to an address signal and a control signal supplied from processor 25.

[0149] Subbank 4eis coupled to processor 25 via internal read/write data bus 5ca, subbank 4fis coupled to processor 25 via internal read/write data bus 5da, and subbanks 4gand 4hare coupled to processor 25 respectively via internal read/write data buses 5cband 5db. Since internal read/write data buses 5caand 5cbalternatively supply and receive valid data, those are connected to each other in an interconnection region 25cin processor 25. Since internal read/write data buses 5daand 5dbtransfer data of banks A and B respectively, they are coupled to each other in an interconnection region 25dof processor 25. Since internal read/write data buses 5caand 5dasimultaneously transfer valid data, internal read/write data buses 5caand 5daare not required to be connected to each other and not required to cross processor 25. Further, since internal read/write data buses 5cband 5dbtransfer data of bank B and simultaneously transfer valid data, there is no need to arrange internal read/write data buses 5cband 5dbcrossing over processor 25 for interconnection thereof.

[0150] Processor 25 is arranged in region 21 extending in the column direction of semiconductor integrated circuit device 1. By providing interface sections in interconnection regions 25cand 25d, other circuit portions (e.g. register, arithmetic logic operation circuit, and the like) can be appropriately distributed so that an influence of these interconnection regions 25cand 25don the layout of processor 25 can be sufficiently suppressed . Bank control circuits 27aand 27beach should drive both of bank A and bank B and the area occupied by the control circuits increases compared with the configuration in which banks A and B are each driven by the dedicated control circuits. However, there exists no crossing of the control circuits and internal read/write data buses 5ca, 5cb, 5daand 5db, the layout can be optimized without considering the crossing and increase of the occupied area of the circuits can be limited to the minimum.

[0151] It is noted that in the configuration shown in FIG. 15, interconnection regions 25cand 25dmay be provided outside processor 25.

[0152] According to the sixth embodiment of the invention as described above, since the processor is provided in a region between subbanks and the bank control circuit is arranged such that it does not cross the internal read/write data bus, increase of the area can be limited to the minimum.

[0153] FIG. 16 schematically shows an entire configuration of a semiconductor integrated circuit device according to the seventh embodiment. Semiconductor integrated circuit device 1 shown in FIG. 16 includes processor (CPU) 2 and DRAM macro 3. DRAM macro 3 contains four subbanks 4r, 4s, 4tand 4u. Subbank 4rincludes a memory array 4rahaving a plurality of dynamic memory cells arranged in rows and columns, a row decoder 4rbfor selecting a row in memory array 4ra, and a column decoder 4rcfor selecting a column in memory array 4ra. Subbank 4sarranged in alignment with subbank 4rin the row direction includes a memory array 4sa, a row decoder 4sband a column decoder 4sc. Subbank 4uarranged in alignment with subbank 4sin the column direction includes a memory array 4ua, a row decoder 4ub, and a column decoder 4uc. Subbank 4tdarranged in alignment with subbank 4uin the row direction includes a memory array 4ta, a row decoder 4tband a column decoder 4tc.

[0154] A preamplifier.write driver block 44vfor writing and reading internal data into and from subbanks 4rand 4tis provided commonly for them, and a preamplifier.write driver block 44w is provided commonly for subbanks 4sand 4u.

[0155] Subbanks 4rand 4sconstitute bank A and subbanks 4tand 4uconstitute bank B. Preamplifier.write driver blocks 44vand 44ware selectively coupled to a selected bank when data is written/read. Preamplifiere.write driver block 44vis coupled to a selected memory cell of memory array 4raof subbank 4rvia an internal IO bus 4rd, and coupled to a selected memory cell of memory array 4taof subbank 4tvia an internal IO bus 4td. Preamplifier.write driver block 44wis coupled to a selected memory cell of memory array 4saof subbank 4svia an internal IO bus 4sd, and coupled to a selected memory cell in memory array 4uaof subbank 4uvia an internal IO bus 4ud. Preamplifiere.write driver block 44vis coupled to processor 2 via internal data bus 5eextending in the row direction, and preamplifier.write driver block 44wis coupled to processor 2 via internal data bus 5fextending in the row direction.

[0156] A bank A control circuit 37ais arranged in a region between subbanks 4rand 4s, a write.read control circuit 37cis arranged between preamplifier.write driver blocks 44vand 44w, and a bank B control circuit 37bis arranged between subbanks 4tand 4u. An address signal and a control signal from processor 2 is supplied to control circuits 37a-37bvia control bus 12c. Control circuits 37a-37cdo not cross internal read/write data buses 5eand 5f. Accordingly, layout of control circuits 37a-37cwith an efficient use of the area is possible. Further, since banks A and B share preamplifier.write driver blocks 44vand 44w, an occupied area of the circuits can be dramatically reduced. Particularly, internal IO buses 4rd, 4sd, 4tdand 4udhave large bit width of 64 to 512 bits, and therefore the number of preamplifier circuits and write drive circuits included in preamplifier.write driver blocks 44vand 44wcan be significantly reduced and an occupied area of the circuits can be significantly decreased.

[0157] FIG. 17 schematically shows a configuration of a one bit circuit portion, included in preamplifier.write driver blocks 44vand 44wshown in FIG. 16, for an internal read/write data bus line RWa (5eaor 5fa). In FIG. 17, the one bit circuit of preamplifier.write driver blocks 44vand 44wincludes a preamplifier circuit PA which is coupled to internal read/write data bus line RWa, amplifies an internal data signal supplied upon activation thereof and outputs the amplified signal, a write drive circuit WD which is coupled to internal read/write data bus line RWa, amplifies data on bus line RWa upon activation thereof and outputs the data, and a selector SEL which selectively connects preamplifier circuit PA and write drive circuit WD to an internal IO bus line IOa (4rdaor 4sda) of bank A and an internal IO bus line IOb (4tdaor 4uda) of bank B according to a selection signal φBA.

[0158] Selection signal XBA is generated according to a bank address signal when data is written/read. When bank A is selected, selector SEL couples preamplifier circuit PA and write drive circuit WD to internal IO bus line IOa. When bank B is designated, selector SEL couples preamplifier circuit PA and write drive circuit WD to internal IO bus line IOb. Selector SEL may be constituted of, for example, a CMOS transmission gate and separates an internal IO bus line of a bank which is not selected from preamplifier circuit PA and write drive circuit WD. Accordingly, even if preamplifier circuit PA and write drive circuit WD are shared between two banks A and B, data can be written or read into or from a selected bank accurately.

[0159] FIG. 18 illustrates a configuration of the first modification of the preamplifier.write driver block of the seventh embodiment of the invention. FIG. 18 shows one (44x) of preamplifier.write driver blocks 44vand 44wshown in FIG. 16. A configuration of a portion for transfer of one bit data in preamplifier.write driver block 44xis representatively shown.

[0160] Referring to FIG. 18, preamplifier.write driver block 44xincludes a preamplifier circuit PAa for amplifying data read from bank A (corresponding subbank), a write.drive circuit WDa generating internal write data to be written into a selected memory cell for bank A, a preamplifier circuit PAb provided for bank B (corresponding subbank) for amplifying data of a selected memory cell of bank B upon activation thereof, and a write.drive circuit WDb generating data to be written into a selected memory cell of bank B.

[0161] Preamplifiere.write driver block 44xfurther includes a bidirectional latch 44xalatching a supplied data responsive to clock signal T1, a selector 44xbselecting one of a set of preamplifier circuit PAa and write.drive circuit WDa and a set of preamplifier circuit PAb and write.drive circuit WDb and couples the selected one to latch 44xaaccording to bank designation signal φBA, a bidirectional latch 44xccoupled to preamplifier circuit PAa and write.drive circuit WDa for latching a supplied data responsive to clock signal T2a, and a bidirectional latch 44xdcoupled to preamplifier circuit PAb and write.drive circuit WDb which is set in a latching state responsive to clock signal T2b. Latch 44xcis connected to internal IO bus line IOa of bank A (corresponding subbank), and latch 44xdis connected to internal IO bus line IOb of bank B (corresponding subbank).

[0162] Clock signals T1, T2aand T2bdefine an internal data transfer cycle. Suppose that a synchronous DRAM is provided as a DRAM implemented by the DRAM macro, which operates in synchronization with a clock signal (e.g. a system clock signal which defines an operation speed of the processor). In the synchronous DRAM, incorporation of a control signal as well as input/output of data are carried out in synchronization with the clock signal. Timing margin for the control signal may be considered relative to the clock signal only, and skew between control signals needs not to be considered and an internal operation can be started at a faster timing. Since writing/reading of data is executed synchronously with the clock signal, writing/reading of data as well as transfer of data at a high speed are possible.

[0163] In some synchronous DRAM, a path through which internal data is transferred is constituted of a plurality of pipeline stages and data is transferred in a pipelined manner through the internal data transfer path to increase transfer rate of data effectively. In this case, the pipeline stages are implemented with latches 44xa, 44xcand 44xdshown in FIG. 18.

[0164] Clock signal T2ais driven into an active state when bank A is designated, and clock signal T2bis activated when bank B is selected. When data is read, data latched by the sense amplifier is selected by the column decoder and transmitted to a corresponding internal IO bus line. When data is latched by latch 44xcor 44xd, preamplifier circuit PAa or PAb is activated and the data is transferred to latch 44xavia selector 44xb. When latch 44xatransfers the latched read data onto internal read/write data bus line RWa, preamplifier PAa or PAb again amplifies data supplied from latch 44xcor 44xd. Accordingly, in one clock cycle, data are present respectively on internal IO bus line IOa or IOb, preamplifier circuit PAa or PAb, and internal read/write data bus line RWa. Transfer rate of data is increased by transferring data successively every clock cycle.

[0165] Selector 44xbwhich selects bank A or bank B responsive to bank designation signal φBA allows the banks A and B to share latch 44xain preamplifier.write driver block 44x. Therefore, it is not necessary to provide latch 44xato each of banks A and B in the pipeline configuration, and reduction of the number of components as well as reduction of an occupied area of the circuits are possible. Internal read/write data bus lines 5eand 5fshown in FIG. 16 are shared by banks A and B. Therefore, an additional latch may be provided in this portion and the number of pipeline stages may be increased.

[0166] In the configuration shown in FIG. 18, latches 44xa, 44xcand 44xdare described as bidirectional latches. However, these latches 44xa, 44xcand 44xdmay be respectively provided for writing of data and reading of data separately. Latches 44xa, 44xcand 44xdmay include transfer gates and inverter latches, and an equivalently unidirectional latch may be implemented by making the timings of conduction/non-conduction of the transfer gates of an input/output portion of the inverter latch different in writing of data and in reading of data.

[0167] FIG. 19 illustrates a configuration of the second modification of the preamplifier.write driver block of the seventh embodiment. A configuration of a portion provided for one bank out of the preamplifier.write driver blocks shown in FIG. 16 is illustrated in FIG. 19. In FIG. 19, preamplifier.write driver block 44xincludes a preamplifier circuit PA and a write drive circuit WD provided commonly to banks A and B (corresponding subbanks), latch 44xaprovided between preamplifier circuit PA and write drive circuit WD, and internal read/write data bus line RWa and responsive to clock signal T1 to attain a latching state, latch 44xcconnected to internal IO bus line IOa and responsive to clock signal T2ato attain the latching state, and latch 44xdconnected to internal data bus line IOb and responsive to clock signal T2bto attain the latching state, and a selector 44xeresponsive to bank designation signal φBA to connect one of latches 44xcand 44xdto preamplifier circuit PA and write drive circuit WD. Configurations of latches 44xa, 44xcand 44xdare identical to those shown in FIG. 18.

[0168] According to the configuration of preamplifier.write driver block 44xof FIG. 19, preamplifier circuit PA, write.drive circuit WD and latch 44xaare shared by banks A and B. Therefore, the configuration shown in FIG. 19 accomplishes further reduction of an occupied area by the circuits compared with the configuration of FIG. 18. The operation of preamplifier.write driver block 44xof FIG. 19 is the same as the circuit operation shown in FIG. 18. The only difference is that preamplifier circuit PA and write.drive circuit WD are in the active state when data is read and written, whichever bank of bank A and bank B is selected.

[0169] FIG. 20 schematically shows a configuration of the third modification of the seventh embodiment. In FIG. 20, one preamplifier.write driver block 44xis representatively illustrated. In the configuration of FIG. 20, a circuit portion for transferring one bit data is shown. In FIG. 20, preamplifier.write driver block 44xincludes latch 44xaconnected between preamplifier circuit PA and write.drive circuit WD, and internal read/write data bus line RWa and responsive to clock signal T1 to attain the latching state, latch 44xfconnected to preamplifier circuit PA and write.drive circuit WD and responsive to clock signal T2 to attain the latching state, and selector 44xgwhich connects latch 44xfto one of internal IO bus lines IOa and IOb according to bank designation signal φBA.

[0170] According to the configuration of preamplifier.write driver block 44xshown in FIG. 20, latches 44xaand 44xf, preamplifier circuit PA and write.drive circuit WD are shared by banks A and B. As a result, the circuit configuration of FIG. 20 accomplishes further reduction of an occupied area by the circuits compared with the preamplifier.write driver block of FIG. 19.

[0171] Depending on the location of the critical path in the pipeline stage regarding data transfer rate, one of the preamplifier.write driver blocks shown in FIGS. 18 to 20 is appropriately selected. If a data transfer path from the memory array to the preamplifier is the critical path, the preamplifier circuit and write.drive circuit are provided as close as possible to the sense amplifier to shorten the length of internal data bus line IO as shown in FIG. 18. In this case, the configuration illustrated in FIG. 18 is employed. On the other hand, if the data transfer path from the preamplifier and the write.drive circuit to the data input/output section is the critical path, the configuration of FIG. 20 is employed to shorten the length of the internal read/write data bus RWa as much as possible.

[0172] It is noted that in the configurations of FIGS. 19 and 20, latches 44xa, 44xc, 44xd, and 44xfmay be separately provided for reading of data and for writing of data.

[0173] According to the seventh embodiment of the invention, those subbanks facing to each other in the column direction are allocated to different banks and share the preamplifier.write driver block, so that an occupied area by the circuits can be dramatically decreased. Since the bank control circuit is provided between subbanks of the same bank arranged in alignment with each other in the row direction, the internal read/write data bus line and the control circuit never cross to achieve an efficient layout of the bank control circuit.

[0174] FIG. 21 shows schematically an entire configuration of a semiconductor integrated circuit device according to the eighth embodiment. Semiconductor integrated circuit device 1 shown in FIG. 21 includes processor (CPU) 2 and DRAM macro 3. DRAM macro 3 includes four subbanks 4a, 4b, 4cand 4d. Subbanks 4aand 4cconstitute bank A and subbanks 4band 4dconstitute bank B. In the eighth embodiment, in each of subbanks 4a144d, a write driver for writing internal data into a selected memory cell and a preamplifier for reading data from a selected memory cell are separately provided. Specifically, a write driver 4awis arranged adjacent to a column decoder 4acplaced at one end of a memory array 4aain subbank 4a, and preamplifier 4apis arranged opposite to column decoder 4acand write driver 4awat the other end of memory array 4aain subbank 4a. Similarly in subbank 4b, a column decoder 4bcand a write driver 4bware arranged at one end of memory array 4ba, and a preamplifier 4bpis arranged opposite to column decoder 4bcand write driver 4bwat the other end of memory array 4ba.

[0175] In subbank 4c, a column decoder 4ccand a write driver 4cware placed at one end of memory array 4ca, and a preamplifier 4cpis arranged at the other end of memory array 4ca. In subbank 4d, a column decoder 4dcand a write driver 4dware arranged opposite to preamplifier 4dprelative to a memory array 4d.

[0176] In a central region with respect to the column direction of DRAM macro 3, bank A control circuit 7aand bank B control circuit 7bare arranged. An internal write data bus 5awis arranged in a region between control circuits 7aand 7band subbanks 4aand 4b, and is extended in the row direction to be coupled commonly to write drivers 4awand 4bw, and an internal read data bus 5aris arranged at the periphery of DRAM macro 3 in the row direction to be commonly coupled to preamplifiers 4apand 4bp.

[0177] For subbanks 4cand 4d, an internal write data bus 5bwcommonly coupled to write drivers 4cwand 4dwas well as an internal read data bus 5brcommonly coupled to preamplifier 4cpand preamplifier 4dpare provided. These buses extend in the row direction and are coupled to processor 2 provided on the outside of DRAM macro 3. Processor 2 provides an address signal and a control signal to bank A control circuit 7aand bank b control circuit 7bvia a control bus (not shown).

[0178] As shown in FIG. 21, since the write driver is located adjacent to the column decoder and the preamplifier is located opposite to the column decoder, when data is read from a selected memory cell, the time required for reading of data from a selected memory cell which is most distant from the column decoder can be made equal to the time required for reading data from a selected memory cell which in turn is closest to the column decoder, and high speed reading becomes possible. The high speed reading is now described below.

[0179] FIG. 22 schematically shows a configuration of one subbank. In FIG. 22, a memory array includes a plurality of subarray blocks MC#(1, 1)-MC#(m, n) arranged in rows and columns. Each of subarray blocks MC#(1,1)-MC#(m, n) includes dynamic memory cells arranged in rows and columns. Subarray blocks MC#(i, 1)-MC#(i, n) arranged in alignment with each other in the row direction constitute a row block RG#i, and memory subarray blocks MC#(1, j)-MC#(m, j) arranged in alignment with each other in the column direction constitute a column block CG#j. The memory array is divided into row blocks RG#1-RG#m in the column direction and divided into column blocks CG#1-CG#n in the row direction. A group of word lines WLS is commonly provided for memory subarray blocks MC#(i, 1)-MC#(i, n) included in row block RG#i, and a group of column selection lines CSLS is provided commonly for subarray blocks MC#(1, j)-MC#(m, j) of column group CG#j.

[0180] Corresponding to memory subarray blocks MC#(1, 1)-MC#(m,n), sense amplifier bands SA#(1, 1)-SA#(m+1, n) are arranged to be shared between adjacent memory subarray blocks. Each of sense amplifier bands SA#(1, 1)-SA#(m+1, n) includes a sense amplifier circuit provided corresponding to each column in a corresponding memory subarray block, and senses and amplifies memory cell data on a corresponding column upon activation thereof, and an IO gate which selects a column in a corresponding memory subarray block in accordance to column selection line CSL included in the group of column selection lines CSLS.

[0181] The memory array shown in FIG. 22 has a configuration of the shared sense amplifier in which only a subarray block including a selected memory cell is coupled to a corresponding sense amplifier band and sensing and amplification of data are carried out. A subarray block which constitutes a pair with a selected subarray block (the block sharing a sense amplifier) is separated from a corresponding sense amplifier band. Remaining memory subarray blocks that are not selected keep a precharge state and are connected to corresponding sense amplifier bands. Accordingly, upon array activation (when word line is selected), a non-selected subarray block which constitutes a pair with a selected memory block is separated from a corresponding sense amplifier band, and only a sense amplifier band provided corresponding to the selected subarray block is activated.

[0182] Row decoder RD selects one row group out of row blocks RG#1-RG#m, and drives one word line WL out of the group of word lines WLS included in the selected row group into a selected state.

[0183] Column decode circuits CD1-CDn are provided respectively corresponding to column groups CG#1-CG#1n. Column decode circuits CD1-CDn simultaneously each drive a prescribed number of column selection lines CSL of corresponding group of column selection lines CSLS into the selected state. In FIG. 22, 8-bit memory cells are simultaneously selected in one subarray block. One column selection line CSL may select eight columns in the subarray block, or one column selection line CSL may select two columns in a corresponding memory subarray block, and four column selection lines CSL may be simultaneously driven into the selected state.

[0184] When this column selecting operation is carried out, the column selecting operation is executed in each of column groups CG#1-CG#n.

[0185] Each of sense amplifier bands SA#(1,1)-SA#(m+1, n) is coupled to a local IO bus line pair LIO via a corresponding IO gate. Local IO bus line pairs LIO are provided corresponding to memory subarray blocks respectively. In FIG. 22, a configuration is shown as one example in which four IO bus line pairs LIO are connected to one sense amplifier band, and four-bit memory cell data is read from one sense amplifier band.

[0186] Global IO buses GIOS are provided corresponding to respective column groups CG#1-CG#n and extend in the column direction. In FIG. 22, global IO bus GIOS includes eight global IO line pairs GIO in each of column groups CG#1-CG#n. Global IO bus GIOS is coupled to each local IO bus line pair LIO of a corresponding column group via block selection gates BSG. Global IO buses GIOS are coupled to a group of write drive circuits WDS1-WDSn as well as a group of preamplifier circuits PAS1-PASn arranged opposite to each other on both sides of the memory array.

[0187] Each of write drive circuits WDSl-WDSn includes a write drive circuit which is provided corresponding to each global IO line pair of global IO bus GIOS and amplifies data supplied via internal write data bus WB upon activation thereof and transmits the amplified data onto a corresponding global IO line pair. Each of preamplifier circuits PAS1-PASn includes a preamplifier circuit which is provided corresponding to each global IO line pair of global IO bus GIOS and is activated upon activation of preamplifier activation signal PAE, to amplify potentials of signals on a corresponding global IO line pair for reading onto internal read data bus RB. Write drive circuits included in write drive circuits WDSl-WDSn are all simultaneously activated when data is written, and preamplifier circuits included in preamplifier circuits PAS1-PASn are all simultaneously activated when data is read. An operation is now briefly described.

[0188] Row decoder RD decodes a row address signal (or row predecode signal) (not shown), selects one row group of row groups RG#1-RG#m, and drives one word line WL of word line group WLS included in the selected row group into a selected state. Accordingly, local IO line pair LIO provided corresponding to a selected row block is connected to a corresponding global IO line pair, and data of a memory cell connected to the selected word line WL in the selected row group RG#i is read onto a corresponding column.

[0189] Sense amplifier bands SA#(i, 1)-SA#(i, n) and SA#(i+1, 1)-SA#(i+1, n) provided for the selected row group RG#i are coupled to corresponding subarray blocks MC#(i, 1)-MC#(i,n). Row group RG#(i+1) which shares a sense amplifier band with the selected row group RG#i and a subarray block included in row block RG#(i−1) are separated from a corresponding sense amplifier band. Remaining row groups that are not selected are maintained in a precharge state (non-activation state). Sense amplifier bands SA#(i, 1)-SA#(i, n) and SA#(i+1, 1)-SA#(i, n) are then activated, and data of a memory cell connected to the selected word line WL of selected row group RG#i is sensed, amplified, and latched. A column selecting operation for writing or reading data is next carried out.

[0190] When the column selecting operation is executed, column decode circuits CD1-CDn each decode a column address signal (predecode signal), and drive a prescribed number of column selection line(s) CSL of a corresponding column selection line group CSLS into a selected state. Responsively, IO gate included in the sense amplifier band is rendered conductive and couples a column in a corresponding memory subarray block to a local IO line pair LIO. Local IO line pair LIO provided for a selected row group RG#i is coupled to global IO line pair GIO of a corresponding global IO bus GIOS via block selection gate BSG. A selected memory cell column in each of column groups CG#l-CG#n is connected to a corresponding global IO bus GIOS. When data is written, write drive circuit group WDSl-WDSn is activated, and data is written into a selected memory cell via global IO bus GIOS. When data is read, preamplifier circuit group PAS1-PASn is activated, and memory cell data read onto global IO bus GIOS is amplified to be transferred onto internal read data bus RB. Accordingly, data of 32 bits to 256 bits is collectively written/read.

[0191] A reason for high speed reading achieved by arranging preamplifier circuit group PAS1-PASn opposite to column decoder circuits CD1-CDn is described below.

[0192] FIG. 23 shows delay time of interconnection line of each signal line in connection with reading of data. Global IO line pair GIO generates delay time ATW when a signal is propagated from one end to the other end thereof. Column selection line CSL also generates the same interconnection line delay ΔTW when a column selection signal from a column decode circuit is propagated from one end to the other end thereof. When data is read, data latched by sense amplifier SA is transferred to local IO line pair LIO via IO gate IOG which is rendered conductive in response to a column selection signal on column selection line CSL, and the data on local IO line pair is transferred to global IO line pair GIO via block selection gate BSG.

[0193] When a column is selected, sense amplifier SA has already amplified and latched data stored in memory cell MS, and potential of a signal on bit line BL (or /BL) is maintained at a potential level corresponding to that of stored data. Suppose that interconnection line delay of ΔT1 is generated at a path through which a signal is propagated via IO gate IOG, local IO line pair LIO and block selection gate BSG when a column is selected.

[0194] If the preamplifier and the column decoder are provided on the same side, interconnection line delay time of global IO line pair GIO needs not be considered when a sense amplifier band closest to the column decoder (sense amplifier columns SA#(m+1, 1)-SA#(m+1, n) of FIG. 22) is selected, and data of a memory cell is transmitted to the preamplifier with delay time ΔT1. If a sense amplifier band most distant from the column decoder (sense amplifier bands SA#(1, 1)-SA#(1, n) of FIG. 22) is selected, a column selection signal arrives after delay time ΔTW has passed and column selecting operation is carried out. After the column selecting operation, memory cell data reaches the preamplifier via global IO line pair GIO after interconnection line delay ΔTW has passed. In this case, a delay time of ΔT1+2·ΔTW is required from rising of a signal at an output section of the column decoder to arrival of memory cell data at the preamplifier.

[0195] In other words, there is a time difference of 2·ΔTW for memory cell data to reach the preamplifier between a case in which a sense amplifier band closest to the column decoder is selected and a case in which a sense amplifier band most distant from the column decoder is selected. The preamplifier should be activated when data on global IO line pair is in a definite state. In this case, timing of the activation is determined according to the worst case. Therefore, the preamplifier cannot be activated until time of 2·ΔTW+ΔT1 has passed after rising of a signal at the output section of the column decoder.

[0196] In order for the preamplifier to execute amplifying operation accurately, the preamplifier should be maintained in the active state for a certain period. Therefore, a signal on column selection signal line CSL should be maintained in the active state such that data on global IO line is in a defined state when the preamplifier is in the active state. In this case, time period of 2·ΔTW+ΔTp or more is required as an activation period for column selection signal line CSL. ΔTp indicates a minimum activation period necessary for the preamplifier.

[0197] Referring to FIG. 24, reading operation carried out when a preamplifier is arranged opposite to a column decoder is described. In FIG. 24, GIO (min) indicates a waveform of an input portion of the preamplifier when a sense amplifier band closest to the column decoder is selected, and GIO (max) indicates a waveform of the preamplifier when a sense amplifier band most distant from the column decoder is selected. CSL (min) indicates a signal waveform for, when a sense amplifier band closest to the column decoder is selected, the selected sense amplifier band, and CSL (max) indicates a column selection signal which reaches, when a sense amplifier band most distant from the column decoder is selected, the selected sense amplifier band. PAE is a preamplifier activation signal.

[0198] At time Ta, an output from the column decoder is in a defined state and potential of a corresponding column selection signal for a selected column rises. If a sense amplifier band closest to the column decoder (sense amplifier columns SA#(m+1, 1)-SA#(m+1, n)) is selected, there is no interconnection line delay for the column selection signal and column selection signal CSL (min) for the closest sense amplifier band rises at time Ta. Data held by the selected sense amplifier band is transmitted to global IO line pair GIO via local IO line pair. When data of a memory cell is transmitted to global IO line pair, delay time ΔT1 is required. Next a sense amplifier band most distant from the preamplifier is selected, and memory cell data reaches the preamplifier after interconnection line delay time ΔTW. Accordingly, input signal GIO (min) for the preamplifier attains a definite state after time ΔT1+ΔTW has passed from time Ta.

[0199] When an output from the column decoder falls at time Tb, column selection signal CSL (min) for the closest sense amplifier band is responsively inactivated. Input GIO (min) of the preamplifier returns to a prescribed precharge potential level after time ΔT1+ΔTW has passed from time Tb.

[0200] If a signal of an output portion of the column decoder rises at time Ta and a sense amplifier band most distant from the column decoder is selected, column selection signal CSL (max) for the most distant sense amplifier band is activated after delay time ATW has passed. Next, memory cell data selected by column selection signal CSL (max) is transmitted onto global IO line pair GIO after delay time ΔT1 has passed. The sense amplifier band most distant from the column decoder is a sense amplifier band closest to the preamplifier and interconnection line delay of global IO line pair can be ignored. Therefore, memory cell data signal GIO (max) for the preamplifier attains a definite state after time ΔTW+ΔT1 has passed from time Ta.

[0201] For the input portion of the preamplifier, delay time when a sense amplifier band most distant from the column decoder is selected equals to delay time when a sense amplifier band closest to the column decoder is selected. In other words, for the preamplifier, read data is supplied at the same timing regardless of the position of the selected sense amplifier band. Preamplifier activation signal PAE is activated whenever an input signal of the preamplifier is in a definite state whichever sense amplifier band is selected. Accordingly, preamplifier activation signal PAE can be activated at time Tc shown in FIG. 24. The preamplifier should complete its amplifying operation before time Tb+ΔT+ΔTW at which global IO line pair returns to a prescribed precharge state if a column selection signal of the output portion of the column decoder is inactivated at time Tb. A relation of the times is expressed as follows.

[0202] Ta+ΔT1+ΔTW<Tc,

[0203] and

[0204] Tc+ΔTP<Tb+ΔT1+ΔTW

[0205] Therefore, a period in which the column selection line should be maintained in the active state, or Tb-Ta satisfies the following relation represented by the following expression.

[0206] Tb-Ta>ΔTP

[0207] The activation period of the column selection line can be reduced to be approximately equal to the time period for which the preamplifier is activated, and the operation frequency can be increased. Particularly, if a clock synchronous type (synchronous) DRAM is employed, column selection signal lines are successively driven into a selected state in synchronization with a clock signal and the frequency of the clock signal can be increased. Accordingly, a synchronous DRAM operating at high speed can be obtained.

[0208] When data is written, a column selection signal from the column decoder and a write data from the write driver are transferred in the same direction. In this case, if the write driver is activated at a timing approximately equal to the activation timing of the column decoder, write data is transmitted at a timing which is identical to that for the column selection signal even if a sense amplifier band of any position in the memory cell array is selected. Therefore, the write driver may be activated for a minimum necessary period for writing data into a memory cell.

[0209] As shown in FIGS. 21 and 22, since the preamplifier is arranged opposite to the column decoder and the write driver is arranged adjacent to the column decoder, the column selection line may be driven into a selected state for a minimum period necessary for writing/reading of data when data is written and read, and high speed operation is implemented.

[0210] FIG. 25A schematically shows a configuration of a main portion of a semiconductor integrated circuit device according to the ninth embodiment. FIG. 25A shows an arrangement of power supply lines of DRAM macro 3. DRAM macro 3 in FIG. 25A includes memory arrays 54a, 54b, 54cand 54darranged in four divided regions respectively. These memory arrays 54a-54drespectively constitute subbanks, however, banks may be allocated according to any of the embodiments described above.

[0211] Power supply pads PS1-PS8 are arranged adjacent to memory arrays 54a-54dalong the periphery of a rectangular region such that two pads are provided for one memory array. Power supply pads PS1 and PS8 are provided adjacent to memory array 54a, power supply pads PS2 and PS3 are provided adjacent to memory array 54b, power supply pads PS4 and PS5 are provided adjacent to memory array 54d, and power supply pads PS6 and PS7 are provided adjacent to memory array 54c.

[0212] Power supply pads PS1 and PS8 are coupled to each other by a main power supply line MV1 formed in the lateral U shape along the periphery of memory array 54a. Power supply pads PS2 and PS3 are coupled to each other by a main power supply line MV2 arranged in the opposite lateral U shape along the periphery of memory array 54b. Power supply pads PS4 and PS5 are coupled to each other by a main power supply line MV4 arranged in the opposite lateral U shape to surround memory array 54d. Power supply pads PS6 and PS7 are coupled to each other by a main power supply line MV3 arranged in the lateral U shape on the outside of memory array 54c.

[0213] Each of main power supply lines MV1-MV4 has opposite portions coupled to each other by a sub power supply line SV extending over a corresponding one of memory arrays 54a-54d. Main power supply lines MV1-MV4 are coupled to an internal power supply circuit 56 and supply the power supply voltage to internal power supply circuit 56. Internal power supply circuit 56 includes internal voltage down converters provided corresponding to respective main power supply lines MV1-MV4, down-converts the power supply voltage supplied from the main power supply lines and supplies the voltage to array power supply lines AP1-AP4 as the power supply voltages for the arrays. This power supply voltage for the array is used as operation power supply voltage for the sense amplifiers.

[0214] Array power supply lines AP1-AP4 are arranged on the respective memory arrays 54a-54din a meshed shape. Array power supply lines AP1-AP4 arranged in the row direction are formed of, for example, a second level aluminum interconnection layer. Main power supply lines MV1-MV4 and sub power supply line SV are formed of a third level aluminum interconnection layer at a higher level. Array power supply lines AP1-AP4 are used for generating an operation power supply voltage for the sense amplifiers provided for memory arrays 54a-54dand an intermediate voltage such as bit line precharge voltage. Voltage on main power supply lines MV1-MV4 and sub power supply line SV is used as operation power supply voltage for the array peripheral circuit and the bank control circuit.

[0215] Internal power supply circuit 56 is shown being collectively arranged at the central portion of DRAM macro 3. However, internal power supply circuits 56 are provided adjacent to memory arrays 54a-54drespectively, each includes an internal voltage down converter (VDC) which supplies the array power supply voltage from a corresponding main power supply line to a corresponding array power supply line, and the internal voltage down converter is provided adjacent to a corresponding memory array.

[0216] FIG. 25B shows an arrangement of main power supply line MV and sub power supply line SV. As shown in FIG. 25B, power supply pads PS (PS1-PS8) each include a power supply voltage pad PSc receiving power supply voltage Vcc and a ground voltage pad PSs supplying ground voltage Vss. Main power supply line MV includes a main power supply voltage transmission line MVc coupled to power supply voltage pad PSc, and a main ground voltage transmission line MVs which is coupled to ground voltage pad PSs and transmits ground voltage Vss. Sub power supply line SV accordingly includes a sub power supply voltage transmission line SVc coupled to main power supply voltage transmission line MVc, and a sub ground voltage transmission line SVs coupled to main ground voltage transmission line MVs. Main power supply voltage transmission line MVc and main ground voltage transmission line MVs have the lateral U shape and formed in the same interconnection layer. In order to prevent collision of interconnection lines at a crossing of main ground voltage transmission line MVs and main power supply voltage transmission line MVc, one of the voltage transmission lines is interconnected by another interconnection layer, for example, the second level aluminum interconnection layer. Similar arrangements are provided for a crossing of sub power supply voltage transmission line SVc and main ground voltage transmission line MVs and a crossing of sub ground voltage transmission line SVs and main power supply voltage transmission line Mvc.

[0217] Main power supply line MV and sub power supply line SV are coupled to power supply pads PS, have a sufficiently large interconnection line width, and have ability of supplying large amount of current.

[0218] FIG. 26 illustrates how the power supply is distributed in one subbank. In FIG. 26, main power supply line MV connected to power supply pad PS supplies operation power supply voltage to peripheral circuitry 57 including a preamplifier, a write driver, a column decoder and a bank control circuit, and is also connected an array power supply circuit (VDC) 56aincluded in internal power supply circuit 56. Array power supply circuit 56adown-converts a power supply voltage on main power supply voltage transmission line MVc included in main power supply line MV, generates the power supply voltage for array and transmits it onto array power supply line AP.

[0219] A group of sense amplifiers 58 operates using the power supply voltage on the array power supply line AP as its operation power supply voltage.

[0220] Peripheral circuitry 57 includes the preamplifier, the write driver and the like as described above. In the preamplifier and the write driver, a number of circuits operate simultaneously so that a large amount of current is consumed. If a bank included in the DRAM macro is the clock synchronous type DRAM operating synchronously with a clock signal, a large number of circuits operate simultaneously at a high speed (Writing or reading of data is carried out in each clock cycle). Similarly if column selecting operation is carried out in the column decoder, a large number of column selection lines should be driven simultaneously into a selected state as shown in the array arrangement of FIG. 22.

[0221] The width of main power supply line MV connected to power supply pad PS is increased for supplying current stably. Further, main power supply line MV is coupled by the sub power supply line to have its width equivalently increased to reduce an interconnection line resistance thereof. If any noises are generated on the sub power supply line, the noises are absorbed by the main power supply line and are prevented from transmitting to another sub power supply line. Transmission of the power supply noises is thus prevented to stabilize the power supply voltage. Accordingly, current is supplied stably even if the peripheral circuitry operates at high speed, generation of the power supply noises is suppressed, and the peripheral circuitry operating accurately at high speed is thus implemented.

[0222] Array power supply circuit 56asupplies operation power supply voltage to sense amplifier group 58 via array power supply line AP arranged in the mesh shape. The sense amplifiers group 58 senses and amplifies data of memory cells connected to one row simultaneously and consumes relatively large amount of current. The array power supply line AP arranged in the mesh shape has interconnection line resistance reduced, so that current is stably supplied in the sensing operation by array power supply line AP having its width made larger equivalently even when a large amount of current is consumed in the sense amplifier operation, and thus the sensing operation is stabilized.

[0223] As the peripheral circuitry, FIG. 26 shows a preamplifier and a write driver and a bank control circuit. However, all of the peripheral circuits such as the column decoder and the row decoder are included.

[0224] FIG. 27 schematically shows an arrangement of interconnection lines in the memory array portion, and a cross sectional structure in the column direction is schematically illustrated therein. In FIG. 27, a memory cell includes high concentration N type impurity regions 61aand 61bformed at a surface of a semiconductor substrate region 60, and a gate electrode layer 62 formed of a first level polysilicon layer formed above a channel region between impurity regions 61aand 61bwith a gate insulating film (not shown) underlaid. Impurity regions 61aand 61band gate electrode layer 62 constitute a memory cell transistor (access transistor). Impurity region 61ais electrically connected to a conductive layer 63 which is a bit line formed of a second level polysilicon layer.

[0225] Impurity region 61bincludes a plug layer extending beyond conductive layer 63, a conductive layer 64 formed of a third level polysilicon layer connected to the plug layer, and a conductive layer 65 formed of a fourth level polysilicon level layer formed above conductive layer 64 with capacitor insulating film underlaid. Conductive layer 64 is a storage node for storing information of the memory cell. Conductive layer 65 is the other electrode of the memory cell capacitor and is a cell plate electrode layer receiving a constant intermediate voltage.

[0226] A first level aluminum interconnection layer 66 is formed above conductive layer 65 formed of the fourth level polysilicon layer in parallel with gate electrode layer 62, and a second level aluminum interconnection layer 67 is formed above those layers. The first level aluminum interconnection layer 66 is electrically connected to gate electrode layer 62 in a word line shunt region and reduces electrical resistance of gate electrode layer 62.

[0227] As shown in FIG. 27, an interconnection layer located in the uppermost layer is the second level aluminum layer 67 in the memory array. Therefore, even if sub power supply line SV for reinforcing main power supply line MV is provided on the memory array, sub power supply line SV is the third level aluminum interconnection layer and can be arranged without affecting the interconnection in the memory array.

[0228] FIG. 28 schematically shows a cross sectional structure in the row direction in the memory array portion. In FIG. 28, a conductive layer 70 formed of the first level aluminum interconnection layer which constitutes, for example, a local IO line (LIO) extending in the row direction, as well as upper interconnection layers are shown. The second level aluminum interconnection layer formed above conductive layer 70 is provided. The second level aluminum interconnection layer includes a conductive layer 71 which forms a global IO line (GIO) placed in the word line shunt region, a conductive layer 72 which forms a ground voltage transmission line transmitting round voltage Vss, a conductive layer 73 which forms a power supply voltage transmission line transmitting array power supply voltage Vcc, and conductive layers 74aand 74bthat form column selection lines (CSL). Conductive layers 72 and 73 constitute array power supply line AP and transmit array power supply voltages Vss and Vcc.

[0229] The third level aluminum interconnection layer is formed for arranging main power supply line MV and sub power supply line SV above the second level aluminum interconnection layer. The third level aluminum interconnection layer includes a conductive line 75 transmitting ground voltage Vss and a conductive line 76 transmitting power supply voltage Vcc.

[0230] A conductive layer which forms power supply lines for array transmitting ground voltage Vss and power supply voltage Vcc may be arranged between conductive layers 74aand 74bwhich form column selection lines (CSL) in the arrangement of interconnection lines shown in FIG. 28. Since one column selection line usually selects a plurality of memory cell columns simultaneously, the column selection line has a pitch wider than that of the memory cell column so that a power supply line can be appropriately arranged between column selection lines. Conductive layer 71 which forms global IO line (GIO) may be formed in the third level aluminum interconnection layer as shown by the broken line in FIG. 28. The third level aluminum interconnection layer has its purity higher than that of the second level aluminum interconnection layer and is superior in electrical characteristics, so that interconnection line delay of the global IO line can be reduced.

[0231] As shown in FIG. 28, the employment of conductive layers 75 and 76 of the third level aluminum interconnection layer allows the easy arrangement of the sub power supply line for reinforcing the power supply line or the main power supply line without adversely affecting components in the memory array, and the peripheral circuits can be stably operated at high speed.

[0232] FIG. 29 schematically illustrates how power supply lines are connected in the ninth embodiment. In FIG. 29, for word line WL, the gate electrode layer in the first level polysilicon layer and the first level aluminum interconnection layer located above the polysilicon layer are electrically connected in a contact region CTa in the word line shunt region WLST (word line shunting (staking)). Since the word line shunt region WLST is located in a boundary region between subarray blocks, there is no memory cell therein. Global IO line pair GIO is placed in word line shunt region WLST. Global IO line pair GIO is constituted of the second level aluminum interconnection layer or the third level aluminum interconnection layer. Global IO line pair is electrically connected to local IO line pair LIO extending in the row direction via a block selection gate (not shown). Local IO line pair LIO is located in the first level aluminum interconnection layer (1A1).

[0233] A sense amplifier power supply line APa supplying power supply voltage to sense amplifier SA extends in the row direction, and is constituted of the first level aluminum interconnection layer.

[0234] Column selection line CSL constituted of the second level aluminum interconnection layer (2A1) is arranged in the column direction. A column selection signal on column selection line CSL connects sense amplifier circuit SA electrically to local IO line pair LIO via IO gate (not shown). Sense amplifier power supply line APb formed of the second level aluminum interconnection layer is arranged extending over the memory array in the column direction, and sense amplifier power supply line APc formed of the second level aluminum interconnection layer is arranged in word line shunt region WLST in the column direction. Sense amplifier power supply lines APb and APc formed of the second level aluminum interconnection layer are electrically connected to sense amplifier power supply line APa formed of the first level aluminum interconnection layer in contact regions CTb and CTc.

[0235] Sub power supply line SV formed of the third level aluminum interconnection layer is arranged over the memory array in the column direction. Sense amplifier circuit SA is connected to bit line BL and memory cell MS is arranged corresponding to a crossing of bit line BL and word line WL.

[0236] Since sense amplifier power supply line APa extending in the row direction is formed of the first level aluminum interconnection layer, collision between sense amplifier power supply line APa and column selection line CSL formed of the second level aluminum interconnection layer can be prevented. Further, collision between sense amplifier power supply line APa and global IO line pair GIO can be prevented even when global IO line pair GIO is formed of the second level aluminum interconnection layer. The first level aluminum interconnection layer has a resistivity a little higher than that of the second level aluminum interconnection layer, however, the resistance value of sense amplifier power supply line APa formed of the first level aluminum interconnection layer is reduced by electrically connecting sense amplifier power supply lines APc and APb formed of the second level aluminum interconnection layer at certain intervals.

[0237] Further, by electrically connecting sense amplifier power supply lines APb and APc formed of the second level aluminum interconnection layer to sense amplifier power supply line APa at a prescribed interval, reduction of the resistance value of the sense amplifier power supply line APa as well as increase of an ability of supplying current are accomplished. Variation of power supply voltage of sense amplifier power supply line APa is thus suppressed to implement a stable sensing operation. Sub power supply line SV is formed of the third level aluminum interconnection layer located in a higher layer, so that power supply voltage for the peripheral circuits can be supplied without adversely affecting arrangement of power supply lines in the memory array.

[0238] FIG. 30 schematically shows a modification of the semiconductor integrated circuit device of the ninth embodiment. FIG. 30 only shows an arrangement of main and sub power supply lines. Four memory arrays 54a-54dare placed in FIG. 30. Banks are arbitrarily allocated to memory arrays 54a-54d. Power supply pads PSa, PSb, PSc and PSd are arranged adjacent to memory arrays 54aand 54cin alignment with each other. Power supply pad PSa is coupled to main power supply line MVa arranged extending in one direction over memory arrays 54aand 54b, and power supply pad PSb is connected to main power supply line MVb placed opposite to main power supply line MVa relative to memory arrays 54aand 54b. Main power supply lines MVa and MVb are connected to each other by sub power supply line SV arranged extending over memory arrays 54aand 54b.

[0239] Power supply pad PSc is electrically connected to main power supply line MVc arranged extending over memory arrays 54cand 54d, and power supply pad PSd is electrically connected to main power supply line MVd arranged opposite to main power supply line MVc relative to memory arrays 54cand 54d. Main power supply lines MVc and MVd are coupled to each other by sub power supply line SV arranged extending over memory arrays 54cand 54dat prescribed intervals.

[0240] As shown in FIG. 30, power supply pads PSa-PSd are arranged on only one side of memory arrays 54a-54d, main power supply lines MVa-MVd for supplying power supply voltage from those power supply pads to peripheral circuits and control circuits are provided, and main power supply lines MVa-MVd are reinforced by sub power supply lines SV. If the power supply arrangement of the ladder shape as shown in FIG. 30 is used, an effect similar to that of the configuration in which two power supply pads are provided for each memory array can be obtained. As main power supply lines MVa-MVd and sub power supply line SV, as described regarding the embodiment above, an interconnection layer higher than any interconnection layer formed at the memory array is used (if global IO line is constituted of the second level aluminum interconnection layer).

[0241] In the configuration of FIG. 30, an array power supply circuit for generating the supply voltage for array such as a sense amplifier power supply voltage is placed adjacent to each respective memory array, and the array power supply voltage is generated for a corresponding memory array. In this case, power supply voltage is supplied from an appropriate main power supply line to the array power supply circuit.

[0242] According to the ninth embodiment as described above, power supply lines coupled to power supply pads are configured in the ladder shape by the main power supply lines and the sub power supply lines arranged extending over the memory array, and the power supply lines are utilized as a source for supplying an operation power supply voltage for peripheral circuits such as the preamplifier, the write driver and the bank control circuit. Accordingly, power supply voltage can be stably supplied even under high speed operation, and reading and writing of, for example, 1024 bits data can be executed at high speed without generating power supply noises when data is input/output.

[0243] FIG. 31 shows an entire configuration of a semiconductor integrated circuit device according to the tenth embodiment schematically. Semiconductor integrated circuit device 1 of FIG. 31 includes processor 2 and DRAM macro 3. DRAM macro 3 includes subbanks 4v, 4w, 4xand 4yarranged in four regions respectively. Subbank 4vincludes a memory array 4va, a row decoder 4vb, a column decoder 4vc, a preamplifier 4vdarranged adjacent to column decoder 4vc, and a write driver 4veplaced opposite to column decoder 4vcrelative to preamplifier 4vd.

[0244] Subbank 4wincludes a memory array 4wa, a row decoder 4wb, a column decoder 4wc, a preamplifier 4wdand a write driver 4we. Subbank 4xincludes a memory array 4xa, a row decoder 4xb, a column decoder 4xc, a preamplifier 4xdand a write driver 4xe, and subbank 4yincludes a memory array 4ya, a row decoder 4yb, a column decoder 4yc, a preamplifier 4ydand a write driver 4ye. Subbanks 4v-4yhave the same configuration in which the preamplifier is placed adjacent to the column decoder, and the write driver is arranged opposite to the column decoder relative to the preamplifier. Subbanks 4vand 4waligned in the row direction constitute bank A, and subbanks 4xand 4yconstitute bank B.

[0245] A DRAM control circuit (bank control circuit) 7 extending in the column direction is arranged in a central region with respect to the row direction of DRAM macro 3. DRAM control circuit 7 includes a bank A control circuit and a bank B control circuit.

[0246] DRAM control circuit 7 is coupled to processor 2 via a control bus 12eextending in the column direction, and receives necessary control signal and address signal.

[0247] An internal read/write data bus 5xextending in the column direction is arranged commonly to subbanks 4vand 4x, and an internal read/write data bus 5yextending in the column direction is arranged commonly to subbanks 4wand 4y. Internal read/write data buses 5xand 5yare arranged extending over respective subbanks 4vand 4w, and formed of, for example, the third level aluminum interconnection layer higher than the interconnection layers in the memory array.

[0248] Internal read/write data bus 5xincludes a read data bus line 5xrcommonly coupled to preamplifier 4vdof subbank 4vand preamplifier 4xdof subbank 4x, and an internal write data bus line 5xwcommonly coupled to write driver 4veof subbank 4vand write driver 4xeof subbank 4x. Internal read/write data bus 5yincludes an internal read data bus line 5yrcommonly coupled to preamplifier 4wdof subbank 4wand preamplifier 4ydof subbank 4y, and an internal write data bus line 5ywcommonly connected to write driver 4weof subbank 4wand write driver 4yeof subbank 4y. These read data buses and write data buses are separately provided to reduce the load of the data bus for transferring data at high speed.

[0249] FIG. 32 schematically shows an arrangement of the interconnection lines in the memory array portion. In FIG. 32, the uppermost interconnection layer in the memory array is the second level aluminum interconnection layer. The second level aluminum interconnection layer includes a conductive layer 80 forming global IO line (GIO), a conductive layer 81 forming array power supply line (Vcc/Vss), and a conductive layer 82 forming column selection line (CSL) which transmits a column selection signal from the column decoder.

[0250] Conductive layers 83aand 83bconstituting bus lines (R/W) of internal read/write data buses 5xand 5yare arranged in a layer higher than the second level aluminum interconnection layer. The third level aluminum interconnection layer is an interconnection layer higher than the uppermost interconnection layer in the memory array. Accordingly, the internal read/write data buses can be arranged extending over the memory array, the column decoder, the preamplifier and the write driver without affecting the layout of the subbank (Transistors constituting the column decoder, the preamplifier and the write driver are formed of an interconnection layer lower than the second level aluminum interconnection layer).

[0251] As shown in FIG. 31, since the read/write data bus is arranged utilizing an interconnection layer higher than the uppermost interconnection layer of the memory array, such as the third level aluminum interconnection layer, the internal read/write data buses can be arranged over the subbank. Therefore, a region for arranging the internal read/write data buses is unnecessary viewed in a planar layout to dramatically reduce an area occupied by the DRAM macro.

[0252] In the arrangement of the array shown in FIG. 32, global IO line pair may be placed in the third level aluminum interconnection layer as shown by the block surrounded by the broken line in FIG. 32.

[0253] The processor includes data input buffer and data output buffer provided separatedly. By providing read data bus lines 5xrand 5yrand write data bus lines 5xwand 5ywseparately for internal read/write data buses 5xand 5y, the data buses can be easily arranged according to the arrangement of the input buffer and the output buffer of the processor. In this case, although the number of necessary bus lines is doubled compared with the case in which the read data bus line and the write data bus line are made common, the number of memory cells in one row is sufficiently larger than the number of memory cells (columns) that are simultaneously selected and have data written/read in one memory array as clearly shown by the array arrangement of FIG. 22. Accordingly, even if the number of read/write data bus lines is large, those can be arranged over the memory array with a sufficient margin.

[0254] The write driver and the preamplifier are arranged facing to a central region in the column direction of DRAM macro 3 as shown in FIG. 31. Subbanks aligned in the row direction are included in the same bank. Since writing/reading of data is started from a region adjacent to the central region in the column direction of DRAM macro 3 in both of the access to bank A and access to bank B, and since the read/write data buses are formed of the third level aluminum interconnection layer having a superior electric characteristics, delay on interconnection lines can be substantially ignored. As a result, access times in both of the bank A access and the bank B access are made equal and a DRAM operating at high speed can be obtained.

[0255] By providing the read data bus and the write data bus separately, load of each data bus line decreases and high speed access is accomplished. The internal read/write data bus may be constituted of read/write data bus lines which transmits both of write data and read data. The positions of the preamplifier and the write driver may be exchanged in each of subbanks 4v-4y. Further, the preamplifier may be arranged opposite to the column decoder relative to the memory array in each of subbanks 4v-4y(IO separation structure).

[0256] According to the tenth embodiment as described above, since the internal read/write data bus is configured utilizing an interconnection layer higher than the interconnection layers of the subbank, the area for interconnection of internal read/write data bus can be effectively eliminated. An area occupied by the DRAM macro is thus reduced and accordingly a chip occupation area of a semiconductor integrated circuit device is reduced. If an aluminum interconnection layer is employed as the uppermost interconnection layer, the uppermost aluminum interconnection layer has the highest aluminum purity and is most superior in electric characteristics. Accordingly, data can be transferred at high speed and a semiconductor integrated circuit device operating at high speed can be implemented.

[0257] FIG. 33 schematically shows an entire configuration of a semiconductor integrated circuit device of the eleventh embodiment. Semiconductor integrated circuit device 1 shown in FIG. 33 is different from the semiconductor integrated circuit device of FIG. 31 in the following point. Specifically, a testing circuit 100 is placed between processor 2 and DRAM macro 3. Testing circuit 100 is coupled to internal read/write data buses 5xand 5yas well as a control bus 12econnecting processor 2 and DRAM macro 3. Accordingly, internal read/write data buses 5xand 5yand control bus 12eare shared by testing circuit 100 and processor 2. The configuration of testing circuit 100 is approximately similar to that of testing circuit 15 shown in FIGS. 5-9. Semiconductor integrated circuit device 1 of FIG. 33 is identical to that of FIG. 31 except for that testing circuit 100 is provided, and corresponding components are indicated by the same reference characters.

[0258] Since testing circuit 100 is arranged between DRAM macro 3 and processor 2 as shown in FIG. 33, processor 2 and testing circuit 100 can share internal read/write data buses 5xand 5yas well as control bus 12e. Therefore, testing circuit 100 for conducting a function test for DRAM macro 3 can be easily incorporated without adversely affecting the layout of DRAM macro 3 and processor 2.

[0259] In testing circuit 100, the read data bus and the write data bus are separately provided such that internal read/write data bus 5xand 5yinclude read data bus lines 5xrand 5yrand write data bus lines 5xwand 5yw, and the bus selector in the testing circuit shown in FIG. 7 is provided for each of the write data bus and the read data bus.

[0260] FIG. 34 schematically shows an entire configuration of the semiconductor integrated circuit device according to the eleventh embodiment on a semiconductor chip. Semiconductor integrated circuit device 1 includes pads 101a-101iarranged along the periphery thereof. Pads 101a-101iare coupled to processor 2 by an internal bus interconnection line IL. Processor 2 provides and receives data and a control signal to and from an external unit via pads 101a-101i. A specific pad 101hamong these pads 101a-101iis used as a pad for input of DRAM test activation signal TE. Testing circuit 100 is coupled to prescribed pads among pads 101a-101i, electrically couples internal read/write data buses 5xand 5yas well as control bus 12eto the prescribed pads upon activation of DRAM test activation signal TE supplied to pad 101hto enable external supplying and receiving of testing data and control signals.

[0261] FIG. 35 schematically shows a configuration of the testing circuit shown in FIGS. 33 and 34. Testing circuit 100 includes a select/connect circuit 100awhich couples internal nodes to pads 101x-101yaccording to activation of test activation signal TE supplied to pad 101h, and an actual testing circuit 100bwhich inputs and outputs testing data via pads connected by select/connect circuit 100aand conducts a prescribed test. In FIG. 35, as an internal component of select/connect circuit 100a, a transfer gate provided for each of the pads and rendered conductive in response to activation of test activation signal TE is representatively shown. FIG. 35 exemplarily shows that a control signal is supplied to pad 101yin the testing mode of DRAM macro.

[0262] Further, as data bus lines coupled to testing circuit 100, a read data bus line r, a write data bus line w and a control bus line c are representatively shown.

[0263] Read data bus line r is coupled to a processor read data bus line pr connected to processor 2 as well as a test read data bus line tr connected to actual testing circuit 100bat a node 103r. Write data bus line w is coupled to a processor write data bus line pw which transmits write data from processor 2 as well as a test write data bus line tw which transmits testing data from actual testing circuit 100bat a node 103w. Control bus line c is coupled to a processor control bus line pc which transmits a control signal from processor 2 as well as a test control bus line tc which transmits a test control signal from testing circuit 100 at a node 103c. Buffer circuits are provided for those bus lines respectively for waveform shaping.

[0264] Processor 2 includes an input buffer 2awhich buffers a signal on processor read data bus line pr, and output buffers 2band 2cfor driving processor write data bus line pw and processor control bus line pc respectively. Actual testing circuit 100bincludes an output buffer 100bbfor transmitting testing data onto the test write data bus line tw, and a buffer 100bafor buffering testing data supplied from the DRAM macro or transmitting data of the test result to pad 101xor the like.

[0265] Test control bus line tc is provided with a buffer 100cfor buffering a control signal supplied from an external testing unit via pad 1l0y.

[0266] In the arrangement shown in FIG. 35, testing circuit 100 and the bus lines from processor 2 are coupled at nodes 103r, 103wand 103c. Read data bus line r is coupled to actual testing circuit 100band processor 2 via bus lines tr and pr respectively. Therefore, collision of read data does not occur in testing circuit 100 and processor 2. If processor 2 is brought in a non-operation state by a control signal applied to a non-used pad in testing mode of the DRAM, an adverse effect of a signal necessary for testing circuit 100 in the testing mode exerted on processor 2 can be prevented, and contrarily, an adverse effect of processor 2 on test result of the DRAM macro can be prevented. In the normal operation mode, testing circuit 100 is in the non-operation state. If the selector is in a nonconductive state as shown in FIGS. 7 and 8, a reading portion of the testing data of testing circuit 100 is separated from processor 2 and the reading portion does not exert any negative influence.

[0267] In the normal operation mode and the testing mode for the DRAM macro, collision of writing data on write data bus line w and control bus line c is generated by buffers 100bb, 100c, 2band 2cand an adverse influence may caused in both operation modes. Therefore, testing circuit 100 should be separated from the write data bus and the control bus in the normal operation mode and processor 2 should be separated from the write data bus and the control data bus in the testing mode for the DRAM macro.

[0268] In this case, when DRAM testing mode activation signal TE is activated, buffers 2band 2ccould be brought into output high impedance state in processor 2, and buffers 100bband 100ccould be brought into the output high impedance state when testing mode activation signal TE is inactivated. However, this approach is not preferable since the output buffer of processor 2 should be comprised of a tristate buffer for implementation of testing of the DRAM macro and therefore a signal irrelevant to processor's inherent operation mode should be internally used.

[0269] If the interconnection line capacitance of DRAM macro 3 and the capacitance of interconnection line passing over testing circuit 100 should be driven at high speed, driving power of output buffers 2band 2cof processor 2 as well as output buffers 100bband 100cof testing circuit 100 should be increased. In this case, tristate buffers for the testing circuit and for the processor might be provided at nodes 103r, 103wand 103c.

[0270] FIG. 36 shows a configuration of a conventional tristate inverter buffer. The tristate inverter buffer of FIG. 36 includes p channel MOS transistors PQ1 and PQ2 and n channel MOS transistors NQ1 and NQ2 connected in series between power supply node Vcc and ground node Vss. DRAM test activation signal TE is supplied to the gate of MOS transistor PQ1 via an inverter, and is also supplied to the gate of MOS transistor NQ2. Input signal IN is supplied to the gates of MOS transistors PQ2 and NQ1, and output signal OUT is output from a connection node between MOS transistors NQ1 and PQ2.

[0271] If such tristate inverter buffer is used, output signal OUT is charged or discharged via two MOS transistors connected in series to each other. Accordingly, channel resistance of the MOS transistors increases and high speed charging and discharging becomes impossible. Further, since four MOS transistors are connected in series, the area occupied by those increases. Since diffusion capacitance of two MOS transistors (junction capacitance between a substrate and an impurity region) is connected to a signal line in operation, interconnection line capacitance increases and high speed charging and discharging becomes impossible. If the state of input signal IN is fixed at an H level or an L level in non-activation state, one of MOS transistors PQ2 and NQ1 is in a conductive state. For example, if MOS transistor NQ1 is in an ON state, diffusion capacitances of the source and the drain of MOS transistor NQ1 and diffusion capacitance of the drain of the MOS transistor NQ2 are connected to an output node. Diffusion capacitance of the drain of MOS transistor PQ2 is further connected to the output node. Accordingly, if such a tristate inverter buffer is utilized, parasitic capacitance of the output node increases and high speed transmission of a signal becomes impossible. This is because the parasitic capacitances of both of the tristate inverter buffer in the active state and the tristate inverter buffer in the inactive state are connected to the write data bus line or the control bus line.

[0272] In order to transmit a signal at high speed to the DRAM macro without increase of an area occupied by a circuit, following configuration is employed.

[0273] FIG. 37 shows a configuration of a modification of the eleventh embodiment. In FIG. 37, a selection gate GPW which is rendered non-conductive upon activation of DRAM test activation signal TE is arranged between node 103wand processor write data bus line pw, a selection gate GTW which is rendered conductive upon activation of DRAM test activation signal TE is arranged between node 103wand test write data bus line tw, a selection gate GPC which is rendered non-conductive upon activation of DRAM test activation signal TE is provided between node 103cand processor control bus line pc, and a selection gate GCW which is rendered conductive upon activation of DRAM test activation signal TE is provided between node 103cand test control bus line tc. These selection gates GPW, GTW, GPC and GCW are each constituted of a CMOS transmission gate.

[0274] Drivers DRW and DWC having a large driving capability are arranged at nodes 103wand 103c. Since selection gates GPW, GTW, GPC and GCW are each constituted of the CMOS transmission gate, they can transmit signals to be transferred with low impedance. Further, corresponding signal lines and source/drain regions can be arranged in parallel with each other so that these selection gates each can be arranged in a smaller area compared with the tristate inverter buffer. Further, the pitch of the gates (distance between adjacent gates) can be reduced sufficiently compared with the tristate inverter buffer. Parasitic capacitance associated with the output node (nodes 103wor 103c) of the CMOS transmission gate in nonconductive state is only the junction capacitance at one side of the CMOS transmission gate in the non-conductive state. Accordingly, the parasitic capacitance can be made sufficiently smaller compared with the case in which the tristate inverter buffer is utilized.

[0275] In addition, MOS transistors are connected in parallel with each other to effectively increase the gate width compared with the MOS transistors connected in series, resulting in high speed transmission of signals. By providing drivers DRW and DRC for write data bus line w and control bus line c, write data can be transmitted at high speed without increasing driving capabilities of output buffers of the processor and the testing circuit, even if interconnection line capacitance of the DRAM macro is large. Write data can be transmitted at high speed in the testing mode and the normal operation mode with a small occupied area.

[0276] Such a selection gate is not provided for read data bus line r. Alternatively, driving capability of the preamplifier of the DRAM macro can be made sufficiently large instead of providing the selection gate, or two stages of preamplifiers (configuration of preamplifier and main amplifier) are provided to drive interconnection line capacitance of DRAM macro and interconnection line capacitance of the read data bus line at high speed with a sufficiently large driving power. However, a selection gate with its conduction/non-conduction controlled responsive to DRAM test activation signal TE may be used for read data bus lines pr and tr.

[0277] FIG. 38 schematically shows an entire configuration of a modification of the eleventh embodiment. In FIG. 38, read data bus PRBx from processor 2 and test read data bus TRBx of testing circuit 100 are coupled to each other to become read data bus 5xrcoupled to DRAM macro 3. Write data bus PWBX from processor 2 and test write data bus TWBx from testing circuit 100 are coupled to selector.driver 110 and one of them is selected to become write data bus 5xwvia the driver. Control bus PCB from processor 2 and control bus TCB from testing circuit 100 are coupled to selector.driver 110. One of those is selected according to an operation mode to be coupled to control bus 12evia the driver. Write data bus PWBy from processor 2 and test write data bus TWBy from testing circuit 100 are both coupled to selector.driver 110. One of those is selected according to an operation mode to be coupled to write data bus 5yW via the driver. Read data bus PRBy from processor 2 is coupled to read data bus TRBy from testing circuit 100 to become read data bus 5yR skipping selector.driver 110.

[0278] As shown in FIG. 38, selector.driver 110 including a selector and a bus driver is used to select write data and control signal from one of testing circuit 100 and processor 2 for transmission to DRAM macro 3 via the driver. Accordingly, even if interconnection line capacitance within the DRAM macro and interconnection line capacitance on interconnection lines passing over testing circuit 100 are increased, write data buses 5xW and 5yW and control bus 12ecan be driven at high speed by the driver and the DRAM macro can be driven at high speed in both of the normal operation mode and the DRAM testing mode.

[0279] According to the eleventh embodiment as described above, the testing circuit is arranged between the processor and the DRAM macro, and the read data bus, the write data bus, and the control bus to the DRAM macro are shared between the processor 2 and the testing circuit. As a result, the testing circuit for the DRAM macro can be easily arranged within the semiconductor integrated circuit device without affecting the layout of the processor and the DRAM macro. Further, the bus of one of the processor and the testing circuit is selected according to operation mode using the selector and the driver, and the write data bus and the control bus for the DRAM macro are driven via the driver. As a result, write data, the control signal and the address signal can be transmitted at high speed without increase of an area occupied by the circuit, to implement a semiconductor integrated circuit device operating at high speed.

[0280] FIG. 39 schematically shows an entire configuration of a semiconductor integrated circuit device according to the twelfth embodiment of the invention. Referring to FIG. 39, a semiconductor integrated circuit device (semiconductor memory device) 3 which is a DRAM macro is divided into four regions by a first central region 201 arranged extending in a column direction at a central region with respect to a row direction of a rectangular region of the device, and a second central region 203 arranged extending in the row direction in a central region with respect to the column direction of the rectangular region. Memory arrays 204aa, 204ba, 204caand 204daare respectively arranged in the four regions. Each of memory arrays 204aa-204daincludes a plurality of memory cells arranged in rows and columns.

[0281] Row-related array circuits 204ab-204dbperforming an operation related to row selection of a corresponding memory array upon activation thereof, and column-related array circuits 204ac-204dcperforming an operation related to column selection (including data writing/reading) of a corresponding memory array upon activation thereof are provided to memory arrays 204aa-20darespectively. Row-related array circuits 204ab-204dbare arranged facing to the first central region 201, and column-related array circuits 204ac-204dcare arranged facing to the second central regions 203.

[0282] An interconnection-dedicated region 210ais arranged between row-related array circuits 204aband 204bbthat are arranged opposite to each other with respect to the first central region 201, and an interconnection-dedicated region 210bis arranged in the first central region 201 between row-related array circuits 204cband 204db. An interconnection line for transmitting a signal voltage as well as a decoupling capacitor for suppressing noises are arranged in these interconnection-dedicated regions 210aand 210b(the configuration is described later in detail). A logic circuit performing a prescribed processing on a signal is not provided in interconnection-dedicated regions 210aand 210b.

[0283] In the second central region 203, peripheral circuitry such as a control circuit for controlling operations of row-related array circuits 204ab-204dband column-related array circuits 204ac-204dc, and an internal voltage generation circuit generating internal voltage (reference voltage such as internal power supply voltage, internal negative voltage, internal high voltage and the like) is placed. Peripheral circuitry placing region 205 is arranged in an almost entire portion of the second central region 203. Further, a region between column-related array circuits 204acand 204bcas well as a region between column-related array circuits 204ccand 204dcare used as peripheral circuitry placing region 205. No bonding pad for externally supplying/receiving a signal and voltage is placed in interconnection-dedicated regions 210aand 210b.

[0284] FIG. 40 schematically shows a configuration of the row-related array circuit shown in FIG. 39. Referring to FIG. 40, a configuration of a row-related array circuit provided corresponding to one memory array is schematically shown. Memory arrays (204aa-204da) are each divided into a plurality of array blocks 215a-215min the column direction. Between these array blocks 215a-215m, sense amplifier bands 220b-220mincluding sense amplifiers placed corresponding to respective memory cell columns are provided. On the outside of array blocks 215aand 215m, sense amplifier bands 220aand 220nare further provided. Sense amplifier bands 220b-220mare each shared by adjacent array blocks.

[0285] Row decoders.drivers 222a-222mfor decoding an address signal (not shown) and driving an addressed word line WL to a selected state are provided corresponding to array blocks 215a-215mrespectively. Each of row decoders.drivers 222a-222mincludes a decode circuit provided corresponding to each word line and decoding an address signal, and a word line drive circuit for driving a corresponding word line to a selected state according to an output signal from the decode circuit.

[0286] Sense control circuits 224a-224nare provided corresponding to sense amplifier bands 220a-220nrespectively. Sense control circuits 224a-224neach include a bit line isolation control circuit for controlling connection/isolation of each column of an adjacent array block and a sense amplifier of a corresponding sense amplifier band, as well as a sense amplifier activation control circuit for activating a sense amplifier included in a corresponding sense amplifier band.

[0287] As shown in FIG. 40, in each of row-related array circuits 204ab-204db, the same circuit is repeatedly arranged in the column direction correspondingly to an array block. A storage capacity of a memory array is increased by repeatedly placing sense amplifier band 220 and array block 215 as one unit in the column direction. Correspondingly, sense control circuit 224 and row decoder.driver 222 are repeatedly placed.

[0288] FIG. 41 illustrates one example of a configuration of row decoders.drivers 222a-222m. In FIG. 41, row decode/drive circuits 230aand 230bprovided corresponding to two word lines WLa and WLb are representatively shown. In FIG. 41, row decode/drive circuit 230aincludes an NAND circuit 230aareceiving an internal row address signal Ada and an inverter 230abinverting a signal output from NAND circuit 230aafor transmission onto word line WLa. Row decode/drive circuit 230bincludes an NAND circuit 230bareceiving a row address signal Adb and an inverter 230bbinverting a signal output from NAND circuit 230bafor transmission onto word line WLb. Row decode/drive circuits 230aand 230bare each constituted of the NAND circuit and the inverter. Circuits 230aand 230bmerely differ from each other in the point of receiving different address signals. Therefore, the same circuit pattern is repeatedly arranged in the column direction in each of row decoders.drivers 222a-222m.

[0289] NAND circuits 230aaand 230baeach are an NAND type decode circuit, and inverters 230aband 230bbeach are an inverter type word line drive circuit. As for a configuration of the word line drive circuit, a configuration for driving a corresponding word line to a selected state according to a predecode signal or any other configuration may be employed. In any case, each row decode/drive circuit has the same circuit configuration and the same layout pattern.

[0290] FIG. 42 shows one example of a configuration of each of sense control circuits 224b-224millustrated in FIG. 40. FIG. 42 representatively shows a configuration of a sense control circuit 224 for sense amplifier band 220jplaced corresponding to array blocks 215iand 215j. Between sense amplifier band 220jand array block 215i, a bit line isolation gate 219jafor connection/isolation of a bit line pair of array block 215iand a sense amplifier included in sense amplifier band 220jis provided. A bit line isolation gate 219jbis arranged between sense amplifier band 220jand array block 215i. Bit line isolation gates 219jaand 219jbhave the same circuit configuration.

[0291] Sense control circuit 224 (224b-224m) includes an OR circuit 232areceiving array block designation signals BSi and BSj, an AND circuit 232breceiving a signal output from OR circuit 232aand main sense amplifier activation signal MSA to output local sense amplifier activation signal LSA to sense amplifier band 220j, an NAND circuit 232creceiving internal row address strobe signal RAS and array block designation signal BSj to output bit line isolation instruction signal BLIa to bit line isolation gate 219ja, and an NAND circuit 232dreceiving internal row address strobe signal RAS and array block designation signal BSi to output bit line isolation instruction signal BLIb to bit line isolation gate 219jb.

[0292] Array block designation signals BSi or BSj is driven to an H level of a selected state when array blocks 215ior 215jincludes selected word lines. Internal row address strobe signal RAS attains the H level at the start of a memory cycle, and is at an L level in a standby state. Main sense amplifier activation signal MSA is driven into an activation state of the H level at a prescribed timing after the memory cycle starts.

[0293] According to the configuration of sense control circuit 224, in the standby state, internal row address strobe signal RAS is at the L level, bit line isolation instruction signals BLIa and BLIb output by NAND circuits 232cand 232dare at the H level, so that a sense amplifier included in sense amplifier band 220jis connected to a corresponding bit line pair in each of array blocks 215iand 215j.

[0294] When the memory cycle starts, internal row address strobe signal RAS rises to the H level. When array block 215iincludes a selected word line, array block designation signal BSi is driven to the H level, bit line isolation instruction signal BLIb goes down to the L level, bit line isolation gate 219jbis set to a non-conductive state, and array block 215jis isolated from sense amplifier band 220j. On the other hand, array block designation signal BSj is at the L level, bit line isolation instruction signal BLa maintains the H level, bit line isolation gate 219jamaintains a conductive state, and array block 215iis connected to sense amplifier band 220j.

[0295] According to the configuration of sense control circuit 224, gate circuits having the same configuration are arranged for respective sense amplifier bands and bit line isolation gates. The only difference is a supplied array block designation signal. Accordingly, in row-related array circuits 204ab-204dbshown in FIG. 39, components having the same circuit pattern (or circuits having the same configuration) are repeatedly arranged in the column direction. If the number of array blocks of the memory array in the configuration shown in FIG. 40 is increased, the sense amplifier band, the sense control circuit and the row decoder.driver are repeatedly arranged in the column direction. In this case, interconnection-dedicated regions 210aand 210bare arranged between row-related array circuits 204aband 204bb, and between row-related array circuits 204cband 204db. Only an interconnection line and a decoupling capacitor are placed in each of interconnection-dedicated regions 210aand 210b, and no control circuit for internal circuit control is arranged therein. Therefore, the storage capacity of the memory array can be easily increased just by extending the signal interconnection line in the column direction.

[0296] FIG. 43 schematically shows a configuration of an interconnection-dedicated region. Referring to FIG. 43, a signal line group SGL formed of signal lines transmitting signals, and a reference voltage transmission line RVL transmitting prescribed voltage are arranged in interconnection region 210 (210a, 210b). A decoupling capacitor DCP for reducing noises is connected to reference voltage transmission line RVL. Signal line group SGL includes signal lines transmitting an address signal and a control signal. Reference voltage transmission line RVL includes transmission lines transmitting a power supply voltage, ground voltage, an internal high voltage higher than the supply voltage, an intermediate voltage between the supply voltage and the ground voltage, and a negative voltage. No circuit such as a logic circuit carrying out a prescribed processing is provided in interconnection-dedicated region 210. A buffer circuit for buffering a signal may be provided correspondingly to a signal line of signal line group SGL in interconnection-dedicated region 210. Such a buffer circuit just performs waveform shaping of a supplied signal and does not perform any logical processing, so that the buffer circuit does not affect other circuit portion. The buffer circuit is provided just for compensating for attenuation of a signal and may be placed in the interconnection-dedicated region similarly to decoupling capacitor DCP.

[0297] In other words, no circuit which processes a supplied signal to affect an operation of any other circuit as a result of the process is provided in the interconnection-dedicated region. In addition, no bonding pad for an external electrical connection is placed in the interconnection-dedicated region 210. If the DRAM is employed as a DRAM macro, the DRAM is integrated formed with other functional blocks (macro) on a semiconductor chip. A bonding pad is appropriately arranged in a suitable region of the semiconductor chip as a substrate, according to a scale of a circuit device implemented by the semiconductor chip.

[0298] FIGS. 44A and 44B illustrate one example of the way in which storage capacity of a semiconductor memory device is changed. In FIG. 44A, four memory arrays 204aa-204daare arranged in semiconductor integrated circuit device (DRAM macro) 3. Each of memory arrays 204aa-204dais divided into eight array blocks #0-#5, #a and #b. Suppose that the storage capacity is decreased by removing two array blocks #a and #b in each of memory arrays 204aa-204da.

[0299] Column-related array circuits 204ac-204dcare respectively provided commonly to array blocks of corresponding memory arrays 204aa-204da. Row-related array circuits 204ab-204dbeach has circuit portions provided corresponding to respective array blocks as illustrated in FIGS. 41 and 42. Interconnection-dedicated region 210ais arranged between row-related array circuits 204aband 204cb, and interconnection-dedicated region 210bis arranged between row-related array circuits 204bband 204db. Peripheral circuitry placing region 205 is arranged in a central region of integrated circuit device 3. If array blocks #a and #b are just removed from each of memory arrays 204aa-204da, circuit portions provided corresponding to these array blocks #a and #b are removed from each of row-related array circuits 204ab-204db. Only the interconnection lines and the decoupling capacitor are arranged in each of interconnection-dedicated regions 210aand 210b, and there is no need to change a circuit arrangement according to elimination of array blocks #a and #b. In this case, the length of interconnection lines is just decreased.

[0300] If memory arrays 204aa-204daeach includes array blocks #0-#5 as shown in FIG. 44B, the layouts of column-related array circuits 204ac-204dcare not changed. Further, peripheral circuitry arranged in peripheral circuitry placing region 205 is not required to change its arrangement. In this case, just the area of interconnection-dedicated regions 210aand 210bis reduced according to reduction of memory arrays 204aa-204dain the column direction. In row-related array circuits 204ab-204db, circuit portions having the same circuit pattern are removed and there is no need to change the arrangement of an internal circuit portion provided corresponding to each array block. If the storage capacity is to be decreased as shown in FIG. 44A and FIG. 44B, only these circuit portions having a periodic circuit pattern of row-related array circuits 204ab-204db, each of which is a repeated pattern circuit where a circuit pattern is repeatedly arranged in the column direction, are eliminated, in order to decrease the storage capacity, or reduce the number of array blocks.

[0301] On the contrary, if the storage capacity is to be increased by changing the semiconductor integrated circuit device shown in FIG. 44B to the one shown in FIG. 44A, just some array blocks and basic circuit blocks each having a repeated pattern in the row-related array circuit are added. In other words, the same circuit pattern is again repeatedly arranged to easily achieve increase in the storage capacity.

[0302] As heretofore described, an interconnection-dedicated region is arranged in a region between row-related array circuits placed facing to each other, and the interconnection lines and the decoupling capacitor are exclusively arranged in the interconnection-dedicated region. Consequently, there is no circuit portion having its position changed according to increase/decrease of a capacity of a memory array in the interconnection-dedicated region, so that the storage capacity can be easily increased/decreased. In addition, peripheral circuitry having various circuit patterns is arranged in peripheral circuitry placing region 205. Consequently, the position at which the peripheral circuitry is arranged in peripheral circuitry placing region 205 is not required to be changed according to increase/decrease in the capacity of the memory array. Accordingly, a highly universal DRAM macro can be provided in which increase/decrease in the storage capacity is easily achieved by implementing an optimum arrangement of peripheral circuitry in peripheral circuitry placing region 205.

[0303] Where the storage capacity is increased/decreased by increasing/decreasing the size of the memory array in the row direction, an additional configuration is required for selecting an internal read/write circuit in order to fix the number of data bits selected by the column-related array circuit if the number of input/output data bits is fixed. As a result, change of arrangement of circuits is necessary in the peripheral circuitry placing region. The number of data bits selected by the column-related array circuit can be made unchanged by increasing/decreasing the size of the memory array in the column direction (suppose that the number of array blocks simultaneously driven into the selected state is the same). Accordingly, any configuration for modifying the number of input/output data bits is unnecessary in the peripheral circuitry replacing region, so that increase/decrease in the storage capacity can be easily achieved.

[0304] FIG. 45 schematically shows a configuration of the DRAM macro according to the twelfth embodiment of the invention, wherein the DRAM macro is applied to the semiconductor integrated circuit device (DRAM mixed with logic) shown in FIG. 31. In FIG. 45, components corresponding to those configuration of FIG. 31 have the same reference characters and detailed description of those components is not repeated here. According to the configuration illustrated in FIG. 45, an interconnection-dedicated region 310ais arranged between row decoders 4vband 4wbprovided for banks 4vand 4w, and an interconnection-dedicated region 310bis arranged between row decoders 4xband 4ybfor banks 4xand 4y. In a peripheral circuitry replacing region 305, a DRAM control circuit (bank control circuit) 7 is placed in the central region between banks 4v, 4w, 4xand 4y. A circuit which generates an internal high voltage, a negative voltage and the like is also arranged in peripheral circuitry placing region 305.

[0305] A bus 12eincluding an address signal line and a control signal line from a CPU 2 is arranged extending over interconnection-dedicated region 310a, and connected to DRAM control circuit (bank control circuit) 7. Each signal line of address/control signal bus 12eis formed at, for example, a third level aluminum interconnection layer higher than the layer where signal interconnection lines 4vxand 4wxplaced between DRAM control circuit (bank control circuit) 7 and row decoders 4vband 4wbare formed. Consequently, signal interconnection lines for CPU 2 and internal signal interconnection lines for row decoders 4vband 4wband the like are never made complicated in interconnection-dedicated region 310a, and signal lines can be easily arranged.

[0306] If the storage capacity of semiconductor integrated circuit device 3 which is a DRAM macro is to be increased in the DRAM mixed with logic as shown in FIG. 45, the size of semiconductor integrated circuit device 3 is increased or decreased only in the column direction and the arrangement of the peripheral circuitry including DRAM control circuit (bank control circuit) 7 placed in peripheral circuitry placing region 305 is not changed. Accordingly, a DRAM mixed with logic having an optimum storage capacity can be easily implemented using a DRAM macro registered as a library.

[0307] In FIG. 45, row decoders 4vb, 4wb, 4xband 4ybare shown as row-related array circuits. Memory arrays 4va, 4wa, 4xaand 4yaare each divided into a plurality of array blocks as shown in FIG. 44. A sense amplifier control circuit, a bit line isolation control circuit and the like are arranged corresponding to each array block in a region where the row decoders are placed.

[0308] FIG. 46 schematically shows a configuration of the semiconductor integrated circuit device shown in FIG. 33 with the twelfth embodiment of the invention applied thereto. In a semiconductor integrated circuit device 1 illustrated in FIG. 46, a testing circuit 100 is placed between CPU 2 and DRAM macro 3. The configuration of DRAM macro 3 is identical to that shown in FIG. 45. In semiconductor integrated circuit device (DRAM macro) 3 shown in FIG. 46, interconnection-dedicated regions 310aand 310bare arranged between row decoders of respective banks, and DRAM control circuit (bank control circuit) 7 is arranged in remaining peripheral circuitry placing region 305.

[0309] If such a testing circuit 100 is contained, bus 12efrom CPU 2 is connected to testing circuit 100 and thereafter connected to DRAM control circuit (bank control circuit) 7 via interconnection-dedicated region 310a. Therefore, in semiconductor integrated circuit device 1 containing such a testing circuit 100, the storage capacity of DRAM macro 3 can be increased or decreased by increasing or decreasing the size of the memory array in the column direction utilizing DRAM macro 3 according to the twelfth embodiment of the invention.

[0310] FIG. 47 schematically shows a configuration of a modification of the twelfth embodiment of the invention. In FIG. 47, a difference between a shown semiconductor integrated circuit device 3 and the semiconductor integrated circuit device shown in FIG. 39 is as follows.

[0311] Specifically, an interconnection-dedicated region 410ais arranged between column-related array circuits 204acand 204ccplaced opposite to each other with respect to the second central region 203, and an interconnection-dedicated region 410bis arranged between column-related array circuits 204bcand 204dc. A peripheral circuitry replacing region is provided in the first central region 201. Interconnection lines and a decoupling capacitor are formed in interconnection-dedicated regions 410aand 410beach as described above, and no circuit for performing a logical processing operation is placed therein.

[0312] FIG. 48 schematically shows a configuration of column-related array circuits 204ac-204dc. FIG. 48 representatively illustrates a configuration of one column-related array circuit 204. In FIG. 48, a memory array MAA (204aa-204da) is divided into a plurality of column blocks CB#0-CB#n. Each of column blocks CB#0-CB#n share a row. Column decoder blocks CD#0-CD#n, preamplifier blocks PA#0-PA#n, and write driver blocks WD#0-WD#n are provided corresponding to respective column blocks CB#0-CB#n. Preamplifier blocks PA#0-PA#n and write driver blocks WD#0-WD#n are respectively connected to selected columns of corresponding column blocks CB#0-CB#n via internal data buses (global data buses) GB#0-GB#n. Each of preamplifier blocks PA#0-PA#n includes preamplifier circuits of the same number as that of memory cells selected simultaneously in a corresponding column block. Each of write driver blocks WD#0-WD#n also has write drive circuits of the same number as that of memory cells simultaneously selected in a corresponding column block.

[0313] Column decoder blocks CD#0-CD#n include the same components, and select a prescribed number of columns in corresponding column blocks CB#0-CB#n.

[0314] FIG. 49 schematically shows a configuration of a column-related array circuit corresponding to one column block. In FIG. 49, a column decoder block CD# (CD#0-CD#n) includes column decode circuits CDK0-CDKx which decode address signals (not shown) and drive corresponding column selection signal lines CSL0-CSLx into a selected state. Column selection signal lines CSL0-CSLx are arranged extending in the column direction in a corresponding column block. In column decoder block CD#, AND type decode circuits are just arranged corresponding to respective column selection signal lines, and the same circuit pattern is repeatedly placed in the row direction.

[0315] A preamplifier block PA# includes preamplifier circuits PAK0-PAKy provided corresponding to respective bus lines of a global data bus GB# (GB#0-GB#n). Preamplifier circuits PAK0-PAKy are activated in response to activation of preamplifier activation signal PAE (not shown), and amplify data read on global data bus GB#. Preamplifier circuits PAK0-PAKy also have the same circuit configuration and the same circuit pattern is arranged in the row direction.

[0316] A write driver block WD# includes write drive circuits WDK0-WDKy placed corresponding to respective bus lines of global data bus GB# and transmitting internal write data onto corresponding global data bus lines. Write drive circuits WDK0-WDKy are each activated in response to activation of a write drive activation signal (not shown) to execute a data writing operation. These write drive circuits WDK0-WDKy in write driver block WD# have the same circuit configuration, and the same circuit pattern is repeatedly arranged in the row direction.

[0317] Accordingly, in the column-related array circuit, the same circuit pattern is repeatedly arranged in the row direction.

[0318] FIG. 50A schematically shows a configuration of one memory array portion. The memory array shown in FIG. 50A corresponds, for example, to memory array 4vaor 4washown in FIG. 45. The memory array is divided into eight column blocks CB#0-CB#7. A row-related array circuit 412 is commonly provided to column blocks CB#0-CB#7. A column-related array circuit 414 having circuit blocks respectively corresponding to column blocks CB#0-CB#7 is provided to column blocks CB#0-CB#7. A read/write data bus 415 which externally supplies and receives data is arranged in the column direction over the memory array as shown in the configuration of FIG. 31. Regarding read/write data bus 415, a read data bus transferring read data and a write data bus transferring write data may be placed separately. Read/write data bus 415 includes a read/write bus RWB# provided to each of column blocks CB#0-CB#7. Suppose that read/write bus RWB# has a 16-bit width. In this case, 128-bit data is simultaneously transferred in a single memory array.

[0319] Suppose that the bit width of read/write data bus 415 is reduced from 128 bits to 64 bits. In this case, the number of column blocks is reduced by half as shown in FIG. 50B. Specifically, column blocks CB#4-CB#7 are removed, leaving column blocks CB#0-CB#3 only. In column-related array circuit 414, circuit blocks corresponding to removed column blocks CB#4-CB#7 are also removed. The configuration of row-related array circuit 412 is not changed. In this case, the region adjacent to column-related array circuit 414 is the interconnection-dedicated region as shown in FIG. 47, and any peripheral circuitry having an irregular circuit pattern is not arranged in that region. Therefore, elimination of column blocks and reduction of column-related array circuit 414 can be easily achieved to change the data bus width.

[0320] FIG. 51 shows a configuration of the semiconductor integrated circuit device shown in FIG. 31 with the modification of the twelfth embodiment of the invention applied thereto. In FIG. 51, interconnection-dedicated region 410ais arranged in a region between write drivers 4veand 4xeincluded in column-related array circuits of banks 4vand 4xarranged facing to each other in the column direction. Further, interconnection-dedicated region 410bis placed in a region between write driver 4weand 4yeof banks 4wand 4yplaced facing to each other.

[0321] DRAM control circuit (bank control circuit) 7 is arranged in peripheral circuitry replacing region 405 arranged extending in the column direction.

[0322] In the case of the DRAM mixed with logic shown in FIG. 51, data buses 5xand 5yfrom CPU 2 is arranged extending over memory arrays 4vaand 4waof banks 4vand 4w. When the bus widths of data buses 5xand 5yare changed, the column-related array circuit including the column decoder, the preamplifier and the write driver can be reduced/extended in the row direction to achieve change of the bus widths of data buses 5xand 5yeasily.

[0323] In the configuration of FIG. 51, signal lines that transmit control signals for DRAM control circuit (bank control circuit) 7 are placed in interconnection-dedicated regions 410aand 410b. Data buses 5xand 5yare formed at the uppermost layer, for example, at the third level aluminum interconnection layer. An internal interconnection layer in interconnection-dedicated regions 410aand 410bis a lower interconnection layer. Therefore, a column-related control signal from DRAM control circuit (bank control circuit) 7 can be transmitted to each bank via interconnection-dedicated region 410aor 410bregardless of the data buses.

[0324] In the case of the configuration shown in FIG. 51, DRAM control circuit (bank control circuit) 7 can be arranged extending in the column direction. As a result, there is a sufficient area to allow an internal circuit arrangement therein to be optimized and allow peripheral circuitry to be optimally arranged therein.

[0325] The bank described above is constituted of the DRAM. However, the present invention is applicable to a configuration in which data access is made alternatively to two memory arrays coupled commonly to a data bus, and a flash memory or the like can be employed for example. The bank herein described is intended to include a configuration in which activation/inactivation of the array is independently carried out, and a configuration in which data access is made alternatively.

[0326] According to the present invention, a control circuit is provided in a prescribed region among a plurality of memory arrays, and an internal data bus commonly connected to a prescribed number of memory arrays of the plurality of memory arrays is arranged in a region different from the region where the control circuit is placed. As a result, there is no crossing of the control circuit and the internal data bus so that an efficient layout of the control circuit is possible and a semiconductor integrated circuit device with reduced occupation area of a chip can be implemented.

[0327] Although the present invention has been described and illustrated in detail, it is clearly understood that the same is by way of illustration and example only and is not to be taken by way of limitation, the spirit and scope of the present invention being limited only by the terms of the appended claims.