JPH0690072A - Substrate packaging system and electronic system packaging system - Google Patents

Abstract

PURPOSE:To provide a substrate packaging system with an expandibility in two-dimensional and three-dimensional directions by connecting systems consisting of a number of parts closely. CONSTITUTION:A main substrate 5 is fixed to a power supply bar 1 and a horizontal connection substrate 4 is stack-connected so that two main substrates which are adjacent in horizontal direction are crossed for constituting a two-dimensional module and a vertical connection substrate 7 is erected vertically on the surface of each main substrate 5 and the main substrate 5 is piled up above it for constituting a two-dimensional module, thus achieving a flexible expandibility in a three-dimensional direction and connecting the parts of a large-scale electronic equipment with a number of wirings with improved frequency characteristics and a high reliability.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a mounting system for general electronic equipment such as a massively parallel computer which requires high density mounting.

[0002]

2. Description of the Related Art In recent years, with the development of computer systems, high density mounting of electronic devices has been required. Conventionally, as such a mounting method, a mounting method having a spread to a two-dimensional method using a printed circuit board is widely used, and a backplane and a cable are widely used for connecting between the boards.

However, since the size of a printed circuit board that can be manufactured is limited, the number of components that can be mounted on one board is limited by the area of the board, and the number of inter-board connection wirings by backplane is one side of the board. The number of substrates that can be bonded by the backplane is limited by the size of the backplane substrates.

As described above, it has been difficult for the conventional two-dimensional mounting method to tightly couple a large-scale system composed of an extremely large number of components.

On the other hand, a method of stacking substrates for the purpose of improving the number of inter-substrate connecting wirings and a three-dimensional mounting method using optical wiring have already been proposed. However, although these have expandability in the direction of stacking the substrates, they are poor in expandability in the remaining two-dimensional direction of the substrate surface. Therefore, for example, when implementing a three-dimensional mesh-coupled massively parallel computer or the like, in order to increase the number of processors, there is a drawback that the configuration must be long in only one of the three directions. In addition, the expansion of the number of boards in the one-dimensional direction only increases the number of processors in proportion to the number of boards, and thus has little effect on the increase in the number of processors.

On the other hand, when a parallel computer having a large number of processors is constructed, two-dimensional mesh connection and its connection are easy because it is easy to implement, it is easy to map problems in the natural world, and the utilization efficiency of wiring is high. A two-dimensional torus connection in which the ends are connected in a donut shape has been widely used. In particular, when the torus connection is used, the average communication distance of the connection network and the average communication distance are half that of the mesh, and the communication time becomes uniform even in the simulation using the periodic boundary condition, so that the performance is high.

Conventionally, when a parallel computer having a mesh connection or a torus connection is mounted, the wiring between all the boards is connected between the boards mounted with one or several processors, or the backplane and the cables are connected. It was common to implement them together.

In the case of the torus coupling, it is necessary to connect the end portions of the mesh in a donut shape. Therefore, in a large-scale system that cannot be accommodated by only one backplane, the cable length is increased by that portion, and the transfer speed is increased. Was hindering the improvement of In some cases, a simple mesh is used to prioritize the improvement of transfer speed.

Therefore, the highly parallel computers PAX128 and QC
In some cases, such as DPAX, the housing is mounted in a donut shape so that the cable length does not become long at the ends. However, it is difficult to extend this method to a three-dimensional torus.

That is, it is difficult to raise the frequency above a certain level by the conventional mounting method using a cable, and
Even if an attempt is made to improve the transfer rate by increasing the bit width of the parallel communication path, a driver for driving the cable is required, which is disadvantageous in terms of power consumption and packaging density.

[0011]

As described above, in the conventional board mounting method, it is difficult to tightly couple a large-scale system including a large number of parts with a large number of wirings. In the past, it was not possible to expand the scale of the parts where the parts were tightly coupled, or it was limited to one direction.

On the other hand, in the conventional electronic system mounting method, a cable has to be used in order to construct a large-scale torus coupling network, and it has been difficult to form a high-frequency, highly reliable, multi-bit width communication path. .

The present invention has been made to solve such a conventional problem, and a first object thereof is to enable a large-scale system composed of many parts to be closely coupled with a large number of wirings. It is to provide a board mounting method having expandability in two-dimensional and three-dimensional directions.

A second object is to realize a large-scale high-speed torus coupling network which is expandable, has a substantially uniform wiring distance between nodes, eliminates long-distance wiring, and does not require a cable. It is to provide an electronic system mounting method.

[0015]

In order to achieve the above-mentioned object, the first invention of the present application is a board mounting system in which a plurality of main boards are arranged in a horizontal direction to form a board of a predetermined size. A horizontal connection board having a first connector that is attachable to and detachable from the main board surface in a vertical direction in a peripheral portion on the board surface, and has a second connector that can be fitted with the first connector on the surface periphery. Is used to connect the adjacent main substrates to each other to form the substrate of the predetermined size.

Further, in the second invention of the present application, in the board mounting method of the first invention, a third connector is provided on the front surface and the back surface of each main board, and a plurality of main boards are provided in a direction normal to the main boards. Are arranged in parallel at a predetermined interval, and the respective main boards arranged in parallel are connected by using a vertical connection board having a fourth connector on the surface periphery that can fit with the third connector. Characterize.

A third invention of the present application is an electronic system mounting method for mounting a circuit module configured by torus-coupling a plurality of circuit blocks having nodes in an n-dimensional direction (n ≦ 3), wherein the circuit blocks are Terminals T1 and T2 connected to both ends of the node and a terminal T of a short-circuit wire arranged in parallel between the terminals T1 and T2 for each one dimension direction.
3 and T4, and terminals T1 and T3 of the circuit block which are not located at the ends are terminals T of the circuit block adjacent to one side.
4 and T2 respectively, and terminals T2 and T
4 is connected to the terminals T3 and T1 of the circuit block adjacent to the other, respectively, and the circuit block located at the end short-circuits the terminal T1, T3 or T2, T4 on which the adjacent circuit block does not exist. And

The fourth invention of the present application is 2 ⁿ pieces (n is the number of dimensions)
A circuit block having nodes of n direction (n ≦ 3)
In an electronic system mounting method for mounting a circuit module configured by torus coupling to a circuit block, the circuit block has inter-node connection terminals of each node for each one-dimensional direction, and the node of the circuit block not located at an end portion. Each of the inter-connection terminals is connected to one of the inter-node connection terminals of the adjacent circuit blocks, and the circuit blocks located at the ends form the torus of the inter-node connection terminal where the adjacent circuit block does not exist. Characterized by short-circuiting each other.

[0019]

In the first invention of the present application configured as described above, the main board is provided with the stacking connector (first connector) used for horizontal connection in the peripheral portion of the board. When the stacking connectors are arranged, the stacking connectors in the peripheral portion of the adjacent main boards are arranged in a region close to each other. For this reason, by stacking a horizontal connection board having a stack connector (second connector) on the surface facing the main board so as to straddle the two main boards and combining the connectors,
It becomes possible to connect the connectors of the adjacent main boards in an extremely short distance without using a cable.

According to the present invention, it is possible to combine the connectors without moving the main board, and the assembly can be performed with the main board fixed, so that a structurally stable module can be constructed.

Further, since a single connector merging operation can be closed by merging only the connectors in the local area equal to or less than one side of the main board, the connector insertion force is determined regardless of the number of main boards. Therefore, the constraint can be eliminated by the connector insertion force to the expandability of the main board in the two-dimensional direction.

As described above, it is possible to expand a large number of substrates in the two-dimensional direction while keeping the wiring between the substrates at a constant wiring density. Since there is no restriction on the number of expansions in the two-dimensional direction, it is possible to construct a two-dimensional module almost equivalent to a huge substrate by using a small main substrate without forcing the substrate to be enlarged. Therefore, it is possible to save the design cost of the board, and even when a failure occurs, only a small board including the failure portion needs to be replaced, and therefore damage at the time of failure repair is small.

Further, in the second invention of the present application, in the inter-board connection system using the first invention, first, a surface mount type connector used for vertical connection at positions symmetrical to the front surface and the back surface of the main board with respect to the board surface. (Third connector). Further, a vertical connection board having a connector (fourth connector) for connecting to the surface-mounted connector on the main board on two opposite sides of the board is prepared.

The assembling method is as follows. First, a two-dimensional module composed of a main board and a horizontal connection board is made, and the surface mount type connector on the surface of this main board and the connector on the vertical connection board are combined to form the main board. Vertically stand the vertical connection board. The surface mounting type connector on the back surface of the main board is united by covering the connector on the other side of the vertical connection board in this state with the main board forming the second two-dimensional module. Since the work of covering the main board can be performed independently for each small main board, the insertion force required at the time of assembly is determined by the number of connectors for vertical connection on one main board. It does not depend on the number of substrates.

The main connecting boards constituting the second two-dimensional module stacked on the first two-dimensional module are connected by the horizontal connecting board as in the method of forming the first two-dimensional module. . In this way, the two-story three-dimensional module is completed. It is possible to further expand in the three-dimensional direction by repeating the method of expanding from one-story to two-story.

As described above, when only the first invention is applied, the expandability can be achieved in another direction, which is not expandable, and the expandability can be achieved in all three-dimensional directions as a whole without using a cable. .

Further, in the third invention, one or more circuit blocks are mounted in an electronic circuit module that constitutes a system having a ring-shaped structure in a node coupling mode, and a node in the circuit block is physically arranged. One jump wiring is provided for coupling with a node adjacent to the above adjacent node (that is, a node next to the adjacent node). This wiring is provided in the circuit module when there are a plurality of circuit blocks in the circuit module.

In the case of connecting the nodes extending across the circuit modules, for this purpose, a connecting portion is provided around the end portion of the circuit module for pulling it out of the circuit module. In other words, if these circuit modules are arranged regularly, the coupling parts at the ends of the circuit blocks come close to each other. In this coupling section, there are wiring that is directly connected to the node in the adjacent circuit block on the physical layout from the coupling section and wiring that is led from the circuit block that is further adjacent to the circuit block (one-step wiring). It is connected.

Specifically, the terminal T of the first circuit block
1 is connected to the terminal T4 of the second circuit block adjacent to this circuit block, and similarly, the terminal T3 of the first circuit block is connected to the terminal T2 of the second circuit block. The terminals T2 and T4 of the first circuit block are
Of the third circuit block adjacent to the second circuit block on the side opposite to the second circuit block. Then, the circuit blocks existing at the ends short-circuit the terminals T1 and T3 or the terminals T2 and T4 which have no adjacent circuit blocks. As a result, one jump wiring is possible, and all nodes can be connected in a ring shape by relatively short wirings having substantially equal distances. Further, if the above ring-shaped structure is expanded in two-dimensional and three-dimensional directions, a two-dimensional and three-dimensional torus can be constructed.

In the fourth invention of the present application, in a system having a ring-shaped structure in which nodes are connected to each other, the minimum unit circuit block includes a plurality of nodes, and all the connection wirings from the nodes in the unit circuit block are included. Is output outside the unit circuit block, and the node in the unit circuit block is connected to the node in the adjacent circuit block. In addition, a circuit module having one or more unit circuit blocks each including a plurality of nodes includes a coupling portion arranged around an end portion.

When the coupling between the nodes on the adjacent unit circuit blocks extends between the circuit modules, they are coupled by the coupling portions which are close to each other when the circuit modules are regularly arranged and the coupling means for coupling the coupling portions. To be done.

In the state where the circuit modules are regularly arranged and the connecting portions and the connecting means arranged around the end portions of the circuit modules are connected between adjacent circuit modules as described above, except for the end portions of the array of circuit blocks. All the nodes in 1 are connected to the nodes in the adjacent unit circuit block.

In this state, there is a coupling portion left without coupling the first connecting means at the end portion of the array of circuit modules, and in the wiring inside the coupling portion, a predetermined portion in the unit circuit block at the end portion is present. A connecting means for short-circuiting the wirings of the nodes is coupled to this coupling portion. By doing so, all the nodes can be coupled in a ring shape by relatively short wirings having substantially equal distances.

In addition, the above ring-shaped structure is
A two-dimensional or three-dimensional torus can be constructed by expanding in the three-dimensional direction.

[0035]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a configuration diagram showing an embodiment of a mounting board to which the board mounting method according to the first and second inventions of the present application is applied.
FIG. 2 is an explanatory view when assembling a substrate in the first embodiment to which the first invention method of the present application is applied.

As shown in FIG. 2, two types of boards, a main board 5 and a horizontal connection board 4, and a power supply bar 1 are used for horizontal mounting. 3 shows a sketch of the main board 5, FIG. 4 shows a sketch of the horizontal connection board 4, and FIG. 5 shows a sketch of the power supply bar 1.

As shown in FIG. 3, a board stack connector 3 is mounted on the periphery of the four sides of one side of the main board 5, and a connecting portion is formed in a direction perpendicular to the board surface. This connector is a through hole type. However, it may be a surface mount type. One connector 3 may be mounted on each side, and if the number of pins is required, the mounting area of other components will be reduced, but by arranging a plurality of connectors on each side, it is possible to connect more pins. .
For example, if a plurality of small surface mount type connectors with a pitch of 0.5 mm available at the present time are used, if a connector area of about 10 square cm can be provided on each side of the main board, it will range from 300 to
It is also possible to provide signal lines of about 400 pins in four directions.

An electronic component 10 is mounted on the area of the main board other than the area on which the connector 3 is mounted, for example, a processing element of a massively parallel computer is mounted. Also,
The circuit mounted on the main board 5 does not necessarily have to be common in the system as long as it can interface with the connector 3.

On the other hand, as shown in FIG. 4, on one surface of the horizontal connection board 4, a connector 3 which can be combined with the board stack connector 3 in the peripheral portion of the main board 5 is mounted. Like the main board 5, the connector 3 may be a through hole type or a surface mount type. An electronic circuit may be mounted on the horizontal connection board 4, or an electronic component may not be mounted. The wiring in the horizontal connection board 4 has a very short wiring distance, and since it is a board, it is easy to match the impedance with the main board and the influence from reflection and external noise can be suppressed. The reliability is high when performed, and higher frequency transmission becomes possible. In particular, when the number of signals is large, the driver can be omitted, and the effect of improving the mounting density of this mounting method is great.

Further, as shown in FIG. 5, the power supply bar 1
Is used for supplying power to the main substrate 5 and for holding the main substrate 5. For this reason, the main board 5 is provided with power supply pads 11 and 12 and has holes 13 for passing screws (see FIG. 3). The power supply bar 1 is provided with holes 15 which are threaded in accordance with the positions of the power supply pads 11 and 12 and the screw holes 13 of the main board 5. The power supply bar 1 may be a linear metal rod provided with a screw hole 15 as shown in FIG. 6, but if the power supply pad has a sufficient power supply capability, as shown in FIG. It is possible to increase the component mounting area by making 14 a projection 16.

At the time of assembly, first, in a state where they are regularly arranged as shown in FIG. 7, the main board 5 is lightly fixed to the power supply bar 1 fixed to the case support 8 and the power supply bar support 9 with the fixing screw 6. Fix it with screws and fix it in a semi-fixed state. Next, the horizontal connection substrate 4 is used to connect two main substrates adjacent to each other in the horizontal direction to the substrate surface. FIG. 8 shows a side view when the two main boards 5 and the horizontal connection board 4 are connected. By repeating the operation of combining the horizontal connection substrates 4 as described above, a two-dimensional module as shown in FIG. 2 is formed. Since the screws 6 are joined together in a semi-fixed state, the horizontal connection boards 4 are combined, so that a slight positional deviation is corrected when the horizontal connection boards 4 are combined. When the main board 5 group in an appropriate range is horizontally connected, the fixing screws 6 for the power supply bar 1 are successively tightened to a fixed state to reduce the stress on the connector portion and stabilize the structure. Since the board is held by the power supply bar 1 rather than the connector 3, stress on the connector 3 is small, so that a surface mount type connector having a characteristic of low mechanical strength but high pin density can be used. it can.

By using the method of this embodiment, the connectors 3 can be combined without moving the main board 5, and the main board 5 can be assembled with the main board 5 fixed, so that a structurally stable module is constructed. can do.

Further, since a single joining operation of the connectors 3 can be performed by closing only the connectors 3 in a local area of one side of the main board 5 or less, the connectors 3 can be irrespective of the number of the main boards 5. The insertion force is determined. Therefore, it is possible to eliminate the restriction of the connector insertion force on the expandability of the main board 5 in the two-dimensional direction.

As described above, a large number of substrates can be expanded in the two-dimensional direction while maintaining a constant wiring density of the wiring between the substrates. At this time, since there is no limitation on the number of expansions in the two-dimensional direction, it is possible to construct a two-dimensional module that is substantially equivalent to a huge substrate by using a small main substrate without forcibly enlarging the substrate.

For this reason, it becomes possible to save the board design cost, and even when a failure occurs, only a small board including the failed portion needs to be replaced, so that damage at the time of failure repair is small.

It should be noted that the particularly large-scale two-dimensional module 1
9 and so on, as shown in FIG. 9, the power supply bar 1
Since the area 20 where the hand does not reach exists when is arranged in the horizontal direction with respect to the ground, the assembly work can be facilitated by arranging in the vertical direction as shown in FIG. Further, if the power supply bar 1 is arranged in the vertical direction, there is an effect that deformation of the power supply bar 1 due to gravity can be reduced.

As described above, the free expandability in the two-dimensional direction can be realized. However, if it is used to implement a massively parallel computer, for example, as shown in FIG. 11, it is easy to implement a massively parallel computer in which the processing elements 24 are two-dimensionally mesh-connected with the communication path 23, but the processing elements are The maximum inter-processor distance becomes too large in a machine with more than one machine. For example, in the case of 16384, the maximum inter-processor distance of the two-dimensional mesh is 256,65.
In the case of 536 units, the number is 512, which is not practical except for the application that requires only communication between adjacent processing elements. Therefore, the range of application will be limited only by implementing the two-dimensional mesh connection.

Further, the size of the two-dimensional module is such that, even if one processing element can be realized on a 10 cm square substrate, the number of 16384 modules is 12.8 m square, 6
In the case of 5,536 units, the size is 25.6 m square, which requires a very large space. Therefore, next, by further applying the second invention of the present application and implementing a three-dimensional module, it is shown that a massively parallel computer with three-dimensional mesh connection can be realized.

In order to apply the second embodiment according to the second invention of the present application, a main substrate 5 as shown in FIG. 12 is used. That is, in addition to the connector for horizontal connection, the connector 2 for vertical connection is provided. The connector 2 is mounted on the surface on which the connector 3 for horizontal connection is mounted, and as shown in FIG. 13, the connector 2 for vertical connection is also mounted on a position symmetrical with respect to the substrate surface on the back surface. For this reason, the surface mounting type is used as the connector 2 for vertical connection. As a connector that can be used for such purposes, a connector having a pin mounting density equivalent to that of horizontal connection can be used. That is, it is possible to provide about 300 to 400 pins per 10 square cm in two directions perpendicular to the substrate surface.

It is possible to combine with the main substrate 5 having the vertical connection connector 2 on the front and back as shown in FIG.
A vertical connection board 7 as shown in is prepared. Angle-type connectors 25 are mounted on two opposite sides of the vertical connection board 7.

Next, a method for assembling a three-dimensional module using the above-mentioned three types of substrates, that is, the main substrate 5, the vertical connection substrate 7, and the horizontal connection substrate 4, will be described.

First, the main boards are connected by the horizontal connection board 4 and the two-dimensional module is assembled in the same manner as in the above example. Next, for this 2D module,
As shown in FIG. 15, the vertical connection board 7 is inserted into the vertical connection connector 2 on the main board 5. Power supply bar 1 in this state
The power supply bar 1 as shown in FIG.
Secure it to the fixed part of. Next, as shown in FIG. 1, the vertical connection board 7 and the new main board 5 are combined so as to be covered with the connector 2 from above. By repeating the above operation, the two-dimensional modules can be stacked and expanded in the direction perpendicular to the main substrate surface.

By the mounting method as described above, it becomes possible to construct a three-dimensional module in which boards of about 10 cm square in each of the six directions of north, south, east, west, and north are connected by about 300 printed circuit board wirings. For example, if this is used to implement a massively parallel computer, a three-dimensional mesh connection with a super-high-speed communication capability that secures a sufficient transfer width even if a high-speed and highly-reliable communication path that uses half the wiring as a ground line is constructed. A net can be constructed.

By applying the second invention, by changing from the two-dimensional mesh connection to the three-dimensional mesh connection, the maximum transfer distance, which has been a problem in a large-scale two-dimensional mesh, is a parallel root of the number of processors. What is proportional to is only proportional to the cube root, so the range of application is wide. For example, the maximum transfer distance is 256 even with 16384 units in the two-dimensional mesh, whereas the maximum transfer distance is only 96 in the three-dimensional mesh with 32768 units, which has twice the number of processors, and processing elements that are not adjacent to each other. A great effect can be expected in mitigating collisions between communications.

Further, the number of transfer paths per processing element is four in the two-dimensional mesh, whereas in the three-dimensional mesh to which the present invention is applied, the number of transfer paths is 1.5 times as many as six, that is, six. Therefore, the peak transfer rate can be increased 1.5 times.

The three-dimensional structure is advantageous in terms of the size of the system. For example, the main substrate 5 is 10 cm
Even if the distance between the corners and the connector of the vertical connection board 7 is 3 cm,
3.2mx in 32768 (32 x 32 x 32) configuration
The same number of two-dimensional meshes (256 × 12) mounted on a size of 3.2 m × 1 m and having the same size of main board 5
Compared with 8) of 25.6 m × 12.8 m, it is possible to install in a space where the three-dimensional mesh is much smaller. The fact that it is mounted in a small space also has an advantage in terms of clock synchronization and the like.

The layout of the main substrate 5 shown in FIG. 12 corresponds to the third embodiment of the present invention. With such an arrangement of the vertical connection board, an appropriate space is secured between the stacked main boards by the vertical connection board, and a space penetrating the three-dimensional module in the direction AB or the direction CD. Can be secured. That is, in the inter-board connection method of the second invention, the vertical connection board 7 is arranged so that the space between the adjacent two-dimensional modules formed by the vertical connection board 7 is not closed, and this space is used as the heat exhaust path. It is a thing.

FIG. 17 is a block diagram showing a fourth embodiment of the present invention, in which the space between the substrates is used as a heat exhaust path.

When the power consumption of the electronic device mounted on the main board 5 is small, or when the number of main boards 5 arranged in the direction AB or CD is small, the three-dimensional module is mounted in that direction. Natural convection in the penetrating space,
It is possible to cool the system by forced air cooling as in. At the time of air cooling, it is possible to enhance the cooling capacity by disposing the cooling fan 26 in the space that penetrates the three-dimensional module.

When the power consumption of the electronic device mounted on the main board 5 is large, or when the number of main boards 5 arranged in the direction AB or CD is large, insulation is performed as shown in FIG. It is possible to perform cooling by immersing it in a cooling liquid 28 having a property to circulate it between the space passing through the three-dimensional module and the heat exchanger 32.

That is, the insulating primary cooling liquid 28 injected into the water tank 27 absorbs heat from the substrate and then the circulation pump 30.
The heat is exchanged by the secondary cooling water that is circulated by the cooling fins 32 of the heat exchanger 32 and injected from the water injection port 33, and is returned to the water tank 27 through the distribution pipe 29. In addition, the secondary cooling water is drained from the drain port 34. Unlike indirect cooling, this method can eliminate the thermal resistance of the cooling module, so that extremely efficient cooling is performed.

Further, since it is possible to adjust the space between the stacked main boards by changing the size of the vertical connection board, an indirect water cooling module as shown in FIG. 19 is provided between the main boards adjusted according to its thickness. Although the cooling efficiency is slightly inferior to that of the direct immersion cooling, it becomes possible to use inexpensive water as the cooling liquid instead of the relatively expensive insulating cooling liquid by inserting the water into the space.

That is, the metal hat 35 with which the cooling water pipe 36 is in contact with the groove 37 is mounted on the substrate, and the portion 41 where the metal hat 35 and the electronic component on the substrate contact each other is vertically moved by the spring 39. The piston 40 is provided. As a result, heat from the electronic component can be absorbed and cooling can be performed with cooling water. Note that the metal hat 35 of the indirect water cooling module may have a structure that also serves as the power supply bar 1.

In the fifth embodiment of the present invention, the sub-board is arranged in the direction perpendicular to the main board surface to expand the mounting area.
FIG. 20 is a configuration diagram showing such a fifth embodiment,
By providing the child board 71, more components can be mounted by effectively utilizing the limited space, and high-density mounting can be achieved.

The child board 71 may itself be a processing element of a parallel computer, or may be used for various purposes such as an additional memory board. The sub board 71 is arranged such that the space between the adjacent two-dimensional modules formed by the vertical connection boards is not closed in at least one direction. By doing so, this space can also be used as a cooling path.

FIG. 21 is a block diagram showing a sixth embodiment of the present invention. In this embodiment, twelve main boards 5 are arranged three-dimensionally (2 * 2 * 3), and the horizontally adjacent main boards 5 are connected using two horizontal connection boards 44. The vertically adjacent main substrates 5 are connected using four vertical connection substrates 43, and four main substrates 5 vertically adjacent to each other are provided for supplying power and physically fixing the main substrates 5. A metal spacer 42 that also serves as a power supply is used. 22 shows the configuration of the main substrate, FIG. 23 shows the configuration of the horizontal connection substrate 44, FIG. 24 shows the configuration of the vertical connection substrate 43, and FIG. 25 shows the configuration of the metal spacer 42 that also serves as a power supply.

The horizontal connection board 4 is used as the main board in FIG.
4 is mounted with a socket type connector 46, a vertical connection board 43 is connected with a right angle type connector 45 and an electronic circuit, and a horizontal connection board 44 shown in FIG. 23 is mounted with a right angle type connector 45. Then, a socket type connector 46 is mounted on the vertical connection board 43 in FIG. The metal spacer 42 also serving as a power supply in FIG. 25 is preferably one having a high conductivity, a high strength and a small mass, and a female screw 47.
And the male screw 48.

The horizontal main boards 5 are connected by two horizontal connection boards 44, and the vertical main boards 5 are connected by four vertical connection boards 43. By connecting the horizontal connection board 44 and the main board 5 by using the right-angle connector 45, the horizontal connection board 44 is mounted upright with respect to the plane of the main board 5.
The relatively small horizontal connection board 44 can be easily inserted and removed.

Further, since the connector for the horizontal connection board 44 is arranged so as to be orthogonal to the edge of the main board 5 in the main board 5, as compared with the case where the same connector is arranged horizontally with respect to the edge of the board. And high density mounting becomes possible. Needless to say, higher density mounting is possible compared to the case where a cable is used for connection between boards as in the past.

FIG. 26 shows a cross-sectional view of the connecting portion between the vertical connection board 43 and the main board 5. In FIG. 26, the vertical connection substrate 4
3 shows that the connector 46 of No. 3 is inserted into the connector 45 from the horizontal direction with respect to the main board 5 and connected. Compared to a method in which the vertical connection board 43 is sandwiched from above and below and the vertical connection boards 43 are connected at the same time in the vertical direction, the vertical connection board 43 is inserted horizontally to check the connection. Since it is possible to confirm by using the same method, it is possible to improve the reliability of the mounting and to connect one vertical connection board 43 to the main board 5 at a time. It becomes possible to connect to the main board 5.

In the metal spacer 42 also serving as the power supply shown in FIG. 25, the power is supplied to each main substrate 5 and the main substrate 5 is supported, so that the main substrate 5 is stably supported in this embodiment. In order to do so, four metal spacers 42 also serving as a power supply are used for each main substrate 5. The use of the metal spacer 42 that also serves as a power supply is not limited to the power supply, but can be used for other signals (especially for large current). In the embodiment shown in FIG. 21, the number of main substrates 5 is 12, but the present invention is applicable regardless of the number of main substrates 5.

Further, the present invention can be applied regardless of the number of horizontal connection boards 44 and vertical connection boards 43 in connection between the main boards 5. In the embodiment, four power supply / metal spacers 42 are used in one first substrate, but in the present invention, any number of power supply / metal spacers 42 may be used.

The seventh embodiment is one in which the horizontal connection board 44 shown in FIG. 23 is provided with flexibility, and it is possible to absorb a positioning error when the connector is mounted. When using a high density connector, there is almost no play in the connector itself, and when using a surface mount connector, the current technology has a tolerance of about 0.1 mm for component placement and about 0.1 mm for reflow. An error of up to about 0.2 mm can occur even with only one substrate. Considering that two substrates are connected on top of that, an error of up to about 0.4 mm may occur. As described above, it is almost impossible to construct a very large-scale system by using a connector having no play.

Therefore, when a connector having no play is used, the horizontal connection board 44 using a rigid flexible board having a rigid portion 49 and a flexible portion 50 as shown in FIG. 27 is used, or as shown in FIG. This is solved by using the horizontal connection board 44 using the plastic board 52. Of course, the rigid flexible board and the plastic board are not applied only when using a connector having no play.

The eighth embodiment of the present invention is implemented by providing the vertical connection board 43 shown in FIG. 24 with flexibility. In order to actually perform this, a vertical connection board 43 using rigid flexible boards 49 and 50 as shown in FIG. 29 is used, or a vertical connection board 43 and a reinforcing plate 51 are used as shown in FIG. This is solved by using the connection board 43.

When the vertical connection board 43 does not have flexibility, the space between the main boards 5 must be increased in order to mount a medium scale or more. Vertical connection board 43
Since the main board 5 and the vertical connection board 43 are connected on a plane which is actually mounted by providing flexibility in FIG.
It is possible to crush to the mounting plane as shown in. As a result, the adjacent main boards 5 to which the main board 5 and the vertical connection board 43 are to be connected are raised by one step, so that the insertion of the connector can be facilitated even if the space between the main boards 5 is reduced as much as possible.

In each of the above embodiments, the main substrate 5 has a rectangular shape, but it may have various shapes such as a triangle, a hexagon and an octagon, or an octagon and a square. It may be mixed like a combination of.

FIG. 32 is a block diagram showing a system to which the electronic system mounting method according to the ninth embodiment of the present invention is applied.

In this embodiment, one circuit module corresponds to one circuit block BK0 to BK3, and one circuit block shows one-dimensional torus coupling in which one node (N0 to N3) is mounted. Plural circuit blocks may be arranged in one circuit module, or plural nodes may be mounted in one circuit block.

Specifically, an electronic circuit such as a processing element of a massively parallel computer corresponds to the node, but all the circuits of the node may be the same or different circuits. FIG. 33 is a diagram showing a physical configuration example of the circuit module of the processing element of the massively parallel computer used in this embodiment. The processing element of the massively parallel computer includes a CPU 58 in charge of calculation, a memory 59 for storing data and programs, a router LSI 57 in charge of communication with other processing elements, a power supply unit 56, and the like.

A coupling portion A1 and a coupling portion B1 for connecting to another circuit block are provided in the peripheral portions of the two opposite sides of the circuit module. In FIG. 32, the connecting portion A1 and the connecting portion B1 on one circuit block are shown as being divided into two parts, but physically, they may correspond to one connector as shown in FIG. , It may be compatible with three or more connectors.

Further, as shown in FIG. 32, on the circuit block, in addition to the wiring for connecting the node mounted therein to the coupling portion A1 and the coupling portion B1, the coupling portion A1 (terminals T1, T3) to coupling portions are connected. It is provided with one jump wiring which connects between the parts B1 (terminals T2 and T4) and bypasses the node of the circuit block.

The circuit blocks are regularly arranged such that the connecting portion A1 and the connecting portion B1 are close to each other. In the embodiment of FIG. 32, the number of nodes is four and the number of circuit blocks is four, but these numbers can be arbitrarily selected as long as there are two or more nodes. For the sake of explanation, the four circuit blocks are named from the left as circuit block BK0, circuit block BK3, circuit block BK1, and circuit block BK2, and the nodes are named from the left as node N0, node N3, node N1, and node N2.

The first connecting means 55 connects the connecting portions A1 and B1 of the adjacent circuit blocks.
FIG. 34 shows the first connecting means 55 using the printed circuit board 60.
An example of is shown. In this example, a connecting portion connector 61a and a connecting portion connector 61b by stack connecting connectors are attached to both ends of the printed circuit board 60, the connecting portion connector 61a corresponds to the connecting portion A1 connector, and the connecting portion connector 61b is formed. Corresponding to the connector for connecting portion B1 and wiring is provided between the connector for connecting portion 61a and the connector for connecting portion 61b, the first connecting means 55 is formed. If the wiring of the first connecting means 55 is composed of a printed circuit board like the circuit module, it is possible to use a shield layer as an inner layer, so that the impedance of the line can be easily matched, and unnecessary radiation noise and external noise The influence can be suppressed.

In FIG. 32, the wiring extending from the node N0 to the right is the connection portion A1 of the node N0 → the block BK0.
(Connector 61a for connecting portion) → first connecting means 55 between blocks BK0 to BK3 → connecting portion B1 (connector 61b for connecting portion) of block BK3 → block BK3
Wiring of one block → connecting part A1 of block BK3 (connector 61a for connecting part) → block BK3 to block BK1
First connecting means 55 between → connecting portion B1 of block BK1
(Connector 61b for coupling unit) → Node N0 is coupled to node N1 along the path of node N1, and node N2 to node N3 are coupled in the same manner.

Second ends of the array of circuit blocks are joined by second connecting means 54 which provide a connection between the end circuit block and its adjacent circuit blocks. For example, in FIG. 32, the wiring extending to the left from the node N0 has a node N0 → the connecting portion B1 of the block BK0 →
Second connection means 54 → joint portion B1 of block BK0 → one-step wiring of block BK0 → joint portion A1 of block BK0 (connector 61a) → first connecting means 55 between block BK0 to block BK3 → block The node N0 to the node N3 are coupled along the route of the coupling portion B1 (coupling connector 61b) of the BK3 → the node N3, and the nodes N2 to N1 are also coupled in the same manner.

In the array of circuit blocks connected in the above-mentioned configuration, the nodes are connected by the link 64 as shown in FIG. 35. If this is easily changed without changing the connection relationship, as shown in FIG. 36. It can be seen that they are connected in a ring (one-dimensional torus) shape.

All inter-node links are physically composed of one jump wiring and two wirings of the first connecting means 55 or the second connecting means 54, and the wiring length is almost uniform. In addition, the length of these wirings is relatively short, which is about two circuit blocks, and does not depend on the total number of arrays. Therefore, the timing adjustment during the transmission of the high frequency signal becomes easy.

Moreover, the first connecting means 55 and the second connecting means 54 do not necessarily have to be cables as in the prior art, but rather a printed circuit board 60 as shown in FIG. 34 to which a stack connecting connector is attached. It is possible to achieve both the characteristics of high-density wiring and high-frequency wiring that is less affected by noise.

Of course, the present invention is not applicable only when a cable is not used, and even if the first connecting means 55 and the second connecting means 54 are cables, they are arranged as shown in FIG. 35 according to the present invention. It is more effective to connect the cables in a relatively short length as compared with the case of simply arranging and connecting the cables as shown in FIG. 36. Also, when adding a circuit block, the second connecting means 54 at the end is removed, and the circuit block is added to the structure by using the first connecting means 55, and the second connecting means is added to the end. Since it can be easily made by connecting 54, it has excellent expandability.

In the connection for the one-dimensional direction as in this embodiment, the modules A1 and B1 are directly coupled between the modules by using an angle type or card edge type connector as shown in FIG. It is possible to connect the parts A1 and B1 and it is possible to simplify the first connecting means 55. In this case, the twisted wiring on the first connecting means 55 shown in FIG. 32 is installed in the portion of the circuit module body.

FIG. 38 is a configuration diagram showing a tenth embodiment of the present invention, and is a diagram showing an implementation example of two-dimensional torus coupling in which m = n = 2, that is, circuit blocks are two-dimensionally arranged. . In this embodiment, one circuit module corresponds to one circuit block, and one circuit block is shown to have one node mounted, but one circuit module has a plurality of circuit blocks arranged therein. Alternatively, one circuit block may be equipped with a plurality of nodes.

Specifically, an electronic circuit such as a processing element of a massively parallel computer corresponds to the node, and the circuits of the node may be the same or different. FIG. 39 is a diagram showing a physical configuration example of the circuit module of the processing element of the massively parallel computer used in this embodiment. The processing element of the massively parallel computer is composed of a CPU 58 in charge of calculation, a memory 66 for storing data and programs, a router LSI 65 in charge of communication with other processing elements, and the like.

In the peripheral portions of the four sides of the circuit module, a connecting portion A2, a connecting portion B2 for connecting to another circuit block,
A coupling portion C2 and a coupling portion D2 are provided. Figure 38
On the upper side, the connecting portion A2 and the connecting portion B on one circuit block
2, the coupling portion C2 and the coupling portion D2 are described separately, but physically, they may correspond to one connector or the like as shown in FIG. 39, or may correspond to three or more connectors. It may be.

Further, on the circuit block, from the node mounted therein to the connecting portion A2, the connecting portion B2, and the connecting portion C.
2 and the wiring connected to the coupling portion D2, the coupling portion A2
.About.coupling section B2 and between coupling section C2 and coupling section D2, and one jump wiring for bypassing the node of the circuit block is provided.

The circuit blocks are regularly arranged such that the coupling section A2 and the coupling section B2 are close to each other and the coupling section C2 and the coupling section D2 are close to each other. In the embodiment of FIG. 38, the number of nodes is 1
6, the number of circuit blocks is 16, but these numbers can be arbitrarily selected if the number of nodes is an even number. For the sake of explanation, the sixteen circuit blocks are defined as the circuit block BK00 from the upper left, and the circuit block BK03, the circuit block BK01, the circuit block BK02, and the circuit block BK3 are arranged in the downward direction from the right.
0, the circuit block BK31, the circuit block BK32, the circuit block BK30, and the circuit block BK10, the circuit block BK13, and the circuit block BK1 from the right downward.
1. Circuit block BK12, circuit block BK10, one circuit block BK20, circuit block BK23, circuit block BK21, circuit block BK2
2 nodes, and the 16 nodes are designated as node N00 from the upper left, and nodes N03, N01, and N0 in the downward direction.
2. From the right side of the node N00 to the node N30,
One of the nodes N31, N32, and N30 from right to downward, the node N10, the node N13, and the node N1.
1, the node N12, the node N10, and the node N20, the node N23, the node N21, and the node N in the downward direction.
Name it 22.

The connecting portions A2 and B2 of adjacent circuit blocks are connected by the first A connecting means 55a,
The coupling portion C2 and the coupling portion D2 are coupled by the first B connecting means 55b. As an example of the first A connecting means 55a and the first B connecting means 55b, the printed circuit board 60 shown in FIG. 34 can be used. In the embodiment of FIG. 34, a connector 61a for a connecting portion and a connector 61b for a connecting portion by stack connecting connectors are provided on both ends of the printed circuit board 60.
Is attached, the connector 61 for the connecting portion corresponds to the connector for the connecting portion A2 and the connector for the connecting portion C2, and the connector 61b for the connecting portion corresponds to the connector for the connecting portion B2 and the connector for the connecting portion D2. Wiring is provided between the connector 61a and the connector 61b for the coupling portion, so that the first A connecting means 55a and the second B connecting means 54b are formed. If the first-A connecting means 55a and the second-B connecting means 54b are composed of a printed circuit board like the circuit module, it is possible to use a shield layer as an inner layer, so that the line impedance can be easily matched. It is possible to suppress the effects of unnecessary radiation noise and external noise.

In FIG. 38, the wiring extending from the node N00 to the right has a node N00 → the coupling portion A2 of the block BK00 (the coupling portion connector 61a) → the blocks BK00 to BK00.
First A connecting means 55a between the blocks BK30 → connecting portion B2 of the block BK30 (connecting portion connector 61b) →
One jump wiring of the block BK30 → the coupling portion A2 of the block BK30 (the coupling portion connector 61a) → the block BK
First-A connecting means 55a between 30 and the block BK10 →
The coupling portion B2 of the block BK10 (the coupling portion connector 61
The node N00 to the node N10 are coupled by the route of b) → node N10. Node N01 to Node N11
, Node N02 to node N12, node N03 to node N13, node N20 to node N30, node N21 to node N31, node N22 to node N32.
The nodes and the nodes N23 to N33 are also coupled in the same manner.

In FIG. 38, the wiring extending downward from the node N00 has a node N00 → the coupling portion C2 of the block BK00 (the coupling portion connector 61a) → the blocks BK00 to BK00.
First B connecting means 55b between the blocks BK03 → connecting portion D2 (connecting portion connector 61b) of the block BK03 →
One jump wiring of the block BK03 → the coupling portion C2 of the block BK03 (the coupling portion connector 61a) → the block BK
03-block BK01 first B connection means 55b →
The coupling portion D2 of the block BK01 (the coupling portion connector 61
b) → Node N01 is coupled to the nodes N00 to N01. Node N10 to Node N11
, Node N20 to node N21, node N30 to node N31, node N02 to node N03, node N12 to node N13, node N22 to node N23.
The nodes and the nodes N32 to N33 are also coupled in the same manner.

Both ends of the junction A2 and the junction B2 of the array of circuit blocks are connected by the second A connecting means 54a, and both ends of the junction C2 and the junction D2 are connected to the second B.
They are joined by the connecting means 54b. These provide the coupling between the end circuit block and its adjacent circuit blocks.

In FIG. 38, the wiring extending to the left from the node N00 is the node N00 → the connecting portion B2 of the block BK00 → the connecting means 54a of the second A → the connecting portion B2 of the block BK00 → the one jump wiring of the block BK00 → Combined part A2 of block BK00 Block BK00 to block BK
The first A connecting means 55a between the nodes 30 → the connecting portion B2 of the block BK30 → the node N30 along the route from the node N00 to the node N00
The nodes N30 are connected. Node N01 to node N
31, node N02 to node N32, node N03
-Node N33, node N20-node N10, node N21-node N11, node N22-node N
12 and between node N23 and node N13 are also coupled in a similar manner.

In FIG. 38, the wiring extending upward from the node N00 is the node N00 → the connecting portion D2 of the block BK00 → the second connecting means 54b → the connecting portion D2 of the block BK00 → the one jump wiring of the block BK00 → Combined portion C2 of block BK00 → block BK00 to block B
First B connecting means 55b between K03 and block BK03
Node D00 → node N03 in the path of node N00
~ Nodes N03 are coupled. Between node N10-node N13, between node N20-node N23, node N3
0 to node N33, node N02 to node N01,
The nodes N12 to N11, the nodes N22 to N21, and the nodes N32 to N31 are also coupled in the same manner.

In the array of circuit blocks connected in the above-mentioned configuration, the nodes are connected as shown in FIG. 40.
It can be seen that they are connected in a ring (two-dimensional torus) like No. 1.

All inter-node links are physically one jump wire and the first A connecting means 55a, the first B connecting means 55b, the second A connecting means 54a or the second B connecting. It is composed of two wires of the means 54b,
The wiring length is almost uniform. Moreover, the wiring lengths in these four directions are relatively short, that is, about two circuit blocks, and do not depend on the total number of arrays. Therefore, the timing adjustment during the transmission of the high frequency signal becomes easy.

Moreover, the first A connecting means 55a, the first B connecting means 55b, the second A connecting means 54a and the second B connecting means 54b need not be cables, but rather as shown in FIG.
The printed circuit board 60 with the stack connection connector 61 as shown in FIG. 4 can achieve both high-density wiring and high-frequency wiring with less noise influence.

Of course, the present invention is not applicable only when the cable is not used, and the first A connecting means 55 is used.
a, first B connecting means 55b, second B connecting means 54a
Even if the connecting means of the second B 54b is a cable, the cable length is relatively shorter when arranged and connected as shown in FIG. 40 according to the present invention than when simply arranged and connected as shown in FIG. It is effective because it can be aligned.

Also when the circuit block is added, the second A connecting means 54a or the second B connecting means 54 at the end portion is added.
b, and the first A connecting means 55a or the first B
By connecting the circuit block with the structure using the connecting means 55b of No. 2 and connecting the connecting means 54a of the second A or the connecting means 54b of the second B to the end of the circuit block, it is possible to easily add more and expand. Are better.

Here, the first A connecting means 55a and the first B connecting means 55b in the tenth embodiment shown in FIG. 38 are treated as different means for the sake of explanation, but in reality, they are exactly the same means (the same thing). Similarly, the second A connecting means 54a and the second B connecting means 54b are also treated as different means for the sake of explanation, but in reality they are exactly the same means (the same thing). Things are possible.

As an eleventh embodiment of the present invention, examples of application to three-dimensional torus coupling and n-dimensional torus coupling (n is an arbitrary natural number) can be given. The three-dimensional torus combination is shown in FIG.
And a method of arranging the circuit modules in the coupling of the embodiment shown in FIG. 38 in one dimension (that is, m = 1) and then mounting the two-dimensional circuit modules in one circuit module, m = 2) and then a one-dimensional circuit module is mounted in one circuit module, or one-dimensional circuit module is arranged in three dimensions (that is, m = 3) It is realized by the method of mounting less than the required number of circuit blocks.

Similarly for the n-dimensional arrangement, the circuit modules are arranged one-dimensionally and then (n-
1) A method of mounting a circuit block corresponding to a dimension, a method of arranging a circuit module in two dimensions and then mounting a circuit block corresponding to an (n-2) dimension on one circuit module, and a method of tertiary circuit module It is realized by using a method in which the circuit blocks corresponding to the (n-3) dimension are mounted on one circuit module after being originally arranged.

Further, n-dimensional circuit blocks may be mounted on one circuit module as long as the size of the circuit module is not limited.

FIG. 42 is a diagram showing an electronic system mounting method according to the twelfth embodiment of the present invention. In this embodiment, one circuit module corresponds to one circuit block, and one circuit block shows two-dimensional torus coupling in which two nodes are mounted. However, two circuit nodes are mounted in one circuit block. If so, a plurality of circuit blocks may be arranged in one circuit module. In the present invention, in order to configure the one-dimensional torus coupling, it is necessary that one circuit block always includes two nodes.

Specifically, an electronic circuit such as a processing element of a massively parallel computer corresponds to the node, but all the circuits of the node may be the same or different circuits. FIG. 43 is a diagram showing a physical configuration example of the circuit module of the processing element of the massively parallel computer used in this embodiment. The processing element of the massively parallel computer is a CPU 58 in charge of calculation, a memory 59 for storing data and programs, and a router LSI 65 electrode 56 in charge of communication with other processing elements.
Etc.

A coupling portion A3 and a coupling portion B3 for connecting to another circuit block are provided on the peripheral portions of the two opposite sides of the circuit module. In FIG. 42, the connecting portion A3 and the connecting portion B3 on one circuit block are shown divided into two parts, but physically, they may correspond to one connector as shown in FIG. , It may be compatible with three or more connectors.

The circuit blocks are regularly arranged such that the connecting portion A3 and the connecting portion B3 are close to each other. In the embodiment of FIG. 42, the number of nodes is 8 and the number of circuit blocks is 4, but these numbers can be arbitrarily selected except that two nodes must be mounted in one circuit block. For the sake of explanation, the four circuit blocks in the embodiment of FIG. 42 will be referred to from the left as circuit block BK0, circuit block BK1,
Named circuit block BK2 and circuit block BK3, and the nodes from the left are node N00, node N01, and node N1.
0, node N11, node N20, node N21, node N30, and node N31.

The first connecting means 67 connects between the connecting portions A3 and B3 of the adjacent circuit blocks. As an example of the first connecting means 67, the printed circuit board 60 shown in FIG. 34 can be used. In this example, a connecting portion connector 61a and a connecting portion connector 61b by stack connecting connectors are attached to both ends of the printed circuit board 60, the connecting portion connector 61a corresponds to the connecting portion A3 connector, and the connecting portion connector 61b is Corresponding to the connector for connecting portion B3, wiring is provided between the connector for connecting portion 61a and the connector for connecting portion 61b, so that the first connecting means 67 is formed. If the wiring of the first connecting means 67 is composed of a printed circuit board like the circuit module, it is possible to use a shield layer as an inner layer, so that the impedance of the line can be easily matched, and unnecessary radiation noise and external noise The influence can be suppressed.

The wiring extending from the node N00 to the right has a node N00 → the connecting portion A3 of the block BK0 (connecting portion connector 61a) → the first connecting means 67 between the blocks BK0 to BK1 → the connecting portion of the block BK1. The node N00 to the node N10 are coupled by a route of B3 (connector 61b for coupling unit) → node N10. Node N1
0 to node N20, node N20 to node N30,
The nodes N01 to N11, the nodes N11 to N21, and the nodes N21 to N31 are also coupled in the same manner.

Second ends of the array of circuit blocks are joined by second connecting means 68 which provide a connection between the end circuit block and its adjacent circuit blocks. For example, the wiring extending to the left from the node N00 is node N00 → joint B3 of block BK0 → second connecting means 68 → joint B3 of block BK0 → node N01.
The node N00 to the node N01 are coupled to each other by the route. The nodes N30 to N31 are also coupled in the same manner.

In the array of circuit blocks connected in the above-described configuration, the nodes are connected as shown in FIG. 44.
As can be seen from Fig. 5, they are connected in a ring (one-dimensional torus) shape.

All the inter-node links are physically composed of one wiring line of the first connecting means 67 or the second connecting means 68, and the wiring length is substantially uniform. Moreover, the wiring length of these is relatively short, that is, one circuit block, and does not depend on the total number of arrays. Therefore, the timing adjustment during the transmission of the high frequency signal becomes easy.

Moreover, the first connecting means 67 and the second connecting means 68 do not necessarily have to be cables as in the conventional case, but rather a printed circuit board 60 as shown in FIG. 34 to which a stack connecting connector is attached. It is possible to achieve both the characteristics of high-density wiring and high-frequency wiring that is less affected by noise.

Of course, the present invention is not applicable only when a cable is not used, and even if the first connecting means 67 and the second connecting means 68 are cables, they are arranged according to the present invention as shown in FIG. It is more effective to connect the cables by simply connecting them by simply arranging them as shown in FIG. Also, when adding a circuit block, the second connecting means 68 at the end portion is removed, and the circuit block is replenished with the structure using the first connecting means 67, and the second connecting means is provided at the end portion. Since it is easily possible by connecting 68, it has excellent expandability.

In the connection in the one-dimensional direction as in this embodiment, the modules A1 and B3 are directly coupled between the modules by using an angle type or card edge type connector as shown in FIG. It is also possible to simplify the first connecting means 67 by combining the parts A3 and B3.

FIG. 46 shows an electronic system mounting method according to the thirteenth embodiment of the present invention, showing an embodiment of the two-dimensional torus coupling in which m = n = 2, that is, the circuit blocks are two-dimensionally arranged. Is. In this embodiment, one circuit module corresponds to one circuit block, and four circuit nodes are mounted in one circuit block. However, in a two-dimensional torus coupling, one circuit block has four nodes. A plurality of circuit blocks may be arranged in one circuit module as long as is mounted.

Specifically, the node corresponds to an electronic circuit such as a processing element of a massively parallel computer, and the circuits of the node may be the same or different circuits. FIG. 47 is a diagram showing a physical configuration example of the circuit module of the processing element of the massively parallel computer used in this embodiment. The processing elements of the massively parallel computer are composed of a CPU in charge of operations, a memory for storing data and programs, a router LSI in charge of communication with other processing elements, and the like.

In the peripheral portion of the four sides of the circuit module, a connecting portion A4 and a connecting portion B4 for connecting to another circuit block,
A connecting portion C4 and a connecting portion D4 are provided. Figure 46
And in FIG. 47, the connecting portion A on one circuit block is shown.
4, the connecting portion B4, the connecting portion C4 and the connecting portion D4 are written as one, but physically one connector or the like may be used, or two or more connectors may be used. good.

Further, on the circuit block, there are provided wirings for connecting the four nodes mounted therein to the coupling section A4, the coupling section B4, the coupling section C4 and the coupling section D4. The circuit blocks are regularly arranged such that the coupling part A4 and the coupling part B4 are close to each other and the coupling part C4 and the coupling part D4 are close to each other. In the embodiment shown in FIG. 46, the number of nodes is 16 and the number of circuit blocks is 4. However, these numbers can be selected arbitrarily while strictly observing that four nodes are arranged on one circuit block. For explanation here,
The four circuit blocks are defined as the circuit block BK0 from the upper left, and the circuit blocks BK3 and BK0 are arranged in the downward direction.
To the right of one of the circuit blocks BK1 and BK
1 is named a circuit block BK2 in the downward direction, and 16 nodes are named as a node N00, a node N30, a node N03, and a node N33 from the top of the circuit block BK0.
From the top of the circuit block BK1 to the node N10 and the node N2
0, node N13 and node N23, node N11, node N21, node N12 and node N22 above circuit block BK2, node N01, node N31 and node N above circuit block BK3.
02 and node N32.

The connecting portion A4 and the connecting portion B4 of the adjacent circuit blocks are connected by the first A connecting means 69a,
The connecting portion C4 and the connecting portion D4 are connected by the first B connecting means 69b. As an example of the first A connecting means 69a and the first B connecting means 69b, the printed circuit board 60 shown in FIG. 34 can be used. In the embodiment of FIG. 34, a connector 61a for a connecting portion and a connector 61b for a connecting portion by stack connecting connectors are provided on both ends of the printed circuit board 60.
Is attached, and the connector 61a for the connecting portion is connected to the connecting portion A4.
Connector for connecting portion and connector for connecting portion C4, connector for connecting portion 61b corresponds to connector for connecting portion B4 and connector for connecting portion D4, and wiring is provided between the connector for connecting portion 61a and the connector for connecting portion 61b. As a result, the first A connecting means 69a and the first B connecting means 69b are formed. If the first A connecting means 69a and the first B connecting means 69b are constituted by the printed circuit board 60 like the circuit module, it is possible to use a shield layer as an inner layer, so that the line impedance can be easily matched. Moreover, it is possible to suppress the effects of unnecessary radiation noise and external noise.

In FIG. 46, the wiring extending from the node N00 to the right is the connection portion A of the node N00 → the block BK0.
4 (connecting portion connector 61a) → first A connecting means 69a between block BK0 to block BK1 → block BK
1 coupling part B4 (connecting part connector 61b) → node N
The node N00 to the node N10 are coupled by the route of 10. Between node N01 and node N11, node N02
-Node N12, node N03-node N13, node N20-node N30, node N21-node N
31, node N22 to node N32, and node N
23 to the node N33 are also coupled in the same manner.

In FIG. 46, the wiring extending downward from the node N00 is the node C00 → the coupling portion C of the block BK0.
4 (connector 61a for coupling portion) → first B connecting means 69b between blocks BK0 to B3 → block BK3
Connection part D4 (connection part connector 61b) → node N0
The nodes N00 to N01 are connected by the route of 1. Between node N03 and node N02, between node N30 and node N31, between node N33 and node N32, between node N10 and node N11, node N13 and node N12
Between nodes N20 and N21 and between nodes N23
~ Nodes N22 are also coupled in the same manner.

Both ends of the junction A4 and the junction B4 of the array of circuit blocks are connected by the second A connecting means 70a, and both ends of the junction C4 and the junction D4 are connected to the second B.
They are joined by the connecting means 70b. These provide the coupling between the end circuit block and its adjacent circuit blocks.

In FIG. 46, the wiring extending to the left from the node N00 is a node N00 along the route of the node N00 → the connecting portion B4 of the block BK0 → the connecting means 70a of the second A → the connecting portion B4 of the block BK0 → the node N30. ~ The nodes N30 are coupled (x = 0). Between node N03 and node N33 (x = 3), node N01 to node N31
Between the nodes N02 and N32 (x = 1)
2), node N10 to node N20 (x = 0), node N13 to node N23 (x = 3), node N11 to
The nodes N21 (x = 1) and the nodes N12 to N22 (x = 2) are also coupled in the same manner.

In FIG. 46, the wiring extending upward from the node N00 has a path of the node N00 → the connecting portion D4 of the block BK0 → the connecting means 70b of the second B → the connecting portion D4 of the block BK0 → the node N03. ~ Nodes N03 (y = 0) are connected. Between node N30 and node N33 (y = 3), node N10 to node N13
Between nodes (y = 1), between node N20 and node N23 (y =
2), nodes N01 to N02 (y = 0), nodes N31 to N32 (y = 3), nodes N11 to N11
The node N12 (y = 1) and the node N21 to the node N22 (y = 2) are also coupled in the same manner.

In the array of circuit blocks connected in the above-described structure, the nodes are connected as shown in FIG. 48.
It can be seen that they are connected in a ring (two-dimensional torus) like.

All inter-node links are physically a single first A connecting means 69a, first B connecting means 69b,
Second A connecting means 70a or second B connecting means 70b
The wiring length is substantially uniform. Moreover, the total line length in these four directions is relatively short, that is, approximately one circuit block, and does not depend on the total number of arrays. Therefore, the timing adjustment during the transmission of the high frequency signal becomes easy.

Moreover, the first A connecting means 69a, the first B connecting means 69b, the second A connecting means 70a and the second B connecting means 70b do not have to be cables, but rather as shown in FIG.
The printed circuit board 60 with the stack connection connector as shown in FIG. 4 can achieve both the characteristics of high-density wiring and high-frequency wiring with less noise influence.

Of course, the present invention is not applicable only when the cable is not used, and the connecting means 69 of the first A is used.
a, first B connecting means 69b, second B connecting means 70a
Even if the second connecting means 70b of the second B is a cable, the cable length is relatively shorter when arranged and connected as shown in FIG. 48 according to the present invention than when simply arranged and connected as shown in FIG. It is effective because it can be aligned in length.

Also when the circuit block is added, the second A connecting means 70a or the second B connecting means 70 at the end portion is added.
b and remove the first A connecting means 69a or the first B
Since the circuit block is replenished by the structure using the connecting means 69b and the end portion thereof is connected to the second A connecting means 70a or the second B connecting means 70b, it is possible to easily expand the number, and thus the expandability is excellent. ing.

Here, the first A connecting means 69a and the first B connecting means 69b in the embodiment of FIG. 46 are treated as different means for the sake of explanation, but in reality they are exactly the same means (the same thing). Is possible, as well as the second A
The connecting means 70a of No. 2 and the connecting means 70b of No. 2B are handled as different means for the sake of explanation, but in reality, it is possible to use completely the same means (the same thing).

As a fourteenth embodiment, which is an extension of the thirteenth embodiment in the n-dimensional direction, there are three-dimensional torus coupling and n-dimensional torus coupling (n is an arbitrary natural number). The three-dimensional torus coupling is a method of arranging the circuit modules in the coupling of the embodiments shown in FIGS. 42 and 46 in one dimension (that is, m = 1) and then mounting the two-dimensional circuit modules in one circuit module, The circuit module is two-dimensional (ie m
= 2) and then mounting a one-dimensional circuit module in one circuit module or a circuit module arranged three-dimensionally (that is, m = 3) and then one circuit module in one dimension. It is realized by a method of mounting less than the required circuit blocks.

Similarly for the n-dimensional arrangement, the circuit modules are arranged one-dimensionally and then (n-
1) A method of mounting a circuit block corresponding to a dimension, a method of arranging a circuit module in two dimensions and then mounting a circuit block corresponding to an (n-2) dimension on one circuit module, and a method of tertiary circuit module It is realized by using a method in which the circuit blocks corresponding to the (n-3) dimension are mounted on one circuit module after being originally arranged.

Further, n-dimensional circuit blocks may be mounted on one circuit module as long as the size of the circuit module is not limited. Generally, when n-dimensional circuit blocks are arranged to form an n-dimensional torus combination, the number of nodes per circuit block is 2 n.

FIG. 50 is a diagram showing an electronic system mounting method according to the fourteenth embodiment of the present invention. In this embodiment, six main substrates 80A, 80B, 80C, 80D, 80E, and 80F arranged in parallel in the normal direction of the main substrate at four intervals are four vertical connection substrates 83A, 83B, 83C, and 83D, and They are connected by two vertical short-circuit boards 84A and 84B, and power supply and physical fixing to the main board 80 are performed by using four power supply / metal spacers 82 between the main boards. The shape of the metal spacer 82 also serving as a power supply may be an octagonal column or a column as shown in FIG. At this time, generally, a prism such as an octagonal prism is easier to work. Further, in the present embodiment, a substrate for vertically connecting the adjacent main substrates 80 with one or more main substrates 80 interposed therebetween is referred to as a vertical connection substrate 83, and a substrate for vertically connecting the immediately adjacent main substrates 80 is used. The vertical short circuit board 84 is used. Therefore, as shown in FIG. 54, which will be described later, it can be connected to any main board 80, and the configuration of the torus network becomes easy.

FIG. 51 shows a main substrate 80 used in this embodiment.
Shows the configuration of. Although the main board 80 of this embodiment is specifically assumed to be a processing element of a massively parallel computer, the electronic circuit of the main board 80 may be another electronic circuit. The processing element of the massively parallel computer is provided with a CPU in charge of operations, a memory for storing data and programs, and a router for processing communication with other processing elements. The main board 80 includes a main board in addition to the processing elements. A right-angle type connector 85, which is connected to the vertical connection substrate 83 or the vertical short circuit substrate 84, is provided on the front and back surfaces of the main substrate 80 at the edge of the main substrate 80.

FIG. 52 shows the structure of the vertical connection substrate 83 used in this embodiment, and FIG. 53 shows the structure of the vertical short circuit substrate 84. The vertical connection board 83 and the vertical short-circuit board 84 are made of a flexible board such as a flexible board or a plastic board, and are socket type connectors to be connected to the connectors on the front and back surfaces of the main board at the positions in the figure. 86
Is provided. The vertical connection board 83 and the vertical short-circuit board 84 are made to have flexibility, and the vertical connection board 83 and the vertical short-circuit board 84 are slightly bent at the time of mounting to prevent misalignment between the main boards arranged in parallel. Mounting error can be absorbed. Further, the vertical connection substrate 83 is bent in a valley fold at the broken line portion in the figure.

Further, when the connection is not made in the horizontal direction, it is not necessary to fold the substrate, and it suffices to direct the bent portion in this embodiment to the center of the substrate. However, as in the sixth embodiment. When assembling a three-dimensional torus network system, horizontal connections are essential. As shown in FIG. 50, the horizontal connection is made at an arbitrary position near the center of the side of the main substrate 80 as in the sixth embodiment.
This horizontal connection is performed using the horizontal connection board 44 connected to the right angle connector 45. Therefore, by bending the vertical connection board 83, interference with the horizontal connection board 44 can be easily avoided, and the degree of freedom in design can be increased.

Further, at this time, the part of the periphery of the main board 80 that can be used for vertical connection is mainly a corner, and therefore, by bending the vertical connection board 83, the length of one side of the main board 80 is longer than that when not bent. Can be shortened, and the overall outer diameter can be reduced.

To assemble the electronic system shown in FIG. 50, one connector 86R of the vertical connection board 83A is connected to the connector 85R on the back surface of the first main board 80A, and the connector on the front surface of the first main board 80A is connected. One connector 86 of the vertical short circuit board 84A is connected to 85S. Next, the vertical connection substrate 8 is attached to the connector 85R on the back surface of the second main substrate 80B.
After connecting one connector 86R of 3B, the first main board 80A and the second main board 80B are physically connected and fixed using the power supply pad 81 and the metal spacer 82 for both power supply and It should be electrically connected to supply power. Next, the connector 85 on the surface of the first main board 80A
The remaining other connector 86 of the vertical short circuit board 84A connected to S is connected to the connector 85S on the surface of the second main board 80B.

When connecting the third main board 80C,
After connecting one connector 86R of the vertical connection board 83C to the connector 85R on the back surface of the main board 80C, the second main board 80B and the third main board 80C are connected to the power supply pad 81.
And a metal spacer 82 for both power supply and power supply are used to physically connect, fix, and electrically connect to supply power. Next, the connector 8 on the back surface of the first main board 80A
The remaining connector 86S of the vertical connection board 83A connected to 5R is connected to the connector 85S on the surface of the third main board 80C.

Similarly, when connecting the main board 80, one connector 86 of the new vertical connection board 83 is connected to the connector 85R on the back surface of the main board 80, and the upper and lower adjacent main boards are connected to each other on the main board 80. The power supply pad 81 is used to physically connect and fix the metal spacer 82 for both power supply and secure the power supply to the main board 80, and jump to one and connect to the connector 85R on the back surface of the lower main board 80. The system can be constructed by connecting the other connector 86 of the remaining vertical connection board 83 to the surface of the main board 80.

However, regarding the last two main substrates 80, the adjacent short main substrates 80 must be connected by using the vertical short circuit substrate 84 instead of the vertical connection substrate 83. For example, in the case of FIG. 50, when mounting the fifth main board 80E,
One connector 86 of the vertical short circuit board 84B is connected to the connector 85R on the back surface of the fifth main board 80E, and the fourth main board 80D and the fifth main board 80E are connected by using the metal spacer 82 that also serves as a power supply. Physically fixed and main substrate 8
Connect power to 0. Further, the connector 85R on the back surface of the fifth main board 80E is connected to the remaining one connector 86S of the vertical connection board 83C connected to the connector 85R on the back surface of the third main board 80C. When connecting the final sixth main board 80F, first, the fifth main board 80E and the sixth main board 80F are physically and electrically fixed to each other by using the metal spacer 82 also serving as a power supply. Fifth main substrate 8
The remaining connector 86 of the vertical short circuit board 84B connected to the connector 85S on the front surface of 0E is connected, and the connector 85S on the surface of the sixth main board 80F is connected to the fourth main board 80D.
This is performed by connecting the remaining connectors 86S of the vertical connection board 83D that are connected to the connectors 85R on the back surface of the.

FIG. 54 is a view showing the connection relationship between the main boards of the electronic system shown in FIG. 50. The first main board 80 is shown in FIG.
A, third main substrate 80C, fifth main substrate 80E, sixth main substrate 80F, fourth main substrate 80D, second main substrate 80
In the order of B, they are connected as a torus network.

When the main board 80 to be connected in this embodiment is added, the vertical short circuit board 84B is removed and the vertical connection board 83 is connected instead, and the main board 80 is connected to the vertical connection board 8.
3 to connect one by one, and when the last two main boards 80 are connected, the vertical short circuit board 84 is used instead of the vertical connection board 83 to connect the last two main boards. By doing so, the torus network of the main substrate is completed. As described above, in this embodiment, the main substrate 80 can be connected by a torus network of any size. In addition, since errors due to the mounting accuracy of the connector and the like are absorbed by the flexible vertical connection board and the vertical short-circuit board, mounting can be performed without problems even in a large-scale system.

[0154]

As described above, according to the first and second inventions of the present application, the flexible expandability in the three-dimensional direction is realized, and the components of a large-scale electronic device are compared with the conventional system. It enables to connect with a large number of wirings. By using abundant inter-module wiring using this, it is possible to easily implement a massively parallel computer with a connection network such as a three-dimensional mesh connection with high interprocessor communication capability. Further, according to the present invention, efficient cooling and power supply are not hindered. Furthermore, since the wiring is mounted by the connector and the printed circuit board, it is possible to realize a high-frequency and highly reliable connection as compared with a cable or pressure bonding using an anisotropic conductive rubber.

Further, in the third and fourth inventions, when a large-scale parallel computer constituted by a system having a ring-like structure in the connection form of nodes (n-dimensional torus connection) is mounted, long-distance wiring is implemented. It is easy to construct a high-speed link that has a wide transfer width and is not easily affected by noise without degrading the performance of the entire system. Furthermore, the system can be modularized to facilitate expansion, and the circuit modules can be arranged three-dimensionally (three-dimensionally), which makes it possible to construct a massively parallel computer with a compact size. .

[Brief description of drawings]

FIG. 1 is a configuration diagram showing a substrate to which a substrate mounting method according to a first exemplary embodiment of the present invention is applied.

FIG. 2 is an explanatory diagram showing an example of connecting substrates in a horizontal direction.

FIG. 3 is a diagram showing a configuration example of a main board used in the first embodiment.

FIG. 4 is a diagram showing a configuration example of a horizontal connection board.

FIG. 5 is a diagram showing a power supply bar having a protrusion.

FIG. 6 is a view showing a power supply bar having no protrusion.

FIG. 7 is a diagram showing a state in which a power supply bar is connected to a support.

FIG. 8 is an explanatory diagram showing a state in which two main substrates are connected by a horizontal connection substrate.

FIG. 9 is an explanatory diagram showing a region out of reach of a worker when the power supply bar is arranged horizontally.

FIG. 10 is a diagram showing a state in which a power supply bar is arranged vertically.

FIG. 11 is an explanatory diagram showing a massively parallel computer coupled with a two-dimensional mesh.

FIG. 12 is a configuration diagram of a main board in which horizontal connection board connectors are provided in four directions.

FIG. 13 is a diagram showing the back surface of the main substrate.

FIG. 14 is a configuration diagram of a vertical connection board.

FIG. 15 is an explanatory diagram showing a state of connection between a vertical connection board and a vertical connection connector.

FIG. 16 is an explanatory diagram showing a power supply bar of two layers or more.

FIG. 17 is a configuration diagram according to a fourth embodiment of the present invention.

FIG. 18 is a configuration diagram showing a modification of the fourth embodiment.

FIG. 19 is a configuration diagram showing a modification of the fourth embodiment.

FIG. 20 is a configuration diagram showing a fifth embodiment of the present invention.

FIG. 21 is a configuration diagram showing a sixth embodiment of the present invention.

FIG. 22 is a diagram showing a configuration example of a main board according to a sixth embodiment.

FIG. 23 is a diagram showing a configuration example of a horizontal connection board.

FIG. 24 is a diagram showing a configuration example of a vertical connection substrate.

FIG. 25 is a view showing a metal spacer that also serves as a power supply.

FIG. 26 is a diagram showing how the main board and the vertical connection board are connected.

FIG. 27 is an explanatory diagram when a horizontal connection board is configured using a rigid flexible board.

FIG. 28 is an explanatory diagram when a horizontal connection substrate is formed using a plastic substrate.

FIG. 29 is an explanatory diagram when a vertical connection board is configured using a rigid flexible board.

FIG. 30 is an explanatory diagram when a vertical connection substrate is formed using a plastic substrate.

FIG. 31 is an explanatory diagram showing a connection state between a main board and a vertical connection board.

FIG. 32 is a configuration diagram of a system to which the electronic system mounting method according to the ninth exemplary embodiment of the present invention is applied.

FIG. 33 is a configuration diagram showing a circuit module according to a ninth embodiment.

FIG. 34 is a configuration diagram showing a first connection board.

FIG. 35 is a diagram showing a physical link connection between nodes according to the ninth embodiment.

FIG. 36 is a diagram showing logical link connection of inter-node connection according to the ninth embodiment.

FIG. 37 is an explanatory diagram showing a state where boards are connected by an angle type connector.

FIG. 38 is a configuration diagram showing a tenth embodiment of the present invention.

FIG. 39 is a configuration diagram showing a circuit module according to a tenth embodiment.

FIG. 40 is a diagram showing physical link connection of inter-node connection according to the tenth embodiment.

FIG. 41 is a diagram showing logical link connection of inter-node connection according to the tenth embodiment.

FIG. 42 is a configuration diagram showing a twelfth embodiment of the present invention.

FIG. 43 is a configuration diagram showing a circuit module according to a twelfth embodiment.

FIG. 44 is a diagram showing physical link connection of inter-node connection according to the twelfth embodiment.

FIG. 45 is a diagram showing logical link connection of inter-node connection according to the twelfth embodiment.

FIG. 46 is a configuration diagram showing a thirteenth embodiment of the present invention.

FIG. 47 is a configuration diagram showing a circuit module according to a thirteenth embodiment.

[Fig. 48] Fig. 48 is a diagram showing physical link connection of inter-node connection according to the thirteenth embodiment.

FIG. 49 is a diagram showing logical link connection of inter-node connection according to the thirteenth embodiment.

FIG. 50 is a mounting view of the electronic system according to the fourteenth embodiment of the present invention.

FIG. 51 is a diagram showing a configuration example of a main board according to the fourteenth embodiment.

FIG. 52 is a diagram showing a configuration example of a vertical connection board according to a fourteenth embodiment.

FIG. 53 is a diagram showing a configuration example of a vertical short circuit substrate according to a fourteenth embodiment.

FIG. 54 is a diagram showing logical connections in the mounting diagram of the electronic system according to the fourteenth embodiment.

[Explanation of symbols]

 1 Power Supply Bar 2 Vertical Connection Connector 3 Horizontal Connection Connector 4 Horizontal Connection Board 5 Main Board 7 Vertical Connection Board 16 Projection 17 Main Board Side Connector 18 Horizontal Board Side Connector 23 Communication Path 24 Processing Element 25 Angle Type Connector 26 Cooling Fan 28 Insulating primary cooling liquid 29 Distribution pipe 30 Circulation pump 31 Cooling fin 32 Heat exchanger 33 Secondary cooling water injection port 34 Secondary cooling water drainage port 35 Metal hat 36 Cooling water pipe 37 Cooling water pipe crimp groove 40 Piston 41 Chip crimping surface 43 Vertical connection board 44 Horizontal connection board 45 Right angle connector 46 Socket connector 49 Rigid part 50 Flexible part 51 Reinforcing plate 52 Plastic board 53 Circuit block 54 First connecting means 55 Second connecting means 64 Link 67 sixth Connecting means 68 the second connection means 69 first connection means 70 the second connecting means 71 children substrate 80 main board 81 power supply pads 82 power supply combined metal spacer 83 vertical connection board 84 vertically short board 85 connector 86 connector

Continuation of the front page (51) Int.Cl. ⁵ Identification number Reference number within the agency FI Technical indication location H01R 23/68 303 H 6901-5E 7165-5B G06F 1/00 360 C

Claims (4)

[Claims] 1. A board mounting method for forming a board of a predetermined size by arranging a plurality of main boards side by side in a horizontal direction, wherein the main board surface is attached to and detached from a peripheral portion in a direction perpendicular to the main board surface. A possible first connector is provided, and adjacent horizontal main boards are connected using a horizontal direction connection board having a second connector which can be fitted with the first connector on the periphery of the surface, and the main boards adjacent to each other are connected to each other. A board mounting method characterized by creating a board. 2. The board mounting method according to claim 1,
A third connector is provided on the front surface and the back surface of each main board,
A plurality of main boards are arranged in parallel in a normal direction of the main board at a predetermined interval, and each of the main boards arranged in parallel has a fourth connector which can be fitted with the third connector on the surface periphery. A board mounting method characterized by being connected using a vertical connection board. 3. A plurality of circuit blocks each having a node
In an electronic system mounting method for mounting a circuit module configured by torus coupling in the dimension direction (n ≦ 3),
The circuit block has terminals T1 and T2 connected to both ends of a node and a terminal T for each dimension.
Terminals T3, T of the short-circuit wire arranged in parallel with the terminals T1, T2
4 and the terminals T1 and T3 of the circuit block which are not located at the ends are respectively connected to the terminals T4 and T2 of the circuit block adjacent to one side, and the terminals T2 and T4 are the terminals of the circuit block adjacent to the other side. The electronic system mounting method is characterized in that the circuit blocks respectively connected to T3 and T1 and located at the ends short-circuit the terminals T1, T3 or T2, T4 on the side having no adjacent circuit block. 4. An electronic system mounting method for mounting a circuit module configured by torus-coupling circuit blocks having 2 ⁿ (where n is the number of dimensions) nodes in an n-dimensional direction (n ≦ 3), Each block has an inter-node connection terminal for each node in each dimension, and the inter-node connection terminals of circuit blocks that are not located at the ends are connected to one of the inter-node connection terminals of adjacent circuit blocks. An electronic system mounting method characterized in that the circuit blocks located at the ends are short-circuited to each other to form the torus of the inter-node connection terminal where the adjacent circuit block does not exist.