FIELD OF THE INVENTIONThe present invention relates to technology for manufacturing a semiconductor and, particularly, to technology that can be effectively applied to highly densely mounting the semiconductor chips on a memory-module.[0001]
BACKGROUND OF THE INVENTIONThe technology described below is the one discussed by the present inventors in studying and accomplishing the present invention, and is roughly as described below.[0002]
A memory-module is one of the module products mounting a plurality of semiconductor devices.[0003]
The memory-module includes a plurality of semiconductor devices having memory chips that are mounted on one surface or on both the front and back surfaces of a module board. In mounting the memory on a personal computer or a work station, the memory-module mounts the memory by being mounted on a mother board provided in the personal computer or the work station with each module as a unit.[0004]
As the semiconductor device mounted on the memory-module, there has been used the one of the surface mount type called SMD (surface mount device) having a semiconductor chip sealed with a resin and having lead terminals (external terminals) for drawing the electrodes to the outside of the resin-sealed portion, as represented by TSOP (thin small outline package) and TCP (tape carrier package).[0005]
Module products of various structures have been disclosed in, for example, Japanese Patent Laid-Open Nos. 209368/1998, 258466/1989 and 86492/1995.[0006]
Japanese Patent Laid-Open No. 209368/1998 discloses a CPU (central processing unit) module, and Japanese Patent Laid-Open No. 258466/1989 discloses a memory-module mounting SMD parts having a DRAM (dynamic random access memory) chip. Japanese Patent Laid-Open No. 86492/1995 discloses technology for applying an underfiller resin in the MCM (multi-chip module).[0007]
SUMMARY OF THE INVENTIONThe SMD parts to be mounted on the above-mentioned conventional memory-module have a large package size compared with the chip size due to the package body (semiconductor device body) that is sealed and the outer leads.[0008]
As a result, limitation is imposed on the number of the semiconductor chips that can be mounted on the module board.[0009]
There further arouses a problem in that due to the inductance added as a result of the sealing, it becomes difficult to design a memory-module having a high-speed interface to meet a high-speed CPU.[0010]
The object of the present invention is to provide a memory-module that offers an increased module capacity as a result of enhancing the density for mounting the semiconductor chips and is capable of coping with a high-speed bus, and a method of manufacturing the same.[0011]
The above and other objects as well as novel features of the present invention will become obvious from the description of the specification and the attached drawings.[0012]
Briefly described below are representative examples of the inventions disclosed in this application.[0013]
That is, the memory-module of the present invention comprises protruded terminal semiconductor devices having protruded terminals as external terminals, mounted via the protruded terminals, and are provided with wiring portions for expanding the pitch among the protruded terminals to be wider than the pitch among the bonding electrodes of semiconductor chips; lead terminal semiconductor devices having outer leads as external terminals, and are mounted via the outer leads that are electrically connected to the bonding electrodes of the semiconductor chips; and a module board supporting the protruded terminal semiconductor devices and the lead terminal semiconductor devices; wherein the protruded terminal semiconductor devices and the lead terminal semiconductor devices are mounted in a mixed manner on the module board.[0014]
Further, the memory-module of the invention comprises protruded terminal semiconductor devices of a chip size having protruded terminals as external terminals, mounted via the protruded terminals, and are provided with rewirings which are wiring portions for expanding the pitch among the protruded terminals to be wider than the pitch among the bonding electrodes in the areas of semiconductor chips; lead terminal semiconductor devices having outer leads as external terminals, and are mounted via the outer leads that are electrically connected to the bonding electrodes of the semiconductor chips; and a module board supporting the protruded terminal semiconductor devices and the lead terminal semiconductor devices; wherein the protruded terminal semiconductor devices and the lead terminal semiconductor devices are mounted in a mixed manner on the module board.[0015]
In mounting the protruded terminal semiconductor devices together with the lead terminal semiconductor devices in a mixed manner, therefore, the mounting is accomplished requiring mounting areas nearly equal to those of the semiconductor chips.[0016]
Therefore, the semiconductor chips can be mounted requiring the least areas, making it possible to increase the density for mounting the semiconductor chips.[0017]
This makes it possible to increase the module capacity of the memory-module.[0018]
The method of manufacturing a memory-module according to the present invention comprises a step for preparing protruded terminal semiconductor devices having protruded terminals as external terminals, and wiring portions for expanding the pitch of the protruded terminals to be wider than the pitch of the bonding electrodes of semiconductor chips; a step for preparing lead terminal semiconductor devices having outer leads which are the external terminals electrically connected to the bonding electrodes of the semiconductor chips; a step for arranging the protruded terminal semiconductor devices and the lead terminal semiconductor devices on a module board; and a step for simultaneously reflowing the protruded terminal semiconductor devices and the lead terminal semiconductor devices to mount them on the module board; wherein the protruded terminal semiconductor devices and the lead terminal semiconductor devices are mounted in a mixed manner on the module board.[0019]
Further, the method of manufacturing a memory-module of the present invention comprises a step for preparing protruded terminal semiconductor devices of a chip size having protruded terminals as external terminals, and rewirings which are wiring portions for expanding the pitch of the protruded terminals to be wider than the pitch of the bonding electrodes in the areas of semiconductor chips; a step for preparing lead terminal semiconductor devices having outer leads which are external terminals electrically connected to the bonding electrodes of the semiconductor chips; a step for arranging the protruded terminal semiconductor devices and the lead terminal semiconductor devices on a module board; and a step for simultaneously reflowing the protruded terminal semiconductor devices and the lead terminal semiconductor devices to mount them on the module board; wherein the protruded terminal semiconductor devices and the lead terminal semiconductor devices are mounted in a mixed manner on the module board.[0020]
BRIEF DESCRIPTION OF THE DRAWINGSFIGS. 1A, 1B and[0021]1C are diagrams illustrating the structure of a memory-module according to anembodiment 1 of the present invention, wherein FIG. 1A is a plan view, FIG. 1B is a side view, and FIG. 1C is a sectional view along the A-A section of FIG. 1A;
FIG. 2 is an enlarged partial sectional view illustrating, on an enlarged scale, a portion B in the sectional view of FIG. 1C;[0022]
FIG. 3 is a view of block circuits of the memory-module shown in FIG. 1;[0023]
FIG. 4 is a perspective view illustrating the appearance of the structure of a wafer process package (protruded terminal semiconductor device) mounted on the memory-module shown in FIG. 1;[0024]
FIGS. 5A and 5B are diagrams illustrating an SMD (lead terminal semiconductor device) mounted on the memory-module shown in FIG. 1 and the structure of a wafer process package, wherein FIG. 5A is a plan view of the SMD and FIG. 5B is a plan view of the wafer process package;[0025]
FIG. 6 is a process flow illustrating the steps for manufacturing the wafer process package mounted on the memory-module shown in FIG. 1;[0026]
FIGS. 7A, 7B,[0027]7C,7D,7E and7F are enlarged partial sectional views illustrating the structure of the semiconductor wafer corresponding to the major steps in the process flow shown in FIG. 6;
FIG. 8 is a basic mounting flow illustrating the procedure for mounting the wafer process package and the SMD on the module board so as to be mounted on the memory-module shown in FIG. 1;[0028]
FIG. 9 is a mounting flow illustrating the procedure for mounting the wafer process package on the module board so as to be mounted on the memory-module shown in FIG. 1;[0029]
FIG. 10 is an enlarged partial perspective view illustrating a method of applying an underfiller resin onto the wafer process package mounted on the memory-module shown in FIG. 1;[0030]
FIGS. 11A, 11B,[0031]11C,11D,11E,11F,11G and11H are views illustrating the permeation of the underfiller resin that is applied as shown in FIG. 10, and wherein FIGS. 11A, 11C,11E and11G are perspective views and FIGS. 11B, 11D,11F and11H are plan views showing the semiconductor chip in a see-through manner;
FIGS.[0032]12 is a plan view illustrating a modified structure of the memory-module of theembodiment 1 of the present invention;
FIG. 13 is a plan view illustrating a modified structure of the memory-module of the[0033]embodiment 1 of the present invention;
FIGS. 14A, 14B,[0034]14C,14D,14E,14F,14G and14H are views illustrating the permeation of the underfiller resin that is applied according to a modified example of theembodiment 1 of the invention, and wherein FIGS. 14A, 14C,14E and14G are perspective views and FIGS. 14B, 14D,14F and14H are plan views illustrating a semiconductor chip in a see-through manner;
FIGS. 15A and 15B are view illustrating a modified structure of the memory-module according to the[0035]embodiment 1 of the present invention, wherein FIG. 15A is a plan view and FIG. 15B is a side view;
FIG. 16 is a side view illustrating the memory-module of FIG. 15 in a warped state;[0036]
FIG. 17 is a plan view illustrating a modified structure of the memory-module according to the[0037]embodiment 1 of the present invention;
FIG. 18 is a side view illustrating the memory-module of FIG. 17 in a warped state;[0038]
FIG. 19 is a plan view illustrating the structure of the memory-module according to an[0039]embodiment 2 of the present invention;
FIGS. 20A and 20B are views illustrating the structure of a memory-module according to an[0040]embodiment 3 of the present invention, wherein FIG. 20A is a plan view and FIG. 20B is a side view;
FIG. 21 is a diagram of block circuits of the memory-module shown in FIG. 20;[0041]
FIG. 22 is a bottom view illustrating the structure of a wafer process package (protruded terminal semiconductor device) mounted on the memory-module shown in FIG. 20;[0042]
FIG. 23 is a diagram of wirings on the side of the board illustrating an example of wirings on the module board at a portion C in the memory-module shown in FIG. 20;[0043]
FIG. 24 is a diagram of wirings illustrating a modified example of the bump arrangement on the wafer process package in the memory-module according to the[0044]embodiment 3 of the invention and a modified example of the wirings on the side of the board corresponding thereto;
FIG. 25 is a diagram of wirings illustrating a modified example of the bump arrangement on the wafer process package in the memory-module according to the[0045]embodiment 3 of the invention and a modified example of the wirings on the side of the board corresponding thereto;
FIG. 26 is a diagram of wirings illustrating a modified example of the bump arrangement on the wafer process package in the memory-module according to the[0046]embodiment 3 of the invention and a modified example of the wirings on the side of the board corresponding thereto;
FIG. 27 is a diagram of bump arrangement and wirings illustrating a further modified example of the bump arrangement on the wafer process package and of the wirings on the side of the board shown in FIG. 25;[0047]
FIGS. 28A, 28B and[0048]28C are views illustrating the structure of a CSP which is a modified example of the protruded terminal semiconductor device mounted on the memory-module of the present invention, wherein FIG. 28A is a plan view, FIG. 28B is a sectional view and FIG. 28C is a bottom view;
FIGS. 29A and 29B are views illustrating the structure of a BGA of the chip face-up mounting system which is a modified example of the protruded terminal semiconductor device mounted on the memory-module of the present invention, wherein FIG. 29A is a perspective view illustrating the appearance and FIG.[0049]29B is a sectional view; and
FIGS. 30A, 30B and[0050]30C are views illustrating the structure of the BGA of the chip face-down mounting system which is a modified example of the protruded terminal semiconductor device mounted on the memory-module of the present invention, wherein FIG. 30A is a plan view, FIG. 30B is a sectional view and FIG. 30C is a bottom view.
DESCRIPTION OF THE PREFERRED EMBODIMENTSEmbodiments of the present invention will now be described in detail with reference to the drawings.[0051]
(Embodiment 1)[0052]
FIG. 1 is a diagram illustrating the structure of a memory-module according to an[0053]embodiment 1 of the present invention, wherein FIG. 1A is a plan view, FIG. 1B is a side view and FIG. 1C is a sectional view along the A-A section of FIG. 1A, FIG. 2 is an enlarged partial sectional view illustrating, on an enlarged scale, a portion B in the sectional view of FIG. 1C, FIG. 3 is a view of block circuits of the memory-module shown in FIG. 1, FIG. 4 is a perspective view illustrating the appearance of the structure of a wafer process package (protruded terminal semiconductor device) mounted on the memory-module shown in FIG. 1, FIG. 5 is a diagram illustrating an SMD (surface mount-type semiconductor device having lead terminals, which is hereinafter referred to as lead terminal semiconductor device) mounted on the memory-module shown in FIG. 1 and the structure of a wafer process package, wherein FIG. 5A is a plan view of the SMD and FIG. 5B is a plan view of the wafer process package, FIG. 6 is a process flow illustrating the steps for manufacturing the wafer process package mounted on the memory-module shown in FIG. 1, FIGS. 7A, 7B,7C,7D,7E and7F are enlarged partial sectional views illustrating the structure of the semiconductor wafer corresponding to the major steps in the process flow shown in FIG. 6, FIG. 8 is a basic mounting flow illustrating the procedure for mounting the wafer process package and the SMD on the module board so as to be mounted on the memory-module shown in FIG. 1, FIG. 9 is a mounting flow illustrating the procedure for mounting the wafer process package on the module board so as to be mounted on the memory-module shown in FIG. 1, FIG. 10 is an enlarged partial perspective view illustrating a method of applying an underfiller resin onto the wafer process package mounted on the memory-module shown in FIG. 1, FIGS. 11A, 11C,11E and11G are perspective views illustrating the permeation of the underfiller resin that is applied as shown in FIG. 10 and FIGS. 11B, 11D,11F and11H are plan views showing the semiconductor chip in a see-through manner, FIGS. 12 and 13 are plan views illustrating the modified structures of the memory-module of theembodiment 1 of the present invention, FIGS. 14A, 14C,14E and14G are perspective views illustrating the permeation of the underfiller resin that is applied according to modified examples of theembodiment 1 of the invention and FIGS. 14B, 14D,14F and14H are plan views illustrating a semiconductor chip in a see-through manner, FIG. 15 is a view illustrating a modified structure of the memory-module according to theembodiment 1 of the present invention, wherein FIG. 15A is a plan view and FIG. 15B is a side view, FIG. 16 is a side view illustrating the memory-module of FIG. 15 in a warped state, FIG. 17 is a plan view illustrating a modified structure of the memory-module according to theembodiment 1 of the present invention, and FIG. 18 is a side view illustrating the memory-module of FIG. 17 in a warped state.
A memory-[0054]module100 of theembodiment 1 shown in FIG. 1 comprises protruded terminal semiconductor devices having protruded terminals as external terminals, mounted via the protruded terminals and having wiring portions for expanding the pitch among the protruded terminals to be wider than the pitch among the bonding electrodes la of thesemiconductor chips1; TSOPs (thin small outline packages) which are leadterminal semiconductor devices20 havingsemiconductor chips1, outer leads21 as the external terminals and are mounted via the outer leads21 which are electrically connected to thebonding electrodes1aof thesemiconductor chips1; and amodule board2 for supporting the protruded terminal semiconductor devices andTSOPs20; wherein the protruded terminal semiconductor devices and theTSOPs20 are mounted in a mixed manner on themodule board2 by the simultaneous reflowing.
Here, the protruded terminal semiconductor device has a plurality of bump electrodes[0055]11 (protruded terminals) that serve as the external terminals arranged in an area of a package body13 (semiconductor device body), and wiring portions for expanding the pitch among thebump electrodes11 to be wider than the pitch among thebonding electrodes1aof thesemiconductor chip1.
The lead terminal semiconductor device has a plurality of[0056]outer leads21 serving as the external terminals that are arranged protruding from the package body22 (semiconductor device body).
In the protruded terminal semiconductor device and the lead terminal semiconductor device, the[0057]bonding electrodes1aof thesemiconductor chip1 are formed by using, for example, aluminum or the like, and are electrically connected to the bonding wires when the wires are to be bonded.
The external terminals of the protruded terminal semiconductor device and of the lead terminal semiconductor device are electrically connected to the connection electrodes on the side of the[0058]module board2 when the semiconductor devices are mounted on the mounting board such as themodule board2.
The[0059]embodiment1 deals with the case where the protruded terminal semiconductor device is a wafer process package (hereinafter abbreviated as WPP)10 which is a small semiconductor device of a chip size.
Therefore, the memory-[0060]module100 of theembodiment 1 includesWPPs10 which are the protruded terminal semiconductor devices of a chip size,TSOPs20 which are SMD (surface mount type package) parts and are lead terminal semiconductor devices, and an EEPROM (electrically erasable programmable read-only memory)5 which is a nonvolatile read-only memory as an example of another lead terminal semiconductor device, that are mounted in a mixed manner on themodule board2.
Here, as shown in FIG. 4, the[0061]WPP10 is a protruded terminal semiconductor device havingbump electrodes11 which are the protruded terminals serving as the external terminals, and is mounted on themodule board2 viabump electrodes11, and is provided withrewirings12 which are wiring portions for expanding the pitch among thebump electrodes11 to be wider than the pitch among thebonding electrodes1ain an area of thesemiconductor chip1.
The[0062]bump electrodes11 used for theWPP10 have little dispersion in the height, decreasing percent defective when it is mounted on the board and, hence, improve the mounting yield. Besides, thebump electrodes11 have a mounting height of about 0.13 mm, which makes it possible to decrease the mounting height.
Referring to FIG. 1, on the memory-[0063]module100 are further mountedcapacitors3, small surface-attachedresistors4 and other electronic parts in addition to theWPPs10,TSOPS20 andEEPROM5.
That is, the memory-[0064]module100 of theembodiment 1 includes 18WPPs10, twoTSOPs20, 18capacitors3, 36 small surface-attachedresistors4 and oneEEPROM5 that are mounted on either the front surface or the back surface thereof, and 18WPPs10 that are mounted on the surface on the opposite side thereof.
In the memory-[0065]module100 of theembodiment 1, there are arranged the WPPs10 in a total number of 18 in a sequence on both sides of the two TSOPs20 (ten on one side and eight on the other side with theTSOPs20 being sandwiched therebetween) on one surface of themodule board2.
Between the two[0066]TSOPs20, the one (TSOP20 arranged on the upper side in FIG. 1) is a PLL (phase-locked loop)6 which is a frequency control means and the other one (TSOP20 arranged on the lower side in FIG. 1) is aregister8 having a register function.
That is, in the memory-[0067]module100 of theembodiment 1, both thePLL6 and theregister8 are the lead terminal semiconductor devices.
Each[0068]capacitor3 is arranged to correspond to eachWPP10 close thereto.
Further, a total of 36 small surface-attached[0069]resistors4 are arranged in sequence; i.e., two for eachWPP10. The small surface-attachedresistors4 are provided to correspond to the I/Os of the memory-module100. In the memory-module100 of theembodiment 1, there are provided 36 I/Os on one surface and, hence, the surface-attachedresistors4 are mounted in a number of 36. The small surface-attachedresistors4 of the number of 36 are arranged in a sequence nearly along and near theconnection terminals2awhich are the external terminals of themodule board2.
Referring to FIG. 1A, the[0070]module board2 of the memory-module100 measures, for example, L=133.35 mm and M=38.1 mm, and the mounting height (max) is N=4 mm as shown in FIG. 1B.
In the memory-[0071]module100 of theembodiment 1, further, theTSOPs20 and theWPPs10 are mounted by the simultaneous reflowing. As shown in FIG. 2, however, theWPP10 is sealed with the underfiller resin after the reflow, so that a sealingportion14 is formed.
That is, the surrounding of the[0072]bump electrodes11 between thepackage body13 ofWPP10 and themodule board2 is sealed with a resin thereby to form the sealedportion14.
The memory-[0073]module100 shown in FIG. 1 uses theWPPs10 as DRAMs and further uses themodule board2 of a bus of a width of 72 bits with error code correction.
Therefore, the memory-[0074]module100 mounts a total of 36 DRAMs (WPPs10) on both the front and back surfaces of themodule board2. When the DRAM has, for example, 64 megabits (16 M×4), the DRAM module has a constitution of 16 words×72 bits×2 banks.
FIG. 3 is a diagram of block circuits of the memory-[0075]module100 shown in FIG. 1, i.e., the diagram of block circuits of the DRAM module of the constitution of 16 words×72 bits×2 banks.
In the structure of FIG. 3, the RS0 system and the RS2 system of the[0076]bank1 operate simultaneously, and the RS1 system and the RS3 system of thebank2 operate simultaneously. Thebank1 or thebank2 is selected by aregister8. When thebank1 is read out, thebank2 is not read out. Similarly, when thebank2 is read out, thebank1 is not read out.
A terminal A (S[0077]0 to S3) of theregister8 is connected to a chip select (CS) terminal of the DRAM (WPP10) of either thebank1 or thebank2. The bank that is selected by theregister8 forms an input to the CS terminal of the selectedsemiconductor chip1.
D[0078]0 to D35 of each chip represent theWPPs10 of the number of36, and the [I (input)/O (output)0 to I/O3] terminals of each chip are connected to theconnection terminals2aof themodule board2 as independent terminals.
In all DRAMs, the I/Os used as data consist of 64 bits of from DQ[0079]0 to DQ63, and the I/Os used as check consist of 8 bits of from CB0 to CB7. The sum of the two constitutes a two-bank constitution of 72 bits.
Symbols attached to the terminals shown in FIG. 3 are described below. [A[0080]0 to A11] are address inputs, [DQ0 to DQ63] are data inputs/outputs, [CB0 to CB7] are check bits (data inputs/outputs), [S0 to S3] are chip select inputs, [RE] is a row enable (RAS) input, [CE] is a column enable (CAS) input, [W] is a write enable input, [DQMB0 to DQMB7] are bite data masks, [CK0 to CK3] are clock inputs, [CKE0] is a clock enable input, [WP] is a write protection for serial PD, [REGE] is a register enable, [SDA] is a data input/output for serial PD, [SCL] is a clock input for serial PD, [SA0 to SA2] are serial address inputs, [Vcc] is a power source of the high-potential side, [Vss] is a ground, and [NC] is a non-connection.
Next, the structure of the[0081]WPP10 will be described in detail. Referring to FIG. 4, thebonding electrodes1aof thesemiconductor chip1 in theWPP10 are electrically connected to thesolder bump electrodes11 which are the external terminals throughrewirings12.
That is, the[0082]bonding electrodes1aarranged at a narrow pitch are expanded by therewirings12 to a pitch of thebump electrodes11 that are electrically connected thereto.
This is to form a package of a chip size by forming functional portions of the elements in a unit of the wafer and, then, effecting the dicing to divide into[0083]individual semiconductor chips1.
Therefore, the device is efficiently produced at a low cost compared with the small packages assembled by a method of production similar to that of manufacturing packages of the SMD (surface mount type) parts.[0084]
FIG. 5 illustrates the[0085]TSOP20 which is an SMD part and theWPP10 which is the protruded terminal semiconductor device of the chip size, from which a difference in the size can be comprehended.
FIG. 5A is a plan view of the[0086]TSOP20 mounted on the memory-module100 shown in FIG. 1, and FIG. 5B is a plan view of theWPP10 mounted on the memory-module100 shown in FIG. 1.
As shown in FIG. 5, compared to the DRAM of the SMD (surface mount) type package such as[0087]TSOP20, theWPP10 can be realized in a small size since it has neither the inner leads nor the outer leads21.
By mounting the DRAMs in the form of WPPs[0088]10 on themodule board2 as in the memory-module100 of theembodiment 1, therefore, the mounting area can be greatly reduced as compared to when theTSOPs20 that are formed by being individually treated are mounted.
That is, by mounting the[0089]WPPs10, the mounting area can be minimized so far as thesemiconductor chips1 are mounted and, hence, the module capacity can be greatly increased.
The same capacity can be realized even by mounting the flip chips which is the mounting of bare chips. In mounting the flip chips, however, there is formed no[0090]rewiring12. Accordingly, the pitch is small among the external terminals, and it is not allowed to accomplish the mounting by reflow simultaneously with the SMD type parts. Therefore, the parts mounting the bare chips must be mounted one by one by using a flip chip bonder, which is inferior in efficiency to mounting theWPPs10.
In other words, the[0091]WPPs10 are mounted without using any special mounting device such as the flip chip bonder, and makes it possible to decrease the number of the steps for mounting.
Further, the[0092]WPPs10 can be mounted permitting the pitch among thebump electrodes11 that are the external terminals to be expanded to be broader than the pitch of when the flip chips are mounted, enabling the wiring rule to be broadened on themodule board2. This does not drive up the cost of themodule board2, and makes it possible to realize the memory-module100 of a highly dense mounted form suppressing the cost.
In the[0093]WPP10, further, the wiring lengths from thebonding electrodes1aof thesemiconductor chip1 to thebump electrodes11 which are the external terminals become shorter than the wiring lengths from thebonding electrodes1ato the ends of the outer leads21 of the SMD part such asTSOP20, making it possible to transfer signals at a high speed.
This enables the memory-[0094]module100 to operate at high speeds and, hence, to cope with a high-speed bus.
Described below is the reason why the semiconductor devices (packages) mounted on the memory-[0095]module100 of theembodiment 1 are not all in the form of WPPs10, i.e., why theWPPs10 which are the protruded terminal semiconductor devices of the chip size and the SMD parts (TSOPs20 in the embodiment 1) are mounted in a mixed manner.
The[0096]WPPs10 are formed by treating the wafers in the preceding steps. In the subsequent steps, therefore, they are all treated in a unit of the wafer even in a step of forming the devices one by one.
When the number of non-defective products is small in a piece of wafer, the defective products must be worked, driving up the cost.[0097]
As a result, for some kinds of products for which the yields of the semiconductor wafers are not so high, no cost merit is obtained.[0098]
Further, a reticle for exposure to light must be prepared for every kind of products. For the products that are not produced in large quantity, therefore, a material having general applicability is used for the semiconductor devices (packages) that are incorporated in the lead frames. Therefore, the products that are not produced in large quantity are better not in the form of the[0099]WPPs10.
Besides, physical conditions play important roles. From a relationship between the number of the terminals to be drawn out and the chip size, the logical devices in the form of small chips but having many terminals to be drawn out, are better not in the form of the[0100]WPP10. This is because the electrode pads (diffusion-preventingadhesion layers7cshown in FIG. 7) and thebump electrodes11 cannot be formed after therewirings12 are formed from thebonding electrodes1a.
Therefore, the devices that are better formed as WPPs[0101]10 are those chips produced maintaining a high yield and obtained in a large number per a wafer and, particularly, are those small memory devices.
On the other hand, the devices that are not better in the form of WPPs[0102]10 are those chips produced maintaining a low yield and obtained in a small number per a wafer and, particularly, are those large chips, end devices or devices produced in small quantity. Further, when an ASIC (application specific integrated circuit) having many external terminals compared to the chip area, is obtained in the form of theWPP10, a sufficiently large pitch is not often maintained among thebump electrodes11. In this case, too, the package should be in a conventional form for easy mounting.
Next, described below is a method of manufacturing the[0103]WPP10 with reference to a process flow of WPP10 (see FIG. 1) shown in FIG. 6 and sectional views of the wafer shown in FIG. 7 corresponding to the principal steps in the process flow.
First, the wafer is subjected to a pre-treatment at step S[0104]1 shown in FIG. 6. The bonding electrode la is exposed on the main surface of thesilicon board7 shown in FIG. 7A thereby to form an inorganic insulatingprotection film7a.
Then, a WPP first insulating layer is formed at step S[0105]2. That is, as shown in FIG. 7B, the first insulatinglayer7bof polyimide or fluorine-contained resin is formed on the inorganic insulatingprotection film7aof thesilicon board7.
Then, at step S[0106]3, a WPP rewiring layer is formed. That is, as shown in FIG. 7C, arewiring12 is formed on the first insulatinglayer7bbeing electrically connected to thebonding electrode1a.
Then, at step S[0107]4, a WPP second insulating layer is formed. That is, as shown in FIG. 7D, a secondinsulating layer7dcomprising polyimide or epoxy is formed on therewiring12.
Then, at step S[0108]5, a WPP-UBM (underbump metal) is formed. That is, as shown in FIG. 7E, a diffusion-preventingadhesion layer7cwhich is a UBM is formed being electrically connected to therewiring12.
Then, at step S[0109]6, the wafer is inspected (W-test). This is to inspect whether the wafer has been treated as contemplated relying on the electric characteristics by bringing a probe needle into contact with the electrode pad formed on a scribe area of the semiconductor wafer (silicon board7).
Then, at step S[0110]7, thesilicon board7 is inspected by using the probe (P-test 1). This is to detect defective portions by inspecting whether thesemiconductor chip1 electrically works properly by bringing the probe needle into contact with thebonding electrode1aof thesilicon board7.
Then, at step S[0111]8, the defective portions are relieved; i.e., laser blown fuses is executed. This is to relieve defective portions by cutting the fuse in a redundancy circuit by laser beam.
Then, at step S[0112]9, test is effected by using probe (P-test 2). This is to make sure whether the defective portion relieved by the P-test 1 has been corrected.
Then, at step S[0113]10, marking is effected on the back surface of the wafer to attach a predetermined mark to the back surface of thesilicon board7.
Then, a bump is formed at step S[0114]11. That is, as shown in FIG. 7F, a bump electrode11 (protruded terminal) which is an external terminal of theWPP10 is formed on the diffusion-preventingadhesion layer7cwhich is the UBM provided at an end drawn out from thebonding electrode1aon therewiring12.
Here, the[0115]bump electrode11 is formed by, for example, a printing method. A metal mask corresponding to the bump-forming position is disposed on the wafer (silicon board7), a solder paste is applied, the metal mask is removed, followed by reflowing at one time to form thebump electrodes11 at one time on the wafer.
Then, at step S[0116]12, the semiconductor wafer, i.e., thesilicon board7 is cut by dicing, thereby to form theWPP10 as shown in FIG. 4.
Then, at step S[0117]13, theWPP10 is subjected to the aging, i.e., to the burn-in (BI) testing.
At step S[0118]14, the single products are sorted out to selectnon-defective WPPs10.
Thus, the fabrication of the[0119]WPP10 is completed.
In the procedure of production shown in FIG. 6, no back grinding step (hereinafter abbreviated as BG) for grinding the back surface of the[0120]silicon board7 was executed after the test by using the probe (P-test 2) at step S9. However, the BG step may be executed after step S9 of test by using the probe (P-test 2) but before step S10O of marking the back surface of the wafer.
Here, the BG step is to decrease the height of the[0121]WPP10 by decreasing the thickness of thesilicon board7 by grinding the back surface of thesilicon board7.
In other words, this is to decrease the thickness of the[0122]semiconductor chip1 in order to decrease the thickness of theWPP10.
Upon executing the BG step, it is allowed to decrease the height of mounting the WPP[0123]10 (e.g., to decrease to 1 mm or less).
Through the BG step, further, the thickness of the[0124]silicon board7 can be decreased. Even when the scribing width on thesilicon board7 is decreased at the time of dicing to obtain an increased number of the chips, the dicing is effected without hindering the infiltration of the cooling water at the time of dicing into the scribe grooves.
This prevents damage to the[0125]silicon board7 at the time of dicing, and enhances the yield of thesilicon boards7. This is particularly effective at the time of dicing thesilicon board7 having a diameter of 300 mm.
Further, steps S[0126]6 to S9 (test of wafer (W-test), test using probe (P-test 1), relief by laser, test using probe (P-test 2)) in the procedure of production shown in FIG. 6 may be executed between step S1 of putting the wafer to the pre-treatment and step S2 of forming the WPP first insulating layer.
That is, steps S[0127]6 to S9 are executed after the step S1 of putting the wafer to the pre-treatment.
This makes it possible to execute a series of tests using the probe prior to forming the insulating film on the[0128]silicon board7 and to assemble theWPP10 without leaving damage even in case thebonding electrode1ais damaged.
Next, described below with reference to FIGS. 8 and 9 is a method of manufacturing the memory-[0129]module100 shown in FIG. 1 of theembodiment 1.
The memory-[0130]module100 shown in FIG. 1 is obtained by mounting theWPPs10 on both the front and back surfaces of themodule board2 and mounting theTSOPs20 on one surface thereof.
First, the[0131]WPPs10 are produced in compliance with the process flow shown in FIG. 6.
That is, the WPPs[0132]10 (protruded terminal semiconductor devices) of the chip size shown in FIG. 4 are prepared through the pre-treatment of the wafer (prepared in a number of 18×2=36), theWPPs10 having bump electrodes11 (protruded terminals) serving as external terminals, and rewirings12 (wiring portions) in the areas of thesemiconductor chips1 for expanding the pitch among thebump electrodes11 to be wider than the pitch among thebonding electrodes1a.
In the[0133]embodiment 1, thesemiconductor chip1 possessed by theWPP10 is a DRAM.
In addition to the[0134]WPPs10, there are assembled the lead terminal semiconductor devices which are SMD parts to be mounted on themodule board2.
There are prepared two TSOPs[0135]20 (one being aPLL6 and the other being a register8) which are lead terminal semiconductor devices having outer leads21 that are external terminals electrically connected to the bonding electrodes la of thesemiconductor chips1, an EEPROM5 (lead terminal semiconductor device), and small surface-attachedresistors4 of a number of 36×2=72.
The mounting procedure will be roughly described in compliance with the basic flow for mounting the parts shown in FIG. 8.[0136]
At step S[0137]15, first, the solder is printed on themodule board2 to form terminals (land pads) for electric connection to the ends of the outer leads21 of the lead terminal semiconductor devices and to thebump electrodes11 of theWPPs10.
Thereafter, SMDs are mounted at step S[0138]16 andWPPs10 are mounted at step S17.
Then, at step S[0139]18, reflowing is effected at one time in order to electrically connect the outer leads21 of the lead terminal semiconductor devices to the land pads, and thebump electrodes11 of theWPPs10 to the land pads.
Then, washing is effected at step S[0140]19. The washing, however, may not be effected.
Further, at step S[0141]20, a resin is underfilled to effect the sealing.
Next, the method of manufacturing the memory-[0142]module100 will be described in detail by using the mounting flow closely illustrated in FIG. 9.
At step S[0143]21 shown in FIG. 9, first, the solder is printed on predetermined portions on themodule board2.
Then, at step S[0144]22, the parts are mounted on the surfaces of the module. Here, predetermined numbers of WPPs10 (of a number of 18), TSOPs20 (of a number of 2), small surface-attached resistors4 (of a number of 36) and EEPROM5 (of a number of 1) are arranged on the front surface of themodule board2 using a mounting machine.
Then, at step S[0145]23, all of the above-mentioned parts on the front surface of themodule board2 are mounted by the batchwise (simultaneous) solder reflowing.
Then, at step S[0146]24, the parts are mounted on the back surface of the module. Here, the parts are arranged on the back surface of themodule board2 by using the mounting machine in the same manner as on the front surface.
Then, at step S[0147]25, all of the above-mentioned parts on the back surface of themodule board2 are mounted by the batchwise (simultaneous) solder reflowing.
Thus, the memory-[0148]module100 is fabricated mounting (in a mixed manner) predetermined numbers of WPPs10 (of a number of 18×2), TSOPs20 (of a number of 2), small surface-attachedresistors4 andEEPROM5 on both the front and back surfaces of themodule board2.
Then, washing is effected at step S[0149]26.
Washing, however, may not be effected.[0150]
Then, at step S[0151]27, the module is tested. That is, the memory-module100 is inspected in a predetermined manner to detect defective chips.
Then, at step S[0152]28, the defective chips are repaired and are exchanged. In this case, the solder is melted by being heated again, the defective chip (defective semiconductor device) is removed and is replaced by a non-defective chip (non-defective semiconductor device).
Then, at step S[0153]29, all parts are mounted by effecting the reflowing again.
Thereafter, washing is effected at step S[0154]30.
Washing, however, may not be effected.[0155]
Then, at step S[0156]31, theWPPs10 are sealed by being underfilled with the resin. The underfilling is that when theWPP10 has a relatively large chip size like DRAM and fails to exhibit a sufficient function for buffering stress to thebump electrodes11, theresin9 is applied between thepackage body13 of theWPP10 and themodule body2 to decrease the stress exerted on thebump electrodes11.
That is, the underfilling is a sealing with resin between the[0157]package body13 of theWPP10 and themodule board2, in order to solidify and protect the surrounding of thebump electrode11 with theresin9.
To effect the underfilling, the[0158]liquid resin9 is applied onto themodule board2 one surface by one surface from anozzle60aof adispenser60 shown in FIG. 10. That is, theresin9 is applied onto theWPPs10 on the front and back surfaces of themodule board2 one surface by one surface.
After the application has been finished on both surfaces, the front and back surfaces of the[0159]module board2 are heated at one time to simultaneously cure theresin9 on the front and back surfaces. That is, after the application of theresin9 on both surfaces has been finished, the two surfaces are simultaneously cured (hardened) by baking by heating the atmosphere or by the like means.
Then, at step S[0160]32 as shown in FIG. 9, casing is effected, and the module is finally tested at step S33.
Predetermined data are written into the[0161]EEPROM5 by using a special writer.
Thus, the assembling of the memory-[0162]module100 shown in FIG. 1 is completed.
Here, mounting the bare chip that requires the same area as when the[0163]WPP10 is mounted will be described for the purpose of comparing the two.
First, in mounting the bear chip, the[0164]bonding electrodes1aare mounted on the mounting board without being rearranged byrewirings12. Therefore, the pitch is narrow among the external terminals, the wiring rule becomes strict on the mounting board, and the cost of the mounting board is driven up. In assembling the module, further, it becomes necessary to add a mounting step by using a flip chip bonder of a relatively slow processing speed in addition to the step of mounting the SMD parts by reflowing the solder.
Accordingly, the[0165]WPPs10 on the memory-module100 of theembodiment 1 exhibit much effect in their mounting (decreases the number of the mounting steps since no special mounting device such as flip-chip bonder is used) than mounting the bare chips.
Next, described below is the underfilling method in the method of manufacturing the memory-[0166]module100 of theembodiment 1.
FIG. 10 is a view illustrating a method of applying the resin for underfilling the[0167]WPPs10 mounted on the memory-module100 of FIG. 1, and FIG. 11 is a view illustrating the permeation of theresin9 that is applied by the method of application shown in FIG. 10.
In FIG. 10, arrows indicate the direction in which the[0168]nozzle60atravels. Thedispenser60 and thenozzle60amove on the short sides of theWPPs10 along the arrows.
According to the method of applying the resin of the[0169]embodiment 1, thedispenser60 is moved intermittently and nearly linearly along the direction of short sides of theWPPs10 having a rectangular shape on a plane, and theresin9 is successively dripped on the short sides of theWPPs10 from the upper side of theWPPs10 through thenozzle60a. That is, when the application on oneWPP10 has finished, thenozzle60ais moved to an end on the front side of the short side of anext WPP10, and thenozzle60ais once stopped at this position.
Thereafter, the[0170]resin9 is dripped while moving thenozzle60afrom the end of the front side of the short side of theWPP10 toward the end of the rear side thereof and, at this position, motion of thenozzle60aand dripping of theresin9 are once halted.
Then, in a state where the[0171]resin9 is no longer dripped, thenozzle60ais moved to an end on the front side of the short side of the neighboringWPP10, and theresin9 is similarly dripped and, thus, theWPPs10 are successively underfilled.
FIG. 11 illustrates the spreading of[0172]wet resin9 that is applied by the method shown in FIG. 10 to theWPPs10 of DRAMs in which thebump electrodes11 are arranged in 15 rows×4 columns, wherein FIGS. 11A and 11B illustrate a state right after the resin is applied onto the short side, FIGS. 11C and 11D illustrate a state when a predetermined period of time (short time) has passed after the application, FIGS. 11E and 11F illustrate a state when a predetermined period of time (long time) has passed after the application, and FIGS. 11G and 11H illustrate a state where theresin9 is applied by moving thenozzle60aone turn to form afillet9aalong the periphery after the predetermined period of time (long time) has passed after the application.
Referring to FIGS. 11E and 11F, it is also allowable to move again the[0173]dispenser60 and thenozzle60aabout thepackage body13 of theWPP10 after thewet resin9 has spread throughout between theWPP10 and themodule board2 to reliably form thefillet9ashown in FIG. 11G, so that theWPP10 is secured to themodule board2 more strongly.
FIG. 12 illustrates the structure of a memory-[0174]module200 which is modified from the memory-module100 of theembodiment 1 of the invention.
The memory-[0175]module200 includes 18 WPPs10 (protruded terminal semiconductor devices) that are mounted being arranged in a sequence maintaining an equal pitch on one surface of themodule board2, and one TSOP20 (lead terminal semiconductor device) mounted near theWPPs10, the TSOP20 (lead terminal semiconductor device) being arranged near the center of theWPPs10 that are arranged in a sequence.
That is, a plurality of (10 and 8) WPPs[0176]10 are arranged in a sequence on both sides of oneTSOP20.
Further, nine SOPs (small outline packages)[0177]61 (registers8) which are the lead terminal semiconductor devices are mounted on themodule board2 on the side of theconnection terminals2awhich are the external terminals, and 18WPPs10 are mounted on the side opposite to theconnection terminals2a(on the side remote from theconnection terminals2a), theindividual WPPs10 being underfilled.
In the memory-[0178]module200 of this type of mounting, theresin9 for underfilling theWPPs10 is applied nearly linearly onto the short sides of 18WPPs10 arranged in a sequence.
This makes it possible to efficiently apply the[0179]resin9.
FIG. 13 illustrates the structure of a memory-[0180]module300 which is modified from the memory-module100 of theembodiment 1 of the invention.
In the memory-[0181]module300 shown in FIG. 13, 18WPPs10 are mounted on themodule board2 in the form of groups (masses) each consisting of two or four in a two rows×two columns matrix arrangement.
Further, the[0182]individual WPPs10 are mounted with their lengthwise direction in parallel with the direction of the short sides of themodule board2 of the memory-module300.
Described here is a method of efficiently applying the[0183]resin9 for underfilling the WPPs10 in a state where theWPPs10 are mounted.
When the temperature for applying the[0184]resin9 is low, theresin9 is applied to thepackage body13 of theWPP10 on the side of the long side, since theresin9 may be infiltrated into between thepackage body13 and themodule board2 over a short distance. This makes it possible to shorten the time for application.
It is therefore desired to maintain space for moving the[0185]nozzle60aalong the long sides of at least one side of thepackage bodies13, and to arrange the long sides thereof of the side where nonozzle60amoves as close to other parts as possible from the standpoint of highly densely mounting the parts.
When the[0186]semiconductor chip1 has many bits and many DRAMs are connected to the same I/Os of the memory-module300, a great advantage is obtained by applying the resin onto the DRAMs which are arranged as close as possible to each other in a 2×2 arrangement on the same plane.
With the[0187]WPPs10 being arranged as shown in FIG. 13, therefore, it is desired to apply theresin9 to the outer peripheries of the long sides of the outer side along the outer long sides of theWPPs10. When theresin9 is applied along the outer long sides of 2×2 DRAMs (WPPs10) according to this method of application, theresin9 does not flow onto thepackage bodies13 to which the resin is not to be applied on the side opposite to thepackage bodies13 to which the resin is to be applied. Or, theresin9 does not leak or spread, either.
In the memory-[0188]module300, it is desired that theWPPs10 having I/O of (×4) constitution are collected in a number of four to obtain a 16-bit constitution, and are mounted as a group. In mounting theWPPs10 as shown in FIG. 13, therefore, it is desired to apply theresin9 along the locus as indicated by arrows.
FIG. 14 is a diagram illustrating the permeation of the[0189]resin9 when it is applied according to a modified embodiment.
That is, in mounting the[0190]WPPs10 on a memory-module400 as shown in FIG. 15, theunderfiller resin9 is applied to the outer peripheries along the two opposing sides of thepackage bodies13 as shown in FIG. 14. Here, theresin9 is applied to both short sides of theWPPs10 from the ends on one side to the ends on the opposite side.
Arrows shown in FIGS. 14A and 14B indicate the loci of motion of the[0191]dispenser60. FIGS. 14C and 14D illustrate a state right after the application on both short sides (two sides), FIGS. 14E and 14F illustrate a state when a predetermined period of time (intermediate time) has passed after the application, and FIGS. 14G and 14H illustrate a state of infiltration of theresin9 of when a predetermined period of time (long time) has passed after the application.
In the[0192]embodiment 1, theresin9 permeated starting from both short sides is still in a separated state leaving an intermediate region where noresin9 exists even in the final stage of infiltration of theresin9. The difference in the thermal expansion between thepackage body13 and themodule board2, and stress acting on thebump electrodes11 due to the warping of themodule board2, increase with an increase in the distance from the center of thepackage body13 and become a maximum at thebump electrodes11 at the corners. Therefore, if theresin9 is permeating near both short sides of thepackage body13 of a rectangular shape, the stress acting on thebump electrodes11 can be decreased to some extent even if there exists an intermediate region where theresin9 is not existing.
Thus, the effect close to that of when the resin is applied to the whole surface of the[0193]package body13 is obtained requiring a decreased amount of theresin9 and a shortened operation time.
In other words, it is made possible to shorten the time for application and to decrease the amount of application.[0194]
Further, the[0195]resin9 may simply be applied to the four corners of thepackage body13. In this case, a decreased stress acts on thebump electrodes11 arranged along the outermost circumference and, hence, thebump electrodes11 feature extended life for connection.
FIGS. 15A and 15B illustrate the structure of a memory-[0196]module400 modified from the memory-module100, and in which 16WPPs10 are mounted in a sequence maintaining an equal pitch on one surface of themodule board2. In this memory-module400, theunderfiller resin9 is linearly applied to 16WPPs10 arranged in a sequence. Arrows shown in FIG. 15A indicate the locus of motion of the dispenser60 (see FIG. 10).
FIG. 16 illustrates a state where the memory-[0197]module400 shown in FIG. 15 is deflected. This happens when the ends of themodule board2 are held at the time of inserting the socket for inspecting the memory-module400.
That is, as shown in FIG. 16, when the memory-[0198]module400 shown in FIG. 15 is deflected in the lengthwise direction thereof, the stress is nearly uniformly dispersed over the whole memory-module400 since the memory-module400 as a whole is deflected unless the sealingportions14 of the neighboringWPPs10 are integrally fabricated without coming into contact with each other.
This structure withstands the load from the outer side and, hence, improves the reliability of the memory-[0199]module400.
In a memory-[0200]module500 shown in FIGS. 17 and 18, theWPPs10 of a number of 16 are mounted being divided into four regions each including four of them along the direction in which the plurality ofconnection terminals2aare arranged on themodule board2, the sealingportion14 continuing over the fourWPPs10 in each region.
That is, the[0201]WPPs10 are mounted being divided into groups (masses), and are sealed by being underfilled as designated at the sealingportion14 in a continuing manner with the group as a unit. Therefore, the portions of the groups (masses) of the memory-module500 exhibit increased rigidity apparently including theWPPs10.
Therefore, the bending stress of the[0202]module board2 concentrates at the gaps among the groups of WPPs10.
That is, the neighboring sealing[0203]portions14 may often become continuous due to some factors affecting the application of theunderfiller resin9, such as gaps among the WPPs10. Even in such a case, the memory-module500 that includes thenon-mounting portions2bthat are not partly continuous as shown in FIGS. 17 and 18, deflects at thenon-mounting portions2bwhen an external force is exerted, preventing the stress from being applied to the connection portions ofbump electrodes11 of theWPPs10 or to thesemiconductor chips1.
Since the stress is dispersed, the[0204]WPPs10 on the memory-module500 feature improved reliability in the connection.
In the memory-[0205]modules100,200,300,400 and500 of theembodiment 1, theWPPs10 are sealed by underfilling, and the whole surfaces of the chips or the major portions are secured more strongly. As a result, shock resistance is improved and moisture resistance is improved, too.
In the module product, a TCP (tape carrier package) may be laminated as another means for highly densely mounting the parts. According to this technology, however, the chips are often cracked as their thickness is decreased. In the memory-[0206]modules100,200,300,400 and500 according to theembodiment 1, on the other hand, shock resistance is improved by securing the chips relying on the underfilling, preventing the chips from being cracked.
Further, the[0207]WPPs10 are sealed by underfilling and are mounted on themodule board2 with the main surfaces of thesemiconductor chips1 and the surfaces (back surfaces) of the opposite side being exposed. Moreover, the whole main surfaces or the main portions of thesemiconductor chips1 are secured to themodule board2 by being underfill-sealed, making it possible to decrease the heat resistance.
This helps improve the heat-radiating performance of the memory-[0208]modules100,200,300,400 and500 and lengthen the life.
(Embodiment 2)[0209]
FIG. 19 is a plan view illustrating the structure of the memory-module according to an[0210]embodiment 2 of the present invention.
The memory-[0211]module600 of theembodiment 2 includes 72 WPPs10 (protruded terminal semiconductor devices) which are DRAMs mounted in a matrix arrangement. Connection of input/output signals to theWPPs10 is accomplished in a manner that a group (mass) includes a total of 9WPPs10 consisting of one for ECC and eight of two rows (in the memory-module600 of FIG. 19, the direction in parallel with the short sides of themodule board2 is referred to as row and the direction at right angles therewith is referred to as column, which, however, may be reversed to the above), and nine FET (field-effect transistor)-bus switches15 (lead terminal semiconductor devices) which are memory selection means are mounted for theWPPs10 of each of the groups to switch each of the groups.
That is, in the memory-[0212]module600, the connection of input/output signals to nineWPPs10 of two rows is switched within the group (8 WPPs) by a corresponding FET-bus switch15, making it possible to increase the number of theWPPs10 without increasing the number of theconnection terminals2aof themodule board2.
Therefore, the memory-[0213]module600 mounts theWPPs10 of a number four times as great as that of the memory-module100 of theembodiment 1.
That is, the memory-[0214]module600 separately switches the I/Os using the FET-bus switches15, so that an increased number of DRAMs can be mounted.
In appearance, the FET-[0215]bus switches15 of the memory-module600 are, for example, those of the SOP type, which are lead terminal semiconductor devices.
The structure of the memory-[0216]module600 of theembodiment 2 in other respects and the method of manufacturing the memory-module600 are the same as those of the memory-module100 of theembodiment 1, and are not described here again.
(Embodiment 3)[0217]
FIG. 20 is a view illustrating the structure of a memory-module according to an[0218]embodiment 3 of the present invention, wherein FIG. 20A is a plan view and FIG. 20B is a side view, FIG. 21 is a diagram of block circuits of the memory-module shown in FIG. 20, FIG. 22 is a bottom view illustrating the structure of a wafer process package (protruded terminal semiconductor device) mounted on the memory-module shown in FIG. 20, FIG. 23 is a diagram of wirings on the side of the board illustrating an example of wirings on the module board at a portion C in the memory-module shown in FIG. 20A, FIGS. 24, 25 and26 are diagrams of wirings illustrating modified examples of the bump arrangement on the wafer process package in the memory-module according to theembodiment 3 of the invention and modified examples of the wirings on the side of the board corresponding thereto, and FIG. 27 is a diagram of bump arrangement and wirings illustrating a further modified example of the bump arrangement on the wafer process package and of the wirings on the side of the board shown in FIG. 25.
The memory-[0219]module700 of theembodiment 3 shown in FIGS. 20A and 20B is an unbuffered SDRAM (static DRAM)-DIMM (dual in-line memory-module) of 8 bytes having 168 pins, and includes 8 WPPs10 (protruded terminal semiconductor devices), small surface-attachedresistors4,capacitors3 and anEEPROM5 that are mounted in a mixed manner on one surface thereof.
The memory-[0220]module700, however, does not mount theregisters8 that are mounted on the memory-module100 of FIG. 1.
FIG. 21 is a diagram of block circuits of the memory-[0221]module700 shown in FIG. 20, constituting two banks.
Symbols attached to the terminals shown in FIG. 21 are the same as those described with reference to the block circuit diagram of the memory-[0222]module100 of theembodiment 1, and are not described here again.
In the memory-[0223]module700 shown in FIG. 21, whether the S0 system of thebank1 or the S1 system of thebank2 be read out, is directly determined by a signal since noregister8 has been mounted. That is, since the memory-module is of the unbuffered type, a signal directly enters either bank to select asemiconductor chip1 of either bank.
The chips D[0224]0 to D15 representWPPs10 of a number of 16 on both surfaces, and [I/O0 to I/O3] terminals of each chip are connected as independent terminals to theconnection terminals2aof themodule board2.
The DRAMs as a whole have I/Os of 64 bits from DQ[0225]0 to DQ63 that are used as data, constituting two banks.
The memory-[0226]module700 shown in FIG. 20 is inexpensive compared to the memory-module100 shown in FIG. 1.
The[0227]module board2 of the memory-module700 has a size of, for example, P=133.35 mm and Q=33.02 mm, and the mounting height (max) is R=4 mm as shown in FIG. 20B.
Referring to FIG. 20A, the memory-[0228]module700 includes 8 WPPs10 (protruded terminal semiconductor devices) which are DRAMS arranged in a sequence on one surface thereof, as well ascapacitors3 at portions among the neighboringWPPs10 or by theWPPs10 nearly at the centers in the lengthwise direction.
This is to minimize the wiring length between the WPPs[0229]10 and thecapacitors3.
Here, FIG. 22 illustrates the structure of the[0230]WPP10 used for the memory-module700.
In the[0231]semiconductor chip1 ofWPP10 shown in FIG. 22, afree space1bwithoutbump electrode11 is formed near the center in the lengthwise direction thereof.
This is done by partly changing the pitch among the[0232]bump electrodes11 byrewirings12 so as to form thefree space1b, i.e., to form thefree space1bwithoutbump electrode11 near the center of theWPP10 in the lengthwise direction thereof.
FIG. 23 illustrates the wirings on the side of the[0233]module board2 at the portion C in FIG. 20A.
Referring to FIGS. 22 and 23, the capacitor[0234]3 (lead terminal semiconductor device) is mounted neighboring thefree space1bof thesemiconductor chip1, and power source wirings2cof thecapacitor3 are formed as surface-layer wirings2hon the surface layer opposing thefree space1bof thesemiconductor chip1 on the module board2 (they, however, may be formed as inner-layer wirings2gin the inner layer).
That is, as shown in FIG. 22, the[0235]free space1bwithoutbump electrode11 is formed near the center of thesemiconductor chip1 in the lengthwise direction thereof. Therefore, the connection can be accomplished without drawing the signal lines of theWPP10 to the portions corresponding to the center of the chip on themodule board2 and, hence, thecapacitor3 can be mounted at a portion closest to theWPP10.
Accordingly, the wirings become the shortest between the WPPs[0236]10 and thecapacitors3 to improve the operation characteristics.
Referring to FIG. 23, the[0237]module board2 is formed by a total of six metal layers including two core layers Vcc, a GND layer, and two signal line layers on each surface.Common wirings2eof the address/functional system connect thelands2don the surface layer to which thebump electrodes11 of theWPP10 are connected, to the layer which is just thereunder through via-holes2f, and are connected to the inner-layer wirings2gthat extend in the lengthwise direction of themodule board2.
The I/O wirings are connected to the[0238]connection terminals2adisposed nearby through the surface-layer wirings2hof themodule board2. This suppresses an increase in the inductance that results when the via-holes2fare passed through.
In the wirings shown in FIG. 23, the Vss (GND) and Vdd are extending sideways from the[0239]capacitor3, which, however, may be readily connected to the core layers through via-holes2f.
FIGS. 24, 25 and[0240]26 are diagrams illustrating modified examples of the bump arrangement of theWPP10 in the memory-module700 of theembodiment 3 and modified examples of the wirings on the side of the board corresponding thereto, and FIG. 27 illustrates a further modified example of the bump arrangement of the wafer process package shown in FIG. 25 and of the wirings on the side of the board.
In the WPPs[0241]10 in FIGS. 24, 25,26 and27, there are separately provided a group of common bump electrodes (group of common protruded terminals)1cwhich is a group of common electrodes such as of addresses, functions, power source and GND that can be connected in common among theWPPs10, and a group of independent bump electrodes (group of independent protruded terminals)1dthat is a group of independent electrodes such as of I/Os independently connected for each of theWPPs10.
In the[0242]WPP10, further, the group ofindependent bump electrodes1dis arranged at an end on one side which is the short side of thepackage body13. On one surface of the memory-module700, eightWPPs10 are mounted with their groups ofindependent bump electrodes1dbeing directed to the side of theconnection terminals2aof themodule board2.
On the[0243]module board2 are therefore formed surface-layer wirings2hwhich arecommon wirings2efor connecting the groups ofcommon bump electrodes1cof eight WPPs10.
Here, the pitch is broadened among the group of[0244]common bump electrodes1c, i.e., among thebump electrodes11 of the address system and functional system. In particular, the pitch is expanded in the lengthwise direction of the chip so that many wirings can be formed in the direction at right angles with the lengthwise direction of thepackage body13 passing among the bump electrodes.
Further, the pitch is decreased among the group of[0245]independent bump electrodes1d, i.e., among thebump electrodes11 of the I/O system, and the bump electrodes are arranged in the outer periphery on one side of thepackage body13.
This makes it possible to form[0246]common wirings2erelying on the surface-layer wirings2honly and, hence, to decrease the number of the wiring layers in themodule board2.
In the WPPs[0247]10 shown in FIG. 24, the groups ofcommon bump electrodes1care regularly arranged byrewirings12 being inclined with respect to thepackage bodies13.
This makes it possible to form the plurality of[0248]common wirings2ein parallel in the lengthwise direction of thepackage bodies13 to connect common electrodes such as of addresses, functions, power source and GND.
As a result, the wiring density of the[0249]module board2 can be maximized and the lengths of thecommon wirings2ecan be minimized.
When the number of the[0250]bump electrodes11 of theWPP10 is relatively small compared to its chip size or when themodule board2 involves fine wiring rules like an additive board, the GND and Vcc layers of the surface layer and of the inner layer are partly used as signal layers to produce themodule board2 of four layers and, hence, to assemble the memory-module700 by using thismodule board2.
In this case, the[0251]independent wirings2iof the I/O system are connected from thebump electrodes11 provided on the side of theconnection terminals2a, and the plurality ofcommon wirings2econnecting the common electrodes such as of addresses, functions, power source and GND are so formed as to pass among the chips.
In the WPPs[0252]10 as shown in FIG. 25, further, the groups ofcommon bump electrodes1care arranged like a grid using rewirings12 (see FIG. 22). In this case, as shown in FIG. 22, therewirings12 are used for distributing the power source and GND wirings in the chip, and onebump electrode11 is electrically connected to a plurality ofbonding electrodes1athrough therewirings12 to decrease the number of the bump electrodes11 (to decrease the number of the external terminals).
In the wirings on the side of the board shown in FIG. 25, the connection is made using the surface layer only of the[0253]module board2, and thebump electrodes11 are arranged without being inclined. Therefore, the wirings are accomplished by utilizing the bending and inclination of the wirings on the side of the board.
In the WPPs[0254]10 shown in FIG. 26, the pitch among thebump electrodes11 is slightly expanded to be larger than that of the bump arrangement of theWPPs10 shown in FIG. 25, and the bump electrodes are arranged on themodule board2 being inclined in the lengthwise direction or in the direction of the short side.
Therefore, the[0255]common wirings2eon the side of themodule board2 are inclined relative to the lengthwise direction of thepackage bodies13. As a result, thecommon wirings2eare formed straight like thecommon wirings2eshown in FIG. 24.
FIG. 27 illustrates a further modified example in which the pitch among the[0256]bump electrodes11 is slightly expanded to be larger than that of the bump arrangement of theWPPs10 shown in FIG. 25. In this modified example, independent pins other than those of the I/O system are drawn out from the lower side. This is an example in which the bits are specially constituted to decrease the number of pins to thereby increase the gap among thecommon wirings2e, the I/O pins and other independent pins having a narrow pitch (d1>d2 in FIG. 27).
The modified example shown in FIG. 27 exhibits such an effect that an increased number of wirings can be drawn among the pins since the gap is broadened among the common wirings. Therefore, the wirings on the[0257]module board2 can be used in common using the surface-layer wirings2honly, without using the inner-layer wirings2g(see FIG. 23) of themodule board2. The I/O pins and independent pins such as of the power source have a narrow pitch. These pins may have a narrow pitch since the wirings are drawn downward, i.e., drawn to theconnection terminals2awithout passing among the pins.
In FIG. 27, three surface-[0258]layer wirings2hrun between the pins when the wiring layout D is employed, and four surface-layer wirings2hrun between the pins when the wiring layout E is employed.
In FIGS. 24, 25,[0259]26 and27, the mounting lands are not particularly indicated on themodule board2 and the slit-like openings in the resist at right angles with thecommon wirings2eare regarded to be false lands for connection by soldering, in order to increase the wiring density on themodule board2 up to its limit.
The structure of the memory-[0260]module700 of theembodiment 3 in other respects and the method of manufacturing the memory-module700 are the same as those of the memory-module100 of theembodiment 1, and are not described here again.
In the foregoing was concretely described the invention accomplished by the present inventors by way of embodiments. However, the present invention is in no way limited to the above-mentioned embodiments only but can be modified in a variety of ways without departing from the spirit and scope of the invention.[0261]
In the memory-[0262]modules100 to700 of the above-mentionedembodiments 1, 2 and 3, for example, theEEPROM5 was the lead terminal semiconductor device having outer leads21. However, theEEPROM5 which is a nonvolatile read-only memory may be formed in the same structure as the protruded terminal semiconductor device, i.e., as theWPP10, and may be mounted.
In this case, however, the[0263]EEPROM5 of the WPP structure is not sealed by underfilling but theWPPs10 which are the DRAMs only are underfilled.
That is, the[0264]EEPROM5 of the WPP structure is detachably mounted on themodule board2.
This is because the[0265]EEPROM5 is produced maintaining a low yield and when it is detected to be defective upon electrically writing data therein, theEEPROM5 is better replaced by a non-defective one. TheEEPROM5 has a small chip size compared to the DRAM, causes small stress to exert on thebump electrodes11, and maintains reliability to a sufficient degree even without being underfilled. Upon mounting theEEPROM5 of the WPP structure, the mounting area can be decreased compared to when the SOP type device is mounted and the cost can be decreased to be lower than that of the SOP type device.
The above-mentioned[0266]embodiments 1, 2 and 3 have dealt with the memory-modules of the type mounting theWPPs10 on both the front and back surfaces of themodule board2. However, the memory-module may be the one of the type mounting theWPPs10 on one surface only.
The lead terminal semiconductor device mounted together with the WPPs[0267]10 (protruded terminal semiconductor devices) is not limited toTSOP20 but may be such a semiconductor device as QFP (quad flat package) or TCP (tape carrier package) in addition toTSOP20.
The above-mentioned[0268]embodiments 1, 2 and 3 have dealt with the case where the protruded terminal semiconductor devices are the WPPs10. However, the protruded terminal semiconductor devices may be any other semiconductor devices provided their external terminals arebump electrodes11 and are equipped with wiring portions for expanding the pitch among thebonding electrodes1aof thesemiconductor chips1 to be wider than the pitch among thebump electrodes11.
FIGS. 28, 29 and[0269]30 illustrate modified examples of the protruded terminal semiconductor device other than theWPP10.
FIGS. 28A, 28B and[0270]28C illustrate a CSP (chip scale package)30 as a modified example of the protruded terminal semiconductor device.
The[0271]CSP30 has a chip size nearly equal to, or slightly larger than, thesemiconductor chip1, and is of the fan in structure that supports thesemiconductor chip1 by atape board32 by interposing anelastomer31.
Further, a plurality of bump electrodes[0272]34 (protruded terminals) of solder or the like are formed as external terminals within an area of thesemiconductor chip1, the connection leads32aprovided on thetape board32 are electrically connected to thebonding electrodes1aof thesemiconductor chip1, and terminal pitch-expandingwirings32bare formed on thetape board32 to expand the pitch among thebump electrodes34 to be wider than the pitch among thebonding electrodes1aof thesemiconductor chip1.
A sealing[0273]portion33 is formed on thebonding electrodes1aof thesemiconductor chip1.
FIGS. 29A and 29B illustrate a BGA (ball grid array)[0274]40 of the chip face-up mounting system as a modified example of the protruded terminal semiconductor device.
The[0275]BGA40 is the one in which thesemiconductor chip1 is secured to aBGA board42 in a face-up manner through a die-bonding material45, and thebonding electrodes1aof thesemiconductor chip1 are electrically connected to theboard electrodes42fof theBGA board42 throughbonding wires41 of gold or the like material.
Further, a plurality of bump electrodes[0276]44 (protruded terminals) of solder or the like material are arranged as external terminals like a grid on the back surface of theBGA board42, and terminal pitch-expandingwirings42aare formed on theBGA board42 to expand the pitch among thebump electrodes44 to be wider than the pitch among thebonding electrodes1aof thesemiconductor chip1.
The terminal pitch-expanding[0277]wirings42aincludesignal wirings42b, GND plane42c,Vdd plane42dand throughholes42e.
Further, a molded[0278]portion43 is formed for sealing thesemiconductor chip1 and thebonding wires41 with a resin.
FIGS. 30A, 30B and[0279]30C illustrate a BGA (ball grid array)50 of the chip face-down mounting system as a modified example of the protruded terminal semiconductor device.
The[0280]BGA50 is of the flip-chip structure in which thesemiconductor chip1 is mounted on theBGA board52 in a face-down manner viasmall bumps51, and thebonding electrodes1aof thesemiconductor chip1 are electrically connected to the electrodes of theBGA board52 through the small bumps51.
Further, the bump electrodes[0281]54 (protruded terminals) of solder or the like material are arranged as external terminals like a grid on the back surface of theBGA board52, and terminal pitch-expandingwirings52a(see FIG. 30C) are formed on theBGA board52 to expand the pitch among thebump electrodes54 to be wider than the pitch among thebonding electrodes1a(see FIG. 29) of thesemiconductor chip1.
A gap between the[0282]semiconductor chip1 and theBGA board52, i.e., the periphery of thesmall bumps51, is underfilled with a resin to form a sealedportion53.
In the[0283]CSP30 shown in FIG. 28,BGA40 shown in FIG. 29 andBGA50 shown in FIG. 30, too, the terminal pitch-expandingwirings32b,42aand52aare provided, respectively, to expand the pitch among thebump electrodes34,44,54 to be wider than the pitch among the bonding electrodes la of thesemiconductor chip1, which, therefore, can be mounted by reflowing on themodule board2 or the like.
Briefly described below are the advantages obtained by the representative examples of the inventions disclosed in this application.[0284]
(1) Upon mounting the protruded terminal semiconductor devices on the module board of the memory-module, it becomes possible to greatly decrease the mounting areas compared to mounting the lead terminal semiconductor devices having semiconductor chips that are individually treated. This makes it possible to effect the mounting requiring the least areas so far as the semiconductor chips are mounted and, hence, to greatly increase the module capacity.[0285]
(2) The WPPs are mounted as the protruded terminal semiconductor devices while expanding the pitch among the bump electrodes which are the external terminals to be wider than the pitch among those of the flip chips, making it possible to expand the wiring rules on the module board and, hence, to realize a highly densely mounted memory-module suppressing the cost.[0286]
(3) The bonding electrodes of the semiconductor chip can be connected to the bump electrodes which are the external terminals of the WPPs through wirings of lengths shorter than those of the SMD parts such as TSOPS. This enables the memory-module to cope with high-speed operations and, hence, to cope with high-speed buses.[0287]
(4) Since the WPPs in the memory-module are sealed by underfilling, the whole chip surfaces are strongly secured to exhibit improved shock resistance. Therefore, the chips are prevented from being cracked.[0288]
(5) The WPPs are sealed by underfilling and are mounted on the module board in a state where the back surfaces of the semiconductor chips are exposed and, besides, the whole main surfaces of the semiconductor chips are secured to the module board by underfill-sealing, enabling the heat resistance of the memory-module to be decreased. As a result, the memory-module exhibits improved heat-radiating performance and extended life.[0289]