This application is a continuation of application Ser. No. 09/466,454, filed Dec. 17, 1999, now U.S. Pat. No. 6,222,265, issued Apr. 24, 2001, which is a continuation of application Ser. No. 09/233,997, filed Jan. 19, 1999, now U.S. Pat. No. 6,051,878, issued Apr. 18, 2000, which is a divisional of application Ser. No. 08/813,467, filed Mar. 10, 1997, now U.S. Pat. No. 5,994,166, issued Nov. 30, 1999.
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to an apparatus and a method for increasing semiconductor device density. In particular, the present invention relates to a stacked multi-substrate device using a combination of flip chips and chip-on-board assembly techniques to achieve densely packaged semiconductor devices.
2. State of the Art
Chip-On-Board techniques are used to attach semiconductor dice to a printed circuit board, including flip chip attachment, wirebonding, and tape automated bonding (“TAB”). Flip chip attachment consists of attaching a flip chip to a printed circuit board or other substrate. A flip chip is a semiconductor chip that has a pattern or array of electrical terminations or bond pads spaced around an active surface of the flip chip for face down mounting of the flip chip to a substrate. Generally, the flip chip has an active surface having one of the following electrical connectors: Ball Grid Array (“BGA”)—wherein an array of minute solder balls is disposed on the surface of a flip chip that attaches to the substrate (“the attachment surface”); Slightly Larger than Integrated Circuit Carrier (“SLICC”)—which is similar to a BGA, but having a smaller solder ball pitch and diameter than a BGA; or a Pin Grid Array (“PGA”)—wherein an array of small pins extends substantially perpendicularly from the attachment surface of a flip chip. The pins conform to a specific arrangement on a printed circuit board or other substrate for attachment thereto. With the BGA or SLICC, the solder or other conductive ball arrangement on the flip chip must be a mirror-image of the connecting bond pads on the printed circuit board such that precise connection is made. The flip chip is bonded to the printed circuit board by refluxing the solder balls. The solder balls may also be replaced with a conductive polymer. With the PGA, the pin arrangement of the flip chip must be a mirror-image of the pin recesses on the printed circuit board. After insertion, the flip chip is generally bonded by soldering the pins into place. An under-fill encapsulant is generally disposed between the flip chip and the printed circuit board for environmental protection and to enhance the attachment of the flip chip to the printed circuit board. A variation of the pin-in-recess PGA is a J-lead PGA, wherein the loops of the J's are soldered to pads on the surface of the circuit board.
Wirebonding and TAB attachment generally begin with attaching a semiconductor chip to the surface of a printed circuit board with an appropriate adhesive, such as an epoxy. In wirebonding, bond wires are attached, one at a time, to each bond pad on the semiconductor chip and extend to a corresponding lead or trace end on the printed circuit board. The bond wires are generally attached through one of three industry-standard wirebonding techniques: ultrasonic bonding—using a combination of pressure and ultrasonic vibration bursts to form a metallurgical cold weld; thermocompression bonding—using a combination of pressure and elevated temperature to form a weld; and thermosonic bonding—using a combination of pressure, elevated temperature, and ultrasonic vibration bursts. The semiconductor chip may be oriented either face up or face down (with its active surface and bond pads either up or down with respect to the circuit board) for wire bonding, although face up orientation is more common. With TAB, ends of metal leads carried on an insulating tape, such as a polyamide, are respectively attached to the bond pads on the semiconductor chip and to the lead or trace ends on the printed circuit board. An encapsulant is generally used to cover the bond wires and metal tape leads to prevent contamination.
Higher performance, lower cost, increased miniaturization of components, and greater packaging density of integrated circuits are ongoing goals of the computer industry. As new generations of integrated circuit products are released, the number of devices used to fabricate them tends to decrease due to advances in technology even though the functionality of these products increases. For example, on the average, there is approximately a 10 percent decrease in components for every product generation over the previous generation with equivalent functionality.
In integrated circuit packaging, in addition to component reduction, surface mount technology has demonstrated an increase in semiconductor chip density on a single substrate or board despite the reduction of the number of components. This results in more compact designs and form factors and a significant increase in integrated circuit density. However, greater integrated circuit density is primarily limited by the space or “real estate” available for mounting dice on a substrate, such as a printed circuit board.
One method of further increasing integrated circuit density is to stack semiconductor dice vertically. U.S. Pat. No. 5,012,323, issued Apr. 30, 1991 to Farnworth, teaches combining a pair of dice mounted on opposing sides of a lead frame. An upper, smaller die is back-bonded to the upper surface of the leads of the lead frame via a first adhesively coated, insulated film layer. A lower, larger die is face-bonded to the lower lead frame die-bonding region via a second, adhesively coated, insulative film layer. The wirebonding pads on both upper die and lower die are interconnected with the ends of their associated lead extensions with gold or aluminum bond wires. The lower die must be slightly larger than the upper die such that the die pads are accessible from above through a bonding window in the lead frame such that gold wire connections can be made to the lead extensions. This arrangement has a major disadvantage from a production standpoint as the same size die cannot be used.
U.S. Pat. No. 5,291,061, issued Mar. 1, 1994 to Ball (“Ball”), teaches a multiple stacked dice device containing up to four stacked dice supported on a die-attach paddle of a lead frame, the assembly not exceeding the height of current single die packages, and wherein the bond pads of each die are wirebonded to lead fingers. The low profile of the device is achieved by close-tolerance stacking which is made possible by a low-loop-profile wirebonding operation and thin adhesive layers between the stacked dice. However, Ball requires long bond wires to electrically connect the stacked dice to the lead frame. These long bond wires increase resistance and may result in bond wire sweep during encapsulation. Also, Ball requires the use of spacers between the dice.
U.S. Pat. No. 5,323,060, issued Jun. 21, 1994 to Fogal et al. (“Fogal”), teaches a multi-chip module that contains stacked die devices, the terminals or bond pads of which are wirebonded to a substrate or to adjacent die devices. However, as discussed with Ball, Fogal requires long bond wires to electrically connect the stacked die bond pads to the substrate. Fogal also requires the use of spacers between the dice.
U.S. Pat. Nos. 5,422,435 and 5,495,398 to Takiar et al. (“Takiar”) teach stacked dice having bond wires extending to each other and to the leads of a carrier member such as a lead frame. However, Takiar also has the problem of long bond wires, as well as, requiring specific sized or custom designed dice to achieve a properly stacked combination.
U.S. Pat. No. 5,434,745 issued Jul. 18, 1995 to Shokrgozar et al. (“Shokrgozar”) discloses a stackable packaging module comprising a standard die attached to a substrate with a spacer frame placed on the substrate to surround the die. The substrate/die/spacer combinations are stacked one atop another to form a stacked assembly. The outer edge of the spacer frame has grooves in which a conductive epoxy is disposed. The conductive epoxy forms electric communication between the stacked layers and/or to the final substrate to which the stacked assembly is attached. However, Shokrgozar requires specialized spacer frames and a substantial number of assembly steps, both of which increase the cost of the final assembly.
U.S. Pat. No. 5,128,831 issued Jul. 7, 1992 to Fox, III et al. (“Fox”) also teaches a standard die attached to a substrate with a spacer frame placed on the substrate to surround the die. The stacked layers and/or the final substrate are in electric communication with conductive vias extending through the spacer frames. However, Fox also requires specialized spacer frames, numerous assembly steps, and is limited in its flexibility to utilize a variety of dice.
U.S. Pat. No. 5,513,076 issued Apr. 30, 1996 to Wether (“Wether”) teaches the use of interconnecting assemblies to connect integrated circuits in an integrated manner.
As has been illustrated, none of the cited prior art above uses or teaches flip chip manufacturing methods for attaching dice together in a stacked manner to form a stacked die assembly.
Therefore, it would be advantageous to develop a stacking technique and assembly for increasing integrated circuit density using a variety of non-customized die configurations in combination with commercially-available, widely-practiced semiconductor device fabrication techniques.
BRIEF SUMMARY OF THE INVENTION
The present invention relates to a stacked multi-substrate device using combined flip chips and chip-on-board assembly techniques to achieve densely packaged semiconductor devices, and a method for making same. In this invention, multiple substrates are stacked atop one another. The substrates can include a plurality of semiconductor dice disposed on either surface of the substrates. The substrates can be structures of planar non-conductive material, such as fiberglass material used for PCBs, or may even be semiconductor dice. For the sake of clarity, the term “substrate”, as used hereinafter, will be defined to include planar non-conductive materials and semiconductor dice. The substrates are preferably stacked atop one another by electric connections which are ball or column-like structures. Alternately, solder bumps or balls may be formed on the substrate. The electric connections achieve electric communication between the stacked substrates. The electric connections can be formed from industry standard solder forming techniques or from other known materials and techniques such as conductive adhesives, Z-axis conductive material, flex-contacts, spring contacts, wire bonds, TAB tape, and the like. The electric connections must be of sufficient height to give clearance for the components mounted on the substrates and should be sufficiently strong enough to give support between the stacked substrates.
A preferred embodiment comprises a base substrate, having first and opposing surfaces, and means for electrical connection with external components or substrates, wherein the electrical connection means extends at least from the first surface of the base substrate. The base substrate opposing surface, the other side of the substrate, also includes a plurality of bond pads disposed thereon. Additionally, at least one semiconductor component may be attached to the opposing surface of the base substrate. The semiconductor components are preferably flip chips that are in electrical communication with electrical traces on or within the base substrate with any convenient known chip-on-board (COB) or direct-chip-attachment (DCA) technique (i.e., flip chip attachment, wirebonding, and TAB). Other techniques, such as the use of two-axis materials or conductive epoxies, can also be used for connections between either substrates or substrates and semiconductor chips. The electrical traces form a network of predetermined electrical connections between the base substrate electrical connection means, the base substrate bond pads, and/or the base substrate semiconductor components.
The preferred embodiment further comprises a stacked substrate. The stacked substrate has a first surface and an opposing surface. A plurality of bond pads may be disposed on the stacked substrate first surface and/or the stacked substrate opposing surface. At least one semiconductor component is attached to each of the stacked substrate first surface and the stacked substrate opposing surface. The semiconductor components are preferably flip chips which are in electrical communication with electrical traces on or within the first stacked substrate. The electrical traces form a network of predetermined electrical connections between the stacked substrate first surface bond pads, the stacked substrate opposing surface bond pads, and/or the stacked substrate semiconductor components.
The stacked substrate is attached to the base substrate through a plurality of electric connections. The electric connections can be column-like structures or spherical structures (balls) that support and form electrical communication between the base substrate bond pads and either the stacked substrate first surface bond pads or the stacked substrate opposing surface bond pads (depending upon which stacked substrate surface faces the base substrate first surface). The electric connections are preferably distributed evenly around a periphery of the base and stacked substrates. However, the electric connections may be of any distribution so long as adequate mechanical support exists between the base substrate and the stacked substrate.
In the manner discussed for the stacked substrate, additional stacked substrates may be attached to and stacked above the stacked substrate. Thus, with this technique, a multiple stacked substrate component may be formed. It is, of course, understood that the electrical connection means extending from the base substrate first surface for communication with an outside substrate may not be necessary if the multiple stacked substrate is in and of itself a complete component.
An alternative embodiment comprises substrates of varying size in a single assembly. The variable size substrate assembly is constructed in the manner discussed above. However, the variable size substrate assembly includes smaller sized substrates than the previously discussed base and stacked substrate. The smaller substrate is essentially identical to the previously discussed stacked substrate. The smaller substrate comprises a first surface and an opposing surface with a plurality of bond pads which may be disposed on the smaller substrate first surface and/or the smaller substrate opposing surface. At least one semiconductor component may be attached to the smaller substrate first surface and/or the smaller substrate opposing surface. The semiconductor components are in electrical communication with electrical traces on or within the first stacked substrate. The electrical traces form a network of predetermined electrical connections between the smaller substrate first surface bond pads, the smaller substrate opposing surface bond pads, and/or the smaller substrate semiconductor components.
The smaller substrate may be disposed between the base substrate and the stacked substrate. The smaller substrate is attached to either the base substrate or the stacked substrate through a plurality of electric connections. The electric connections form electrical communication between the base substrate bond pads and the smaller substrate bond pads or between the stacked substrate bond pads and the smaller substrate bond pads (depending upon whether the smaller substrate is attached to the base substrate or the stacked substrate). The smaller substrate may also be attached to the opposite surface of the stacked substrate and multiple smaller substrates may be attached in various positions on any substrate in the variable size substrate assembly.
Thus, the present invention offers the advantages of and achieves superior and improved electrical properties and speed of submodules and the entire module assembly, achieves higher density input/output configurations and locations (array), achieves higher density of devices or complexities of integrated circuits because of optimum input/output locations, results in improved thermal performance, allows easier repair and reusability, and allows easier modification of the package.