Patent Application
20050062144
The present invention relates to aggregating integrated circuits and, in particular, to stacking integrated circuits in chip-scale packages and providing such stacked integrated circuits on boards.
A variety of techniques are used to stack packaged integrated circuits. Some methods require special packages, while other techniques stack conventional packages. In some stacks, the leads of the packaged integrated circuits are used to create a stack, while in other systems, added structures such as rails provide all or part of the interconnection between packages. In still other techniques, flexible conductors with certain characteristics are used to selectively interconnect packaged integrated circuits.
The predominant package configuration employed during the past decade has encapsulated an integrated circuit (IC) in a plastic surround typically having a rectangular configuration. The enveloped integrated circuit is connected to the application environment through leads emergent from the edge periphery of the plastic encapsulation. Such “leaded packages” have been the constituent elements most commonly employed by techniques for stacking packaged integrated circuits.
Leaded packages play an important role in electronics, but efforts to miniaturize electronic components and assemblies have driven development of technologies that preserve circuit board surface area. Because leaded packages have leads emergent from peripheral sides of the package, leaded packages occupy more than a minimal amount of circuit board surface area. Consequently, alternatives to leaded packages known as chip scale packaging or “CSP” have recently gained market share.
CSP refers generally to packages that provide connection to an integrated circuit through a set of contacts (often embodied as “bumps” or “balls”) arrayed across a major surface of the package. Instead of leads emergent from a peripheral side of the package, contacts are placed on a major surface and typically emerge from the planar bottom surface of the package.
The goal of CSP is to occupy as little area as possible and, preferably, approximately the area of the encapsulated IC. Therefore, CSP leads or contacts do not typically extend beyond the outline perimeter of the package. The absence of “leads” on package sides renders most stacking techniques devised for leaded packages inapplicable for CSP stacking.
The previous known methods for stacking CSPs typically present complex structural arrangements and thermal or high frequency performance issues. For example, thermal performance is a characteristic of importance in CSP stacks. To increase dissipation of heat generated by constituent CSPs and the module, the thermal gradient between the lower CSP and upper CSP in a CSP stack or module should be minimized.
Memory expansion is one of the many fields in which stacked module solutions provide advantages. For example, the well-known DIMM board is frequently populated with stacked modules from those such as the assignee of the present invention. This adds capacity to the board without adding sockets.
A memory expansion board such as a DIMM, for example, provides plural sites for memory IC placement (i.e., sockets) arranged along both major surfaces of a board having an array of contacts dispersed along at least one board edge. Although stacking reduces interconnect length per unit of memory, and thus takes advantage of the general rule that interconnects that are less than half the spatial extent of the leading edge of a signal operate as a lumped element more than a transmission line, it does increase the raw number of devices on a DIMM board. Consequently, despite the reduction in interconnect length per unit of memory, signals accessing data stored in memory circuits physically placed on the DIMM board are typically presented with relatively high impedance as the number of devices on the bus is increased by stacking.
What is needed, therefore, is a technique and system for stacking CSPs that provides a thermally efficient, reliable structure that performs well at higher frequencies but does not add excessive height to the stack yet allows production at reasonable cost with readily understood and managed materials and methods and allows significant reductions in interconnect lengths and/or loading when employed in memory expansion boards and design.
The present invention stacks chip scale-packaged integrated circuits (CSPs) into modules that conserve PWB or other board surface area. In another aspect, the invention provides a lower capacitance memory expansion addressing system and method and preferably with the CSP stacked modules provided herein. Although the present invention is applied most frequently to chip scale packages that contain one die, it may be employed with chip scale packages that include more than one integrated circuit die.
Multiple numbers of CSPs may be stacked in accordance with the present invention. A four-high CSP stacked module is preferred for use with the disclosed high performance memory access system while, for many applications, a two-high CSP stack or module devised in accordance with a preferred embodiment of the present invention is preferred. The CSPs employed in stacked modules devised in accordance with the present invention are connected with flex circuitry. That flex circuitry may exhibit one or two or more conductive layers with preferred embodiments having two conductive layers.
The flex circuitry is partially wrapped above a form standard. A form standard is disposed between the flex circuitry and the IC package over which a portion of the flex circuitry is laid. The form standard can take many configurations and may be used where flex circuits are used to connect CSPs to one another in stacked modules having two or more constituent ICs. For example, in stacked modules that include four CSPs, three form standards are employed in preferred embodiments, although fewer may be used. The form standard provides a physical form that allows many of the varying package sizes found in the broad family of CSP packages to be used to advantage while employing a standard connective flex circuitry design. In a preferred embodiment, the form standard will be devised of heat transference material, a metal for example, such as copper would be preferred, to improve thermal performance.
In a preferred embodiment of the present invention, four-high stacked CSP modules are disposed on a memory expansion boards in accordance with the memory expansion system and methods of the present invention which may be employed with CSP or other IC stacked modules. A high speed switching system selects a data line associated with each level of a stacked module to reduce the loading effect upon data signals in memory access. This favorably changes the impedance characteristics exhibited by the board loading. The high speed DQ selection switch may be implemented, in a preferred embodiment, for example, with a high speed FET switch. FET multiplexers for example, under logic control select particular data lines associated with particular levels of the DIMM-populated stacked modules for connection to a controlling chip set in a memory expansion system in accordance with a preferred embodiment.
The invention is used with CSP packages of a variety of types and configurations such as, for example, those that are die-sized, as well those that are near chip-scale as well as the variety of ball grid array packages known in the art. It may also be used with those CSP-like packages that exhibit bare die connectives on one major surface. Thus, the term CSP should be broadly considered in the context of this application. Collectively, these will be known herein as chip scale packaged integrated circuits (CSPs) and preferred embodiments will be described in terms of CSPs, but the particular configurations used in the explanatory figures are not, however, to be construed as limiting. For example, the elevation views of
Typical CSPs, such as, for example, ball-grid-array (“BGA”), micro-ball-grid array, and fine-pitch ball grid array (“FBGA”) packages have an array of connective contacts embodied, for example, as leads, bumps, solder balls, or balls that extend from lower surface 22 of a plastic casing in any of several patterns and pitches. An external portion of the connective contacts is often finished with a ball of solder. Shown in
In
Form standard 34 is shown disposed adjacent to upper surface 20 of each of the CSPs below level four CSP 12. Form standard 34 may be fixed to upper surface 20 of the respective CSP with an adhesive 36 which preferably is thermally conductive. Form standard 34 may also, in alternative embodiments, merely lay on upper surface 20 or be separated from upper surface 20 by an air gap or medium such as a thermal slug or non-thermal layer. However, where form standard 34 is a thermally conductive material such as the copper that is employed in a preferred embodiment, layers or gaps interposed between form standard 34 and the respective CSP (other than thermally conductive layers such as adhesive) are not highly preferred.
Form standard 34 is, in a preferred embodiment, devised from copper to create, as shown in the depicted preferred embodiment of
Preferably, portions of flex circuits 30 and 32 are fixed to form standard 34 by adhesive 35 which is preferably a tape adhesive, but may be a liquid adhesive or may be placed in discrete locations across the package. Preferably, adhesive 35 is thermally conductive.
In a preferred embodiment, flex circuits 30 and 32 are multi-layer flexible circuit structures that have at least two conductive layers examples of which are those described in U.S. application Ser. No. 10/005,581 which has been incorporated by reference herein. Other embodiments may, however, employ flex circuitry, either as one circuit or two flex circuits to connect a pair of CSPs, that have only a single conductive layer.
Preferably, the conductive layers are metal such as alloy 110. The use of plural conductive layers provides advantages and the creation of a distributed capacitance across module 10 intended to reduce noise or bounce effects that can, particularly at higher frequencies, degrade signal integrity, as those of skill in the art will recognize. Module 10 of
Flex 30 is shown in
As depicted in
Respective ones of contacts 28 of second level CSP 16 and first level CSP 18 are connected at the second conductive layer 58 level in flex circuits 30 and 32 to interconnect appropriate signal and voltage contacts of the two CSPs. In a preferred embodiment, respective contacts 28 of second level CSP 16 and first level CSP 18 that convey ground (VSS) signals are connected at the first conductive layer 54 level in flex circuits 30 and 32 by vias that pass through intermediate layer 56 to connect the levels as will subsequently be described in further detail. Thereby, CSPs 16 and 18 are connected. Consequently, when flex circuits 30 and 32 are in place about first level CSP 18, respective contacts 28 of each of CSPs 16 and 18 are in contact with upper and lower flex contacts 42 and 44, respectively. Selected ones of upper flex contacts 42 and lower flex contacts 44 are connected. Consequently, by being in contact with lower flex contacts 44, module contacts 38 are in contact with both CSPs 16 and 18.
In a preferred embodiment, module contacts 38 pass through windows 62 opened in second outer layer 52 to contact lower CSP contacts 44. In some embodiments, as is shown in incorporated U.S. application Ser. No. 10/005,581, module 10 will exhibit a module contact array that has a greater number of contacts than do the constituent CSPs of module 10. In such embodiments, some of module contacts 38 may contact lower flex contacts 44 that do not contact one of the contacts 28 of first level CSP 18 but are connected to contacts 28 of second level CSP 16. This allows module 10 to express a wider datapath than that expressed by the constituent CSPs 16 or 18. A module contact 38 may also be in contact with a lower flex contact 44 to provide a location through which different levels of CSPs in the module may be enabled when no unused CSP contacts are available or convenient for that purpose.
In a preferred embodiment, first conductive layer 54 is employed as a ground plane, while second conductive layer 58 provides the functions of being a signal conduction layer and a voltage conduction layer. Those of skill will note that roles of the first and second conductive layers may be reversed with attendant changes in windowing and use of commensurate interconnections.
Form standard 34 is shown attached to the body 27 of first level CSP 18 through an adhesive. In some embodiments, it may also be positioned to directly contact body 27 of the respective CSP. Form standard 34 may take many different configurations to allow a connective flex circuitry to be prepared exhibiting a single set of dimensions which may, when used in conjunction with form standard 34, be employed to create stacked modules 10 from CSPs of a variety of different dimensions. In a preferred embodiment, form standard 34 will present a lateral extent broader than the upper major surface of the CSP over which it is disposed. Thus, the CSPs from one manufacturer may be aggregated into a stacked module 10 with the same flex circuitry used to aggregate CSPs from another manufacturer into a different stacked module 10 despite the CSPs from the two different manufacturers having different dimensions.
Further, heat transference can be improved with use of a form standard 34 comprised of heat transference material such as a metal or preferably, copper or a copper compound or alloy to provide a significant sink for thermal energy. Such thermal enhancement of module 10 particularly presents opportunities for improvement of thermal performance where larger numbers of CSPs are aggregated in a single stacked module 10.
Chipset 82 depicted in
As shown in the example depicted in
In a preferred embodiment, memory expansion boards 70 are populated with nine four high CSP modules 10 per side. The depiction of
Thus, decode logic 86 may, on the appropriate signal from clock 84, generate a level select signal which, in a preferred embodiment, is a multi-bit signal that controls a multiplexing switch 90 associated with several data lines. Switch 90 is in a preferred embodiment, a high speed switch and a FET muliplexer would provide a preferred multiplexing switch 90 in the practice of a preferred mode of the invention. The fan out of multiplexing switch 90 may be any that provides a selection capability to a variety of device data lines from a DQ line from chipset 82. The DQ lines between chipset 82 and switches 90 are depicted by double-headed arrows 94(1), 94(2), 94(3) and 94(4). As with the depiction of stacked modules 10, only one multiplexing switch 90 is shown per memory expansion board 70, but those of skill will understand that multiple multiplexing switches 90 are employed in practice of the depicted preferred embodiment of the invention. The number of multiplexing switches 90 will depend upon the fan out ratios. For example, use of nine 8:32 multiplexing switches 90 would be preferred (if available) or 4:8 or 1:4 multiplexing switches 90 will also provide advantages as an example. It should be understood that there are merely examples and that a variety of multiplexing switches and ratios may be employed for multiplexing switches 90 although the type of switch and the ratios will affect the loading figures. Consequently, a FET mux is preferred for multiplexing switch 90 and a ratio of 1:4 is one of the preferred ratios to employ.
The depiction in
An exemplar multiplexing switch 90 has multiple inputs 92(a), 92(b), 92(c), and 92(d) to provide independent data lines for each level of an exemplar module 10 populated upon the respective memory expansion board 70. Thus, with a 1:4 switch 90, there will be 18 iterations of multiplexing switch 90, one for each of the 18 four-high module 10's populating memory expansion board 70(1). Thus, the system 80 shown in
The data line of each level of the constituent CSPs of each module 10 is connected to one input 92 of a corresponding exemplar multiplexing switch 90. In response to the CS signal 88 from decode logic 86 on a DIMM expansion board 70, multiplexing switch 90 connects the appropriate one of the DQ signals 94 to one of the four levels of a module 10 on that memory expansion board 70. This switching of the data bus through multiplexing switch 90 may, in some systems, required further control signal connections as those of skill in the art will recognize to accommodate the data latency of one or more clocks cycles, CAS latency, and burst length, for example. In a preferred mode, expansion board 70 may keep all the constituent devices of the modules 10 as if each constituent device of the modules 10 were the target, instead of having to switch terminations each time a different CS is chosen. In some applications it may be preferred to terminate the end of the data line past the last DIMM expansion board 70. Other features may enable improvements to the efficiency of system 80 such as creating more CS banks by decoding the chip select lines.
In the system 80, the capacitive load presented to chipset 82 would be approximately the combination of the input capacitance of switching multiplexer 90 times the number of DIMM slots plus one DRAM device load plus one times the output capacitance of the multiplexing switch 90. In large systems, this will reduce capacitive loading by a notable amount, thus allowing more DIMM slots at higher speeds and/or more densely populated DIMMs. Memory access system 80 provides an opportunity to improve high speed memory performance and allows use of memory expansion configurations that might not otherwise be available due to capacitive loading in conventional DIMM systems.
Although the present invention has been described in detail, it will be apparent to those skilled in the art that the invention may be embodied in a variety of specific forms and that various changes, substitutions and alterations can be made without departing from the spirit and scope of the invention. The described embodiments are only illustrative and not restrictive and the scope of the invention is, therefore, indicated by the following claims.
This application is a continuation of U.S. patent application Ser. No. 10/814,530, filed Mar. 31, 2004, pending, which is a continuation of U.S. patent application Ser. No. 10/453,398, filed Jun. 3, 2003, pending, which is a continuation-in-part of U.S. patent application Ser. No. 10/005,581, filed Oct. 26, 2001, now U.S. Pat. No. 6,576,992. U.S. patent application Ser. Nos. 10/814,530, 10/005,581 and 10/453,398 are incorporated by reference for all purposes.
| Number | Date | Country | |
|---|---|---|---|
| Parent | 10814530 | Mar 2004 | US |
| Child | 10963867 | Oct 2004 | US |
| Parent | 10453398 | Jun 2003 | US |
| Child | 10814530 | Mar 2004 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 10005581 | Oct 2001 | US |
| Child | 10453398 | Jun 2003 | US |
