US RE42252 E1
A multichip module comprises a multilayer substrate circuit having conductive patterns on its surface(s) to which microelectronic device(s) are attached. A part of the substrate is flexible and bifurcated. Two rigid members are attached lengthwise, one on either side of the substrate, and the free ends of the bifurcation are reflexed respectively about these members and bonded to them. Electrodes are located on the bifurcations so that they will be exposed outwardly and/or downwardly after reflexing. The module may further be provided with protective heat spreading covers. The electrodes and rigid members may be configured to engage a mating socket or they may be solderable to a printed circuit board.
1. A multichip module comprising:
a multilayer dielectric substrate comprising two plies, each of said two plies having conductive patterns on both surfaces thereof;
a plurality of microelectronic devices attached to said substrate and operably coupled to said conductive patterns on said two plies;
a bonding layer joining said two plies over a first portion of their respective areas while a second portion of their respective areas is unbonded so that a bifurcated sheet is formed;
two generally elongated rigid dielectric members, one on each side of said multilayer dielectric substrate, configured so that the free ends of said bifurcated sheet may be reflexed respectively about said two rigid members and bonded thereto; and,
electrodes disposed on said bifurcated sheet so that said electrodes are exposed after said reflexing.
2. The multichip module of
3. The multichip module of
4. The multichip module of
5. The multichip module of
6. The multichip module of
7. The multichip module of
8. The multichip module of
9. The multichip module of
10. The multichip module of
11. The multichip module of
12. The multichip module of
13. The multichip module of
This application claims the benefit of Provisional Patent Application No. 61/009,781 by the present inventors, filed on Jan. 2, 2008, the entire disclosure of which is incorporated herein by reference.
1. Field of the Invention
The present invention relates generally to means for packaging microelectronic devices and multichip modules. More particularly, the invention relates to improved flex-circuit solutions for SIMM and DIMM type memory modules and other modular electronic circuits.
2. Description of Related Art
High density computing is continually evolving to offer high performance devices with higher densities and smaller form factors. Servers are designed with these goals in mind and are key enablers for various computing applications. Servers contain, among other electronic components, microprocessors and memory modules, which generate considerable heat during operation. Currently higher performance servers make use of dual and quad core microprocessors. In turn, these microprocessors require memory modules with greater memory capacity for optimal performance.
Another important aspect of server design is that servers are often arranged in closely packed groups set in vertical racks that rely on forced air cooling for heat exhaust. So, air cooling is done for each server unit as well as the entire rack of server units. When multiple memory modules are tightly spaced together inside the tightly packed server chassis, the circulation of air as a cooling fluid becomes ineffective. Air can be restricted, leading to overheating of memory modules which in turn leads to premature failures including device failures and corrupt data streams.
The desirable attributes of the best memory modules include: high memory density, reduced thickness, reduced height, good thermal exhaust, good signal integrity, reliability, manufacturability and reduced cost. These characteristics are inter-related and the optimization of one may adversely impact the other; so, a balance must be found among these various characteristics to determine the most effective solution. The most effective solution may be relative to a given memory density. In other words, the most effective memory module solution at 2 GB or 4 GB may be different than the solution for 8 GB and 16 GB modules. Furthermore, the best DIMM solution for DDR2 DRAMs may be different than the best DIMM solution for DDR3 and DDR4 DRAMs since the standards are slightly different.
Increasing memory density inside a module can be achieved through stacking of chips. However, stacking of chips is accompanied by an increase in cost of the stacked chips. Also, stacking of chips is accompanied by greater heat that is concentrated between the stacked chips and is more difficult to dissipate which leads to an overall hotter module. Additional heat exhaust can be obtained through the use of heat spreaders or heat sinks. However, the addition of larger heat sinks and heat spreaders is accompanied by an increase in thickness. In turn, increasing thickness leads to a pressure drop between adjacent modules which impedes the flow of air and adversely affects cooling. The words heat spreaders and heat sinks are used interchangeably in this invention.
Thickness reduction can be achieved through the use of thin laminates and flexible circuits. However, flexible circuits can be more expensive than standard PCB based on FR4 materials. Furthermore, thickness reduction causes a mismatch between the thin laminate or the flexible substrate and the standard connector.
As previously taught by J. E. Clayton in a series of patents detailing the use of flexible circuits for memory modules (U.S. Pat. No. 6,665,190, U.S. Pat. No. 6,232,659, U.S. Pat. No. 6,091,145, U.S. Pat. No. 6,049,975, U.S. Pat. No. 5,731,633, U.S. Pat. No. 5,751,553, U.S. Pat. No. 5,708,297, U.S. Pat. No. 5,661,339, 2007/0211426A1, 2007/0212902A1, 2007/0212919A1, 2007/0211711 A1 2007/0212906A1, 2007/0212920A1) the design for the optimal thermal path puts the chip in a configuration that optimizes the heat path between the heat generation source and the metallic heat sinks and heat spreaders. However, when the chips are in an optimal heat path configuration, they may not be in an optimal electrical path configuration.
Flexible circuits exhibit many desirable attributes that lend themselves to solving many electronic packaging problems. Because they are constructed using thin flexible laminates, flexible circuits can be adapted into a large variety of three-dimensional configurations. In particular they are uniquely suited for joining two separate circuit components that involve repeated dynamic flexing motions such as when opening and closing cell phone, camera and notebook LCD displays.
Flexible circuits are also used in applications where reduced thickness or curved surfaces are important. Many mobile products produced today are made feasible by the unique characteristics inherent with flexible circuits. For background purposes, a fairly comprehensive description of flexible circuit technology, including construction methods, design and application specific examples may be found in a book authored by Dr. Joseph Fjelstad entitled “Flexible Circuit Technology” (3rd Edition—September 2006) the teaching of which is incorporated herein by reference in its entirety. Another reference on flexible circuits is “Coombs' Printed Circuits Handbook—Fifth Edition” by Clyde F. Coombs, Jr., the teaching of which is incorporated herein by reference in its entirety. Lastly, “Foldable Flex and Thinned Silicon Multichip Packaging Technology” edited by John W. Balde, is incorporated herein by reference in its entirety.
High performance computers, such as server and super computers, involve many dense, high frequency electrical connections where flexible circuits may be advantageously employed. Their thin uniform laminate thickness, ability to be fashioned with fine lined traces and small vias for layer-to-layer interconnections are better suited for higher frequency operation than traditional rigid printed wiring boards (PWB).
As clock frequencies increase with each succeeding generation of microprocessors, there is an increasing need to design circuit motherboards using techniques for controlled impedance and signal integrity. Usually this results in an increase in the layer count of rigid PWB circuit boards. Computer motherboards that could previously be designed with only 4 or 6 laminate layers now require 8 or more layers to properly route traces operating at higher clock rates. This may increase the cost of these special PCBs.
Using finer wiring patterns, flexible circuits can significantly reduce the number of required layers to form the same circuit functions. Rigid PWB motherboards are, nevertheless, presently required for mounting many hardware pieces such as power supplies, disc drives, fans, and component sockets and are therefore in no danger of being eliminated in the foreseeable future. However, the advantages of flexible circuits, as noted above, are leading many engineers to look for creative ways in which to include them in their designs. Examples developed by SiliconPipe are described by co-founder Dr. Fjelstad and illustrated on pages 32 and 33 of the reference text “Flexible Circuit Technology” cited above.
Although the use of flexible circuits in packaging semiconductors and consumer electronics is well known and offers the capability of high density interconnect signal stability at high frequencies, and flexible form factor (connecting sites that are not aligned), the technology based on flexile circuits is not without disadvantages. Flexible circuits present an overall cost disadvantage, they are inherently a non-rigid form factor and need structural members to be incorporated. Furthermore, the technology needs special design talent for high performance, reliability and operation at high frequencies. One of the challenges caused by the use of flexible circuits is that it reduces the thickness of the substrate to the point that it becomes incompatible with current standard connectors.
Multi Chip Modules (MCM) have a known form factor with known advantages and disadvantages. The advantages of thickness reduction in DIMM applications are increase air flow and space savings on the mother board. However, solving the thickness issues create other issues. Thickness reduction can be achieved by the use of Thin PCB, Rigid Flex, Flex circuitry (connected to standard PCB connector), or by using Flex circuit exclusively.
The thickness of a PCB can be reduced as is the practice of companies such as Eastern Company. These thin laminates can reduce thickness of the PCB but may not be good for high density modules and high frequency operations due to the limited number of layers they utilize and their dielectric properties. However, when thin PCB laminates are used a transition between the thin laminate and the wider DIMM connector is needed.
It will be appreciated that the most effective module solution may not necessarily be the best in any one particular performance attribute, but rather offers the best overall solution at a given density and DRAM clock speed. Objects of the present invention include at least the following: providing effective solutions for thickness reduction in electronic modules; providing electronic modules that are thinner and more cost effective; providing electronic modules suitable for placement at higher densities; providing improved memory modules for computing applications; providing improved memory modules for blade servers; and, providing a method for manufacturing electronic modules that is cost effective and allows for rework.
Additional objects of the invention include providing a means for efficient thermal communication between the enclosed integrated circuit devices and the exterior surfaces of the thermally conductive shell in an electronic module; maintaining the shortest thermal path from the surfaces of the enclosed chips to the ambient air outside an electronic module; adapting a portion of an interconnecting thin flexible circuit or thin printed circuit board to provide exterior contact pads for electrical and mechanical mating with existing DIMM type sockets; adapting electronic modules with stackable connections whereby modules may be stacked in vertical and/or lateral configurations; and, providing an improved means for locating passive components in circuit modules. These and other objects and advantages will become apparent from reading the specification in conjunction with the accompanying drawings, which are not necessarily drawn to scale.
In accordance with one aspect of the invention, a multichip module comprises: a multilayer dielectric substrate comprising two plies, each of the two plies having conductive patterns on both surfaces thereof; a plurality of microelectronic devices attached to the substrate and operably coupled to the conductive patterns on the two plies; a bonding layer joining the two plies over a first portion of their respective areas while a second portion of their respective areas is unbonded so that a bifurcated sheet is formed; two generally elongated rigid dielectric members, one on each side of the multilayer dielectric substrate, configured so that the free ends of the bifurcated sheet may be reflexed respectively about the two rigid members and bonded thereto; and, electrodes disposed on the bifurcated sheet so that the electrodes are exposed after reflexing.
In accordance with other aspects of the invention, the multichip module may include logic devices, memory devices, active or passive RF devices and associated antenna structures, optoelectronic devices including LCD displays, and cellular telephone components. The electrodes may be exposed outwardly and/or downwardly after reflexing. The electrodes and the rigid members may be configured to engage a mating socket.
The present invention addresses the designs of effective solutions to optimize the offering and differentiation for thickness reduction, module densities and relative cost effectiveness. These parameters are balanced per density requirement. For instance, a design that can meet and optimize the attributes of a memory module at the low densities (2 GB and 4 GB) can use a central flexible circuit that places the chips in a mirror image configuration and that offers the benefit of using a lesser amount of flexible circuit and hence enables lower cost. Chips in a mirror image around a flex enable simpler electrical design. Furthermore, the centrally-located flex reduces thickness and when coupled with other features in the current invention (including those features dealing with signal length symmetry, grounding strategies, and the connection interface), would produce an extremely thin and relatively cost effective solution at the low densities.
Skilled artisans will appreciate the value of reducing the thickness of an electronic module. Reducing the module thickness allows cooling air to flow more freely between the modules and to pass through and over other components behind the modules that would otherwise be “shadowed” from the cooling fluid. The thickness of the memory module is presently dictated by the thickness of the PWB substrate, the thickness of the memory packages mounted on the substrate's surfaces, and more recently the additional thickness imposed by bulky heat sinks. These heat sinks are typically clipped over both sides of the module.
Reducing IC packaging thickness is a continuing trend, but appears to have reached a practical limit. Another solution presently being developed is to stack multiple memory devices within the same package dimensions. However, assembly yield, and hence cost, are adversely effected, and stacked memory devices tend to cause more heat to be generated and concentrated between the memory devices themselves.
A solution is provided in this invention by substantially reducing the substrate thickness, allowing the memory components to be positioned closer to the centerline of the module, and incorporating the heat sink 16 (herein referred to as a heat spreader or exterior metal shell or metallic enclosure or stiffener plate(s)) as an integral part of the module to compensate for the reduced rigidity of the thinner substrate. The metal walls in the metal shell serve multiple purposes (chip protection, heat spreading, connector, stiffening, interface, grounding and assembly). In addition to enabling a significant reduction in overall thickness, this invention also provides a robust yet lightweight enclosure for protecting minimally packaged and/or unpackaged, “bare” silicon memory devices that otherwise would be exposed to handling damage if assembled on the exterior surfaces of conventional DIMM substrates. The standard IC devices are typically individually pre-packaged devices (e.g. FBGA, μBGA, CSP, etc.), but the use of unpackaged bare silicon chips (e.g. wafer-level RDL, flip chip, DCA, etc.) would be preferred if its use is made possible and practical. Wafer Level Packaging of semiconductors can achieve very small and cost effective semis packaging form factors. Bare silicon memory devices represent the ultimate minimalist design or thinnest form factor since they are not surrounded with conventional molding compounds that are typically used to house and protect IC components. But the industry has grown dependant on pre-packaged components as a means for handling and testing memory devices to various grade levels.
Recent advancements in wafer probing and testing technology may soon enable the long sought goal of a reliable, high volume supply of Known Good Die (KGD) to the industry. The present invention enables the construction of a standard memory module specifically designed to enclose and protect these fragile KDG memory devices in order for the technology to advance. The package-less die leads to savings at the chip level. This saving per unit silicon is multiplied by the number of placements at the module level which leads to significant savings when the module has 18 placements, 36 placements, 72 or 144 placements.
It will be appreciated that although the present invention represents a significant departure from prior electronic module manufacturing practice, it is clearly sufficiently “manufacturable” to become qualified through recognized industry standardization processes, particularly for memory modules. The invention can enable memory device manufacturers and third party module manufacturers to assemble finished module products directly from memory wafers without the necessity of pre-packaging each individual die first, resulting in a considerable savings in cost and process steps. This is a significant cost reduction for DRAM manufacturers compared to the existing art. By reducing cost and assembly time for DRAM manufacturers, the invention is expected to be a significant enabler of a revolutionary “package it once” trend within the industry Furthermore, significant cost savings can be achieved by the DRAM manufacturers from an operational stand point by eliminating the packaging operations within the organization and by aligning themselves closer to a Just-In-Time business model.
Another advantage of the invention is to significantly reduce the mounted height of the module. Reducing the mounted height allows these modules to either fit into products that are very restricted in height, such as blade servers and mobile devices, or to add more memory inside the module as a result of a net gain in component space. The reduction in module height is achieved by means of a unique contact architecture enabled by this invention that allows the seating plane to be reduced in the DIMM socket, as detailed below.
The direct module attachment to mother boards enables the reduction in parasitic signals and high bus speed through reduction of contacts and transitions. Some of the module interfaces described in this invention are applicable to high speed and large bandwidth electronics. The benefit at the mother board level is the reduction of large holes in the motherboard that may inflict speed penalties. By eliminating the standard connectors that are soldered at the mother board level, and by changing the signal routing at the motherboard level, we enable the optimization of high frequency designs that are more stable and scalable with future requirements. It recognizes that the designs of the novel module designs are coupled or done in conjunction with signal routing layout changes at the mother board level. By adopting flexible circuits at the module level, we also enable better signal integrity, wider voltage margins and more thermal stability compared to module on circuit board alternatives.
The present inventions can be expanded beyond memory modules. In some cases the ICs residing in the modules described in the present inventions can be a combination of logic and memory devices. Certain aspects of the present invention teach and describe the concepts of Total System in Module (TSIM) or the “brain”, whereby memory and microprocessors reside in one very thin module/card suitable for the mobile market. The up-grade of a mobile device may then be done through the replacement of the thin module containing the TSIM.
Another advantage of the invention is to achieve higher density through stacking at the module level not just at the chip level. In either the standard height version or the reduced module height version with zero seating plane, the connection between two modules is engineered to enable the stacking and entering into a functional state of electrical continuity when inserted (or engaged) into one another
The solution to the desirable target thickness reduction and the other desirable characteristics can be achieved by two distinctly fundamental ways. In one solution the thickness advantage is achieved and the connector interface is built using a standard PCB. The second solution achieves the goal using an alternate connection method through the use of bifurcated flex as the interface with the standard connectors. Both these innovative aspects are addressed in the current disclosure.
One of the desirable attributes of this invention is to achieve a multichip module of minimum thickness for enclosing integrated circuit (IC) devices. In order to achieve minimum thickness, these devices are preferably interconnected by a thin flexible circuit or thin printed circuit board placed at the center vertical axis of the module. The IC devices can be individually pre-packaged devices (e.g. FBGA, μBGA, CSP, etc.), but are preferably unpackaged bare silicon chips (e.g. wafer-level RDL, flip chip, DCA, etc.) in order to achieve minimum thickness of the module and reduce cost per chip unit as well as per the number of placements on the module.
Another desirable attribute is to provide a rigid, integral, thermally conductive outer shell which forms an interior cavity within the module to protect the integrated circuit devices and interconnecting substrate contained within. It will be appreciated that this outer shell may have multi-purpose functionality. The outer shell can: a) serve electrical functionality for extending the module grounding plane; b) serve as the assembly carrier for the module; and c) be shaped to mate with the connector.
According to one aspect of the present invention, card edge connectors are fashioned using bifurcated sections of flexible circuits, intended for insertion into a socket. A flexible circuit is sandwiched in the middle between two elongated, substantially rigid (preferably dielectric) stiffener plates 16 with bifurcated flex circuit sections extending some distance along the outer surfaces of the stiffener plates. The stiffener plates can either be flush at their respective edges or staggered,
One aspect of the invention involves the use of a bifurcated flex to provide the connecting transition between a thin laminate to a standard DIMM connector 50. This transition is called an Up-transition to reflect going from thin to thick. A standard DIMM PCB substrate 51 thickness for 5V operation is 1.27 mm,
It will be appreciated by skilled artisans that the flex portions used for an up transition or a down transition are made with impedance matching capability between the two portions to be connected or engaged.
It is conceivable that high frequency designs may require an increase in the number of layers in the PCB 53,
The flex bifurcation in this invention may also achieve other purposes as will be described in the following examples.
Bifurcated Flexible Circuit:
In some applications the flexible circuit is a multilayer structure that consists of multiple dielectric core films with patterned copper traces on one or both surfaces of the dielectric core to form a single flex circuit ply. Separate patterned plies of copper/dielectric sheets can then be laminated together with various adhesives materials to form a multilayered flexible circuit. In some applications it is desirable to limit the adhesive joining the separate flexible plies of a multilayer circuit to be discontinuous. This technique is commonly employed to allow individual plies to be separated apart and connected to different locations on the circuit board. For background purposes, examples of discontinuous plies are illustrated in FIGS. 1-4 of page 8, the bottom figure of page 24, FIGS. 5-5 and 5-6 of page 80 and FIG. 6-20 on page 143 of the reference book “Flexible Circuit Technology”.
For purposes of this application Applicants use the term “bifurcated” to distinguish between other methods for separating flexible circuits.
In some applications the flexible circuit is adhesively bonded to a rigid stiffener such as a plastic or metal plate without any direct electrical contact with the object providing a rigid mounting surface. An example would be a thin flex circuit bonded to a metal heat spreader for cooling one or more chips.
The example shown in
Signal Symmetry through Bifurcation
An improvement over the aforedescribed prior art is illustrated in
When a chip 22 is connected to a flexible circuit, a stiffener is needed to give the assembly structural integrity,
When chips 22 are placed in a mirror image configuration of a central flexible circuit, the stiffener on both sides of the flex is designed with a recessed cavity to host one or more chips (see
Electrical ground pins, disposed among the array of external contact pads 18 of the outer copper patterns 7, can bring grounded connections through plated via holes (not shown) to the interior copper patterns 9. Then, when an electrically conductive adhesive is used as the bonding material 14′ for attaching the bifurcated plies 4 to metal stiffeners 16, the metal stiffeners can function as a reference ground plane that provides protection for interior mounted IC devices from Electro-Static Discharge (ESD) and Electro-Magnetic Interference (EMI).
The metal stiffeners also provide a rigid shell for mechanical protection of fragile bare-IC die 22′ or stacked die 22″ disposed on the surfaces of the flexible circuit and a heat spreading surface to conduct heat away from the enclosed IC devices. The metal shell is preferably fashioned to maintain compatibility with standard connectors 50. The dimensions and mechanical sturdiness required for proper operation is designed into the shell.
Additional Symmetry & Grounding Solutions
When a chip is mounted on a flex circuit, another way to achieve symmetry is to have a flex bend on the same side of the chip in two opposite directions as shown in
Since the assembly shown schematically in
When multiple chips are used, a multi-chip module (MCM) is formed using flex bifurcation on the bottom of the module or alternatively at the bottom and the top of the module, 10a,
The connector pads 18′ can be located on the tip of the flex arch,
Since the above-described modules have connections from the top as well as from the bottom, they offer the possibility of module stacking. It will be appreciated that the electrical connection can be fashioned in different ways while remaining within the spirit of the present invention.
Lateral stacking and vertical stacking are both made possible by the present invention.
In yet another embodiment of this invention, the stiffeners may comprise metal heat dissipating plates 16 that are surrounded and bonded to a frame consisting of injection molded plastic 70,
The minimal thickness module design using bifurcated flex as described herein can be implemented in various configurations. As can be seen from
Another significant benefit of the inventive bifurcated flex design is the ability to add passives 35′ in the fold of the plies to achieve capacitive and resistive functions in addition to the ground connection enabled by contacting the metal dam shell as illustrated in the cross-section view
Once the embedded passives 35′ as well as the metallized layers are bonded to the metal enclosure, a laser trimming and splitting operation may be used to split and create the various lateral connection pads. When separated by splitting (not shown), the contact pads are able to independently operate in achieving and maintaining compliant contact with the mating contacts on either a socket or adjacent module. Alternatively, the contact pads may conform over shaped spring contacts (not shown) beneath that are operable by electrical/mechanical forces (e.g. piezo-electrical or mechanical effect) or simple bumped structures as previously described.
Description of the Exterior Shell
A thermally conductive outer or exterior shell 16, or U-shaped stiffener 16′, serves several purposes. It mechanically encloses and protects the devices within an interior cavity to prevent direct contact with the fragile chips during handling. It also provides an electrostatic and electromagnetic shield when properly grounded to prevent possible damage to the contained ICs from electrostatic discharge and/or prevents potential electromagnetic interference with other closely spaced adjacent components. When properly connected in thermal communication with the enclosed devices the outer shell also provides a means for conducting heat from the operating chips to the external surfaces of the outer shell where the heat can be transferred by either static or forced convection air flow. Alternatively, the exterior shell can form a liquid and/or gas tight enclosure through which a circulating fluid or gas can be introduced and removed to provide direct cooling of the devices within the module's cavity.
The exterior shell preferably comprises a thin metal such as aluminum or copper, or metal alloy such as steel or Kovar that is either machined or stamp formed. Many ceramic materials can also be substituted, provided they have sufficient strength and thermal conductivity. Metal can be easily stamped, pressed or cast into a variety of shapes to improve rigidity, resistance to bend or twist, and thermal dissipation. In addition, metal fittings or ports can be fashioned or added to the exterior shell to enable the introduction and removal of circulating liquids or gases within an interior cavity of the module.
When adapted for memory module applications, the exterior shell would preferably have a generally rectilinear shape. If prepackaged memory components (e.g. FBGA, μBGA, CSP, etc.) are used, the combined thickness of these components, when mounted on both front and backside surfaces of a central flex circuit or thin PWB, can exceed the JEDEC (Joint Electronic Device Engineering Council) specification for the module substrate thickness (1.27±0.10 mm) as measured across the width of the edge finger contacts (pad-to-pad thickness). In this instance the exterior shell can have flanges surrounding a pocket shaped area that bulges out slightly to allow for the protruding DRAM packages,
The exterior shell is intended to be an integral part of the module as distinguished from the current practice of adding separate metal heat spreaders on top of traditional surface mounted DRAM components and using metal clips to hold them in place. This current practice adds to the overall mass and thickness of the module (≧6.5 mm) which impedes the flow or circulation of cooling air between adjacent modules that are closely spaced together. Using an integral heat spreading exterior shell in combination with a thin substrate enables the module thickness to be significantly reduced (≦35 mm) and helps achieve a minimum mass. Since no components are exposed on the exterior surfaces, the module's robustness is greatly improved, potentially enabling robotic insertion of the modules into their respective DIMM sockets.
In many applications of the invention, it is contemplated that two separate exterior shells 16 or metal frames 82 are bonded to both sides of a laminated PWB frame 73 or rigid-flex substrate 90 and function as stiffeners to support the thin flex circuit and form a major portion of both front and backside surfaces of the module. However, in another embodiment of this invention, the exterior shells form almost the entire outer surfaces of the module and in particular the bottom edge portion of the module,
In another embodiment of this invention the two separate exterior shells 16 can be joined with metal tabs 125 across the top edges of the module in a manner that enables the two shells to be folded about a center axis between the two halves,
In a similar manner, represented in another embodiment,
Description of Supporting Hollow Frames
In some embodiments,
In one embodiment these supporting frame members 70 comprise two thin mirror-imaged PWB boards that include standard etched-copper metal pads with Ni—Au plated surfaces arrayed along the bottom exterior edges and intended to mate with current DIMM sockets,
In another embodiment the hollow frames 70 consist of plastic molded pieces that can incorporate a variety of different external and/or internal electrical contacts as shown generally in
According to yet another embodiment the hollow frames 70 may comprise plastic molded pieces without any external contact pads, as illustrated generally in FIGS. 15B and 16-22. In this instance the frames 70 can be mated with a bifurcated center flex circuit having contact pads 11 that extend beyond the bottom edges of the molded frames, as illustrated in
The exterior shells 16 can be attached to the hollow frames 70 by several means. In one instance a pressure sensitive adhesive (PSA) perform is pre-applied to either the flanged edges of the exterior shells or the edges of the hollow frames before the pieces are bonded together. Alternatively, the adhesive may be screen printed onto either surface. Alternatively the interior mating surfaces of both flange areas can consist of a metal or metal alloy over the plastic that would enable the parts to be bonded with a fusible solder perform or laser welding or ultra sonic welding. The attachment of the exterior shells to the hollow frames can be performed either before the hollow frames are bonded to the center flex circuit or thin PWB or after.
Description of the Interior Interconnecting Substrate
The interconnecting substrate preferably comprises a thin multilayer flexible circuit or thin multilayer rigid printed wiring board (PWB) that is approximately 0.008″ (200 μm) thick. The interconnect substrate would preferably occupy a center plane within a symmetrically shaped module such that individual integrated devices 22, 22′, 22″ and passive components 35 are mounted on both front and backside surfaces of the substrate and, if needed, electrically connected across the thickness of the substrate. The interconnecting substrate can also incorporate passive components that are integrated 35′ within or on the surface of the substrate itself consisting of thin or thick film resistors or thin metal capacitive plates layered above each other and separated with dielectric films. Integral passive components of this type are well known within the industry.
The interconnecting substrate material may consist of polyimide, liquid crystal polymer (LCP), polyester film and many other materials well-known within the industry. In some applications for the inventions described herein, the flex material may exhibit opto-electric properties that enable optical signals to be transmitted between the integrated devices mounted within the module. The flex circuit material may either be optically transparent to the optical signals, modified with punched or etched holes to allow optical signals to be transmitted through the thickness of the flex material, or contain special coatings with refractive properties suitable for fashioning waveguides across the surface of the flex circuit.
Description of the Integrated Circuit Devices
The Integrated Circuit (IC) devices 22, 22′, 22″ contained within the interior portion of the module can include DRAM, SRAM, PSRAM, Flash, MRAM, and other newer memory type devices currently under development, or logic devices such as Registers, Advanced Memory Buffer (AMB), Phase Locked Loop (PLL), Serial Presence Detect (SPD) and other similar devices intended to function as driver, buffer, control and interface devices. In addition microprocessor devices can be incorporated within the module and inter-connected with memory chips placed in adjacent locations. In the future IC devices with optical emitters and detectors for signal propagation can be employed. It is also contemplated that passive or active RF devices may be included, as well as LCD and other optoelectronic components, making the invention suitable for use in cellular phones, GPS devices, and RFID systems, among others. It will be appreciated that the aforementioned RF devices may have integral antenna structures or the antenna may be a separate component or it may be constructed as part of one of the metallization patterns on the substrate material.
The aforementioned IC devices are preferably attached to the Interconnecting Substrate using flip chip or direct chip attach (DCA) technology. In this manner the IC devices occupy the smallest footprint area on the substrate and are oriented with the active surface of the ICs towards the substrate. The IC devices can occupy either the front and/or backside surfaces of the interconnecting substrate. Alternatively, the IC devices can be individually packaged in molded plastic and attached to the substrate with solder using surface mount technology (SMT).
Description of the Exterior Contact Pads and Means of Connection with the Interior Interconnecting Substrate
In some embodiments,
Contact pads for the module may also be incorporated within extended portions of a flexible interconnecting substrate that is bifurcated and wrapped around the bottom edges of formed portions of the exterior metal shells,
In an alternative embodiment, the contact pads formed on the flexible circuit are fashioned to enable inductive or capacitive coupling of digital signals from the socket or motherboard into the module. Electrodes for this purpose would be printed or etched into specific patterns and layers of the flex circuit to couple RF energy in the form of digital signals from the socket or motherboard to signal traces within the flex circuit.
Description of the Means for Thermal Communication between the Integrated Circuit Devices and Exterior Shell
As noted earlier, a major goal of the present invention is to remove the heat generated by the contained IC devices as efficiently as possible. Therefore the module is designed to enable the shortest thermal path between the backside surfaces of the contained ICs to the heat spreading surfaces of the exterior shell, whereby heat is subsequently dissipated by conduction and/or convection to the surrounding exterior ambient air or cooling fluid. Since air is intrinsically a thermal insulator, it is desirable that the thermal path between the ICs and interior surfaces of the metal shell be filled with a thermal interface material 26 that exhibits properties of good thermal conductivity and elasticity for cushioning the fragile IC devices. Examples of materials suitable for this purpose would include silicone elastomer composites, either in the form of a paste or compliant rubber-like film that is preferably filled with silicon nitride, boron nitride, or other particulate filler with good thermal conductive properties (e.g. diamond, copper, carbon, etc.).
Description of Module Assembly Methods
A common design feature for the exemplary modules of the invention is a centrally placed thin multilayered PWB or flexible circuit. It is contemplated that this interior interconnecting substrate is typically pre-assembled with surface mounted ICs and passive devices prior to attachment of the hollow frames and/or exterior metal shells.
When hollow molded plastic frames 70, metal frames 82, or PWB frames 73 are employed the exterior metal shells 16 can either be pre-attached to the hollow frames as a sub-assembly or adhesively bonded in place after the hollow frames are separately laminated to the interconnecting substrate. Alternately, the metal exterior shells 16 can be added as a last step of the assembly by sliding them over the hollow frame, as shown in
In one embodiment the interconnecting substrate is a flexible circuit that extends some distance beyond the lower edge of the hollowed frame. The extended flex circuit is bifurcated in a manner to enable two separate flex portions to be folded in opposite directions and bonded around the outer bottom edges of the hollow frame. In this example a PSA adhesive, thermoset adhesive such as epoxy, electrically conductive epoxy, or even solder (when joining metal to metal) may be used for this final step of the assembly. Contact pads intended for mating with a standard DIMM socket are exposed and arrayed along the outside surfaces of the folded flex circuit portions. The contact pads are in turn electrically connected through traces to the interior laminated layers of the interconnecting substrate.
In other embodiments the interconnecting substrate is either flush or extends only a short distance beyond the lower edges of the hollowed PWB frames containing DIMM-type copper clad contact pads. The extreme bottom edges of the hollowed frames contain an array of metallized castellations 123 (crescent shaped copper plated through holes) that electrically bridge across the thickness of the frames and enable the exterior PWB contact pads to be electrically connected with interior pads arrayed on the either side of the interconnecting substrate. The metallized castellations are electrically connected to the pads of the interconnecting substrate using either solder or electro-plated connections,
In summary the assembly steps for non-bifurcated flex modules are as follows:
a. Assemble all IC and passive devices to a thin multilayer interconnecting substrate.
b. Form metal exterior shells configured as two symmetrical heat spreading surfaces that are joined together along their lengths with bendable metal tabs or as two individual pieces.
c. Form two symmetrical hollow frames.
d. Adhesively bond the stamped metal exterior shells to the two hollow frames forming either a unified subassembly or two separate subassemblies.
e. Adhesively bond the hollow frames to an interconnecting substrate.
f. Form electrical contact pads with the interconnecting substrate.
In another embodiment of the invention the hollow frames are eliminated and the metal exterior shells are adapted for mating with a thin, flexible, and bifurcated interconnecting substrate as previous discussed above. The bifurcated portions of the flexible interconnecting substrate are folded in opposite directions, as previously described, and bonded around the outer bottom edges of the metal exterior shell.
In summary the assembly steps for this alternative embodiment of bifurcated flex modules are as follows:
a. Assemble all IC and passive devices to a thin multilayer interconnecting substrate.
b. Form metal exterior shells configured as two symmetrical heat spreading surfaces that are joined together along their lengths with bendable metal tabs or as two individual pieces.
c. Adhesively bond the exterior shell(s) to the interconnecting substrate.
d. Form electrical contact pads with the interconnecting substrate.
Another embodiment, shown generally in
The folding process, for forming a multilayered flex circuit from a single sheet of double-sided flex circuit, can be advantageously combined with component assembly through an SMT assembly line. When properly designed, the flex circuit can have all surface mounted components electrically/mechanically attached onto only the top surface of the double-sided flex circuit. Then using a cutting tool (e.g. laser or water-jet) one-half of the flex circuit can be cut around the circuit's perimeter to loosen it from the surrounding flex carrier. This free half can then be folded underneath itself (i.e the portion of the flex circuit that remains attached to the carrier film) such that half of the components are on opposite surfaces of the folded flex circuit. In this configuration, the folded flex circuit is sandwiched in the middle between surface mount components and is folded about a center axis parallel to the length of the circuit. The inner layers of the folded flex circuit would be aligned and selectively bonded together both electrically and mechanically to form the bifurcated or non-bifurcated flex circuits of the inventive module embodiments described herein. This unique assembly process flow would be extremely efficient using either inline, reel-to-reel, assembly equipment or carrier supported transfer of the flex through a conventional SMT assembly line. The typical need to flip the substrate over for a second double-sided pass through the assembly line would be eliminated by this assembly technique, saving time and increasing SMT throughput.
The inner layers associated with this folding process may consist of either conductor trace patterns similar to the outer layers on which the surface mounted components are attached, or may consist of power or ground planes separated with dielectric films that include selective conductor paths through the thickness,
An array of contact bumps 24 or plated bumps 97 or pads 100′ along the bottom edge of the module enables the module to be connected to the motherboard without the necessity for using sockets with pin contacts. This contact means enables the module to be mounted in a vertical orientation with a minimum profile or height above the plane of the motherboard and allows the module to make direct electrical interface with the motherboard, thereby reducing electrical parasitics. Alternatively, the flexible circuit can include standard DIMM contacts 100 arrayed along the bottom external edges of the PCB window-frames 73 and include additional means for connecting contacts of the folded flexible circuit similar to those discussed in the previous Figures. This would thereby enable the module to be interchangeable adapted for either direct connection to the motherboard or into a standard DIMM type socket. The bumps arrayed along the bottom edge of the folded flex circuit may be fashioned as previously described elsewhere.
A summary of the assembly steps for an alternative embodiment of a bifurcated flex module fashioned by folding a double-sided flex circuit into a multilayered circuit is as follows:
a. Fashion a double-sided flex circuit with selective contact points on one surface and on opposite sides of a folding axis.
b. Assemble all IC and passive devices to the surface opposite the contact points.
c. Fold the flexible circuit along the fold axis such that the selective contact points are aligned across from one another on opposite sides of the folding axis.
d. Form electrical connections between the selective contact points that are aligned.
e. Form metal exterior shells configured as two symmetrical heat spreading surfaces that are joined together along their lengths with bendable metal tabs or as two individual pieces.
f. Adhesively bond the exterior metal shells to the interconnecting substrate.
g. Form electrical contact pads with the interconnecting substrate.
Provisions for Repairs
In each of the examples noted above the metal exterior shell can be affixed to the hollow frame or interconnecting substrate using a PSA type adhesive. Therefore this portion of the assembly is able to be removed, if necessary, to enable access to the IC devices contained within the protected cavity for removal and replacement.
In all of the illustrations, it is important to note that the module architectures are compatible with monolithic chips 22″ as well as stacked chips 22′. Also all of the illustrated architectures are compatible with a double row of chips as illustrated in
The connection method shown generally in
FTG. 32 illustrates in cross section yet another embodiment, similar to
In another embodiment of this invention the central cooling core 124 is of sufficient thickness and is extended an additional distance towards the bottom edge sufficient to enable properly dimensioned DIMM contacts pads 100 to be incorporated within the external surfaces of the folded flexible circuit 98. Alternatively, the folded flexible circuit 98 may incorporate bottom contact bumps or pads 24, 97 or 100 as previous described for
Referring now to
Referring now to
Referring now to the cross-section view,