Addressing the flexibility and complexity of accommodating new package technologies on a single machine, while also improving accuracy.
By Luis Gordon
In a relatively short period of time, electronics packaging and assembly technology has become quite large and pervasive, setting a rapid pace of obsolescence and miniaturization. A major trend in the electronics industry is to make products more personal by making them smarter, lighter, thinner and faster while simultaneously making them more friendly, powerful, reliable, robust and increasingly less expensive. To maintain this escalating pace, equipment manufacturers must also conform to the personalization and miniaturization of automation systems while making them more powerful, faster, reliable, accurate and cost-effective.
Figure 1. Matrix/MCM substrate. |
Electronics packaging is the science of establishing interconnections ranging from zero-level packages (chip-level connections) to motherboards. The most prevalent interconnection technologies are face-up wire bonding, tape automated bonding and flip chip. It is important to note that a prerequisite for each of these technologies is a chip-attach process that can rapidly align and affix sorted die circuits to a wide variety of interconnect materials. This process is often overlooked in terms of its importance in contributing to subsequent process yield and package reliability.
Figure 2. Gripper indexer. |
Contemporary chip-attach machines are often designed to accommodate only a small percentage of the materials and package designs used within a single manufacturing facility. Current die placement capability is usually in the range of ±0.002 to 0.005 in. for the XY axis and ±1.5° in theta. Most machines are die size limited in that fixed optics and/or material handling constraints limit the process field of view and material size flexibility. It is important that new die-attach systems address the flexibility and complexity of accommodating new package technologies on a single machine, while also improving accuracy. To handle the latest package types, new generation chip-attach machines should be capable of ±25 micron placement accuracy.
Figure 3. Epoxy writer. |
The advent of chip-scale packaging has forged a new era in chip-attach capability. Chip size has reduced dramatically and I/O per chip has increased, making it even more critical in achieving repeatable die-placement accuracy. Chip size reduction has also opened the doors to increasing the density of the real estate of the chip substrates. Designs accommodating 80 to 150 die per lead frame or substrate are expected to be commonplace. Additionally, it will become increasingly more feasible to assemble multi-chip modules (MCMs) (Figure 1).
Because of the industry requirement to assemble a wide variety of material compositions and outline dimensions, it is necessary to provide a fully automatic machine with recipe programmable materials/recipe conversion capability. Key equipment features that promote rapid changeover and material flexibility include material handling flexibility facilitated through the use of a gripper indexer (Figure 2) that is capable of handling a wide range of semiconductor substrates (metal lead frames, ball grid arrays, chip array ball grid arrays, singulated ball grid arrays, MCM, and other chip-on-board device substrates). The indexer should be self-teaching to reduce operator error and optimize the sequencing of material through the workholder epoxy and bond stations. Automated self-teaching algorithms should also be applied to the input and output modules.
Another essential feature is a die-attach process that uses a versatile high-speed auger dispense mechanism that is volumetrically and pattern writing programmable. Write patterns that can be dispensed at any angle for multiple die sizes make single-pass processing of MCM devices a reality. Additionally, the dispenser should be easily adapted to dedicated epoxy stamp tooling (Figure 3).
New packaging technologies are not without a new set of problems, including an idealistic but challenging goal to reduce cost by increasing system throughput. Variations in packaging materials are a major contributor to placement accuracy and system throughput.
Placement Accuracy/Repeatability vs. Throughput
Variation of the fiducials within and between substrate production lots can cause vision target recognition failures and adversely affect the desired die placement position. New vision engines need to define the centerline of a fiducial's basic shape to eliminate the variations in die placement without being overly sensitive to subtle lighting and shape fluctuations. Implementation of enhanced vision algorithms on new generation equipment can ultimately reduce the inspection failure along with the need to realign.
Figure 4. Moving cameras. |
Desired placement accuracy is directly proportional to system throughput. To achieve tighter placement tolerances, the fiducial variation should be vision compensated. Additionally, vision optics and substrate transport mechanisms must have increasingly higher resolution. Our studies have shown that higher magnification of the objects to be positioned will produce a higher degree of placement accuracy. Hence, it is becoming necessary to use high-speed programmable zoom optics and/or movable cameras to determine object location (Figure 4). However, the added motion of these assemblies can negatively impact the overall system throughput. It is desirable to have a flexible system that can optimize throughput based on programmable placement tolerance and cus tom-made vision alignment schemes. Figure 5 illustrates the various vision target alignment methods that directly impact system throughput and bond placement accuracy.
High accuracy requires high motor resolution for each axis. Higher resolutions can slow the speed of axis motion, especially in the case of DC stepper motors. To meet the challenge of high-resolution and high-speed motion, state-of-the-art machines will need to be improved by incorporating the use of closed-loop brushless servo motors on each critical axis.
Material Variations vs. Throughput
Another variation in substrate materials is related to the substrate exterior dimensions. New substrate materials made of laminated printed circuit board materials cannot achieve the tight tolerances common to metal lead frames. The dimensional tolerances held in stamped metal frames usually are in the range of ±0.0002 to 0.0005 in. Dimensional tolerances of BGA substrates can be as much as ten times greater (Figure 6).
Figure 5. Vision target variation drawings. |
Because of the loose tolerances, several mechanical and vision issues must be considered to minimize the mean time between assists factor caused by vision targeting failures, vision target retries and mechanical jamming of the substrates during transport in and out of the workholder. Automated theta compensation is necessary to align and place the die with respect to the circuit fiducial regardless of the exterior substrate dimensions.
Figure 6. Typical tolerance for BGA substrate materials. |
To achieve the placement accuracy and throughput objective, there are many features to consider. The substrate should be automatically positioned in the Y axis before entering the workholder, and theta offset correction should be achievable via vision system calculation of device fiducial positioning. Additionally a motorized theta bond axis should compensate for the variation (Figure 7).
It is necessary for substrate transport mechanisms to be capable of locating the bond pad area to within ±1 µm of the stated pitch. To accommodate this, a closed-loop DC stepper or servo system is highly recommended. Traditional workholders employ the use of pin indexers to register the bonding pads. These can cause damage to the substrates and contribute to set-up time and assists. A more friendly transport design employs a gripper configuration so that substrates ranging from 0.004 to 0.050-in. thick can be processed without tooling change.
Figure 7. Rotary bondhead. |
New state-of-the-art machines will need to employ non-contact measuring devices, such as reflective fiber optics or laser components, that are inherently fast. The resolution required maintaining the bond-line thickness to within ±10 µm should be at the sub-micron level (0.5 µm or less). Substrate thickness variations are inherent in the manufacturing process of BGA substrates. Therefore, it is important to be able to automatically detect the variation and compensate the Z-axis position at the epoxy deposition and bond stations. Real-time bond line thickness control in conventional die-attach machines uses a mechanical sensing (make/break switch) design that is slow and susceptible to inaccuracies due to contact bounce.
The industry is demanding a completely new generation of chip apply or die attach systems capable of meeting these stringent requirements while also offering a greater degree of flexibility and product versatility. These stringent prerequisites make designing tomorrow's equipment more challenging than ever before.
LUIS GORDON, senior die attach project manager, can be contacted at ESC International, 4 Ivybrook Blvd., Ivyland, PA 18974; 215-682-9300; Fax: 215-682-9318; E-mail: [email protected].