Six process tools enable packaging of high-frequency products.
By Bradley K. Benton
The telecommunications/datacommunications industry is growing at a feverish pace. The demand for Bluetooth, Home radio frequency (RF), Internet connectivity, cellular telephony, cable, microwave and opto-electronics products seems insatiable. Commercial and military requirements have fueled a measurable portion of this growth and the supply and demand curve seems inelastic. Right now, demand far exceeds supply.
Design Implementation
There are three basic styles for designing products that operate at high frequencies. The first style requires a highly trained and educated individual who has developed and manufactured many successful designs. These designs include an artistic component that allows the circuit to perform in a robust fashion even when built with traditional material and process variations. These designers are capable of building 20 GHz circuits that are manually built and use an antennae that is fashioned from the back of a coffee can. They are “senior scientists” that are few in number, kept very happy at their respective companies and seem to be an endangered species.
 Figure 1. Automated eutectic die attach fillet. |
The second style is one that is much more widespread. This style creates a circuit that comes out of engineering much quicker, but is not very robust against material and process variations. It may include the use of multiple lange couplers that are manually removed (plucked) to tune the circuit for acceptable performance. These products are generally not mass-produced because of the manual labor involved and the potential for increased rework cycles.
 Figure 2. Eutectic die collet tool illuminated by LED light clusters. |
The third style differs in that the manufacturing process requires the use of highly accurate, repeatable, assembly equipment. Because the process uses automation to achieve extreme accuracies, consistent designs can be brought to market much faster. The circuit may still be designed by the senior scientists, but can be accomplished in a very quick and less disciplined fashion. Circuit performance and package-to-package repeatability are greatly enhanced. With these approaches in mind, combined with the worldwide demand for telecom/datacom packages, automation becomes a critical success factor.
Electrical Challenges
High-frequency packages would perform best if there were no wires connecting the individual integrated circuits and substrates. Wires generate large amounts of parasitic inductance and capacitance, and increase insertion and return losses. These electrical conditions (circuit responses to added passive elements) are not necessarily show-stoppers. Variations in these conditions (because of variations in die saw cuts, die placement and wire bonds) are equally harmful to consistent circuit response.
Circuits can be “tuned” to compensate for added reactive inductance, but it is nearly impossible to design and manufacture high-frequency circuits if these reactive components are not consistent. It is best that the electrical engineer designing these circuits computer simulate or model the impact; but in general, circuits designed to operate above 20 to 21 GHz are not very forgiving. Bluetooth and Home RF products are usually robust and reliable, and cellular telephony is in the best shape of the bunch due to their relatively low operating frequencies. Yet all packages in these frequency ranges can be built better and more repeatable when built with intelligent automated processes.
 Figure 3. Automated eutectic attach process. The first die in a relative-to placement chain. |
Flip chip methods may someday enable very high frequency products, but even they require interconnects and, unfortunately, all interconnects induce variance. The flip chip process is not nearly as flexible as traditional wire bonding interconnects, and their ultra-short conduction paths are not required at today's operating frequencies. This process has both benefits and problems. Poor thermal conduction/management, thermal mismatch, poor rework capability, high manufacturing temperatures, solder and alternate die issues all weigh in against flip chip.
Wire bonding is a known process that offers many benefits in response to today's needs. At some point, the industry will be forced to embrace a method to reduce reactive, parasitic elements of device interconnection, but this won't likely happen until companies regularly build 50 to 100 GHz products in large volumes. Wire bonds can be modeled on computers, can be compensated for and can be made with round wire or, preferably, ribbon wire.
Some ways to increase usable frequency include reducing the integrated circuit/package size, increasing the power driven through the circuit, and reducing insertion and return losses. To maneuver through these challenges, microelectronics packaging engineers need what can be considered a “telecom tool kit.”
The Telecom Tool Kit
Automatic Eutectic Die Attach: Starting with bare die, substrate, micro stripline and other materials that are die-attached in the package, the thermal challenge arises. High-frequency circuits generate a large amount of heat, and it needs to be efficiently drawn away from the active areas of the package. The best method for transferring this heat is the eutectic attach process (Figure 1).
 Figure 4. Minimal bond deformation of 1 x 2 mil gold ribbon. |
Eutectic die attach has been used for many years. From the 1960s through the early 1980s, integrated circuits were manually attached – sometimes to heat spreaders, such as molybdenum tabs that are epoxied into a package or directly to the package itself. These were the days of tweezers and hotplates. Accuracy was only as good as the operator – and planarity and gaseous voiding under the die were difficult to control (Figure 2).
In the mid- to late-1980s, semiautomatic eutectic assembly equipment was introduced, and the quality of the attach process improved greatly, although accuracy was still controlled by the operator. During this period, the fluids (epoxy) companies introduced a variety of methods to transfer heat out of a die, while maintaining good electrical conductivity; but these fluids were difficult, at best, to control. The packaging community has come full circle and, in the mid-1990s, fully automatic eutectic attach systems became available. With these systems, quality and accuracy (and most importantly, repeatability) of placement can be controlled. This is the first, and very important, tool for the high-frequency packaging engineer.
 Figure 5. 2.5 x 0.5 mil gold ribbon reverse bonded to MMIC with airbridge. |
'Relative To' Die Placement: Placing die, capacitors and substrates with microstrip lines relative to a package is good, but placing subsequent die or substrates relative to the final location of the last placed component is better. The high-frequency conductive pathway can line up better if this is done. Although it may end up slightly farther off target at the end of a very long string of placements, the accuracy from component to component can be enhanced. This buildup is generally negligible because there are usually 20 or fewer components or substrates to line up. The distance from die to capacitors can be made shorter, consistently reducing the length of the interconnect between the active and passive devices (Figure 3).
Ribbon Wire: Ribbon wire can be used as a beneficial replacement for round wire interconnect in high-frequency circuits – basically across the board. The use of ribbon wire reduces the parasitic inductance of the interconnect. Because of its rectangular shape, it is easier to generate very low, repeatable loop profiles with ribbon wire than with conventional round wire bonding.
Bonding with ribbon wire spreads the contact force of the wire across a wider cross-sectional area of the bonding pad, which minimizes aggravation to underlying metallizations (Figure 4). Because of the larger diffusion area of the bond, ribbon wire bonding reduces junction impedance. Ribbon wire can also carry large amounts of current because of the larger cross-sectional area of the wire; this is generally understood as skin-effect conductance. Because of its rectangular shape, ribbon wire is less-susceptible to wire “burnout” caused by constant flexing of the wire during changes in power application.
Until recently, ribbon wire bonding was generally done with manual bonding equipment. Automatic wedge bonders now exist that can bond ribbon wire and control the tail length on the first bond. To gain maximum benefit from the wire interconnect (round or ribbon), automatic bonding is required for consistent loop profiles. Repeatable tail lengths of less than 1 mil are required (less than .5 mil preferred) for circuits that operate in excess of 20 GHz. Longer or inconsistent tails become miniature transmitters and increase insertion loss in the circuit.
Consistency is a critical concept in relation to the manufacturing of high-frequency packages. Even if the package is not built in the best way possible, if it is built consistently, an analog design engineer can model problems and compensate for them. Not all methods or materials are consistent, so it is necessary to then look to the automated manufacturing equipment to compensate as intelligently as it can. Some may consider consistency to be as critical, if not more so, than accuracy.
Constant Wire Length: The length of the first-level interconnect needs to be consistent, even if the distance between the bond points changes. The looping algorithms of the ribbon bonder must automatically adjust for these span changes. As the span between bond points is shortened, the loop must compensate by growing taller. As the span increases, the loop must be lower. While this is not a perfect process, any compensation is better than none.
 Figure 6. High-frequency ultrasonics. |
This process starts by automatically attaching the die because this reduces span variance. Automatically compensating “constant wire length” looping algorithms continue to add consistency to the interconnect length. Because the ribbon wire is wedge-bonded (it cannot be interconnected with a ball bonder), the wedge bonder must have a very robust wire clamping mechanism. This mechanism should be very close to the foot of the wedge bonding tool to minimize the potential for the wire to behave uncontrollably. When done correctly, ribbon loop heights change with span variations, minimizing the change in the interconnect length. High-frequency circuits generally operate best with 10 to 15 mil ribbon spans, but shorter lengths are almost always desired (Figure 5).
Very Short Wires: As the wire length becomes very short (10 mils), we approach interconnects that are known as lange couplers. Three to four years ago, lange couplers were fairly common in microwave circuits. They create jumper paths over other conductors and, because most wire interconnects in high-frequency circuits are considered necessary evils, these interconnects are needed with minimal spans. This means there is a need to generate very short, low-profile loops – even shorter and lower than those used to interconnect the integrated circuits.
The motions in three-dimensional space that a bonding tool goes through to generate a controlled loop look nothing like the resultant loop profile. Quite a bit of bending and working of the wire is required to keep the wire loop from collapsing while remaining constant in length. Normal looping algorithms fail to provide the control needed for very short wire length interconnects. The speed at which the bond tool moves in its 3-D space must be controlled better, and slowed to ensure the exact amount of wire is spooled out during the loop formation.
High-frequency Ultrasonics: There are measurable performance improvements and/or reductions in process capability of packaging associated with any frequency of ultrasonics. 60 KHz is the incumbent frequency and has been used for more than 25 years. Long ago, 40 KHz was used, and it is hard to imagine the difficulty that changing to 60 KHz presented to those early pioneers of wire bonding.
Today, there exists a rather large faction of process engineers who look at new frequencies of ultrasonics as a method to improve difficult, challenging packaging applications. Sometimes they win, sometimes they get lucky – and sometimes they cause more problems than they solve. This is a delicate subject and, if enabled without significant process verification on specific materials, process capability can be reduced. Packaging engineers should not assume that increasing ultrasonic frequency will alleviate most, or all, of their problems.
There are process/material combinations in which higher frequency ultrasonics almost always offer advantages. Low to ambient temperature bonding of gold wire is the first. Ultra fine-pitch applications are next on the list. Both require higher frequency ultrasonics to enable good bonding and reasonable yields. After these two applications, the debate is on. Suffice it to say that, in some cases, this process option enables the packaging engineer to tackle a wider range of complex problems associated with high-frequency packages. Proper frequency selection is critical for this technology to enable increased process capability. Multiple material types may not allow the use of higher frequency ultrasonics without paying a substantial price. Improper frequency selection – and it is highly material-dependent – can decrease process capability by as much as 75 percent (Figure 6).
Conclusion
With these six tools in the packaging engineer's toolbox, it is possible to manufacture higher quality, more repeatable, more robust, higher yield products. These tools are critical to the success of any high-volume manufacturing process and can provide a company a differentiating position in its market. High volume is defined differently at every manufacturing location. The telecom tool kit enables high-frequency packaging at any volume.
The manufacturing systems and intelligent algorithms are currently available. When combined with automated microelectronics manufacturing equipment, a new era opens to high-frequency design engineers and packaging experts. Package frequencies will continue to climb and, as they do, the telecom tool kit will become a more essential portion of the equation for success.
AP
BRADLEY K. BENTON, product marketing manager, can be contacted at Palomar Technologies Inc., 2230 Oak Ridge Way, Vista, CA 92083; 760-931-3600; Fax: 760-931-5191; E-mail: [email protected].