BENEFITS OF AUTOMATED RIBBON BONDING FOR MICROWAVE APPLICATIONS

BRADLEY K. BENTON

Palomar Technologies Inc.

Vista, CA

With today's ever-increasing need for larger bandwidth in telecommunications, datacom, aerospace and defense applications, packaging challenges continue to grow. As the frequency at which the packages are required to operate increases, the wire itself becomes one of the larger problems. Although the impact of first-level interconnect (wire bonding) is fairly well understood, minimizing its influence on circuit performance can be interesting, to say the least.

Operating frequencies are rising faster than technology stocks, and soon it will be commonplace to manufacture circuits that operate in the 50 to 100 GHz range. At some point alternative solutions will have to be examined. This forcing function is not a generation away, but closer to five years. Until then, the challenge of solving today's (and some of tomorrow's) problems with today's technology remains. Fine ribbon wire first-level interconnects are a major part of the solution.

CHALLENGES

First-level interconnects are hazardous to the electrical performance of high frequency circuits! Interconnect variation, both in wire length and loop shape, makes high frequency circuit design all the more critical. For any substantial level of volume manufacturing, packaging engineers work very hard to design a circuit that does not require manual tuning and is robust in the face of uncontrollable variations. This effort can be an artful task, and using ribbon wire to interconnect the microstrip transmission lines, waveguides, and passive and active components, as shown in Figure 1, can assist in making it manageable.

Changing from round wire interconnects to ribbon wire can be considered a technology shift. Wire bonding high frequency circuits is a fairly well understood activity. Circuits are designed as well as possible to include the parasitic reactance functions of the traditional round wire interconnect. Variations associated with interconnect performance become one of the main challenges, and methods to minimize these variations can come at a high fiscal price. Portions of the variation equation can be controlled by attaching components automatically with very high accuracy, high repeatability and, unfortunately, high cost systems. When using round wire, these high accuracy systems are generally a requirement. Ribbon wire enjoys some benefit in this area, but most manufacturers use automated component placement systems regardless of the interconnect media.

Modeling ribbon wire in a high frequency circuit is very similar to methods employed for round wire. Since wires interconnecting microstrip transmission lines can be made basically flat, the inductance of both round and ribbon wire has been expressed as1

Ribbon Wire

Round Wire

where

L = inductance in nanohenries

l = length of the wire

t = ribbon thickness

w = ribbon width

d = wire diameter

m = permeability (assumed to be 1)

e = skin effect correction factor (a function of wire diameter and frequency)

(All of the dimensional units are in micrometers.)

Microwave amplifiers are shrinking not only in surface area, but also in thickness. Twenty-eight gigahertz FETs have pad sizes as small as 1.4 mil square, forcing round wire wedge bonding to be accomplished with 0.0007" Au wire. Just last year, integrated circuits were 4 to 5 mils thick; this year they are as thin as 1 mil. Bonding pads on these active circuits are extremely fragile, especially if they are made of GaAs. Packaging engineers are constantly forced into a trade-off between performance and yield, and they need all the help they can get. Ultra-fine round wire is already being pushed to its limit; however, this is not the case for ribbon wire.

To obtain a high yield with a reasonable process capability on an automatic wire bonder when using ultra-fine round wire (£ 0.0007"), enough ultrasonic power and force must be applied to deform the wire to almost two times the wire diameter, or approximately 1.4 mil. The width of the final deformed bond can be the same size as the bonding pad, if not larger. Figure 2 shows 0.5 mil (thick) ¥ 1.5 mil (wide) ribbon bonds on a FET with 1.8 mil round bond pads. Notice the minimal bond deformation that occurs with 0.5-mil-thick ribbon.

The initial bonder setup is another challenge. To bond the wire, the wire first must be threaded through a titanium carbide wedge tool (for gold wire) with an exceedingly small tool hole. The wire must be capable of being drawn easily through this guide hole in the tool. Threading a 0.7 mil wire through a tight-toleranced wedge tool is almost art in itself .

The smallest tool hole that is considered practical for 0.7 mil round wire is between 1.1 and 1.3 mil. Since the wire can position itself anywhere inside the tool hole, it can be assumed that the wire will be in that range under the bond tool foot when it is time to create the bond. Remember, even with the smallest possible tool hole (1.1 mil), the operator is attempting to put the 0.7 mil wire completely on the bonding pad, deforming the wire to approximately 1.4 mil and doing so on a bond pad that is only 1.4 mil wide. Round wire has provided a solution to this problem for as long as possible, but is at the end of the line. Figure 3 shows 0.5 ¥ 1.5 mil compound ribbon bonds to a 2 mil conductor.

If the wire is against the wall of the tool hole, the resultant bond could be off the bond pad by as much as 0.2+ mil (assuming that the tool foot was perfectly centered on the bond pad). Realistically, there are tolerances in this targeting that add to the problem. Ribbon that is 0.5 mil thick deforms approximately 1.1 to 1.4 times the ribbon width, while 1-mil-thick ribbon wire typically deforms one and a half to two times its width. Half-mil-thick ribbon clearly becomes a possible alternative to ultra-fine round wire bonding. Notice the minimal bond deformation. Figure 4 shows minimal bond deformation on a 4 mil bond pad.

Ribbon wire geometries as small as 0.5 ¥ 1.0 mil are theoretically possible, but, to the author's knowledge, are not being utilized in a production environment at this time. Until recently, creating the bonding tool itself was a difficult task. This geometry allows for the ever-shrinking wire bond pad of the future. Ribbon wire manufacturers can manufacture a wide variety of ribbon wires, including ultra-fine ribbon wire, at various tensile strengths. As the challenge to wire bonding increases over time, ribbon wire will remain a solution.

Fig. 2 0.5 ¥ 1.5 mil gold ribbon bonds on 1.8 mil bond pads.

Fig. 3 Compound ribbon bonds on a 2 mil conductor.

Fig. 4 1 ¥ 2 mil gold ribbon bond deformation on a 4 mil bond pad.

Fig. 5 1 ¥ 10 gold ribbon used for high frequency power connects.

Fig. 6 The effects of impact force and static bonding force on cratering.

FORCING FUNCTIONS

Flip chip is a possible alternative (with its own measurable performance implications), but the cost - both in flexibility and design - suggests that the first-level interconnect will remain wire for an extended period of time. Eventually the industry may be forced into an alternative interconnect technology (possibly flip chip) that has extremely short conductors, but clock frequencies will have to push beyond 50 GHz to force that shift. Performance modeling is not as well defined for this emerging technology.

Substrate materials are constantly evolving and could extend the time of this shift's arrival. Passive components are being embedded with the goal of reducing and controlling parasitic inductances and capacitances associated with surface-mount components. Active circuits are becoming more embedded, and there is a strong desire to have many of today's discrete subsystems integrated into a single integrated circuit. All of these activities are making ribbon wire a reliable, known technology.

PERFORMANCE IMPROVEMENTS

One of the main electrical advantages of ribbon wire vs. round wire is the fact that at high frequencies ribbon wire impedance and inductance can be lower than round wire. (Both are very good qualities.) Skin or surface effects decrease the inductance, but at high frequencies can increase impedance. Close attention is required when balancing the thickness vs. width of the ribbon to minimize or eliminate this potential effect. In general, the higher the aspect ratio the better. This form factor is accomplished by determining the smallest wire bond pad or conductor width and matching it with the largest possible ribbon width. The use of 0.5-mil-thick ribbon is then employed and the best ratio possible is achieved.

Mechanical advantages are measurable as well. Complex microwave hybrids, certain types of multichip modules and most high frequency power amplifiers dissipate large amounts of power. Ribbon wire inherently carries larger amounts of power without forcing the interconnect media into a burnout condition. As power dissipation varies, all wires (both round and ribbon) actually flex. Flexing catalyzes bond heel fatigue (mechanical flexing of the wire at the point where the wire exits the bond site), which is the area that is structurally the weakest point of a wedge-bonded wire. Wire heel cracking with ribbon wire is less pronounced than with round wire, thus extending the life cycle of the ribbon wire interconnect. Figure 5 shows very small deformation of both first and second bonds (both ends of the ribbon) on 1 ¥ 10 mil Au ribbon. Note also the short bond tails.

There are at least two reasons for this low deformation. First, the bond deformation is less for ribbon than for round wire (thus resulting in more thickness of the wire at its weakest point, the first bond heel). Second, due to the rectangular shape of the ribbon, low wire looping profiles are easier to generate, allowing a kinder ascent angle from the first bond and minimizing aggravation to the heel during looping. Less kick-back (reverse loop motion used to help the wire stand up) is required to keep the ribbon's loop from collapsing, allowing lower loops while minimizing the work hardening (flexing) of the heel of the bond. Very low, short wires can be automatically bonded with minimal stress to the ribbon wire itself.

With the reduction in overall wire length, reactive electrical components and impedance, ribbon wire bonded circuits generate lower return and insertion losses.2 In this highly competitive marketplace, hard data to support reports of higher signal-to-noise ratios are difficult to gather in writing from an automated equipment manufacturer's users. However, it has been observed that the vast majority of known users who model the benefits of ribbon vs. round wire make the switch and never go back. It has been reported that a 3 dB gain is very typical when switching from round wire to ribbon, but, again, this is hearsay. Hopefully some of those circuit designers who read this article will take the time to model and measure the performance improvement.

In many ways, ribbon wire wedge bonding has an interesting domino effect of benefits. There are many good reasons to employ ribbon bonding, and very few reasons not to. Best of all, the operator does not have to be working strictly with MMICs, MESFETs or microstrip lines to enjoy the benefits or improvements associated with ribbon bonding.

PROCESS IMPROVEMENTS

It is a rare event in the microelectronics business when a technology shift yields both performance improvements and an increase in process capability; that is, until automated ribbon wire bonding became available. Not only are distinct, measurable electrical performance improvements observed, but process and yield improvements as well.

The lowering of interconnect-related impedance and other process benefits are derived in part due to the large bond area at both ends of a ribbon wire. The large bond area distributes the bonding force over a larger surface area of the bonding pad during the intermetallic formation process. This larger area improves yields by reducing the incidence of cratering (catastrophic bond pad failure) associated with the relatively high point contact force that occurs to the bond pad interface metallization with round wire.

This contact force vs. cratering is different from the potential for cratering caused by too low or too high an impact or static bond force applied during the generation of the bond. Figure 6 shows a relationship for impact force and static bond force with bond pad cratering for thermosonic ball bonding.3 Low bond pad impact force is generally used when round wire wedge bonding delicate ICs such as GaAs, but this process decreases bonding speeds dramatically. When automatically bonding ribbon wire, this speed reduction is not necessarily the case.

Fig. 7 1 ¥ 2 mil gold ribbon on a 4 mil pad.

Fig. 8 Low profile looping.

The wedge bonder applies force through the ribbon by the bond foot (active area) of the wedge tool, as shown in Figure 7. This force is distributed over the entire bond foot of the wedge tool, which has a much larger surface area with ribbon than with round wire. In today's high frequency packages, this force distribution is a very attractive process when dealing with small, fragile bond pads. Since a larger bonded area exists (with minimal deformation), the bond junction impedance also tends to be reduced.

Everything comes at a price, and since less deformation is required to bond ribbon wire, organic contamination at the bond site must be managed. While not a requirement, argon plasma cleaning is recommended to remove or control the organics prior to wire bonding. Surface oxides and contaminations are not scrubbed away as much as they are when utilizing the higher deforming round wires. Wedge tool tip-skid offsets this phenomenon if the bonding tool can continue motion in the z-axis, pivoting after the wire touches down on the bonding pad.

Elevated bonding temperatures are required for a robust process capability when using gold ribbon wire. This characteristic is classical thermosonic wire bonding, and hotter is better up to approximately 160°C. The effects of temperature above 160°C could be debated due to trade-offs in yield, IC performance degradation and various material interactions. Some GaAs circuits have a maximum time at temperature that must be observed.

Aluminum wire is always bonded at room temperature (ultrasonic bonding) unless a true monometallic bonding condition is involved. A true Al-to-Al bond can be created at elevated temperatures, but, if care is not exercised, more problems can be generated than solved. Interestingly, almost all ribbon wedge bonding is done using gold in monometallic and bimetallic modes. The age of ribbon wire is an important variable to track. (Younger is always better.)

Due to ribbon wire's rectangular shape, very consistent loop profiles can be made, as shown in Figure 8. Short wires that have large height differences between the ends of the ribbon can be bonded with exceptional repeatability and control. As with microstrip lines, near-planar ribbon bonding is a very robust process. Having the capability to control loop profiles accurately allows the generation of short, low loops. Both concepts address an improvement in repeatable reactive inductance associated with the interconnecting wire.

CONCLUSION

Some concepts are for certain: Shortening the circuit interconnect wire is, in general, beneficial to high frequency circuits. Making repeatable loop profiles and shapes is another plus when it comes to designing a circuit that is manufacturable. Reducing reactive electrical characteristics associated with the wire interconnect generally lends itself to simpler designs and more repeatable high frequency circuit performance. Finally, yields and reliability improve with ribbon wire.

Ribbon wire bonding improves performance, can carry higher current and is more delicate to fragile GaAs bonding pads. In addition, it generates stronger wire interconnects that last longer. Ribbon wire bonding is a known technology and can be modeled as easily as round wire. Until recently, ribbon interconnects were much larger and had to be soldered or welded into a circuit. Today, fully automatic fine wire ribbon bonders are available to assist circuit and packaging engineers with solutions to their formidable tasks.

Until first-level interconnects are forced into alternate packaging technologies, ribbon wire will bridge the gap. It is rare indeed that a required technology actually improves a process. For the foreseeable future, ribbon wire will remain the dominant high frequency package interconnect of choice. Automated ribbon wire bonding is a true win-win solution. n

References

1. G.G. Harman, Wire Bonding in Microelectronics: Materials, Processes, Reliability and Yield, McGraw-Hill, New York, 1997, pp. 31-32, 275.

2. R. Brown, "Materials and Processes for Microwave Hybrids," ISHM Proceedings, 1991, pp. 219-220.

3. R. McKenna, "High Impact Bonding to Improve Reliability of VLSI Die in Plastic Packages," 39th Proceedings of the IEEE Electronics Components Conference, Houston, Texas, May 21-24, 1989, pp. 424-427.

Bradley K. Benton gained experience in various fields of electrical engineering, computer science and systems engineering at the University of Arizona from 1973 to 1983. From 1983 to 1995, he was a member of the technical staff at Hughes Aircraft Company, specializing in microelectronic assembly equipment design and factory automation. He has a patent in wire bond signature analysis. Currently, Benton is the product marketing manager for Palomar Technologies Inc. and can be reached at 2230 Oak Ridge Way, Vista, CA 92083 (760) 931-3600, fax (760) 931-5191 or e-mail: bkbenton@bonders.com.