Microwave Journal
www.microwavejournal.com/articles/2045-backside-connections

Backside Connections

Various high frequency backside interface designs, which are critical to good RF performance in all connector systems

March 1, 1997

Backside Connections

Bill Oldfield
Wiltron Co.

Morgan Hill, CA

The connector interface and its effect on calibration accuracy have been discussed previously.1 This article focuses on the other side of the connector, the backside, where the connector interfaces with the circuit. It is this area where the design engineer has the most flexibility and where most connector-related problems arise. As frequencies get higher and circuits get smaller, the problem of the backside interface becomes more critical. Consider the problem of connection to a 5 mil microstrip at 85 GHz. This article discusses the pros and cons of various high frequency backside interface designs.

Simple solutions exist for backside interface problems at low frequencies where circuits are approximately the same size as the center conductors of the connectors, which typically are SMA types. However, problems arise as frequencies increase and circuit size decreases. A common circuit board designed for use at higher frequencies is 10 mil (0.25 mm) microstrip. Transitioning from a 10 mil trace to a 50 mil center conductor (SMA size) is not good design practice. Even when the center conductor is reduced to 10 mils, the connection to the trace is not simple. Bonding is difficult and produces poor RF performance due to its inductance. Soldering is difficult because the joint is not flexible and the unit must be heated. Excessive solder or a long pin overlap will degrade the RF performance, and a small amount of solder or small overlap will yield a weak joint. Figure 1 shows a typical high frequency connector system consisting of a connector, glass bead and backside interface. The pin overlaps the microstrip trace and must be attached to it in some manner. The glass bead provides a rigidly mounted, small-diameter pin for attachment to the narrow microstrip trace. Tolerances and assembly techniques are critical to good performance. If good performance is required, assembly can become expensive. Figure 2 shows the range of return loss from such a system.

Fig. 1: A typical high frequency connector system.

Fig. 2: The glass bead interface return loss range.

Overlap Connection Problems

In the case where an overlap pin is used, the pin is set over the microstrip trace and may or may not make mechanical connection to it. Direct mechanical connection must be made using two techniques: solder and ribbon bonding. Figure 3 shows the problems associated with these techniques. The overlap design is capacitive because the pin is larger than the trace and the solder only adds to the capacitance. Also, the solder joint must be kept small for electrical performance and is, therefore, rather weak. Thermal shock can break this bond easily.

Fig. 3: Coax pin-to-microstrip connections.

The gold ribbon wraparound technique is inductive and it is difficult to obtain reasonable performance above 20 GHz. Another disadvantage of the overlap design when used with a glass bead is that the circuit substrate must be mounted before the glass bead is installed. This assembly procedure subjects the circuit substrate to the soldering temperature and flux of the glass bead soldering operation.

Glass Bead vs. No Glass Bead

Glass beads have many advantages: they provide a rigid, hermetic connection point and the fragile backside connection is not disturbed by movement or removal of the connector. However, glass beads also have disadvantages. The added complexity of the bead provides more opportunity for mismatches and, as described previously, the assembly process is complex and potentially damaging to the assembled circuit.

Connections without glass beads are not recommended if the substrate is 10 mil alumina and the trace width is 10 mils because the contact is too fragile and will break if the connector is not perfectly stable. If the substrate is Duroid or another material with a wide trace, enough solder surface exists to provide adequate support to the pin. This design is shown in Figure 4 . Thin alumina substrates such as 10 mil alumina can be used if the trace is widened, the ground plane is removed from the bottom of the substrate and the appropriate air gap is machined in the housing. An example of the relieved ground plane design is shown in Figure 5 . Care must be taken to properly compensate the area where the trace width changes.

Fig. 4: A beadless connection.

Fig. 5: The relieved ground plane design.

Backside Design Practices

There are a number of basic rules for good backside design. The interface area must be reduced to a size commensurate with the substrate used. A 50 mil SMA center conductor should not be connected to a 10 mil alumina substrate. A good rule of thumb is that the pin diameter should be no larger than the substrate width. This small size minimizes the generation of unwanted modes and keeps compensation steps to a minimum. The model should be 50 ohms. The only non-50 ohms sections should be compensation steps, which should be small. All changes in geometry must be compensated. Thermal expansion and contraction must be allowed for. Connection inductances and capacitances must be minimized. (Examples were given previously.) Finally, it should be remembered that someone must assemble this design. Therefore, the assembly procedure should be made as simple as possible.

Sliding Contacts

A major problem with the overlap connection is that the pin must be installed after the substrate is placed into the housing. If a glass bead is used, the substrate is subjected to the soldering and cleaning operations involved. Using a glass bead is a poor practice, especially if the substrate contains devices or delicate wire bonds. Some alternatives to the overlap connection are shown. Figure 6 shows a design using a sliding contact formed from thin, gold-plated BeCu shim stock. The pin is recessed in the housing and, after substrate assembly, the sliding contact is slid over the pin and bonded to the substrate. The sliding contact also provides stress relief since the pin joint can slide during expansion and contraction. The tab of the sliding contact can be made small to accommodate thin microstrip substrates. On the negative side, the sliding contacts are difficult to push into position because they are small and delicate. In addition, bonding can be difficult because the contacts are thick (0.001") compared to the thickness of the gold on the substrate. Typical performance of the sliding contact design is shown in Figure 7 .

Fig. 6: The sliding contact design.

Fig. 7: The sliding contact return loss.

Overlap Connection Alternatives

Axial Pin Design

The axial pin design is shown in Figure 8 . The pin enters the housing cavity but does not overlap the substrate. Instead, the pin is axial to the substrate and the ground plane under the pin is set so that the pin is a 50 ohms section of wire-above-ground transmission line. Connection to the microstrip is made using any bonding technique that accommodates thin gold ribbon. Stress relief is provided by the slight bow of the ribbon. This technique is user friendly. The gap between the pin and ground can be set by placing the proper thickness shim between the pin and ground, and bending the pin to the proper gap. Bonding is accessible and easy. Thin substrates can be accommodated using this technique. The width of the jumper ribbon is selected to match the substrate trace width. The disadvantages to this technique are that a glass bead must be used and the housing must be machined to a fairly tight tolerance, a procedure that is easy with modern computerized numerically controlled machines.

Fig. 8: The axial pin design.

Resilient Axial Pin Design

A newly patented axial design, sometimes called a watermelon seed contact, requires no bonding or soldering for the connection to microstrip or coplanar waveguide (CPW) lines. The design, shown in Figure 9 , uses a special pin inserted into the backside fingered contact of the connector. When axial pressure is exerted on the pin, the tapered pin spreads the fingers of the female contact causing an opposing pressure. The fingers attempt to expel the pin in the same manner as a squeezed watermelon seed. On the circuit side, a gold ribbon is wrapped over the end of the circuit and the pin makes contact with the end of the circuit. A major advantage of this design is that no bonding or soldering is required; the contact is made by resilient axial pressure. The design works best for CPW circuits with no ground on the far side, but can be adapted to microstrip if a section of the ground is removed as shown. Circuit connections up to 60 GHz have been made using this design.

Fig. 9: A watermelon seed contact.

The typical performance of a fixture consisting of a pair of watermelon seed connectors connected to a 10 mil microstrip line is shown in Figure 10 . The ripple line is the performance of the entire fixture and the smooth line is a single connector. The single-connector measurement is made using frequency-gated-by-time techniques.

Fig. 10: Return loss of a fixture containing a pair of watermelon seed connectors.

Conclusion

Backside interface design is critical to good RF performance in all connector systems. The backside design becomes more critical as frequency increases. Many design alternatives are available to deal with backside connection problems for a variety of requirements.

Reference

  1. Bill Oldfield, "The Connector Interface and Its Effect on Calibration Accuracy," Microwave Journal , March 1996, pp. 106-114.