Selecting Circuit Materials for Microwave Power Amps
Demand is growing for microwave power amplifiers (PA) due to an increasing number of applications, from commercial cellular systems to military radars that rely on amplified high-frequency signals. Solid-state microwave PAs depend on the characteristics of their active devices, but they also depend a great deal on the behavior and performance of their printed-circuit-board (PCB) materials for such functions as minimizing loss, maximizing gain and efficiently dissipating heat. Solid-state amplifier designers currently have a wide range of PCB materials to choose from. Sorting through these materials to make a suitable choice for a PA application can require careful comparison of some key PCB material properties, including the relative permittivity (or dielectric constant, εr) and the dissipation factor (tan δ). But many other PCB material parameters must be considered when choosing a substrate for a microwave PA, including thermal conductivity, temperature coefficient of dielectric constant (TCDk), copper surface roughness, tolerance of εr and even the tolerance of the PCB material’s thickness. Understanding how these different PCB material properties relate to microwave PA performance can help in sorting through the many PCB material choices currently available and help enable the design of a microwave PA that not only meets its performance goals, but is dependable and reliable under a wide range of operating conditions.
Design and fabrication of a microwave PA requires control of a number of different parameters, with impedance control among the most important. Since a microwave PA may interact with or include many different types of components, such as impedance matching networks, quarter-wave transformers and 3 dB quadrature couplers as part of a larger circuit or system, the characteristic impedance of the amplifier circuitry, which is usually 50 Ω, should be tightly controlled. The circuit material εr tolerance is one key material parameter to consider when controlling impedance. Although designers may often assume that circuit material εr tolerance is the most critical material parameter for achieving good impedance control, this is typically not the case. In truth, when using a relatively well-controlled, high-frequency circuit laminate, the εr tolerance will generally be fairly tight and one of the lesser concerns in terms of impedance control. Table 1 provides an example, assuming a 50 Ω microstrip transmission line on a high-frequency circuit laminate with nominal εr of 3.5 and εr tolerance of ±0.05.
Table 1 shows the impact of some common variables associated with PCB technology on the impedance of a PCB-based microstrip transmission-line circuit. It is divided into two groups of data, with the top section for a thicker circuit (20 mil thick substrate) and the lower section relating the same information for a thinner circuit (10 mil thick substrate). Comparing the two groups of data shows that thinner circuits are more sensitive to conductor effects, which means that variations in circuit conductor width and copper thickness will also impact the changes in impedance.
For example, a PCB with plated-through-hole (PTH) technology uses additional copper plated on top of the original laminate copper to achieve the PTH features. The amount of additional copper can vary from circuit-to-circuit and from batch-to-batch of fabricated circuits. Copper thickness plating tolerance of ±0.5 mils is common in the high-frequency industry although some circuit fabricators can achieve and maintain even tighter tolerances. Based on this tolerance value, a 1 mil range in different copper thicknesses for the same circuit design is not uncommon and, as noted earlier, the thinner the circuit material, the larger the impact in terms of impedance variations.
In addition to concerns of copper conductor thickness, the tolerance of the width of the conductor is also a concern in maintaining consistent impedance for a PCB-based PA. A tolerance of ±0.5 mils is again not uncommon in the high-frequency industry for the width of a copper conductor on a PCB, although many fabricators can achieve and maintain even tighter tolerances for the width of a copper conductor. For a tolerance of ±0.5 mils, the variation in copper conductor width is 1 mil, and thinner circuits will be more affected in terms of impedance by these conductor-width variations than thicker circuits.
One of the circuit material parameters shown in Table 1 with a significant impact on maintaining consistent impedance is the tolerance of the circuit substrate thickness. For the example in Table 1, the substrate thickness tolerance of ±10 percent has the greatest effect on the variations in circuit impedance. While this value of circuit material thickness tolerance is not uncommon, high-frequency laminates with better thickness tolerance values are available, which can maintain greater control of impedance for circuits such as microwave PAs.
The εr variation shown in Table 1 is one sided and only shows the negative portion of the tolerance. An εr tolerance of ±0.05 for a nominal circuit material εr of 3.5 represents a minus variation in εr of approximately 1.4 percent. But because it is a tolerance range with “plus” and “minus” limits, the actual deviations in εr can be a range that is twice that value, with changes of εr of 0.1 possible, or a percentage of 2.8 percent of the total circuit material εr value of 3.5. This parameter is not easy to portray in Table 1, but it should be noted that it will be the least-significant issue for the thinner (10 mil thick) circuit material and the next-to-least significant issue for the thicker (20 mil thick) circuit material. A circuit material with this nominal εr value and tolerance is considered quite good for many high-frequency circuit applications, although circuit materials are available with even tighter εr tolerance control.
Insertion Loss Variations
Circuit materials with low insertion loss play a major role in achieving maximum gain in a microwave PCB-based microwave PA. PCB insertion loss is made up of four components: conductor loss, dielectric loss, radiation loss and leakage loss. At lower microwave frequencies, dielectric and conductor losses account for most PCB insertion loss; which of the two parameters is dominant depends upon the thickness of the circuit substrate material. Figure 1 shows three insertion loss charts for 50 Ω microstrip transmission-line circuits based on the same circuit material, although with different material thicknesses. The insertion loss models were generated with MWI-2014 circuit material analysis and modeling software from Rogers Corp. which uses closed-form equations from Hammerstad and Jensen1 regarding microstrip impedance and loss predictions.
The three charts in Figure 1 compare measured and computer-modeled results. The measured insertion loss results were collected using a simple microstrip differential length method,2 which minimized the contributions of insertion loss due to connectors and signal launches. The modeled insertion loss consisted of dielectric and conductor losses and, as can be seen from the plotted data, the modeled data for the total loss agreed fairly closely with the measured results. For simplicity, losses due to radiation and leakage were ignored in this comparison.
Figure 1 shows that insertion loss for a thin circuit (a) is dominated by conductor loss. For the thicker circuit (c), the dielectric and conductor loss contributions to circuit material insertion loss are about equal. If data for yet another, thicker (30 mil thick) circuit material had been added to Figure 1, it would have shown that dielectric loss would dominate the contribution to insertion loss. By knowing which loss components impact the insertion loss the most, a designer can select a material (and thickness) that will provide the optimum insertion loss performance for an application. In the 30 mil thick case not shown, changes in the dissipation factor had the greatest effect on the circuit material’s insertion loss properties. For the thin circuit material in Figure 1a, the insertion loss was most affected by conductor loss, reiterating that changes to the circuit material’s conductor properties will have the greatest impact on the circuit’s insertion loss performance.
The copper surface roughness of a high-frequency circuit material can have an impact on the material’s conductor loss.3 A circuit material with a rough copper surface will have increased conductor loss compared to the same circuit material with a smooth copper surface. How much the circuit material’s conductor loss will change with a change in copper surface roughness will depend on how much the conductor losses dominate the material’s insertion loss. For the insertion loss of the thin circuit material shown in Figure 1a, a change in copper surface roughness will have a significant impact on the circuit material’s insertion loss. For a thicker circuit material, the same change in copper surface roughness will not impact the material’s insertion loss performance as much, and adjustments to the thicker material’s dissipation factor would have more impact on the insertion loss of the thicker circuit material.
To better understand how the copper surface roughness of a circuit material can impact the insertion loss for different thicknesses of the same circuit substrate material, Figure 2 shows microstrip transmission line circuits fabricated on the same substrate, but with different material thicknesses and copper surface roughness. The plots show circuit materials with a standard level of copper surface roughness at 2.8 µm RMS compared to LoPro™ ?laminate from Rogers Corp. with copper surface roughness of only 0.8 µm RMS. As shown, the benefits of a smoother copper surface for circuit insertion loss are much greater for the thinner circuits than the thicker circuits, with a difference in insertion loss of about 0.3 dB/in for the 7.3 mil thick circuit (when switching to the smoother copper surface) compared to a difference of about 0.1 dB/in for the 20 mil thick circuits.
Circuit material insertion loss is a concern for PA designers for a number of reasons, one of which is thermal management. A PA built on a PCB with higher insertion loss will generate more heat per applied amount of RF/microwave power than the same PA built on a PA with lower insertion loss. This heat that is generated contributes to a PCB’s maximum operating temperature (MOT) specification, and the PCB’s MOT should not be exceeded for any prolonged period of time. PA designers typically try to minimize insertion loss with consideration of applied power and frequency to ensure that the MOT will not be exceeded. A thinner circuit can have more insertion loss than a thicker circuit; however, it can also enjoy the benefits of shorter heat flow paths to the ground plane and an attached heat sink to offset the heat generated by the higher insertion loss. Additional electrical-thermal interactions which impact thermal management were outlined in a recent article.4
A circuit material’s TCDk is often overlooked by PA designers, and it is a parameter that greatly affects PA performance if not considered when selecting circuit materials. Quite simply, TCDk is the amount that the dielectric constant, εr, will change with a change in temperature. It is often apparent when a circuit has been operated in a controlled environment, in a laboratory, for example, which provides stable performance. When the PA is moved into the field, where the temperature may be changing and spans a much wider range, the performance of the circuit changes with the change in temperature. The result is a change in εr which can cause impedance shifts and performance variations. The TCDk parameter provides a way to compare this behavior for different circuit materials. Many PCB-based PAs will generate heat during normal operation and can create their own temperature change in addition to temperature changes from environmental effects.
Some low-loss polytetrafluoroethylene (PTFE) laminates used for microwave applications exhibit relatively high TCDk values, which can be a concern for applications designed for a wide range of operating temperatures. Ceramic-filled PTFE circuit materials tend to provide better TCDk properties and are better suited for applications that must handle wide operating temperature ranges.
PCB thermal conductivity is an important circuit material parameter for PA designers. Because of the circuit trace heating exhibited by higher-power PAs, they are often mounted on heat sinks. For PAs with high levels of trace heating, the thermal path to the heat sink is often through the circuit substrate material, and the heat flow path will be through the substrate. In such a case, the PCB thermal conductivity will provide a measure of the circuit material’s capability to transfer any heat generated in the circuit traces to the heat sink, with higher values of PCB thermal conductivity sought for PAs operating at higher power levels.
Table 2 shows typical values of circuit materials commonly used for PCB-based microwave PA applications. In general, TCDk numbers closer to zero in a mathematical absolute value sense are more ideal. In terms of thermal conductivity, higher values are better. As a rule of thumb, a good value of thermal conductivity for a PA circuit is 0.50 W/m/K or higher.
ATTACHING A HEAT SINK
Thermal management is a critical component for achieving high reliability in any microwave PA, and an important step in this process involves attaching the microwave PA PCB to a heat sink. Several options are available for attaching the PA PCB’s ground plane to a heat sink, including mechanical screw-down fasteners, sweat soldering and applying thermally and electrically conductive adhesive (TECA). Each approach offers its own set of capabilities and limits. TECA can provide a very uniform and consistent bond without voiding or air gaps. In the past, some TECA materials have had issues with delamination through lead-free solder reflow, or they would suffer changes in electrical properties at elevated temperatures. But the latest generation of TECA materials solved these problems and provided consistent and reliable performance, even at elevated temperatures.
To better understand the use of TECA and its impact on PCB material insertion loss, a simple experiment was performed: a 10 mil thick microstrip transmission-line circuit, TECA, and a thick copper heat sink was assembled in the configuration shown in Figure 3. The transmission-line circuit was initially evaluated for insertion loss using the differential-length method, testing the circuit as a microstrip transmission line without the addition of the TECA or heat sink. The TECA and heat sink were then attached through a lamination process, and the insertion loss was tested with the TECA and heat sink as part of the ground return path, since end launch connectors were used and the ground was being picked up from the bottom of the circuit (the heat sink in this case). Following these measurements, the circuit assembly was subjected to a lead-free solder reflow cycle two times and the insertion loss was tested again. As Figure 4 shows, there is no difference in insertion loss for the microstrip circuit with the addition of the TECA and heat sink, or with the completion of the lead-free solder reflow cycles. The TECA used in this study appears to perform after reflow cycles.
In general, some basic parameters can serve as guidelines when selecting circuit materials for microwave PA applications. A candidate PCB material for a microwave PA should have an εr tolerance that is ±1.5 percent or better, low dissipation factor, low insertion loss (possibly a PCB material with smooth copper), low TCDk and high thermal conductivity. In addition to these recommendations and the circuit fabrication guidelines of Table 1 for minimizing impedance variations, circuits with a conductor width tolerance of ±0.5 mil or better are recommended for microwave PA designs, as well as the same tolerance and value for copper plating thickness.
Selecting circuit materials for microwave PAs is never routine and it is always a good idea to have the circuit material supplier involved in the selection process, helping evaluate basic design and circuit fabrication considerations. A circuit material supplier’s experience with many different high-frequency circuits can lend a great deal of valuable guidance to a specific microwave PA design.
- E. Hammerstad and O. Jenson, “Accurate Models of Microstrip Computer Aided Design,” 1980 MTT-S International Microwave Symposium Digest, May 1980, pp. 407-409.
- John Coonrod, “Methods for Characterizing the Dielectric Constant of Microwave PCB Laminates,” Microwave Journal, May 2011.
- J.W. Reynolds, P.A. LaFrance, J.C. Rautio and A.F. Horn III, “Effect of Conductor Profile on the Insertion Loss, Propagation Constant and Dispersion in Thin High Frequency Transmission Lines,” DesignCon 2010.
- John Coonrod, “The Impact of Electrical and Thermal Interactions on Microwave PCB Performance,” Microwave Journal, February 2014.