>[For a copy of this article in PDF format, which displays figures and equations, click

here

. Requires Adobe Acrobat Reader,

free download.

]

For many years, solid state power amplifiers (SSPA) and traveling wave tube amplifiers (TWTA) have been in competition in communication services. This is particularly so in satellite communication systems in the 6 GHz and 14 GHz satellite uplink frequency bands.

TWTAs have always had the advantage of being able to produce a large amount of microwave energy in a single electron tube. A TWTA is capable of producing power levels in excess of 1 kW from a single tube, up to approximately 500 W/cm². Additionally, the tube is capable of operating at anode temperatures of 800° C. Contrast this to the typical gallium arsenide (GaAs) microwave transistor, which has power densities up to 35 W/cm² and must not exceed a channel temperature of 175° C.

Currently, the highest RF output power achievable from microwave transistors is 60 W at 6 GHz and 20 W at 14 GHz. Therefore, to produce communication amplifiers capable of output power levels greater than 100 W, many transistors must be combined. Furthermore, microwave device DC to RF efficiencies range from 30 to 40%. With such low efficiencies and such a large number of devices, large power losses to heat are inevitable. Not only is the DC to RF efficiency a concern, but the RF combining efficiency is also critical to the success of a solid state power amplifier.

Much research has been done to develop novel, high efficiency microwave power combiners. One general result of all high efficiency combiners however, is the requirement that the microwave transistors be mounted in very close proximity to one another. This results in a very complex, two-fold, heat-spreading problem. The first problem in spreading the heat from the microwave transistor involves transferring the heat from the microwave transistor chip to its mounting flange. The gallium arsenide (GaAs) chip must be mounted to a thermally stable metal, such as kovar. The kovar carrier is then attached to a copper spreader, which is usually the mounting flange of the transistor.

With the extremely small size of the microwave transistor die, this results in a relatively inefficient heat spreader. This problem is the domain of the transistor manufacturer, and there is little the amplifier designer can do about this. The heat spreading resistance, or thermal impedance from the channel to the flange of the microwave device is fairly large. Channel-to-flange thermal resistance is typically 0.6° C/W for a 60 W, 6 GHz device and 1.2° C/W for a 20 W, 14 GHz device. This high thermal impedance of the transistor, along with the high heat density created by the large number of devices, creates a very difficult thermal management problem. The successful realization of high power solid state power amplifiers requires the unique combination of effective microwave combining techniques along with clever thermal design.

Historically, SSPA designers have used brute force techniques to handle the thermal management problem. This involves very large heatsink extrusions and heavy heat spreading plates. The pressure drop and fin length of these heatsinks typically require large fans to produce the required volumetric airflow. This results in solid-state amplifiers that are usually three to four times larger and four to five times heavier than the TWTA equivalent. Even though it is accepted that solid-state amplifiers are much more reliable than TWTAs, the size and weight differential is a difficult hurdle for the SSPA.

Furthermore, TWTA reliability has improved in recent years, making it more difficult to sell SSPAs based on the reliability advantage alone. The reliability advantage of the SSPA can only be realized by careful observation of the maximum device operating temperatures. If the transistor is operated above its maximum channel temperature of 175° C then the reliability decreases. This places the SSPA designer in a very difficult position when trying to reduce the size of high-power amplifiers. In the past, the increased size and weight of the SSPA have completely eliminated their use in mobile and aeronautical applications. It is somewhat less critical in base station applications, but along with the additional size and weight usually come higher initial cost. This higher cost has been yet another dilemma that has been associated with the SSPA.

In an effort to reduce the size and weight of high-power amplifiers, designers have looked toward using higher density finned heatsinks. While these designs achieve high fin density they are typically fabricated by epoxy-glue attachment to a heat spreader plate. This results in a heatsink that is unable to cope with the high heat density encountered in the SSPA.

However, there has been recognition of this problem. A high density heatsink has been developed that uses a fin swaging process to dramatically increase the efficiency of high-density finned heatsinks.

The heatsinks used in this SSPA application were bonded using a metal displacement process referred to as “Swaging.” This process (see figure 1) can be described as a cold forming process, which is used in the fabrication of high fin density heatsinks. Currently, this process involves the placement of a fin assembly with a tapered base into a slotted base plate, then applying rolling pressure on the opposite sides of each fin. This results in vertical and lateral pressure on the base unit material, which tends to push the fin toward the bottom of the groove in the base. This secure connection provides very good thermal contact between the fins and base and also prevents air and moisture from entering the grooves, thereby preventing corrosion and allowing the heatsink to be anodized.

The heatsink base plate area, fin height and fin center-to-center distance are shown in figure 2. The aluminum hollow-fin heatsink was swaged with an average fin center-to-center distance of 3.43 mm. The individual hollow-fin is extruded with a wall thickness of 1 mm and an overall average thickness of 3.8 mm. The hollow-fin is extruded with a tapered thick foot (see figure 1). The thicker fin base helps to secure the connection between the fins and the baseplate and results in good thermal contact through the swaging process. The extrusion process used to produce the aluminum hollow-fins is flexible enough to allow for different fin body and base geometries (see figure 2).

A 400 W, 6 GHz device is of extreme interest for satellite earth station transmitters. Because of the superior distortion characteristics a 400 W SSPA is approximately equivalent to a 1 kW TWTA.

To achieve low distortion from a TWTA, it must be operated well below its maximum output power capability. The SSPA, however, can be operated much closer to its maximum rated output power, resulting in a more efficient amplification system. Even with this efficiency improvement, the SSPA still requires 1.630 kW of DC input power to produce 400 W of microwave power at 6 GHz. The amplifier designer cannot assume that the amplifier will always be operating at maximum output power as this is under control of the earth station engineer. Therefore the heatsink must be designed to handle the entire 1.630 kW dissipation.

The size of the heatsink for such an amplifier is 355 mm wide by 381 mm long. The heatsink must accept two 200 W amplifier modules (see figure 3). The output power from the modules is combined using a wave-guide combiner to achieve the 400 W power level. A parallel set of fans is used in a push-pull arrangement to develop the system airflow.

Starting out using one particular manufacturers fan in each of the four positions resulted in a measured fin velocity of 5 m/s (1,000 LFM). The design goal is to achieve less than 75° C flange temperature on the RF output transistors.

RF designers require quick and accurate heatsink solutions. There are a number of such tools available, both from dedicated CAD developers and on the Web. Such tools generally use thermal modeling based on a set of analytical models for conduction heat transfer in the solid elements, coupled with natural and forced convection heat transfer models in the cooling airflow.

The conduction heat transfer model in the baseplate of the heatsink is based on the steady state solution of the Laplace equation for general rectangular geometry. The analysis is based on a general three-dimensional Fourier series solution, which satisfies the conduction equation in the base plate. For the forced convection air-cooled fins, an analytical model is used to predict the average heat transfer rate. The model used is a composite solution based on the limiting cases of fully developed and developing flow between parallel plates. With analytically-based tools, the solution is usually available within a few seconds, a very short time compared to the several hours required for a full CFD simulation.

The use of analytically-based design tools allows the user to perform the thermal design of the heatsink concurrent with the optimization of the electrical and manufacturing elements prior to any prototype or testing. This approach results in the reduction of design time and better reliability in the finished product.

Figure 4 shows the temperature map on the baseplate of the heatsink. The temperature shown in the figure is the maximum temperature on the heat sink baseplate under each individual microwave transistor. The tools can provide hydraulic parameters for the heat sink performance, such as the pressure drop and Reynolds number. The pressure drop can be used to determine the appropriate fan, which can deliver this volumetric flow rate for the system.

Temperature of the device's channel junction can be calculated, as well. This can be achieved by providing interface thermal resistance R_sc and channel-to-case device thermal resistance R_cc (junction-to-case thermal resistance R_jc). These temperatures are based on the average temperature under the device. The average temperature under the hottest four devices is shown in Table 1.

The software tool's results showed that the hollow-fin heatsink would be capable of dissipating the heat out of the transistors. The case temperature under the hottest devices on the module is less than 75° C and the channel temperature is well below the maximum specified temperature of 175° C.

Table 2 shows a close agreement between the analysis tool's results and measured case temperature.

Once the basic heatsink configuration has been determined using software analysis tools, the detailed heatsink design can proceed. The detailed heatsink design is implemented with thermal design software. Such software is commercially available (a Microsoft Windows version was used for this design).

This particular design software is quite effective in heatsink designs that include a stack of interface materials or plates. The plate stack-up is a typical problem encountered in most RF amplifier designs. It is usually impractical to have the microwave transistors mounted directly to the heatsink. In each 200 W amplifier module, the transistors are mounted to an aluminum housing. The amplifier housing is then mounted to the heatsink. In each case there is a thermal interface material, which must be taken into account (see figure 5).

The completed thermal model is shown in figure 6. The figure shows 16 transistors modeled as heat sources on the amplifier housing floors. The software creates an electrically equivalent network of nodes and resistors throughout the plates. It uses the classic thermal network method of calculating the heat transfer throughout the plates and across the interfaces between plates. The program automatically calculates the resistor and node values based on the plate dimensions and material properties chosen.

The fin linear air velocity is then entered into the program along with the ambient temperature. The program quickly calculates the steady state temperatures throughout the heatsink assembly along with the channel temperature of the transistors. Such software lends itself well to performing what-if analysis.

Plate dimensions, heat source positions, and fin dimensions can be readily changed to determine the optimum heat transfer. Figure 7 shows the computed heat transfer results for one of the amplifier modules. The temperatures directly above the heat source node are the transistor's simulated case temperature and channel temperature respectively. The heatsink is colorized to show the thermal contours across the plate assembly. The actual measured transistor flange temperatures for the hottest four devices are shown in the inset. The difference between the measured flange temperatures and the simulated flange temperatures is no greater than 3° C.

The program is able to achieve very good correlation with the measured results. This is particularly impressive considering the interface stack up and the high heat density created by the close proximity of the RF output transistors. In this case, the design goal of 75° C maximum device flange temperature has been achieved.

While the software tools used in this article are either free or low cost, commercial tools may be needed to work through more involved design projects. With the use of modern software, excellent thermal engineering can be achieved. Reliable thermal design tools can prove invaluable to the design engineer. Costly and time-consuming trial and error techniques can now be replaced with thorough design.

Good thermal design, along with an innovative new heatsink technology has been combined to produce an exciting new series of SSPAs. These devices can now be manufactured with tremendous size and weight reductions. This allows SSPAs to be used in installations that were previously the domain of TWTAs.

To fully appreciate the size and weight reductions, consider the comparison shown in figure 8. The amplifier on the left is the new SSPA using the hollow-fin heatsink technology. The amplifier on the right is its predecessor, which used combinations of extruded heatsinks. Both amplifiers are in standard EIA racks.

The old style amplifier is 425 mm high and weighs 79 kg. The new amplifier is 178 mm high and weighs 38 kg. This represents a 60% reduction in rack height and a 50% reduction in weight. Along with this size and weight reduction comes a comparable decrease in manufacturing cost.

For the first time, SSPA technology is truly comparable in size, weight, and cost to the TWT.

Dr. Ahmed M. Zaghlol received his B.S. and M.S. in Mechanical Engineering from Alexandria University, in Egypt. After receiving his Ph.D. in Engineering Science from the University of Western Ontario, he joined the Microelectronic Heat Transfer Laboratory (MHTL) at the University of Waterloo as postdoctoral Fellow. He joined R-Theta as Applications Engineering Manager in 1999. He manages R&D efforts for new products, development and implementation of R-Tools, on-Line thermal modeling of heatsinks software. A registered Professional Engineer in Ontario (PEO), Dr. Zaghlol is a member of ASME, IEEE and IMAPS. He can be reached at azaghlol@r-theta.com

Stephen Turner received a B.S.E.E. degree from the University of Pittsburgh and a Master of Engineering degree from Penn State University. He has over 20 years of experience in amplifier design from 2 MHz to 20 GHz. Stephen is a member of IEEE MTTS and AMSAT. He is presently the VP of Engineering at Paradise Datacom LLC and can be reached at sturner@paradisedata.com.

Contact the authors for particulars about the software or material used for the heat sink design (R-Tools and Sauna).