The major components of a typical military communication system include: antenna set, antenna switching unit, a receiver, transmitter, a control subsystem, and a remote display interface.  One of the key modules within the transmitter subsystem is the power amplifier module.  This article discusses the challenges of designing a more efficient power amplifier and also highlights techniques for selecting and integrating these amplifiers into military systems.  To balance tradeoffs for efficiency and functional performance, the following factors will be examined:  PA packaging, operating class, stages of the PA, discrete component packaging, gain, thermal management, biasing configurations, and RF input matching techniques.

PA Packaging Requirements
In order to transmit encoded information over long distances to locate survivors, the search and rescue systems are required to transmit at high power levels.  The output power level for an amplifier used in this application needs to be several watts and needs to accommodate various modulation formats such as amplitude modulation, phase shift keying, and on-off keying modulation.  PAs of this type typically have issues with efficiency and are designed to be reliable when operated under specified conditions in harsh environments.  In a PA, many of the radio frequency (RF) drive components will reach very high junction temperatures during the normal course of operation.  Any packaging approach will involve a series of methods which address the mitigation of thermal issues such as thermal runaway.  Some of the key PA parameters to consider when optimizing for efficiency are:  operating class, discrete PA stages, topology, thermal design, transistor packaging, and gain, biasing schemes, and input matching circuitry.

Design Decisions
The first design decision is that of operating class.  For low power levels (less than about 100mW), class C becomes difficult to implement.  It is difficult to maintain good linearity with class B.  Consequently, for their design, engineers at the center selected a class A linear device based on gallium nitride (GaN) technology (Figure 4-1).  The next choice is whether to design your own amplifier or buy an off-the-shelf  module.   The choice is dependent on the eventual production quantities of the project.  If the quantities are small, then the use of a purchased module is probably the best choice, since the increased cost per amplifier will be more than offset by the avoidance of the development costs of doing a discrete design.  For large quantities, a discrete design should be considered and its overall cost compared with the cost of a module.    For high power amplifiers that also require high gain, it is worth considering the use of a PA module as a driver for discrete output stages.

Discrete PA Stages
With a purchased module, much of the design process will have been done for you (though harmonic filters may still need to be added).  One of the first items to decide when designing an RF power amplifier is the choice of single ended or push pull architecture.  A push pull design will have the advantage of a lower second harmonic output level and a higher output capability.  The reduced second harmonic level makes broadband amplifiers simpler since each harmonic filter can be made to cover a wider passband.  The single-ended design has the advantage of fewer components, and therefore, is cheaper and requires less board space.  Once the choice of architecture has been made, the next thing to consider is the load impedance presented to the transistors. 

Power Transistor Packaging
There are many varieties of power transistor packages and new ones are continually being developed. Care should be taken when selecting a TO39 device, as some transistors have the case connected to the collector. This can make construction more difficult since any heat sink used must be electrically isolated from the case.  The ceramic studless package relies partly on the ground plane to conduct away heat via the emitter leads.  For this reason, the emitter leads should connect directly to large areas of copper.  In larger sizes, there are flange-mounted or stud mounted devices (stud mounted devices also overlap with the TO39 transistor).  Devices of the highest dissipation rating are the isolated flange type or one that is used as the ground connection.   When using a PC board with a metal plate backing that doubles as a heat sink and ground plane, the latter is the better choice.  Otherwise, the choice is dependent on mechanical arrangements.  The isolated flange type is preferred in situations where the heat sink is not connected to the ground plane in close proximity to the RF power transistor.  If designing a push pull stage, the dual transistor package is preferable since stray inductance between two devices is reduced.  It also has the advantage that matched pairs are kept together. 

PA Package Types:

•            SO8
•            SOT223
•            TO39
•            Pill (studless package)
•            TO-220
•            8-32-UNC-2A
•            Turnstile Package (stud mount)
•            Turnstile package Flange mount
•            Flange Mount

Gain
The gain specified by manufacturers in their datasheets is the
data measured in the test circuit.  If operating the device in a different class with a different load impedance, or with feedback or extra damping not included in the manufacturer’s circuit, then one can expect the gain to differ.  If the device is characterized for class C operation but is being operated in class B, then the gain will be higher by 2 dB.  Changing to class A operation will yield higher gain.  The choice of load impedance affects gain and efficiency.  It is possible to trade off gain in order to obtain higher efficiency.

Thermal Design
Thermal design is a very important part of PA design and selection. The main source of heat will probably be the power transistors.  To calculate the dissipation of a PA transistor, the simplest approach is to calculate the difference between the power input and the power output.  The power input is simply:

DC collector / emitter voltage x DC collector current = input drive power. 

The power output is simply the RF power delivered to the output load.  The maximum allowable transistor junction temperature and the thermal resistance from the junction to case are usually given in the manufacturer data sheet.  Usually, the manufacturer will quote a maximum dissipation and supply a derating curve.  If this is the case, the maximum junction temperature can be taken as the point on the derating graph where the allowable dissipation is zero.   The thermal resistance can be taken from the slope of the derating graph.  The circuit, in a practical situation ,will probably be more complex with other heat sources summing ( e.g. more than one transistor bolted to the heat sink)  and extra resistances used for mounting brackets.  Contact resistance can also play a significant part.  To minimize this, mating surfaces should be as flat as possible with a very thin layer of heatsink compound.  With this information you will be able to calculate the maximum junction temperature achieved in the device for a particular heat sink.  It is not a good practice to operate the device continually at its maximum temperature, as this will greatly reduce the reliability. 

Biasing
Metal Oxide Semiconductor Field Effect Transistors (MOSFET) are generally easier to bias in PAs than bipolar transistors since they are less susceptible to thermal runaway and do not draw current from their bias circuits.  The disadvantage is that MOSFETs have a very wide tolerance on their gate threshold voltage.  This means that either the circuit must be set up for each device fitted, or some form of active bias control circuit must be used.  This can be adjusted to whatever bias current is required.  The gated threshold voltage changes with temperature; this may be compensated for by adding a thermistor.  Active bias circuits need no alignment to compensate variation in the gate threshold voltage.   This is a good solution for a class A stage, which needs a constant current bias.  Although the circuit is more complex, the cost of the extra components may be offset by reduced alignment costs.  This circuit may also be used  in  a variable class mode, if the set device current is less than that required for class A operation.

Input Matching
In general, the input impedance of a bipolar PA transistor is comprised of a resistive component and a reactive component.  At lower frequencies, the reactive component is capacitive, while at higher frequencies it is inductive.  The crossover point is in the mid-VHF band, when the resistive component is reduced as the power of the stage increases.  At VHF and above, particularly for higher power devices, impedance matching circuits are installed internal to the transistor package.  These do not usually match directly to 50 ohms but they do raise the very low input impedance of the transistor to an impedance that is much easier to match.  The internal matching shunt capacitor has the advantage over the external circuits in that one end is directly attached to the same grounding point as the transistor chip. 

There are many factors to consider when integrating RF packaging technology and maximizing efficiency of electronic RF devices.  By applying proper design criteria to the requirements, it is possible to maximize efficiency of RF power amplifiers by using COTS modules for currently deployed military systems.  For further information regarding packaging of efficient RF devices, please contact the EMPF Helpline at (610) 362-1320.