High power amplifiers (HPAs) are an important part of systems and are widely used. For instance, they are used in weather satellites, radar systems for air traffic control, satellite television broadcasts, satellite radio, military radar, wireless backhaul for internet traffic, mobile phones, mobile phone cell towers, and many more. In addition, the amplifiers must remain reliable for many years (decades, in fact). One of the most important determining factors for reliability is the operating temperature of the devices. To remain reliable, HPAs must be designed so that their operating temperatures (device junction temperature) remain below the required level for reliable operation. This is achieved, in part, by proper heat spreader implementation.
The heat spreader is physically connected to the heat generating devices in the HPAs so they can spread the heat generated in device junctions. The heat density generated in the FET junctions of gallium nitride (GaN) amplifiers can be as high as 3200 W/cm2, which more than two orders of magnitude higher heat density than the average clothing iron turned to its highest temperature. Managing this thermal load requires heat spreaders that are often large compared to the wavelength of operation. As a result, proper heat spreader design is critical to achieve reliable operation.
Prior work has shown electrical packaging can often dictate the performance of a product operating at microwave and millimeter-wave frequencies. For instance, in  it was shown that interconnects between the active devices must be optimized which often means that the connections must be short and small relative to the wavelength of operation. Prior work has also shown the benefits of GaN amplifiers to achieve high RF output power and some thermal management methods. In  it was shown that output power from GaN hybrid amplifiers can be many times higher than competing semiconductor technology such as gallium arsenide (GaAs). In  it was shown that 15 W of output power is possible with GaN even at Ka band. An example of microfluidic cooling methods for GaN was investigated in  which showed that optimum flow rates can significantly improve GaN output power. A more conventional approach to thermal management was taken in  which mounted the GaN amplifiers on copper tungsten (CuW) heat spreaders in a hybrid module for satellite communication systems. This prior work demonstrates the importance of the electrical packaging and thermal design of GaN amplifiers.
However, prior work does not address the challenges of simultaneous thermal design and electrical design of GaN packaging. Thermal and electrical requirements for HPA packaging and modules must be balanced to meet overall system performance goals. This work investigates this balance for a particular HPA package type. Section 2 describes package type being considered in this analysis and the GaN device being used. Section 3 shows that an optimum thickness of the heat spreader exists for some modules, which achieves the minimum device junction temperature. Section 4 describes the electrical interconnect performance as a function of heat spreader thickness and illustrates that the ground inductance increases as the heat spreader thickness increases. Section 5 describes the balanced design solution. Section 6 shows the conclusions of this investigation.
2. GaN Device and Package Type Chosen for Analysis
GaN field effect transistor (FET) devices and monolithic microwave integrated circuits (MMICs) are packaged using a wide variety of materials and physical configurations. For instance, a plastic package was used for a 100 W GaN power amplifier  as a method to achieve low cost. Liquid crystal polymer (LCP) packaging was used in  along with stud bumped GaN devices. The approaches result in acceptable junction temperatures for reliable operation for applications that function at low duty cycle or low RF output power levels.
Another approach to the packaging of GaN devices uses a high thermal conductivity metal base and direct die attach to it. This type of package is described in   and is illustrated in Figure 1.
GaN devices can be individual FET devices or they can be MMICs. The benefit of the MMIC is that it has all, or most, of the required matching and bias circuitry integrated into it. It also can have multiple stages of amplifiers to increase its signal gain. This means that it can generate high output power when driven by lower signal power levels. However, MMICs require more semiconductor area to fit all the required functions. Since semiconductor cost is roughly proportional to its area, MMICs tend to be higher cost than FET devices. Our analysis will focus on using a FET device with the following parameters.
・ FETX = dimension of the GaN FET devices in the x-direction = 0.8 mm.
・ FETY = dimension of the GaN FET devices in the x-direction = 1.5 mm.
・ Max Output Power = 25 W.
・ Material = GaN on SiC base.
・ θjc = junction to case (junction to bottom of die) thermal resistance (˚C/W) = 4.7 to 12.2.
Note that the thermal resistance, θjc, is provided by the fabricator of the GaN FET device.
3. Heat Spreader
A heat transfer analysis study was conducted to determine the optimum heat
Figure 1. Module HPA package used for thermal analysis and electromagnetic simulations.
spreader dimensions. For the purposes of the study, only conduction was considered as the heat transfer mechanism. Radiation and convection were ignored. The choice of dimensions immediately presents the designer who is seeking a balanced designed with several choices.
1) The x-axis dimension of the heat spreader must be chosen to optimize the length of the wire bond interconnect from the device to the package. The optimized size is normally the minimum that can be achieved for GaN FET die fit and die attach so that the wire bonds are as short as possible. Short wire bonds have less inductance than longer ones and they minimize the radiation which causes undesired stray coupling in the package. For optimum heat transfer, the x-dimension may need to be larger than can be accommodated by the wire bond requirement.
2) The y-axis dimension of the heat spreader is also chosen primarily for fit and assembly reasons. That said, a package that is too small in the y-direction can also have non-optimum heat transfer characteristics.
3) The z-axis dimension (thickness) of the package will be considered in detail in the next section. The goal is to determine the optimum thickness, t, of the heat spreader that will provide the lowest thermal resistance.
In practice, the x and y dimensions of the heat spreader are often set by standard package sizes from suppliers. This is especially true if a pre-packaged amplifier (from the manufacturer) is chosen.
4. Optimum Heat Spreader Thickness for Thermal Performance
The optimum heat spreader thickness was investigated using 3D finite element method (FEM) simulations. The cross section of the structure that was analyzed is illustrated in Figure 2. This configuration is important since most high power amplifier packages are mounted into a housing or onto a pallet. The optimum thickness of the heat spreader depends upon the material properties (and physical dimensions) of housing. The heat source is the FET devices with dimensions FETX = 0.8 mm and FETY = 1.5 mm. These dimensions correspond to the size of the GaN die shown in Figure 2.
The FEM simulation assumed a thickness for the housing base of 2.5 mm since this is in the range of typical values. Three housing materials were considered: 6061 aluminum (k = 170 W/mK), cold rolled steel (CRS) (k = 51.9 W/mK), and Kovar (k = 17.3 W/mK). Where k is the thermal conductivity of the material. The thickness of the heat spreader was varied from 0.1 mm to 10 mm
Figure 2. Simplified cross section of the material stack up used in the 3D finite element method (FEM) heat transfer analysis.
and the heat spreader material was taken to be CuW (80% W, 20% Cu, k = 206 W/mK). For the FEM simulations, the mesh control on the heat source was set to 0.01 mm. For the rest of the model, the mesh min element size of set to 0.01 mm and mesh max element size was set to 0.5 mm.
The temperature profiles from FEM simulations are shown in Figure 3. The figure is for the case of an aluminum 6061 base with a 1 mm CuW (20/80) heat spreader. Note that the temperature gradient illustrates a max temperature of 2.36˚C.
The results of the heat transfer analysis are shown in Figure 4. Note from the
Figure 3. Thermal profiles from FEM simulation for the case of an aluminum 6061 base and 1 mm CuW (20%/ 80%) heat spreader.
Figure 4. Heat transfer analysis results showing thermal resistance as a function of heat sink thickness and housing material type.
figure that the thermal resistance decreases as the heat sink thickness increases for thicknesses below approximately 1 - 2 mm. As thickness increases beyond the minimum thermal resistance point, the thermal resistance slowly increases.
5. Optimum Heat Spreader Thickness for Electrical Performance
When the HPA package is integrated into a module, at least one important electrical performance parameter is impacted by the choice of optimum heat spreader thickness. Namely, the ground inductance introduced by the signal path at the I/O tabs of the package. The inductive ground path is illustrated in Figure 5. It shows a cross section of the interconnect from the output of the package to module floor and transmission line in the printed circuit board (PCB) of the module. Note how the ground current must travel down the package heat spreader, through the housing floor, and back up to the transmission line ground of the PCB. This inductance is undesired since it creates an effective series inductance in the signal path. This creates an impedance mismatch and reduces the electrical bandwidth.
As can be seen in Figure 5, as the height of the heat spreader, t, is reduced, the current path is also reduced. The undesired inductance can impact port impedance match, gain, stability, and efficiency. For instance, in   it was shown that proper impedance loads at harmonics affects the efficiency of the high power amplifier. For these reasons, minimization of the ground inductance is an important design consideration. For electrical performance reasons, the ideal thickness of the heat spreader is zero since that will minimize the inductive loop. However, this is certainly not acceptable for thermal reasons. As a result, a balanced design must be achieved.
6. Optimum Design and Conclusions
There are several important conclusions from this work. First, it was shown that an optimum design is required to simultaneously achieve the required junction temperature of the GaN amplifier devices and minimize the ground inductance. Second, it was shown that interconnect ground inductance reduces the operating electrical bandwidth. Third, for the specific case and package type illustrated in
Figure 5. Illustration of the output signal ground current path which adds undesired series inductance.
this work, the optimum spreader thickness is between 1 - 2 mm. These lessons learned can be applied to other package types and at different operating frequencies.