Considerations in Power Amplifier Package Design
By Jerry L. Carter, Senior Applications Engineer, StratEdge, San Diego, CA

Understanding the Role of Compromises and Trade-offs in the Design Process
Electronic packages serve many purposes, including protecting the device, connecting the device and cooling the device. If any of these are done improperly, the chip is rendered inefficient at best or useless altogether. The focus of this article is the effect of thermal dissipation and expansion characteristics on the design of electronic packages. High frequency devices can suffer a loss of efficiency as temperatures at the chip level increase. Every degree of heat that can be dissipated, therefore, adds to the efficiency of the device, as well as to its useful life.

Several types of engineers are typically involved in the electronic package design cycle. Mechanical engineers, materials engineers, electrical engineers and thermal engineers contribute most often, but sometimes manufacturing or industrial engineers might also be included. So the compromises and trade-offs begin at the engineering level, but don’t end there. People from the accounting, quality, purchasing, assembly, sales and marketing departments may also have input.

Within this assortment of people that have such different levels of education, experience and expectations, it can be difficult to please everyone. The person in charge of the final package design, more often than not, feels pulled in a lot of different directions. But if everyone understands that there are many issues beyond their own that have to be considered, the team generally achieves a workable design that meets the customer’s major objectives.

Layers
A useful tool in considering electronic package designs is to consider the package as a series of stacked layers. Working with and thinking about packages in terms of layers can simplify thermal dissipation and expansion analysis. Viewing packages as layers is a useful way to understand 3D interactions in the package structure. The two areas we will touch on here are thermal expansion and thermal conductivity. The following are examples of packages showing the layered nature of the construction.

Package Construction Examples
The leaded power amplifier (LPA) package is an example of layered construction, as shown in Figure 1. It begins with a combination base and heat sink, topped by an alumina insulating layer, a metal leadframe and a lid.

The surface mount power (SMX) package, as shown in Figure 2, is very similar to the LPA package, but has vias to carry the signal from the bottom of the package to the top conductor pads. The leads are on the same layer as the base.

Applying Trade-offs in the Design Process
The LPA and SMX package examples are ideally suited to many power amplifier applications. They use a single heat sink layer of copper composite that has a high thermal conductivity and is also well matched to the temperature coefficient of expansion of most devices. However, suppose one needs to package a small amplifier of a few watts and the application requires a smaller footprint than is provided by the examples. There is another product style well suited to smaller footprints; the molded ceramic package. These packages have a molded ceramic preform between the base layer and the leadframe layer. Another molded ceramic preform layer attaches between the leadframe and the metal seal ring. As with the other packages, the final layer is a lid or cover. All these layers are placed into a carbon mold and run through a furnace where the temperature reaches 900º C. When everything fuses together, the package is removed from the mold, cleaned, and the metal components are plated with nickel and gold. The final steps in the manufacturing cycle include attachment of the IC and other circuit components and placement and sealing of the lid. This package style is shown in Figure 3.

The package in Figure 3 meets the space requirement, but uses an ASTM F-15 base. ASTM F-15 is an iron alloy that has a rather low thermal conductivity. The ability of this package to manage heat from the chip will not be comparable to the power amplifier packages shown in Figures 1 and 2, but a heat spreader can be added between the chip and the base to improve the thermal dissipation. A material with high thermal conductivity would be a good choice for the heat spreader. Copper comes to mind, but it has high thermal expansion relative to the chip and to the ASTM F-15 base of the package. Another choice would be a copper refractory metal composite, such as copper tungsten. This is the same material used for the base in the LPA and SMX packages.

Layers and Thermal Expansion
In Figure 4, we see an illustration of what happens when materials with different coefficients of thermal expansion (CTEs) are joined together with a solder or braze preform at high temperatures. The part on top has a low CTE, while the part on the bottom has a high CTE. As the parts heat up, both parts expand and the braze alloy melts. As they cool, the braze alloy becomes solid and the two parts are frozen together during the brazing process.

This is not a problem while the assembly is in the furnace at brazing temperature. The problem is what happens when the assembly returns to room temperature, illustrated in the bottom drawing in Figure 4. If the materials are fairly ductile, they will actually appear warped in the direction of the material with a higher CTE. If either material is brittle, it will crack or the parts might even separate.

The change in length of a piece of material is proportional to the change in temperature. If we add a constant (K), we get the following equation:

Expanding the delta terms and adding similar equations for two materials, A and B, we get the following equation:

We are interested in just the area where A and B are attached (see top view in Figure 4). This is the case where

The CTE (k) is usually expressed as parts per million per unit length. That is:

To state this in non-mathematical terms, the change in length (DL) of two materials is the initial length (L0), times the difference in constant (kA - kB), times the change in temperature (DT). Table 1 shows how this rule of thumb is applied to the example above. Consider a 3mm gallium arsenide (GaAs) die on a copper block. The objective is to see how much difference there will be between mounting the die on pure copper versus mounting it on 85/15 copper tungsten.

A temperature of 332ºC was chosen because these parts need to be joined with gold germanium (AuGe), which has a reflow temperature of 357ºC. When the parts come out of the furnace, they will cool to 25ºC, making the DT 332. Table 1 shows that the DL for pure copper will be 0.011, a change of 0.37%, and for 85/15 copper tungsten, the DL will be 0.0012, a change of only 0.040%.

The 0.37% error might sound unmanageable, but in fact, small GaAs die can sometimes be mounted onto copper. What this simple analysis excludes is the relative hardness and stiffness of the two materials. In practice, what appears to happen is that the soft copper will move enough to allow the two layers to cool down to room temperature with no damage to the structure.

There is a similar mismatch if the copper heat sink is attached to the ASTM F-15 base of the package, as shown in Figure 5. For the area where the heat sink attaches to the metal base, the stress on the base is transmitted to the molded ceramic walls and the structure may fail. Fabrication and testing of packages shows that a copper tungsten composite heat sink attached to the ASTM F-15 base will be a more robust construction due to its better matched CTE than copper.

Having found the pure copper insert to be impractical, the next question is, how does this smaller package with the layer of ASTM F-15 in the thermal flow path compare to the LPA packages above?

Layers and Thermal Conductivity
There are simple mathematical models that can be expressed as simple formulas to describe the flow of heat through uniform, homogeneous materials. Viewing the package as a stack of layers with the heat flowing from layer to layer allows one to estimate the thermal efficiency of the package with simple formulas that can be calculated on a spreadsheet. Even though it’s simple, the description of the equations takes some time to develop, so for this article, only the concept of heat flow through the layers is presented.

Consider the 3mm GaAs chip from the example above, mounted on a heat spreader that is attached to an ASTM F-15 alloy base in a molded ceramic package as shown in Figure 5. The cross section of the package is shown with the chip and heat spreader mounted into it.

This simple model assumes that a hot area in contact with a cooler plate will radiate out the heat at 45 degrees from the contact area. The heat sources are the transistors on the top of the GaAs die. Heat flows through the die and the layers below it. Even the material that attaches the layers can impact the heat flow. Figure 6 shows just the layers in the heat flow analysis with a vertical exaggeration to show the adhesive layers.

Table 2 shows the temperature drop in each layer. For this example, the difference in the heat flow path for the LPA versus the molded ceramic package with the copper tungsten insert is that the molded ceramic package has the extra layer of the ASTM F-15 base. When comparing the copper tungsten flanged LPA package to the molded ceramic package with copper tungsten insert, there is only a 1.34 degrees (1.29 + 0.05) difference in the junction temperatures. For low power applications, the molded ceramic package with the copper tungsten insert may be good enough.

Conclusion
Many factors need to be considered when designing the optimum power amplifier package, including thermal dissipation and expansion requirements. Usually, no single consideration can dominate, so the design decisions end up as compromises. One simple method to aid understanding of the issues in assembly and function of the package is to view the package as layers. This simple analysis is useful to understand both the CTE and heat flow. Two tools have been provided to analyze the impacts of expansion mismatch and thermal conductivity on the reliability and performance of the final assemblies. These simple tools can be used to estimate if material choices are practical and to what degree a material or design change will impact performance.

StratEdge
www.stratedge.com
TXTLINX.COM103
Email this article to a friend!