X-band: GaAs vs GaN Efficiency Tradeoffs
by Grant Wilcox, Qorvo
As the war between GaN and GaAs rages on, GaN continues to win battles associated with increasing difficulty in SWaP-C (Size, Weight, Power, Cost) requirements. This can be seen across the frequency spectrum in both wideband and narrowband applications. As the power and frequencies increase, GaN really separates itself from the competing technologies. Using X-band as a benchmark, Qorvo is targeting superior performance in output power, gain and power-added efficiency while minimizing product footprint through our portfolio of power amplifiers for commercial and defense-related radar and communications systems, as well as electronic warfare. Figure 1 shows a range of supported power levels for these applications.

For GaAs, the benchmark for X-band power reached 25W with multiple stages of gain and power-added efficiency around 30 percent. Lower power options reached efficiencies of 40 percent. This performance was met by moving to a high-voltage pHEMT technology operating at a drain voltage up to 15V. Power densities associated with this technology reached slightly over 1W/mm. It was a nice improvement over the standard 0.25um pHEMT technology, but didn’t drive the change into the system design as was originally expected.
GaN brought that change, creating options that gave the system designer flexibility to achieve next generation performance, reduce form factors or some combination thereof. Initially, the focus was to drive down form factors while maintaining or increasing power performance. However, as X-band markets transition to GaN, users are noting the lack of efficiency improvements over previous GaAs solutions. That is not a shortcoming of the GaN technology but rather a result of the development focus of the product itself. By targeting high power with minimum form factor, the tradeoff was optimum efficiency. Table 1 compares benchmark Qorvo X-band power amplifiers using both GaAs and GaN technologies.
As you can see in Table 1, at a given power level, PAE was generally unchanged to marginally better with GaN. However, the size reduction was roughly 70 percent by going with the GaN solution. What allows this size reduction is a combination of higher power density and improved thermal management associated with GaN. This is possible due to GaN’s increased reliability at higher junction (Tj) temperatures. Whereas the GaAs Tj benchmark for 1E6 MTTF is 150C, GaN extends that beyond 200C. This adds flexibility and cost savings to a system design that is not achievable with older GaAs solutions.

There is also a push from the market to transition higher power amplifiers from large, flange style packages to surface mount technology (SMT). This is increasingly difficult with large GaAs amplifiers. As the chip size increases, the quality risk will increase due to CTE mismatch sensitivity, voiding and other mechanical robustness concerns. The chip area reduction made possible by GaN makes it more achievable to support high power in a smaller SMT package. The problem then moves to a thermal management issue at the system level as the higher power density within a smaller footprint will significantly increase the heat flux that the cooling system needs to manage.
Therefore, as the resulting efficiency remained somewhat unchanged at X-band, the system level thermal design becomes increasingly important in achieving the targeted performance. This has been met with various levels of success with air and liquid-cooled systems along with various proprietary techniques. As system design constraints are better understood, the focus of the component development will need to adjust to better meet the needs of the system.
PAE can be incrementally improved by reducing drain voltage or targeting an efficiency load with the amplifier design. Unfortunately, this undermines the power density benefit that drove GaN development in the first place. Going down this path will ultimately increase the chip size and further minimize the added benefits of GaN. Lowering the drain voltage might also reduce system efficiencies by increasing I2R losses. However, this may also be the best compromise. By reducing drain voltage, the amplifier designer can target a higher efficiency load. This, in turn, requires more FET periphery to meet a target power level, thereby increasing chip size. The increased chip size for a given power level will reduce the heat flux that the cooling system will need to manage. The downside would be added component size and cost. To better manage the system thermal load while meeting RF performance expectations, it is clear that amplifier development will need to find the right balance between output power, PAE and form factor. The design factors that will be in play will vary depending on the frequency and bandwidth of the application as well as the target output power.
And it’s not just at X-band. As system developers better understand their design constraints, component efficiencies are quickly becoming equally important to the power and gain specification in order meet system performance and cost targets across the frequency spectrum. High efficiency, GaN solutions have proven to support significant cost savings at the system level. Not only in smaller system designs but in less complicated cooling systems. This bodes well for applications with limited options in thermal management such as airborne systems. However, ground-based and shipborne systems will benefit equally with complexity and size reductions and even a reduction in fuel costs.
GaN is becoming the technology of choice across frequency and markets. Initially, higher output power and smaller form factors were the focus of GaN product development. However, the resulting thermal constraints at the system level are swinging the focus to achieve a better balance with efficiency. This will help reduce dissipated power and ease the thermal load at the system level. With GaN leading the way, the next generation systems are becoming a reality. As component and system developers continue to work together, the promise of GaN many years ago will be eclipsed by the reality of what can be achieved.
About the author:

Grant Wilcox has worked in the RF field for over 25 years with experience in analog modeling, MMIC design, account management and product marketing. Grant currently manages Qorvo’s portfolio of high power MMIC amplifiers supporting both commercial and defense related applications.
(3152)