RF Power: GaN Moves In for the Kill

Is gallium nitride (GaN) the wide bandgap material that will turn RF power generation on its head and relegate gallium arsenide (GaAs) and LDMOS (Laterally Diffused Metal Oxide Semiconductor) to the annals of history? Well you'd certainly think so, based on articles in the trade press, symposium papers, analyst reports, and corporate brochures. After all, GaN has at least ten times the power density per millimeter of transistor gate periphery, higher operating voltages (reducing impedance transformation issues), higher efficiency, and the ability to combine high RF power output over broad bandwidths at high frequencies. So is GaN, whether on substrates made of silicon, silicon carbide (SiC) or even diamond a slam dunk in every application?

Well no, or at least not right now. But first, it's important to review the different flavors of GaN: on silicon, silicon carbide (SiC), or diamond substrates:

GaN on silicon: This approach yields lower performance than either of the alternatives but has the potential advantage of using the world's low-cost, large-wafer silicon CMOS and power silicon foundries. As a result, it will soon be price-competitive with existing silicon and GaAs technologies, consequently threatening them in their entrenched markets.

GaN on SiC: The "high-end" version of GaN for RF, GaN on SiC will deliver GaN's highest power levels and other performance metrics that will ensure its place in the most demanding applications.

GaN on diamond: It's not easy to combine these two materials, but the benefits are enormous: Industrial diamond has the highest thermal conductivity (and thus is best able to remove heat) of any material on Earth. Replacing SiC, silicon, or other substrate materials with diamond allows the thermal conductivity benefits of diamond to be realized very close to the active area of the device. GaN on diamond is the focus of DARPA's Near Junction Thermal Transport (NJTT) program, beginning in 2011, in which TriQuint and partners University of Bristol, Group4 Labs, and Lockheed Martin are participants.

The team announced in 2013 that they had achieved an improvement of three-times the RF power density of GaN-on-SiC. This would allow a GaN-on-diamond device to be either three times smaller or achieve three times the RF power density. The program has achieved other program benchmarks as well, and it is possible and perhaps likely that GaN-on-diamond will meet requirements for manufacturability in 5 years.

Where GaN, GaAs, and LDMOS Will Coexist

Here's where GaAs and LDMOS technologies will continue to play a role in the foreseeable future:

Wireless infrastructure, industrial, and some radar applications: LDMOS is a fully-mature technology that is firmly entrenched in these markets thanks to its ability to deliver very high per-device RF power levels (greater than 1 kW CW). LDMOS can withstand almost infinite impedance mismatches without damage and employ advanced low low-thermal-resistance plastic packaging while remaining low in cost. Its limitations are a maximum usable frequency of less than 4 GHz and optimum performance only over narrow bandwidths. LDMOS is still a potential choice in radars that have the space to accommodate pallet-based, multi-stage amplifiers (rather than MMICs) and operate over narrow frequency ranges.

Low-power, battery operated devices: Smartphones, tablets, and virtually every product of this type owes its existence to GaAs MMICs and discrete devices. GaAs well serves both their receive and transmit chains and benefits from 30 years of development, a huge choice of suppliers and range of devices, low cost, and a small form factor.

Small cells, distributed antenna systems, and some microwave links: All of the advantages of GaAs MMICs apply in these markets as well, as their RF lower levels are comparatively low. Its typical GaN competition is represented by the TriQuint Semiconductor T2G4005528-FS (Figure 1) GaN-on-SiC HEMT (High Electron Mobility Transistor) that operates from DC to 3.5 GHz and delivers RF output power at 3-dB gain compression (P3dB) of 64 W at 3.3 GHz.

Figure 1:TriQuint Semiconductor's broadband T2G4005528-FS packaged GaN-on-SiC RF power transistor.

Some Military Radios Operating at HF through UHF Frequencies: These systems will remain viable candidates for LDMOS although as GaN-on-silicon devices cover much broader bandwidths, can deliver competitive CW RF power outputs, gain, efficiency, and linearity, they will become even more appealing as their cost declines.

There are many other applications, such as cable distribution amplifiers, with similar requirements in which the benefits of GaAs and LDMOS will enable them to play a substantial role as well. In short, GaAs and LDMOS technologies are not going away.

Where GaN Wins

Here's where GaN does best:

Active Electronically-Steered Array (AESA) radar and Electronic Warfare (EW) systems: These are key applications in which GaN-on-SiC (or perhaps diamond) transistors and MMICs are poised to become the de facto standard for many, many years. No other current or near-future technology can deliver the power density and other advantages of GaN-on-SiC.

This is graphically depicted by the GaAs and GaN-on-SiC Ka-band MMIC RF power amplifiers shown in Figure 2, both fabricated by TriQuint. Each one delivers 6 W of power at 30 GHz. However, GaN requires far fewer active devices to achieve it so the MMIC requires only a simple four-way power combiner. The GaAs MMIC amplifier requires many more devices and is more complicated, as it must incorporate a 32-way combining network, and plays a major role in the resulting size of the MMIC. The GaAs MMIC is about the size of the top of a pencil eraser while the GaN amplifier is about the size of a gain of uncooked rice.

Figure 2:The GaAs MMIC is about the size of the top of a pencil eraser while the GaN amplifier is about the size of a gain of uncooked rice. Source: http://www.triquint.com/special/gan-for-dummies

Obviously neither of these devices is large in general terms, but when considered in their most likely application--AESA-based radar--GaN's advantage is enormous. In an AESA radar that might have 70,000 elements, each one being served by a MMIC-based Transmit/Receive module, the benefit in size reduction versus GaAs MMICs is obvious. Combine this with GaN's ability to produce much high RF power outputs at this and higher frequencies and GaN MMICs are certain to replace their GaAs counterparts in future AESA-based radars and EW systems.

High-Power, Broadband Systems Operating Above 4 GHz: There are no other technologies except GaN that deliver the performance that these systems require. From Very Small Aperture Terminals (VSATs) for satellite communications to microwave links at higher frequencies, GaN is or will be the obvious (and only) choice.

Some Low-noise Amplifiers (LNAs): While GaN and GaAs have comparable noise performance, GaN can handle signals with greater amplitude with degradation or failure. GaN won't soon replace GaAs, silicon germanium (SiGe), or any other technologies in these in LNAs. However when high signal levels must be handled GaN has a unique advantage.

High-Power RF Switches and Other Control Components: The high breakdown voltage and current handling ability of GaN makes it better suited than GaAs-based MMICs switches. They can operate over wide bandwidths with high efficiency as well.

They have about the same low insertion loss and high isolation of PIN diode switches but handle higher power levels and draw less current. A good example is TriQuint's TGS2354 GaN-on-SiC SPDT reflective switch die (Figure 3) that covers 500 MHz to 6 GHz, handles 40 W of RF power, has switching speed of less than 50 ns, loss of 0.8 dB or less, and isolation greater than 25 dB.

Figure 3:TriQuint's TGS2354 GaN-on-SiC switch die fits in neatly with the requirements of high-power applications.

A Very Bright Future

If GaN's development in the RF world is separated into chapters; after completing initial development in Chapter 1, we have just finished Chapter 2. Thus far, a commercial market has been established, device reliability and manufacturing have confirmed, wafer sizes have reached 6 inches, and the material's potential has been confidently demonstrated by many companies. All this has occurred since the early 2000s, which is as impressive an achievement as the development of GaAs MMICs beginning in the 1980s.

In the ensuing chapters, GaN will begin to achieve more of its potential. Thermal management, a major factor in the technology's progress is being addressed by the use of diamond both as a substrate and as a heat spreader material (in aluminum-diamond matrix composites), advances in heat sinks through the use of materials with higher thermal conductivity, and other techniques. These and other approaches will increase achievable power density, which today in practical terms is below 10 W/mm of transistor gate periphery (GaAs offers no more than 1.5 W/mm) but has been demonstrated in a very simple device to be as high as 50 W/mm. While this will take many years to achieve, getting even halfway there would offer truly astonishing performance.

Figure 4:Cree's large family of GaN-on-SiC HEMTs offer greater power density and wider bandwidths compared to Si and GaAs transistors. The series ranges in bandwidths starting at 10MHz up through 18GHz.

Like GaAs before it, GaN will be essential for defense systems, primarily but not limited to AESA radar and EW, to meet next-generation requirements as there are several very large programs whose future more or less depends on it. Consequently, GaN MMICs will proliferate in the merchant market and defense prime contractors will begin to deploy them. GaN's future in commercial applications such as wireless infrastructure is definitely bright but further off as its acceptance depends to a large extent on cost reduction.

In short, GaN is only now beginning to fulfill its promise and its last chapter will not be written for decades. The whole saga should make for a very good read, however, as GaN moves in for the kill and GaAs and LDMOS become history.