HRL pushes high-frequency envelope
Although there is strong demand for high powers and fast data rates, today s RF engineers are restricted by a lack of devices with operational frequencies beyond the low-gigahertz range at useful power levels. Fortunately, however, the situation is set to change, thanks to the significant amount of research that is focusing on a very promising class of device, GaN transistors.
Probably the most publicized and well funded effort developing this device is the US Defense Advanced Research Projects Agency program, entitled Wide Bandgap Semiconductors for RF Applications, which has focused on frequencies up to 45 GHz. But GaN can also serve far higher frequencies than this. Our development of this technology at HRL Laboratories of Malibu, California, has demonstrated that GaN devices can deliver a power density of 2 W/mm at 80 GHz. In fact, we have fabricated a range of GaN transistors operating in the 50–100 GHz range with similar power densities, which we detailed at last December s International Electron Devices meeting in San Francisco.
The key breakthrough resulting from our GaN development is an eight-fold hike in output power density compared with GaAs and InP technologies. This tremendous increase in output power will have far-reaching effects in the satellite and high-data-rate communications industries. It will also have effects in other applications, such as non-lethal weaponry, a technique that uses electromagnetic radiation at particular frequencies to disable opponents by increasing their skin temperature.
To produce these improvements in GaN HFET s frequency performance required aggressive scaling of the device s dimensions. This included scaling of the gate with electron beam lithography, which led to the production of a 100 nm T-shaped structure with a very low resistance. Another key feature of our transistors is the use of a patented double hetero-structure that is formed during the MBE growth of GaN on SiC. This heterostructure is critical for reducing short-channel effects that can degrade transconductance. A heavily doped cap layer is also used in our design to reduce parasitic contact resistance to 0.2 Ω/mm, which required the development of a recess gate process.
Scaling is not restricted to the device. It also includes the substrate s thickness and the ground interconnect dimensions. SiC substrates are thinned to 50 µm to avoid unwanted modes in the microstrip transmission line of the MMICs. Vertical vias measuring 30 µm are also cut through the substrate to reduce the ground interconnect inductance to less than 10 pH, which is a suitably low value for high-power devices operating at millimeter-wave frequencies.
These modifications were used to produce all of our HFETs with high cut-off frequencies, including devices with an fT of 90 GHz, 6 dB of power gain at 90 GHz and 2 W/mm power. Such devices operated with a high drain voltage of 15 V.
One factor that hampers the output power density of every III-V high-frequency FET is an inherent restriction on the gate width. If the gate is too large, propagation delays prevent the signals from combining in phase so the gate width must be less than a certain value, which decreases as the operating frequency increases. NGST holds the current output-power record for W-band MMICs and its InP HEMTs can produce 0.42 W at a gate width of 1.6 mm. However, we estimate that a switch to our T-gate GaN HFETs with a similar gate width could ramp MMIC output to 3 W.
This eight-fold hike in power density could cause thermal issues. However, the channel temperature in the GaN devices is actually similar to that of their InP or GaAs counterparts, thanks to SiC s far higher thermal conductivity. In addition, GaN is capable of operating at far higher temperatures, which makes the thermal management for this class of device relatively simple.
Our GaN HFETs are also highly reliable. At this year s CS Mantech conference in Austin, Texas, we showed results that demonstrated the robustness of these transistors under tough power compression. Testing under compression generates valuable data because it replicates the typical operating conditions encountered by commercial devices. The 2000 h test at 2 dB power compression and temperatures of 285, 315 and 345 °C produced a value for the activation energy of 1.8 eV, which implies a mean time to failure of more than 100 years at a baseplate temperature of 80 °C. This impressive reliability results from careful encapsulation of the transistor by SiN passivation and demonstrates that this technology is ready for product development.
High-frequency applications
Our GaN HFETs could be used to improve the performance of many different types of product. This includes millimeter-wave solid-state power amplifiers (SSPAs) for military and civilian communications. These amplifiers currently couple several InP MMICs to produce less than 1 W units typically costing $15,000. But the introduction of GaN could slash the component count and overall cost. The high price of the GaAs SSPA is not just caused by the cost of the InP MMICs themselves but also results from addressing combiner losses, the high assembly costs and other manufacturing issues. These include the need to hand-tune the SSPA to ensure that the phases and amplitudes of all of the arms of the waveguide combiner are perfectly matched over a broad range of temperatures, a requirement that ensures an efficient overall output.
Using GaAs and InP MMICs also limits the output power to 1 W. It is impractical to build a 3 W SSPA because this would require a 16-way waveguide and a discrete component count of nearly 200, including almost 50 MMICs. Switching to GaN can cut the number of components to five (a single GaN chip, two electronic probes and two bias capacitors), thanks to the hike in power density. Although there is no getting away from the fact that the price of GaN devices is more expensive, partly because of the higher substrate costs, this dramatic reduction could cut the price for an SSPA to a few thousand dollars, even if the cost of a GaN chip is 10 times that of a state-of-the-art GaAs PHEMT. In addition, the GaN SSPA could deliver a further advantage – greater temperature tolerance – because there would be no temperature-induced phase mismatch between the power-combining arms.
GaN SSPAs could also provide an alternative to traveling-wave tube amplifiers (TWTAs). These devices provide amplification in the W-band (75–110 GHz) but are only available at power levels of 100 W or more. TWTAs have several other weaknesses: they are fragile, very power hungry, take a long time to warm up and can cost over $100,000.
GaAs MMICs are not likely to offer an alternative to this technology because of the lack of power associated with these chips. However, GaN HFETs could be combined into an SSPA to offer an alternative type of amplifier delivering 10 W or more in a relatively small package. The potential for miniaturization also creates new opportunities for millimeter-wave radar, such as miniature satellite transponders.
GaN MMICs are attracting significant interest in the commercial E-band communication market for providing high-speed wireless data links. These links operate at 71–76 and 81–86 GHz, two 5 GHz bands that have been opened up by the Federal Communications Commission. Communications providers, such as Bridgewave, GigaBeam, Loea, E-band Communications and Sophia Wireless, are all seeking business in this market and could benefit from deploying 0.5–2.0 W MMICs in their point-to-point data links. Although relatively low transmission powers are needed for these local links during good weather, heavy rain is highly absorbing at these frequencies and can break up the data link. This is obviously a major downside for businesses using this service, as well as a potential barrier to increasing the number of customers using this wireless connection, but the high transmission powers of GaN could lead to more reliable connections.
GaN could also become the technology of choice in satellite communication, missile systems and fighter jets. These applications demand low weight and high power, which equates to very high efficiencies and minimal heat sinks. GaN MMICs are competitive in terms of efficiency and have relatively small heat-sinking requirements, so we plan to get our devices space-qualified.
There has also been interest in GaN MMICs for active denial systems – non-lethal weapons that transmit electromagnetic waves at 95 GHz. This radiation excites water molecules in the outer layer of the skin and causes a burning sensation. The radiation can be directed at targets up to 0.5 km away and is viewed as a technology that can stop potential terrorists without injuring innocent civilians.
Raytheon has already built a TWTA-based Silent Guardian protection system that has been mounted on a humvee. But the military has been increasingly asking for portable, cheaper and lower-power versions, which could be used as an alternative to electric shock weapons. The potential market for police and prisons is very large, but the current cost for millimeter-wave power is too high. GaN brings this cost down and can even make active denial systems the "killer application" for millimeter-wave devices.
The various applications that we have highlighted illustrate that there are many sectors within the millimeter-wave market that could benefit from the introduction of amplifiers based on GaN HFETs. However, this is not the only potential role for this device, as it could also be used in RF switches. These are an essential component of many systems, such as filters and transmitter–receiver modules.
All types of solid-state switch are cheaper, faster and more reliable than their mechanical equivalents. However, they are limited by their voltage swing before they distort signals or are damaged, and by the RF parasitics that attenuate the signals. GaAs PHEMTs currently have a significant market in switches because they have low-RF parasitics, but their power is limited by their breakdown voltage. A 0.1 µm gate GaAs PHEMT typically has a breakdown voltage of less than 8 V, which is less than one-fifth of an equivalent GaN FET.
Today, GaAs MMICs enjoy tremendous production volumes in commercial devices, such as cellular phones, while their GaN counterparts that can offer greater performance are still in their infancy. Fortunately, specialized foundries, such as ours and Northrop Grumman s, can supply GaN MMICs to the nascent market, which will, we hope, lay the foundations to widespread deployment of this high-performance technology.
