Parent Category: 2017 HFE
By Tim Galla
The rising presence of GaN has been hard to miss. It’s even been called “the technology that will displace all others,” and it not hard to see why: GaN offers the best combination of high power, gain, and efficiency while operating at high bias voltages that reduce current consumption. It also has the highest power density of any semiconductor material used for RF power generation and can operate over multi-octave bandwidths.
That said, LDMOS and GaAs, its more mature competitors, aren’t likely to be run over by the GaN freight train any time soon, nor will GaN be the de-facto choice for every RF power application. To see why, it’s important to illustrate the strengths of each one that in large measure determine the applications for which they are best suited.
LDMOS
LDMOS technology has advanced continuously in the decades since it was first offered commercially. Current LDMOS RF power transistors are the most rugged of the three technologies and some LDMOS RF power transistors can operate into an impedance mismatch greater than 65:1 (GaN and GaAs are limited to about 20:1 or less), when driven with twice their rated RF input power without degradation or damage. A single LDMOS transistor can generate a CW RF power output greater than 1500 W, a metric that is rising every year, efficiency and gain are high, and LDMOS devices operate at voltages up to 50 VDC. The transmitter shown in Figure 1 is a good example of what can be achieved with LDMOS.
Figure 1 • This 250-kW solid-state system based on RF power amplifiers using LDMOS transistors powers NASA’s new Doppler Radar Wind Profiler at Cape Canaveral’s launch facility. Source: Delta-Sigma, Inc.
As a result, LDMOS RF power transistors and RFICs are used almost exclusively in base station amplifiers as they have for many years, and increasingly in small-cell amplifiers as well. Other applications in which LDMOS has a commanding lead include land mobile radio, solid-state broadcast transmitters, and Magnetic Resonance Imaging and other medical systems. They are also used in avionics systems as well as L-and S-band radars whose narrow bandwidths preclude the need for GaN’s broadband performance characteristics. A typical example of an amplifier module using LDMOS transistors is shown in Figure 2.
Figure 2 • Pasternack’s PE15A5028F amplifier is conservatively rated at 25 W saturated power that operates from 150 kHz to 230 MHz. It has small signal gain of at least 43 dB +/-1.5 or better, IP3 of 48 dBm, and includes a regulated power supply and ample cooling supplied by a heat sink and fan.
As LDMOS combines very high RF output power with exceptional ruggedness, it has become viable for use in industrial systems that have always been powered by vacuum tubes. These applications all use RF energy for heating, and range from welding and drying to sealing of plastics, and dozens of other uses. The most recent emerging application for LDMOS is in two cooking applications: replacing magnetrons in microwave ovens and creating an entirely new type of appliance that can simultaneously cook multiple types of foods. The new “cooking appliances” (Figure 3) differ from a conventional microwave oven and benefit from solid-state power in many ways.
Figure 3 • Cooking appliances powered by LDMOS transistors may be the “next big thing” for LDMOS. Source: NXP Semiconductors.
Unlike a magnetron that is either on or off, the output power of an LDMOS transistor can be infinitely varied, delivering a more uniformly cooked result that better retains its nutrients and moisture. It also significantly improves the thawing process, which in a microwave oven produces uneven results with some sections of the food properly defrosted, others somewhat cooked, and others dried out.
While the need for LDMOS technology will continue for many years, once the cellular industry enters its fifth generation and uses frequencies above 3 GHz and eventually the millimeter-wave region, LDMOS will no longer be the only RF power technology used in base station amplifiers. Instead, above about 3 or perhaps 4 GHz, GaN will be the only choice in base stations requiring the same RF power levels currently provided by LDMOS, and in small-cell base stations as well.
GaAs: The Indispensable Workhorse
GaAs is one of the most versatile semiconductor materials, as its applications range from power amplifiers to mixers, switches, attenuators, modulators, and current limiters as well as solar cells, laser diodes, and LEDs. In fact, without GaAs technology, some applications taken for granted today would not have been achievable, or at least nowhere near as rapidly. For example, GaAs was responsible for modernizing active phased-array radars thanks to DARPA’s MIMIC program that from 1987 to 1995 spearheaded the development of GaAs MMICs through the efforts of 26 companies and an expenditure of more than $570 million, which would be more than $1 billion today.
GaAs also played a primary role in enabling the “wireless revolution” that perhaps not coincidentally emerged at about the same time as GaAs devices became widely available. GaAs is the primary power amplifier and LNA technology in mobile phones, covering every band currently used throughout the world, and unlike LDMOS used in base stations should retain its solid position as the cellular industry moves to higher frequencies.
GaAs devices are also used in the distribution amplifiers in cable systems as well as microwave point-to-point links up to about 30 GHz, and because GaAs is resistant to ionizing radiation, it is widely used in solar cells and RF power amplifiers in satellite transponders. It was once even considered for high-speed computing applications, but as clock rates increased so did power consumption, which ultimately made it impractical as adding cores to silicon die is far less expensive.
Although relatively high power amplifier modules can be built using GaAs devices, the technology remains susceptible to the onslaught from GaN, as the latter can produce higher RF power at the same frequencies as GaAs, reducing the number of devices and amplifier stages required to produce a given final power level. Nevertheless, GaAs has undergone decades of development with billions of devices in service, and the sheer number of applications it serves ensure its viability for a very long time.
GaN: What’s Next
As it is hopefully obvious from the previous discussion, LDMOS and especially GaAs are and will continue to be extremely useful technologies for RF power generation. However, there is no denying the awesome potential GaN has to offer that will propel it into applications currently served by its counterparts. Having already gained wide (and increasing) adoption in defense systems, it is now targeting cellular infrastructure, a trend that will grow dramatically as 5G networks start to be deployed.
The primary advantages of GaN can best be understood by starting at the molecular level. In a solid, electrons have energy that combines to form bands, the highest being the conduction band and the one below it the valence band. The region between the two is called the bandgap. If the electrons in the valence band are suitably excited they can move up to the conduction band.
The intensity at which this occurs is typically compared with that of silicon, which has a bandgap of 1.1 electron volts (eV), and gallium arsenide (1.4 eV). In comparison, GaN has a bandgap of 3.4 eV, silicon carbide (SiC) has a bandgap of 3 eV, and diamond comes in at 5.5 eV, so they fall into the category of “wide-bandgap” materials. Semiconductors with wide bandgaps can withstand higher breakdown fields, which contribute directly to its GaN’s power density, which is at least 10 times that of GaAs. In Figure 4, the breakdown voltages of GaN, SiC, and silicon are compared.
Figure 4 • Breakdown voltage comparison of semiconductor materials. Source: Embedded.com
In addition, GaN devices can operate at higher temperatures and higher voltages than either GaAs or silicon so they can deliver higher power. GaN can operate at five times higher voltage and twice the current of GaAs devices, have higher potential power-added efficiency than GaAs above 10 GHz and silicon above 1 GHz.
Most literature, articles, and other sources reference GaN-on SiC and GaN-on-Si substrates as these are by far the most common combinations, although the performance that can be achieved by the two is very different. The thermal conductivity of silicon carbide (SiC) is three times higher than that of silicon (and six times higher than that of GaAs), which makes it possible for smaller die to produce a given RF output power.
So, it’s not surprising that GaN devices achieve their best performance when they’re mounted a substrate material such as SiC that also have wide-bandgap characteristics. GaN devices on SiC substrates are more expensive but their performance can mitigate this for applications in which the highest performance is required. In contrast, GaN-on-Si devices have lower performance but also lower cost as they can be fabricated using silicon processes, so they are currently relegated to less demanding applications. Although these two combinations gained the most attention, copper and copper-molybdenum-copper (copper-moly-copper) are also used as substrate materials by some GaN device manufacturers. Both have high thermal conductivity and can deliver better performance than GaN-on-Si at less cost than GaN-on-SiC.
Figure 5 • The PE15A5033F from Pasternack delivers saturated power of 100 W in Class A or AB operation from 700 to 2700 MHz, with efficiency of 30%, at least 45 dB of gain, and a switching time of 5 µs or less.
As noted earlier, one of GaN’s greatest strengths is its very high power density, which allows it to generate very high RF power levels with a much smaller gate periphery than silicon or GaAs. The power density of a GaN device is typically five times higher than a GaAs device, which translates into the ability to produce five times higher power than GaAs in a die that is 80% smaller. Theoretically, GaN’s power density can be at least 20 W/mm (some have claimed even higher values), but practically speaking, producing higher power in a smaller area presents a significant thermal management problem.
That is, achievable power density depends on how quickly and effectively heat can be dissipated from the die, which is why GaN device manufacturers can typically deliver higher power density than system builders can accommodate. That said, major strides have been made in thermal management in the last few years in removing heat at the die, outward through the substrate, and beyond to heat spreaders, heat sinks, and possibly an external cooling subsystem.
Which brings us to the rarefied realm of GaN on industrial diamond, which while only infrequently mentioned in connection with GaN, is the most promising material for heat dissipation by far, but also the most difficult to make manufacturable. Diamond has the highest thermal conductivity of any material on Earth, so it can insulate very high voltages with only very thin layers of the material, which is shown in Table 1. Diamond has been studied for decades initially with minimal success, but very recently promising results are being obtained by the few companies that have begun to make it manufacturable. Although classified, it appears likely that some next-generation radars are already being designed using GaN-on-diamond substrates.
Table 1 • Thermal conductivity of diamond versus other materials.
Diamond’s potential benefits are enormous and although it is more expensive than the alternatives, this single factor is not the only one that determines its true cost. For example, the higher power density and heat dissipation achievable by GaN-on-diamond allow a specific RF output power to be obtained using fewer amplification stages and thus fewer devices. This is an immense advantage in a radar or EW system that uses the Active Electronically-Steered Array (AESA) architecture, the modern-day definition of the active phased array.
These systems deliver RF power at each antenna element rather than from a single source (such as a vacuum tube), and there are typically hundreds or thousands of these elements in an array. The higher the RF power each amplifier produces, the higher the system’s overall effective radiated power. If this can be achieved using fewer devices as it can with GaN MMICs, the system cost, complexity, bill of materials, and cooling overhead can be dramatically reduced. While GaAs essentially made modern active phased arrays possible in the late 1980s, GaN’s characteristics extend this much further to achieve much higher power levels.
Another diamond-based approach uses aluminum-diamond metal matrix composites (MMCs) as a heat spreader material, which can achieve thermal conductivity higher than its competitors, and is manufacturable at reasonable cost. Consequently, after investigating the various choices beyond GaN on SiC, some manufacturers of GaN devices, RF power amplifiers, and complete systems, are turning to aluminum diamond MMCs as a way to squeeze higher performance from GaN at lower cost. Figure 6 shows aluminum diamond heat spreaders used for dissipating heat from GaN devices.
Figure 6 • Aluminum diamond metal matrix composites are increasingly used as heat spreaders for GaN devices. Source: Nano Materials International Corp.
Summary
In short, GaN in all its forms has the greatest long-term growth potential of any semiconductor technology used for generating RF power, and it is just now beginning to be realized. Unlike LDMOS (but not GaAs), which is essentially a narrowband technology, GaN devices can operate over very wide bandwidths, which has significant benefits for inherently broadband applications such as electronic warfare and to a lesser extent communications systems.
And as many types of radars have bandwidths greater than 1 GHz (and rising) it is well suited for use in the next generations of these systems and upgrades to existing ones as well. With all of its inherent benefits, it would be reasonable to assume that GaN and especially GaN-on-SiC would eliminate the need to consider any other device technologies. However, both LDMOS and GaAs have distinct advantages in a growing number of applications that will ensure their place in the RF power domain.
About the Author
Tim Galla serves as Product Manager for Active RF Components at Pasternack Enterprises.
Comments
Post a Comment