Can CMOS catch up to GaAs in PA applications?
I’m sometimes asked why ANADIGICS focuses so heavily on gallium arsenide (GaAs) and related technologies.
I’m sometimes asked why ANADIGICS focuses so heavily on gallium arsenide (GaAs) and related technologies. Isn’t CMOS riding a wave of performance enhancements and market dominance that will drown other technologies? True, CMOS is the dominant general-purpose semiconductor technology, but it certainly is not the only or even the best choice today or tomorrow for a variety of consumer applications. A familiar example where GaAs dominates is in the power amplifier (PA) for wireless networking and handsets. CMOS proponents believe that advances in CMOS technology and its derivatives (such as silicon germanium) will catch up to GaAs and eventually replace it in most of the applications where GaAs is prominent today. I believe that GaAs still has a lot of gas left and will continue to be the preferred solution for years to come. A brief look at the state of the two technologies shows why I’m bullish on GaAs and why the threat of CMOS doesn’t keep me awake at night.
On the plus side for CMOS are the high levels of integration possible along with the economies of scale-and the potential to integrate the PA into transceivers to achieve single-chip radios. However, practical reasons make PA integration difficult. First, CMOS PAs use a specialized CMOS process tailored for RF applications; the highly integrated baseband and transceiver chips need the most advanced, smallest gate-length CMOS process for integrating functions in the smallest possible die. Also, package requirements for the two types of ICs are quite different. Specialized processes for PAs make integration with the standard CMOS process fairly daunting and a significant technical obstacle to a single-die transceiver/PA. Integration also presents packaging problems. While the low-temperature cofired ceramic carrier (LTCC) is a good choice for a flipped CMOS PA die because of its low coefficient of thermal expansion, LTCC doesn’t make sense for larger dies.
At ANADIGICS, we believe integration should occur in the RF section, bringing together the PA, voltage regulation, antenna switch, low-noise amplifier, and other components onto a single die to create a complete RF front end. This approach simplifies handset/WLAN designs and allows the overall design to be nicely partitioned for maximum performance.
While CMOS amplifiers have steadily increased in frequency and power, they have more difficulty satisfying the stringent performance requirements of each successive generation of wireless applications. Even in applications where both CMOS and InGaP amplifiers are available, the InGaP devices tend to offer lower noise, better linearity at higher power levels, and lower jitter.
GaAs substrates with InGaP circuits hold a performance advantage over CMOS, with high-frequency operation upward of 300 GHz. InGaP offers excellent linearity for WLAN and wireless handsets, setting the standard against which other high-frequency technologies are compared. Today, GaAs PAs achieve the highest power levels and highest power-added efficiencies for use in WLAN and handset applications.
In CDMA wireless applications, where linearity is a critical concern, we haven’t even seen CMOS trying to compete. InGaP-based products far surpass CMOS in high-frequency linearity. We are seeing some CMOS activity in GSM applications, where efficiency and harmonic performance are important. One of the leading CMOS PAs reportedly offers power-added efficiency of 48% for the GSM900 and 40% for the DCS1800 bands. InGaP amplifiers from ANADIGICS, however, offer performance of 55% and 53% for these respective bands. The 13 percentage point increase in the DCS band means much more efficient handset operation for the GaAs amplifiers. At high power levels, a low PAE translates into battery-draining performance. For newer wireless standards (EDGE, WCDMA), GaAs is the only solution being offered. CMOS just doesn’t have the capability to meet the performance required from this class of PAs.
While GSM applications require typical second- and third-order harmonic performance of around -22 to -25 dBm, CMOS PA vendors only publish maximum specs, leading one to suspect embarrassing typical numbers.
InGaP has several physical characteristics that make it attractive in high-frequency applications. Its electron mobility is six times higher than that of CMOS, which means better high-frequency performance. The GaAs substrate is semi-insulating, which permits better signal isolation and high-performance passive elements to be fabricated on the die. With CMOS, it is difficult to build the microwave circuit elements, such as high-Q inductors and lossless transmission lines due to the relatively high conductivity of the substrate. Overcoming these difficulties often requires process changes that raise the cost of CMOS devices and partially offset the main reason for using CMOS in the first place.
Silicon germanium (SiGe) has been touted as an evolutionary step for CMOS and a viable alternative to GaAs but has not yet found the proper combination of low-cost, high-yield production and performance to meet specific application requirements. SiGe has generated more press than products because it has not yet proven to have the robustness that OEMs are looking for in building reliable products. While GaAs and related technologies have a reputation for being brittle materials of suspect reliability, they pass the durability stress test and are field-proven for long-term reliability.
SiGe is finding some application in 802.11b/g PAs, but I don’t believe it offers any advantages over InGaP. And it represents a more complex circuit design. InGaP can offer built-in 50-Ω impedance matching. The need for external matching is eliminated, along with the costs of the required components. An InGaP PA needs fewer than half of the external components of a SiGe PA. These are mostly standard filter capacitors.
InGaP also holds an edge in ruggedness. One way to measure ruggedness is by the breakdown voltage at a given frequency. As shown in the following table, SiGe has a lower breakdown voltage at all frequencies. InGaP maintains a nice margin in breakdown voltage to allow it to withstand the operational stresses of WLAN, CDMA, and GSM applications.
While CMOS will clearly capture some of the low-end PA applications (Bluetooth or ZigBee), InGaP remains the technology of choice for high-frequency applications requiring moderately high power outputs. CMOS is definitely evolving enough to encroach on territory previously held by InGaP but still is playing catch-up as applications requirements evolve. Three recent trends point to InGaP’s continuing leadership.
First, higher levels of integration allow additional functions (such as voltage regulation) to be fabricated on-die to simplify the design, reduce the bill of materials, and save board space. Many InGaP PAs are offered in multichip packages that also include CMOS control circuitry. For WLANs, complete front-end modules-incorporating the PA, low-noise amplifier on the receive end, and RF switches-are available in a compact package.
Second, while industrialized countries have a relatively dense deployment of cell towers that allow handsets to operate at low to moderate power levels, emerging nations do not have the same densities. As a result of wider tower spacing, handsets will be required to operate at higher power levels a higher percentage of the time. InGaP-based PAs maintain efficiency at elevated output power levels while CMOS has yet to prove it can extend battery life as well as InGaP PAs under the same circumstances.
The third trend is the recent introduction of commercially viable BiFET InGaP technology that allows bipolar and field-effect transistors on the same InGaP die. For example, newer PAs for WCDMA applications combine heterojunction bipolar transistors (HBTs) and pseudomorphic high-electron-mobility transistors (pHEMTs) in a single monolithic package. This allows the advantages of each device to be applied while minimizing the disadvantages. For example, HBTs are used for amplifier stages for excellent linear operation and high efficiency; pHEMTs create efficient switches because of their fast, efficient switching and low noise. The low loss and very low power requirements make pHEMTs ideal for switches. Switches based on HBTs cannot match the performance possible with pHEMT-based switches.
BiFET technology can be advantageously applied in a wide range of power amplification applications. ANADIGICS, for example, has introduced PAs based on its patented InGaP-Plus technology, which combines HBT and pHEMT structures on the same die. These PAs offer significant performance and integration enhancements. Using this technology, the company recently introduced a new line of PAs that reduce average power consumption by up to 75% over standard PAs. This can translate into 25% longer battery life under typical operating conditions.
Among other areas we’re using InGaP-Plus technology for GSM and WLAN front-end ICs. For GSM, where ruggedness and harmonic performance are paramount, InGaP-Plus technology increases ruggedness by allowing us to create limiting circuits, while InGaP offers better harmonic performance. In WLANs, a complete front end is built using HBTs for the power amplifier in the transmit circuit and a pHEMT for the low-noise amplifier on the receive side. The transmit/receive switch is also pHEMT.
The bottom line is simple: Although CMOS has made huge technical advances and is a clear winner in baseband applications, I don’t see it matching, let alone exceeding, the performance of GaAs/InGaP for power amplifier applications. And GaAs/InGaP itself has demonstrated the ability to evolve and meet the increasingly stringent requirements of each generation of wireless applications.
Aditya Gupta, PhD, is vice president of technology development at ANADIGICS Inc. (www.anadigics.com).
