The race to 5G appears to have picked up speed, particularly in the United States where major telecommunications companies such as AT&T and Verizon have made announcements to launch 5G services by the end of 2018. Advanced LTE (LTE-A) is already expanding rapidly with upgrades to current base stations (BS). Field trials for LTE-Pro (aka 4.5G) are in full swing with download speeds already reaching 1 gigabit per second (Gbps). Fixed wireless access (FWA) technology has already been through significant field trials as well, displaying early successful utilizations of the millimeter-wave (mmWave) spectrum.
The stringent requirements for 5G not only require densification on the macro scale with more base stations, but also densification of power on the device level. Yole Développement predicts GaN penetrating two markets significantly with a CAGR of 20% in the coming decades: defense and wireless telecommunications. While many other compound semiconductors and process technologies will play a significant role in the evolution of 5G technologies, it is evident GaN will play a key role for high performance wireless solutions with its power/efficiency levels and high frequency performance.
Advanced Modulation Scheme Considerations
As cellular technology evolves, the modulation schemes that are utilized are often defined by non-constant envelopes with a high peak-to-average power ratio (PAPR)—the ratio of the peak power to the average power of a signal. As shown in Figure 1, the PAPR has increased dramatically from approximately 2:1 for 3G (W-CDMA) to 7:1 for 4G (LTE/OFDM). And, while advanced modulations schemes such as OFDM enable very high speeds with limited network resources, the increase in spectral efficiency comes at the cost of reduced energy efficiency of the power amplifiers (PA).
The high PAPR waveforms must be amplified linearly in order to avoid signal distortion. If the signal passes through a nonlinear PA, in-band distortion occurs which, in turn, increases the bit error rate (BER), and out-of-band radiation, causing adjacent channel interference . These high powered amplifiers therefore require a tradeoff between linearity and efficiency. For instance, keeping the amplifier at backoff, or reducing the input power in order to reduce the output power, for achieving alternate-channel power ratio (ACPR) requirements will reduce the output power and damage the PAE. But driving the device into compression, or the point at which the gain decreases as the input power increases, improves PAE but damages linearity.
Carrier Aggregation Considerations
Aside from the design constraints that are present with the increasing PAPR, there is also a need to operate over much wider bandwidths than traditional PAs. Mobile network operators (MNOs) have been faced with the need to achieve higher data rates but are limited severely to bandwidths no larger than 20 MHz. Carrier aggregation was implemented to massively increase the effective bandwidth in an operating region where spectrum was sparse. Carrier aggregation combines radio channels within the same frequency band (intraband) or over different frequency bands (interband) to increase wireless data rates and lower latency. LTE-A allows for a component carrier to have up to 20 MHz of bandwidth with a maximum of five that can be aggregated for up to 100 MHz of bandwidth. Previously, MNOs could get away with systems that covered a single 20 MHz band but a serious boost in network capacity is necessary to support the coming increase in mobile traffic. Current technologies now need to support up 20 times the bandwidth to handle these multi-band and multi-carrier systems.
The need to support these advanced modulation schemes is a multi-faceted issue so there are several known solutions that have been developed. Some include digital predistortion (DPD) for improved linearity as well as Doherty and envelope tracking (ET) techniques for a higher efficiency. Gallium nitride (GaN) high electron mobility transistors (HEMTs) have become strong candidates for BS PAs due to their inherent high breakdown voltage, high power density, large bandwidth and high efficiency. Johnson’s figure of merit (FoM)—a measure of semiconductor applicability for high frequency power transistor applications—for GaN devices is orders of magnitude higher than for Silicon (Si), Gallium Arsenide (GaAs), Silicon Carbide (SiC), and Indium Phosphide (InP).
GaN for Base Stations
According to ABI Research, the base station RF power device market was worth $1.1 billion in 2014 where GaN accounted for an 11 percent share while LDMOS share was 88 percent. This estimate increased to a 25 percent share by 2017 and is trending upward. As shown in Figure 2, cellular BSs account for the lion’s share of the RF power market at more than 50 percent. Some of the requirements for BS PAs for 5G may include operation from 3 to 6 GHz and from 24 GHz to 40 GHz with RF powers between 0.2W and 30W. It is likely that early 5G networks will leverage sub-6 GHz bands due to its favorable propagation properties.
As stated earlier, GaN-on-SiC HEMTs are a primary candidate for BS PAs as they achieve a higher PAE in a Doherty configuration at higher frequencies, over a larger bandwidth than Si LDMOS transistors. GaN HEMT technology can also be quite rugged, operating with severe load mismatches at high powers with minimal performance degradation. This inherently high operating voltage and output impedance is what allows for low loss, wideband matching with large output powers. Furthermore, the large Safe Area of Operation (SAO) mitigates any thermal or electric field breakdown due to power surges, thereby minimizing any maintenance necessary on BS equipment. The low noise figure performance of GaN MMICs in combination with its high power density actually make them a potential choice of substrate for PAs in the transmitter chain and low noise amplifiers (LNAs) for the receiver chain. Several existing GaN LNA implementations are able to achieve low noise requirements while being able to withstand high incident powers without taking damage .
GaN for mmWave
The mmWave spectrum is critical to the realization of 5G; the large amounts of available bandwidth are a strong option to support high data rate applications such as 4K/8K video streaming as well as augmented reality and virtual reality (AR/VR) applications. Small cells are the technology of choice for taking advantage of the available mmWave bands as they can be placed close together in urban environments for line-of-sight (LoS) links, mitigating the lossy propagation properties of high frequency signals. For practical purposes, these small cells must be readily mountable on a structure with high size, weight, and power constraints. The issue of size is somewhat resolved by the fact that transistor and antenna dimensions grow smaller at higher frequencies. Still, components with smaller dimensions generally have poor thermal management characteristics as larger surface areas are able to better spread dissipated power across the device. The SiC substrate has relatively high thermal conductivity (~120 W/mK) so heat can more readily be removed from the transistor to the heat sink. For small cell applications where cost is less of a concern, chemical vapor deposition (CVD) diamond exhibits thermal conductivities much greater than SiC at around 1800 W/mK.
GaN PAs are already being leveraged for the Ka-band transponders in cutting-edge satellite communications. Up-and-coming high throughput satellites (HTS) and low earth orbiting (LEO) small/medium sized satellites demand smaller form factor components for a high level of integration in an extremely power-constrained environment. This technology can be reproduced for 5G mm-Wave bands above 24 GHz. Current 0.2, 0.15 and 0.1 um GaN processes allow for cut-off frequencies well into the W-band with power densities on the order of 2 W/mm. The same qualities of high power density, wide bandwidth operation, good PAE and linearity, and low noise that a GaN PA exhibits at lower frequencies can also be seen at the mmWave frequencies. AlGaN/GaN heterostructures are particularly suited for high frequency performance due to large spontaneous and piezoelectric polarization effects that create an electron channel without requiring any modulation doping, unlike devices based on AlGaAs/GaAs .
GaN for Massive MIMO
While current BS technologies involve MIMO configurations with up to 8 antennas to steer a signal through simple beamforming algorithms, massive MIMO will likely leverage hundreds of antennas to enable the data rates and spectral efficiencies necessary for 5G. Power hungry actively electronically scanned arrays (AESA) utilized in massive MIMO require individual PAs to drive each antenna element—a significant size, weight, power density, and cost (SWaP-C) challenge. This would invariably involve RF PAs that are able to satisfy the power, linearity, thermal, and dimensional requirements of a 64-element and beyond MIMO array with minimal variation across each transmit/receive (T/R) module. Since GaN chips are taking leaps in power density and packaging every year, by the time Massive MIMO systems are commercially viable GaN would likely be a natural choice.
The GaN substrate has been utilized in military radar for decades now, but the confidentiality of such information has somewhat prevented its growth and maturity in the commercial realm—GaN is still more expensive to implement than GaAs-based ICs. Still, as Yole Développement and a number of other organizations predict, there is a shift in demand for this wide bandgap material that will essentially eliminate its exclusivity from the military and integrated device manufacturers (IDM) to independent foundries and design houses. Furthermore, the evolution in cellular communication introduces a very prominent niche for GaN to serve. This demand in the commercial realm will likely catapult the fabrication of GaN-based devices, ultimately driving down bulk prices.
Early implementations of commercial GaN PAs will likely fall, with cellular base stations leveraging advanced modulation schemes and techniques such as carrier aggregation. The outlook for GaN after this is still strong, with mmWave applications and especially with massive MIMO, as there is likely no other candidate technology that fulfills the power density requirements needed for massive AESAs.
1. S. Colangeli, A. Bentini, W. Ciccognani, E. Limiti and A. Nanni, “GaN-Based Robust Low-Noise Amplifiers,” in IEEE Transactions on Electron Devices, vol. 60, no. 10, pp. 3238-3248, Oct. 2013.
3. F. Sacconi, A. Di Carlo, P. Lugli and H. Morkoc, “Spontaneous and piezoelectric polarization effects on the output characteristics of AlGaN/GaN heterojunction modulation doped FETs,” in IEEE Transactions on Electron Devices, vol. 48, no. 3, pp. 450-457, Mar 2001.
4. M. Paredes, M. Garcia, “The Problem of Peak-to-Average Power Ratio in OFDM Systems”