Overcoming Wireless Infrastructure Capacity Challenges with GaN
by Sumit Tomar, General Manager of Wireless Infrastructure
Mobile data usage continues to soar worldwide, driven primarily by the rapid global adoption of smartphones and the rollout of LTE networks in China and other countries. Worldwide LTE subscriptions leapt 151% to 635 million in the 12 months ending in March 2015, according to the GSA mobile industry group. That growth is expected to continue, generating as many as 2.5 billion LTE subscribers by 2020.
Operators of mobile networks face the challenges of quickly expanding capacity to support this growth, while minimizing network disruption and cost. In the long term, 5G networks are expected to provide massive increases in capacity as well as much faster data rates. But, 5G specifications are still being defined and deployments are unlikely for at least five years. In addition, 5G may involve significant changes to network architecture due the very low network latency, ultra-wide bandwidth and extremely high data rates that need to be supported in 5G. That is at least 5 times more backhaul capacity than currently used today in 4G. This will force network operators to look for different front-haul and backhaul solutions as the current backhaul solutions used for 4G wouldn’t be able to scale up to support 5G.
To meet the immediate requirements for much higher capacity well before 5G arrives, operators are scrambling to expand the capacity of their 4G networks without re-architecting infrastructure. They’re focusing on technologies that enable them to extract more capacity from their existing LTE spectrum allocations and to minimize the need for costly purchases of additional spectrum.
Operators are focusing on several key capacity and performance upgrades. Short-term plans center on carrier aggregation, a feature of LTE Advanced. Medium-term enhancements include a collection of enhancements variously described as 4.5G, 4G plus, LTE evolution or pre-5G, including high-order (up to 64X) multi-user multiple-input, multiple-output (MU-MIMO) as well as higher-order modulation and use of unlicensed 5GHz spectrum.
These short- and medium-term capacity improvements, and ultimately 5G networks, will require a unique semiconductor process technology to manufacture base station power amplifiers (PAs) with a higher power density, more efficiency over wider bandwidth, and support higher carrier frequencies ranging from 400 MHz to 100 GHz.

The Promise of GaN on SiC
Historically, base station power amplifiers have primarily used silicon-based Laterally Diffused Metal Oxide semiconductor (LDMOS) technology. However, progressively more demanding requirements have exposed the device physics limitations of LDMOS and resulted in a growing number of BTS (Base Station) OEMs adopting Gallium Nitride (GaN on SiC) as technology of choice for next generation base station PA.
Carrier aggregation is driving the need for multi-band power amplifiers with 2X more power and 3X more video bandwidth compared to what is used in 4G Macro Cell BTS today. LDMOS has device physics limitations to support video bandwidth higher than 300 MHz as well as the ability to support wider video bandwidth of LDMOS PAs decreases even more sharply as the frequency increases beyond 2GHz. While LDMOS is effective only at frequencies up to about 3.5 GHz, GaN PAs already handle higher millimeter wave frequencies. In addition, GaN PAs support greater video bandwidth, even at higher frequencies.
The two leading GaN flavors available today are GaN on Silicon Carbide (SiC) and GaN on Silicon (Si). GaN on Si offers the advantage of a low-cost substrate that can be produced in silicon foundries, with associated economies of scale. But GaN on SiC supports much higher power density, and correspondingly higher power output. This is due to the fact that SiC provides better thermal conductivity: about three times higher than Si. Compared with LDMOS, GaN on SiC offers roughly 7X the power density, at about 5-12 W/mm. Because of this, GaN on SiC PAs can deliver about double the power output in the same footprint. As a result, GaN on SiC has become the technology of choice for high-power RF applications.
The benefits of GaN on SiC PAs directly address operators’ top three concerns, the three “Cs”: cost, coverage, and capacity.
GaN on SiC PA enables higher efficiency, reducing operators’ huge electricity cost. In order to enable higher capacity, operators use higher order QAM modulation. As the QAM order increases, the effective cell radius decreases. Higher peak power enabled by GaN on SiC overcomes the challenges of smaller cell radius and enables better cellular coverage. Higher break down voltage in GaN on SiC process enables wider video bandwidth required to enable higher network capacity.
To examine these benefits in more detail, this article will address the likely role of GaN on SiC at each stage of wireless network evolution, starting with carrier aggregation, followed by 4.5G, and finally 5G.
Near–Term: Carrier Aggregation
Operators are in the early stages of deploying carrier aggregation, a feature of LTE Advanced (3GPP Release 10). With carrier aggregation , operators can increase data capacity and throughput by combining up to 32 component carriers, each between 1.4 and 20 MHz, up to a maximum of 100 MHz total bandwidth.
A key attraction of carrier aggregation is that it lets operators make better use of fragmented spectrum allocations by combining component carriers from multiple bands (inter-band carrier aggregation). Many operators have less than 20 MHz of contiguous spectrum and they need carrier aggregation to support demand for faster data services. Initial deployments typically use carrier aggregation for downlink communications only and combine two 10 MHz component carriers for a total bandwidth of 20 MHz.
Carrier aggregation generally requires wideband PAs in order to avoid the additional cost and complexity of using a separate PA for each component carrier. Common carrier aggregation combinations, such as Band 1 (1800 MHz) with Band 3 (2100 MHz), require PAs with a bandwidth greater than 300 MHz. The ability of GaN PAs to support greater video bandwidth than LDMOS, even at higher frequencies, is a key benefit. The cost-per-die advantage of LDMOS is negated by the greater power efficiency of GaN and the fact that a single GaN PA can support bandwidth that would require multiple narrowband LDMOS PAs. Carrier aggregation also requires higher power output to enable simultaneous transmission on multiple component carriers. GaN on SiC PAs are capable of meeting today’s typical requirements for multi-band PAs delivering 100 W or more in average power output and supporting video bandwidth greater than 300 MHz.
The Impact of Power Efficiency on Operating Cost
The power efficiency of GaN also plays a major role in helping operators control their biggest cost: utility bills. PAs are responsible for a large proportion of the power consumed by base stations. If the PA operates at only 35% efficiency, as was typically the case with older LDMOS PAs, 65% of the energy is wasted as heat. The heat produced also causes reliability issues and requires large heat sinks, which increases the overall product size.
Medium–Term: “LTE-Advanced Pro”
Beyond carrier aggregation, operators are looking at a range of different technologies to improve capacity. These medium-term developments, collectively described as 4.5G or pre-5G, are expected to roll out in 2016 and beyond. 3GPP approved a new marker for LTE evolution system as “LTE-Advanced Pro” in October 2015.
Massive MIMO
MIMO enables operators to increase data rates and network capacity by transmitting multiple spatially separated data streams over the same frequency band, using multiple antennas on the base station and the user’s device. LTE–Advanced Pro defines up to 8×8 downlink MIMO, and up to 4X4 for uplink connections. It also defines Mu-MIMO, which expands capacity by enabling a base station to use each stream to communicate with a different device.
4.5G is expected to usher in much higher-order MIMO in 2.5 to 2.7 GHz, 3.4 to 3.8 GHz and 5 to 6 GHz bands to enable a further leap in network capacity, with base stations handling up to 256 simultaneous data streams. This introduces yet another set of challenges. The base station requires more power to drive 256 channels, so energy efficiency and heat dissipation become even bigger issues and the higher power efficiency of GaN becomes correspondingly more valuable.
The other major challenge with massive MIMO is managing complexity. Squeezing 256 transmit channels into a single base station will require highly integrated subsystems that package PAs, low-noise amplifiers (LNAs), switches, and filters into compact modules. To achieve the greatest performance and power efficiency, these subsystems must combine components based on different process technologies. For example, while GaN PAs provide the required power output and power efficiency, low-noise amplifiers (LNAs) based on CMOS may maximize receive sensitivity while minimizing noise. Advanced filters will be required to avoid interference with adjacent bands. Because base stations are typically mounted in locations where they are highly exposed to the elements, they experience extremes of temperature and humidity. Therefore, they will need BAW and SAW filters that exhibit a stable response to temperature variation. Highly integrated subsystems also offer base station manufacturers the benefits of reduced development and test time, because all the elements within the subsystem are already matched and tested as one.
LTE in 5GHz Spectrum (LTE-U)
LTE-U would allow global operators to boost coverage in their cellular networks, by using the unlicensed 5 GHz band already used by Wi-Fi devices. LTE-U would share space with Wi-Fi equipment already inhabiting that band. LTE-U is intended to let cellular operators boost data speeds over short distances, without requiring the user to log in to a separate Wi-Fi network.
This band is beyond the range of LDMOS PAs, which are limited to frequencies at or below 3.5 GHz. In contrast, 5 GHz is well within the range of GaN PAs, which already operate at mmWave frequencies.
Higher–Order Modulation
Moving to higher-order modulation enables further increases in data rates and network capacity.
3GPP Release 12 defines an increase in complexity from 64 QAM to 256 QAM, which provides peak data rates up to 33% faster by transmitting eight bits instead of six bits per OFDM symbol. However, using a more complex modulation scheme without increasing power output results in a reduction in cell range. To maintain coverage, operators will need higher-power PAs. This is expected to further fuel demand for GaN PAs that can provide the necessary power output.
Long–Term: 5G
5G specifications are still being defined and will not be complete for several years; the current 3GPP proposal is for submission of final specifications in 2020. However, it is widely expected that to achieve multi-gigabit data rates, 5G will utilize frequencies higher than 6 GHz, including use of millimeter-wave frequencies at 60 GHz – 90 GHz. Millimeter-wave spectrum is currently used for a number of military, satellite, and other applications. It is also used for the 802.11ad Wi-Fi standard, but is generally much less crowded than the lower-frequency bands currently used for LTE. Use of millimeter-wave spectrum for LTE will therefore make available a huge amount of additional bandwidth, while reducing concerns about congestion of lower-frequency spectrum.
Historically, the available devices in mmWave frequencies have been either too bulky or too expensive to be used in commercial grade applications. The key innovation in GaAs and GaN technologies has made it possible to transmit required power with a monolithic device. The process advancement in RF-CMOS and SiGe technologies has enabled highly integrated RF TRx up to E-band. The MCM (Multi-Chip Module) integration has enabled sophisticated antenna arrays in relatively small form factor and lower cost.
5G BTS would require massive MIMO, so size of components used in RF frontend will be very critical. GaN on SiC with significantly higher power density compared to GaAs or silicon based devices will make an ideal choice not only for RF power, but also for combining LNA, switch and PA monolithically in very small form factor.
Conclusion
Demand for GaN PAs is expected to increase rapidly over the next three to five years as operators push forward with new LTE capabilities to accommodate the unrelenting growth in mobile data usage. To meet this demand, a growing number of PA suppliers are expanding their portfolios to include GaN products. It is important to remember that PAs used in wireless base stations must meet high requirements for performance and efficiency, as well as high reliability under harsh conditions. Each generation of network capacity enhancements will require new levels of performance and power efficiency. Careful evaluation is required to ensure PA suppliers provide the performance, reliability, process maturity, and in-house manufacturing capacity to meet these exacting requirements.
(748)