A year or two ago, it was thought that the first 5G applications would be sub 6 GHz using technologies such as massive MIMO, high orders of carrier aggregation, heterogeneous networks, etc., to achieve better data rates, latency and capacity. But new semiconductor processes and architectures have rapidly progressed making 5G mmWave technology available today at reasonable costs. Fixed Wireless Access is being targeted as the first 5G type of deployment that will be implemented ahead of 5G systems. In the US, Verizon and AT&T are already doing field trials in several cities and Samsung /Arqiva recently announced the first trials in the UK. How is this possible?
First, low cost Si semiconductor processes have been released that are capable of achieving excellent performance at mmWave frequencies. An example is Global Foundries’ 45 nm RF SOI process released earlier this year, making them the first foundry to announce an advanced, 300 mm RF silicon solution to support mmWave beamforming applications. The technology is optimized for beam forming front-end modules, with back-end-of-line features including thick copper and dielectrics that enable improved RF performance for LNAs, switches and power amplifiers (ft of 305 GHz and fmax of 380 GHz). Another example is Xilinx is using 16 nm FinFET technology to integrate data converters and RF on the same chip as the FPGAs for a complete RF SoC.
Second, several semiconductor device companies have been developing new architectures for highly integrated SoC mmWave solutions using standard and advanced node Si technologies. While many demos at trade events over the past couple of years used SiBeam 60 GHz arrays or UCSD proto-type 28 GHz arrays, this year we started to see low cost, production ready arrays from companies like Anokiwave/Ball Aerospace, Intel, Samsung, ADI, and just announced this week, Movandi.
Earlier this year, Anokiwave/Ball Aerospace announced a 28 GHz phased array for 5G and SATCOM applications followed by a recent release of 39 GHz ICs by Anokiwave. These are highly integrated SoC phased arrays with advanced features such as temperature compensation, build-in self test, and software control. Similarly, Movandi just announced BeamX technology, a scalable RF front-end system solution for both 28 and 39 GHz applications. Movandi’s BeamX front-end integrates RF, antenna, beamforming, and control algorithms into a modular 5G millimeter wave solution targeted for CPE, small cell, and base station applications. They use standard bulk Si technology so is low cost and easily available in high volumes plus a high efficiency antenna technology to achieve its performance levels.
Xilinx jumped into the 5G wireless game by integrating RF-class analog technology into its 16 nm FinFET technology to create the first hardware and software programmable RF SoC. Based on an ARM-class processing subsystem merged with FPGA programmable logic, the all programmable SoC has 12-bit, 4 GSPS RF sampling analog-to-digital converters and 14-bit, 6.4 GSPS direct RF digital-to-analog converters, along with built-in digital down-conversion and up-conversion. RF SoCs enable a 50 to 75 percent reduction in system power and footprint for LTE-A Pro and 5G active antenna systems and massive MIMO radios so is aimed at higher power systems where power consumption is critical.
At IMS in June, Analog Devices demonstrated a complete 26 to 44 GHz solution that combines CMOS and SiGe for highly integrated radio products and has been working with several companies in this area. At the same event, OMMIC discussed their effort to develop a front-end MMIC for 28 GHz phased arrays. The IC integrates a PA, LNA and switch with the PA delivering 37.5 dBm output power with 35 percent power-added efficiency and 22 dB small-signal gain. OMMIC is using GaAs for the core chips that provide the amplitude and phase control for the array and think that arrays based purely on CMOS will not deliver the required EIRP.
Qorvo has the same opinion voiced by OMMIC and announced a GaN front-end module designed for a 39 GHz phased array. The FEM covers 37 to 40.5 GHz and contains two channels, with each channel comprising a multi-function GaN MMIC containing a three-stage PA, three-stage LNA and a SPDT T/R switch, fabricated with Qorvo’s 0.15-µm GaN on SiC process. The PA provides 23 dBm average output power. But the Silicon based semiconductor manufacturers typically find that the array is thermally limited and can achieve the required EIRP (~65 dB). This seems to be a bone of contention and once key metrics and measurements are determined, we should find out for sure. Both approaches seem to work and it could be a matter of performance and cost.
We don’t even know of other projects going on at Intel, Samsung, Huawei and others so it will be very interesting to see how this plays out over the next year. I expect mmWave systems to play a major role in initial 5G deployments. EJL Wireless Research seems to agree and forecasts 5G and fixed wireless access transceivers at 28 GHz to exceed 20 million units by 2021. Unlike 3G and 4G systems where modules and components could be put together to build a system, I think 5G will be different due to the radio complexities and demanding RF performance. I think those companies that supply an optimized solution as a complete array or sub-system will be the winners.