Note that the configuration of beams and streams is not set based on hardware. The OEM can choose to change the configuration in software, assuming that the antenna elements are equipped with analog phase shifter and variable gain components that can be individually controlled. In almost all prototypes, this “hybrid beamforming” approach is used today, as full digital beamforming at very wide bandwidths can be costly in terms of processing power and dollar cost.
Currently, SOI and SiGe semiconductor technologies are used in many base stations in order to achieve high levels of integration and low-cost. GaN also holds great potential for lower power dissipation at high levels of EIRP, using the higher inherent linearity/power of GaN devices to achieve 60 dBm or higher with fewer antenna elements.
![Figure 4](/ext/resources/images/Figures/2019/May/5M27/5m27f4.jpg)
Figure 4 Comparison of power dissipation in GaN, SOI and SiGe arrays.
Based on PA efficiency data and size/efficiency of heatsinks for live demonstrations at MWC Barcelona 2019, the DC power consumption of multiple mmWave arrays was estimated as shown in Figure 4. It appears that GaN has a significant advantage in terms of raw efficiency of a linear power amplifier at 28 GHz. However, all major OEMs have chosen to use SOI or SiGe so far, to take advantage of higher levels of integration, larger wafers and the resulting lower cost profile.
Over the next five years, significant adjustments are expected to occur to the balance between narrow beams (for long range) and wide beams (for better mobility). The optimal tradeoff in a dense urban network is not well understood today, and is likely to break into specific configurations to handle trains/buses/moving vehicles differently than pedestrian users. In particular, the large SOI-based arrays are expected to support the applications that cover dense urban pockets, where both vertical and horizontal steering are required and pedestrian speeds are typical. Other applications with higher mobility and less vertical steering are likely to move toward GaN devices.
![Figure 5](/ext/resources/images/Figures/2019/May/5M27/5m27f5.jpg)
Figure 5 A diagram representing physical packaging/integration for mmWave front ends (source: pSemi).
The physical integration of the RF front-end will also be critical. Very tight integration will be necessary in the 24 to 40 GHz bands to keep insertion losses low, so either LTCC or 3D glass structures will be used to embed the active die and passive elements (see Figure 5).
In the Radio Unit (RU), one convenient arrangement is to use an RFIC device for four antenna elements. From a simple geometric point of view, one RFIC for beamforming (phase and amplitude adjust) can be positioned between four antenna elements, using short traces and vias to route the mmWave signal (see Figure 6)
One open question concerns the use of filters in the mmWave front-end. Currently, no bandpass filters are used at the front-end, and during field trials the spectrum was clean enough to rely on the natural rolloff of the patch antenna and distributed antenna feed to provide out-of-band rejection. In the future, spectrum auctions and multi-operator deployment suggest that interference will arise. In fact, with high EIRP and very narrow beams, the interference will be intense when it unexpectedly pops up. Recent analysis indicates that filters will be introduced into the packaging over the next three years.
![Figure 6](/ext/resources/images/Figures/2019/May/5M27/5m27f6.jpg)
Figure 6 A typical panel with 4 sections, 64 256 dual-polarized antennas total (Source: FCC filing).
RF Implementation - CPEs
In fixed wireless, the Customer Premises Equipment (CPE) is a key part of the system. Initial deployments of 5G mmWave networks rely on high antenna gain and high EIRP from the CPE in order to support the necessary capacity. CPE RF front-ends today are constructed using a method that is similar to the network infrastructure, with a panel of antenna elements supported by beamforming RFICs, up-/down-conversion and then baseband processing. A typical CPE uses 32 dual-polarized antenna elements, supporting 2x2 MIMO with about 20 dBi gain from the antenna system.
Because the CPE is always connected to prime power, the PA efficiency is not a crippling limitation, and the CPE can often achieve high gain and high transmit power (linear EIRP in the range of 40 dBm).
RF Implementation - Handsets and other Mobile Devices
The biggest challenge facing the 5G mmWave link will come from the user’s hand blocking the antennas on a smartphone. In the 28 GHz band, the user’s hand is likely to attenuate the signal by at least 30 to 40 dB, effectively killing the link altogether. There can be multiple strategies to avoid this issue:
1. Multiple antenna sub-arrays on each handset. All 5G mmWave handset prototypes demonstrated over the past year utilize multiple sub-arrays, placed on both sides of the smartphone.
2. Foldable handsets are coming to market such as Samsung’s Galaxy Fold and Huawei’s Mate X. Because a foldable handset would be much larger than a human hand in the unfolded position, the placement of antennas could be more exposed.
3. Mobile hotspots can be used instead of mmWave links directly to the smartphone. This avoids the hand issue altogether, but may incur greater interference in the unlicensed bands. Importantly, the space and battery size constraints of the smartphone do not apply here, so the number of antennas can be increased to achieve much higher EIRP.
![Figure 7](/ext/resources/images/Figures/2019/May/5M27/5m27f7.jpg)
Figure 7 Layout of three mmWave sub-arrays on a handset.
The physical implementation on a handset is limited for cost and space reasons to a few sub-arrays, an RFIC and the modem/beamforming processing. To make this arrangement economical, each mmWave sub-array includes an up-/down-converter to shift the mmWave signal down to an IF frequency at roughly 4 to 6 GHz (see Figure 7). This enables the signals to travel through the PCB to a centralized RF transceiver.
Each mmWave subarray currently uses four dual-polarized patch antennas, each with a transmit/receive switch, low noise amplifier (LNA) and power amplifier (PA) closely integrated using RF-SOI. Each amplifier can only produce about 15 dBm linear power, so as many as eight antennas would be used to reach EIRP levels somewhere above 20 dBm. Three-dimensional beamforming on the smartphone platform is challenging, especially with a cluttered environment with metal surfaces and human hands in very close proximity. Even with eight antennas engaged, prototyping so far suggests antenna gain of only about 5 dBi.
For that reason, we expect much higher performance with hotspot products that utilize 32 antennas or more, achieving gain in the range of 20 dBi in the antenna system (15 dBi from the array and 5 dBi from the patch antenna itself). This type of product should be able to reach roughly 35 dBm linear EIRP or higher. From a system point of view, roughly 35 dBm or higher will be an important level to reach since the 5G link requires a closed loop with TDD channel feedback in order to maintain a continuous connection. Lower EIRP from the client device means a shorter range for the link, and would require the network operator to deploy larger numbers of cell sites in order to blanket a neighborhood with coverage. In short, low transmit power from the client devices would make the 5G business case unworkable for the mobile operator.
Commercial Status
Base station deployment is underway in earnest for the U.S. market this year, and the South Korean market is not far behind. Recent forecasts indicate that more than 600,000 radio heads will be deployed by 2024.
Commercial fixed-wireless services have already been launched in a handful of U.S. cities, with CPEs supported by major OEMs today. A few CPEs have appeared from the ODM community with poor performance, but we expect those to improve quickly to support healthy growth. In the next few years, the fixed-wireless application will account for millions of users.
This generation of technology is also unique in that handsets are coming out very quickly, and smartphones will be available before the network is launched in most countries. The first 5G mmWave handset has already been released (the 5G Moto MOD), and at least eight other mmWave handsets will be released in the second half of 2019.
Summary
5G mmWave radio links are more complex, more expensive and less reliable than LTE connections at 1 to 2 GHz. But mmWave bands will be necessary to keep up with rising demand, so the industry is currently pouring money into deployment of base stations and development of client devices. Initial fixed-wireless performance with CPEs has been surprisingly solid. The migration to mobile 5G usage will be tricky, with tradeoffs on beamwidth, link budget, mobility and cost coming into play. But there is one clear conclusion: 5G mmWave will be a significant part of future mobile networks.