Millimeter Wave Seeks Best Home
Startups, Incumbents See Different Roles for 802.11ad Standard
One year ago, the shotgun marriage of the WiGig Alliance and the Wi-Fi Alliance suggested millimeter-wave radio was off to a slow start as the next Wi-Fi frequency. The placement of millimeter-wave technology in the purview of the IEEE 802.11 working group for wireless LANs was partly a historical accident, since it was unclear whether the technology’s most immediate application would be in LAN topologies for enhancing the successful 802.11a/b/g, 802.11n, and 5GHz 802.11ac standards.
In fact, the first short-range consumer applications have been in serial links as HDMI and USB replacements. Most developers working in 60GHz radio have elected to use the full 802.11ad MAC and PHY specifications in a variety of contexts, some of which scarcely resemble LAN, as Figure 1 shows. For infrastructure, 802.11ad is finding favor as a small-cell backhaul alternative.
Figure 1. 60GHz links will serve in home, enterprise, and infrastructure applications. The 60GHz 802.11ad standard, originally developed for wireless LANs, can serve close to the end user, such as in docking and file synchronization, as well as in backhaul links for small-cell networks. (Source: The Linley Group)
IEEE 802.11’s Very High Throughput study group, formed in 2007, produced both 802.11ac for 5GHz technologies and 802.11ad for millimeter-wave follow-ons. Although the initial developers were uncertain as to whether a 60GHz Wi-Fi market would emerge, alternative applications have blossomed to augment the LAN.
The slow rise of 802.11ad in its original market raises the legitimate question of where and when 60GHz radio topologies will find use in true LANs. But the uncertainty hasn’t stopped semiconductor vendors from offering tri-band MAC and PHY products that support 2.4GHz, 5GHz, and 60Hz in one chipset. In many cases, however, the 60GHz link handles docking and synchronization rather than multipoint data distribution.
Depending on the chip developer’s goals, 802.11ad’s brightest prospects may be in short-range docking and synchronization, in tri-band LAN chipsets, or in wireless-backhaul equipment. Proponents cite healthy markets for wireless infrastructure, the enterprise, digital home, and even factory-floor communications. Underscoring the value of such applications, reports emerged in mid-May that Qualcomm plans to offer $300 million for 802.11ad startup Wilocity, a chipset designer with aggressive growth plans, that has already signed development pacts with both Marvell and Qualcomm’s Atheros group.
Developers of PHY and MAC silicon targeting the fastest and highest returns per unit now look to 802.11ad in small-cell backhaul. Those more interested in the high volumes of consumer devices expect millimeter-wave wireless LAN to be appropriate for tablets and smartphones, because the tiny 60GHz antennas take up much less space and power than antennas for 5GHz 802.11ac.
Chipset Land Grab on the 60GHz Frontier
A few early semiconductor players have staked claims in all portions of the 802.11ad market. Startup Tensorcom announced in February the first single-package hybrid component combining an 802.11ad MAC, PHY, and antennas. Toronto-based Peraso offers small transceiver daughtercards for consumer applications, as well as a suite of chips and modules for backhaul applications.
At February’s ISSCC, Broadcom provided initial details of a dual-chip superheterodyne solution for short-range 802.11ad in which one chip handles baseband tasks while the other, an RF front-end chip, handles RF and IF duties. The company withheld details of how and when it will bring the product to market, but it is feeling little immediate pressure to reduce the current solution to a single integrated CMOS chip.
At February’s Mobile World Congress, Wilocity launched a third-generation transceiver chipset, the Sparrow Wil6300, intended for smartphones. The company’s previous generations, the Wil6100 and Wil6200, combined a baseband/MAC CMOS device in an FPN package with a bare-die radio front end (RFE). Wilocity designed these chips for hybrid integration with antennas (in its reference design, a 16-element printed antenna array). The company has withheld details of its new smartphone-centric Wil6300. Qualcomm’s original development effort with Wilocity sought to combine the Atheros work in lower-frequency wireless LANs with Wilocity’s own work. An acquisition could merge the startup’s effort in smartphones with Qualcomm’s cellular-baseband and Wi-Fi chipsets.
In part, the diverse interest in 802.11ad was due to the fading prospects of 802.11n and the realization that 802.11ac requires too many antennas for platforms smaller than tablets. By default, owing to the limitations of 802.11n and 802.11ac, 802.11ad will serve in many multiband Wi-Fi applications beyond 2.4GHz 802.11a/b/g. Occasionally, it will appear in multiband designs that also support 802.11n and 802.11ac. Following Wilocity’s transition from the Wil6100 to the Wil6200, many vendors expanded IP-address capability to allow easier integration of multiple 802.11 MAC and PHY standards in one platform.
Silicon Image continues to support the informal HD serial-link standard WirelessHD. It still offers the UltraGig 6400 chipset for gaming appliances. WirelessHD supporters hoped to make the standard part of 802.15ac, but the IEEE placed this working group in hibernation mode at the end of 2009. Since then, most millimeter-wave proponents have refocused their interest on 802.11ad.
Big Speed in a Small Range
The 802.11ad standard uses four 2GHz channels in the 60GHz band to achieve data speeds of up to 7Gbps. It relies on an unlicensed ISM band and is the next logical follow-on to 5GHz 802.11ac; its high frequencies create unique design constraints, but they also offer some advantages. Because the line-of-sight 60GHz signal does not penetrate walls, it was seen as an in-room extension of wired interfaces such as HDMI. Some original literature from the WiGig Alliance (now merged with the Wi-Fi Alliance) even suggested PAN (personal-area network) applications.
Before merging with the Wi-Fi Alliance, WiGig developed a series of protocol adaptation layers (PALs) for common interfaces such as HDMI and USB. Developers at 60GHz startups are uncertain as to which PALs will become widely adopted. An USB3.0 interface seems to be a likely candidate for widespread implementation, whereas prospects for the WiGig Display Extension (WDE) remain uncertain owing to the limited development of interfaces for QHD/4K displays.
Beam management also plays a bigger role in 802.11ad than in any other packet-based RF standard because of the high gain necessary to achieve adequate ranges and because the small antennas (roughly one-quarter the size of the signal’s 5mm wavelength) for 60GHz radio allow phased arrays in small packages. The standard specifies such beam-forming protocols as Sector Level Sweep (SLS) and Beam Refinement Phase (BRP). The mandated SLS enables the omnidirectional antenna on a receiving station to search for low-power transmissions from handsets, choosing Tx/Rx sectors to optimize receipt of signals from a given sector. The optional BRP allows further fine-tuning of optimal link budgets in directional communications, particularly when two 802.11ad stations are implementing spatial bandwidth sharing.
The 802.11ad standard mandates “fast session transfers” between PHYs. This term refers to a seamless rate fallback and rate rise between 60GHz and 2.4GHz or 5GHz PHYs, targeting multiband devices in particular (though multiband support is optional in a particular PHY). The 802.11ad MAC layer is backward compatible with 802.11ac and other standards. The MAC for 60GHz radio, however, adds channel access, synchronization, and authentication features for this frequency band.
The physical layer is based on four 2.16GHz channels that reside between 57GHz and 66GHz, with a common default channel defined in all nations and centered on 60.48GHz (other channels differ by geographical regions, per ITU-T standards). It supports a variety of single-carrier and OFDM modulations. In fact, 802.11ad is unique in enabling four PHY signals with different modulation: The control PHY (CPHY), which only handles control messages and supports a more robust error detection and correction than other PHYs, employs BPSK modulation. The single-carrier PHY (SCPHY) employs BPSK, QPSK, or QAM-16. The orthogonal-frequency-division-multiplexing PHY (OFDMPHY) achieves higher data throughput using OFDM but consumes more DSP power than the SCPHY. And for battery-powered applications, a low-power single-carrier PHY (LPSCPHY) saves power by removing QAM-16 capability.
The IEEE mandates only the control-channel CPHY and the SCPHY. To date, most developers have worked with the mandatory SCPHY—and with the LPSCPHY if they are optimizing devices for low power. Vendors see the OFDMPHY as a higher-throughput option for second-generation products. The greater sample rate of the multicarrier OFDMPHY, as Table 1 shows, enables more timing accuracy. The CPHY, SCPHY, and OFDMPHY use low-density parity check (LDPC) for forward error correction (FEC), but the LPSCPHY uses a combination of Reed-Solomon and block codes, because LDPC is one of the SCPHY’s largest contributors to power consumption.
Table 1. Features of 802.11ad PHY types. The CPHY is a low-throughput high-coding PHY for control signals; the SCPHY, LPSCPHY, and OFDMPHY allow data-throughput and power-dissipation tradeoffs. (Source: IEEE, Agilent, and Blu Wireless)
Going the Distance
Owing in part to the unlicensed status of the frequency band and the low cost of beam-forming antennas, some developers have promoted 802.11ad as an alternative to fiber in small-cell backhaul. Although the standard is intended for LANs, at least two PHY types and many MAC features do apply to small-cell backhaul, allowing semiconductor developers to use chips originally designed for a short-reach LAN. Simpler line-of-sight backhaul can employ single-carrier implementations; non-line-of-sight backhaul requires the OFDMPHY and more-complex beam-forming protocols.
Millimeter-wave radio has caught the attention of backhaul specialists because it can integrate with small and inexpensive antenna elements that enable easy phased-array designs. Oxygen attenuation in the 60GHz band—a double-edged sword that limits the range while reducing interference—helped make millimeter-wave radio an early choice for covert satellite communications. The attenuation factor, which causes a 30dB signal to drop to the noise level in less than 2.5km, makes the band equally popular for adaptive mesh networks based on spatial reuse. In addition, phased arrays could implement beam forming to reconfigure the small-cell backhaul links. This capability has attracted both RF and DSP specialists to the redefined 802.11ad backhaul market.
Past backhaul options have included leased T1/E1 lines, analog microwave, digital microwave, and fiber. Fiber can be a fast and secure alternative in macro- and microcell networks where cost is not a concern, but it’s economically unviable for small cells. Analog TDM microwave is obsolete, and digital microwave is proprietary to specific vendors. The frequency band in these cases offers neither the security features of oxygen attenuation nor the ability to achieve spatial reuse through directional antennas, precluding reconfigurable meshes. The arrival of 802.11ad has already prompted design efforts by startups.
Peraso in particular has focused much of its recent development work on backhaul components, though it continues to offer the PRS1021 module: a small card for mobile devices that integrates 15 ICs and an 8.5dBi antenna. This module comes as a wireless USB3.0 adapter or WiGig docking station. More of Peraso’s near-term efforts, however, have focused on developing the PRS1100 transceiver ICs and PRS2100 waveguide modules, both intended for small-cell backhaul applications. The modules’ $60 high-volume price allows a higher profit margin than either standalone transceiver ICs or the packaged consumer-market products deliver.
Small microwave players have created opportunistic transceiver products for small-cell backhaul applications. Hittite Microwave’s HMC6000, which is a SiGe BiCMOS transmitter that converts I and Q signals to 60Hz, and the HMC6001 receiver, which accepts single-ended 60GHz signals and down-converts them to differential baseband I and Q signals, can operate on their own or with third-party devices that integrate the MAC and baseband functions. Some BiCMOS and III-V-semiconductor specialists remain in the market, but the small-cell market is now the target of others more versed in high-integration CMOS design.
For example, ARM founder Robin Saxby is among the founders of Blu Wireless, a U.K. startup working with telecom-card OEM InterDigital on reference boards for small-cell backhaul. That company’s original business plan called for development of IP cores to serve in PHY and MAC products for traditional consumer 802.11ad applications, but the Blu Wireless designers quickly turned to small-cell backhaul because of its more immediate potential. Blu Wireless is also proposing simple backhaul as the first step in creating self-organizing mesh networks among small cells using 60GHz links. Combining point-to-point links with new beam-forming and spatial-sharing technologies could enable backhaul meshes for the first time.
The Hydra transceiver-IP (intellectual property) block from Blu Wireless includes a parallel vector DSP for modem tasks, subcarrier mapping and demapping, and equalization. A heterogeneous processor with fixed DSP functions wraps around the parallel core and handles FFT operations and LDPC encode/decode operations, as Figure 2 shows.
Figure 2. Block diagram of Hydra IP from Blu Wireless. The vector DSP core at the center (shown in purple) handles modulation and demodulation, mapping and demapping, and equalization. The heterogeneous processor handles LDPC code and FFTs. (Source: Blu Wireless)
Two Hydra generations developed for WLAN tasks served as the basis of a separate core for backhaul applications: Hydra BH_1.0, a single-channel-only design supporting data rates of 0.5–2Gbps. This range is less than the 2–7Gbps of the LAN version, which can reach up to 16Gbps when using eight parallel-processing units. Blu Wireless advocates employing many parallel MAC devices in applications where spatial sharing requires multiple parallel modulation channels. The Hydra backhaul core supports phased-array antennas and includes a decision-feedback equalizer.
Opportunity for Mobile Clients
Although chip companies with microwave or DSP expertise focus on the nearer-term backhaul play, development of a tactical backhaul chip still allows an immediate focus on LAN and short-reach serial links. In fact, some companies see an immediate opportunity where others expect longer lag times. Dell integrated Wilocity’s first-generation Wil6100 transceiver chips in its Ultrabooks and docking stations in mid-2013, expecting users to demand multigigabit speeds for docking and synchronization, if not for wireless LAN. These Ultrabooks have been on the market for a year, but those integrating the Wil6100 are selling poorly. Wilocity’s pacts with Marvell and Qualcomm Atheros a year ago indicated how quickly the startup expected the short-reach market for 802.11ad to emerge.
Wilocity intended the Marvell and Qualcomm deals to allow an easy transition from existing 802.11ac designs, though neither of the bigger partners have yet offered practical products. Unfazed, the company announced at MWC its third-generation chipset, the Wil6300, targeting smartphone applications. Although this audacity surprised some developers of backhaul and wireless-access-point products, Tensorcom executives say the market assumptions may be close to the mark. Tensorcom predicted that clients demanding smartphone data throughput of hundreds of megabits per second may want higher docking and synch speeds by 2015; it also said that by 2016, an 802.11ad chipset may become a standard feature in high-end smartphones.
Tensorcom still focuses on more-traditional desktop and tablet platforms, but its TC2522-Y module employs what the company calls “Green” system-on-a-chip (GSoC) technology. By combining low-power design with a conservative 40nm CMOS process, it is opting for a general mobile and low-power market, but may move to smartphones. It presently supports only the mandatory single-carrier and low-power single-carrier PHY modes, but its TC2522-Y consumes peak active power of 150mW; this device is in volume production. Designing the MAC and PHY within such power constraints is more important than offering OFDM support, according to Tensorcom. The company reduces RF-power requirements using passive elements in an asymmetrical pattern where the main lobe of the radiated signal lies along the axis of the array (an “end fire” configuration), or when using patch antennas in vertically-mounted devices.
If Broadcom’s ISSCC paper indicates the target market of larger companies with extensive Wi-Fi experience, the PCI Express integration in the company’s two-chip set suggests a clear focus on consumer applications, given PCIe’s centrality to many consumer products. The 16 transmit chains and 16 receive chains in Broadcom’s radio front end (RFE) support a 16-antenna array, making it larger than most developers’ first-generation 802.11ad chipsets. The company further encourages high-volume laptop applications by supporting a simple coaxial interface between MAC/PHY and RFE chips.
Broadcom’s RFE chip excludes the antenna array, but it’s designed for integration in a module with that array. The RFE performs IF-to-RF up- and down-conversion and includes a 24–28GHz PLL with a x2 multiplier. A mixer/local oscillator, a 1:2 splitter, and prepower amps ensure strong signals for all 16 chains in the architecture. The companion baseband chip integrates a CPU that handles calibration, and it includes hardware support for the 802.11ad beam-forming protocols discussed above.
Hot Air Absorbs RF, Too
We expect at least two or three large communications-IC companies, such as Broadcom, Intel, and Qualcomm, to enter the market—a factor that will help increase the validity of 60GHz radio. Qualcomm would have been a significant player even if it only agreed to an arm’s-length development pact with Wilocity, but its reported outright acquisition of that company will increase its market presence. Vendors will attempt to differentiate their designs on the basis of channel count, power dissipation, antenna-array efficiency, baseband-DSP speed, or any number of additional characteristics bounded by the standards.
The near-term diversion into 4G small-cell backhaul represents an interesting opportunity for some specialists like Blu Wireless and Peraso, though the market for backhaul alone may be unable to sustain multiple players. If its vector-DSP IP can serve meshes as the company claims, Hydra could be a choice core for semiconductor and OEM customers.
In the laptop, notebook, and smartphone segments, 802.11ad could offer higher speeds with fewer of the deployment problems that 802.11n experiences owing to MIMO complexity and that 802.11ac experiences owing to inadequate room for antennas in platforms smaller than tablets. The current interest in tri-band modems seems slow to take off, so expectations of volume sales within two to three years may be optimistic. Nevertheless, growth rates could be healthy enough to support all existing players, as well as the anticipated involvement of Qualcomm, Intel, and others. Few market analysts have considered the possible gaming and HDMI applications for the digital home, provided the standard supports PALs for the more common high-speed serial interfaces.
Several market-research firms have published 802.11ad revenue estimates in the $9 billion to $11 billion range for annual chip and hybrid-module sales by the end of the decade. This revenue would roughly equal the total sales of all Wi-Fi chipsets in 2013 and 2014 for all frequency bands, making the range for 802.11ad alone appear unlikely.
In one study, ABI Research estimates that when tri-band devices (also capable of supporting 802.11n and 802.11ac) enter the mix, the total number of 802.11ad MAC and PHY IC products selling into all market sectors could exceed 1.5 billion units by the end of 2018. Given the multiple markets in which 802.11ad now serves, unit sales of 1–1.1 billion by the end of the decade and revenue of $5 billion to $7 billion annually in that same period seem more viable. But these numbers include backhaul, serial-link, and smartphone applications as well as traditional Wi-Fi uses.
Even as excitement increases in new market sectors, however, it’s important to realize that millimeter-wave radio is primarily a line-of-sight technology originally intended for a few meters’ reach (though it can extend to non-line-of-sight uses at the cost of additional coding complexity and power consumption). Its signal-attenuation properties have yet to receive thorough testing in adaptive meshes. And some medical studies of workers constantly exposed to non-ionizing phased-array beams above 25GHz suggest that dwelling among dozens of 60GHz beams connecting mesh networks may have health consequences.
The 802.11ad space is now splitting into consumer gaming, docking and synchronization, mobile devices, Wi-Fi networking, and small-cell backhaul—all markets with significant growth potential. But just as 802.11n and 802.11ac encountered speed bumps, millimeter-wave radio will see unanticipated market limiters that could be as diverse as the many sources of potential growth for 60GHz radio.
For More Information
For an excellent history of 802.11ad development, written by Eldad Perahia and Michelle Gong of Intel, access www.researchgate.net/publication/220410973_Gigabit_wireless_LANs_an_overview_of_IEEE_802.11ac_and_802.11ad.
A good analysis of PHY and MAC implications for testing 60GHz radio platforms is available at cp.literature.agilent.com/litweb/pdf/5990-9697EN.pdf.