Testing UWB: Don't try this at home!

If many of the best-known electronics companies—and a large number of smaller start-ups—get their way, UWB (ultrawideband), the new, short-range, ultrahigh-speed wireless-communication technology, will soon transform communication among computers and peripherals and among consumer-electronic devices—at least in the United States. What's more, Japan appears close behind. Meanwhile, Europe seems to be taking a wait-and-see position on UWB.

In the United States, the FCC (Federal Communications Commission) has allocated a 7.5-GHz-wide swath of spectrum—3.1 to 10.6 GHz—to UWB. If you haven't yet read about the technology, your reaction may be, "How can it do that? Other services are already using most of that spectrum." The answer is that UWB, by design, sends signals at such low average power (–41.3 dBm/MHz) and in such short bursts that the signals appear as low-level noise to the receivers for the other services. Those receivers have enough noise immunity to be unperturbed by the UWB transmissions.

Two industry groups support competing and incompatible forms of UWB. The WiMedia Alliance, of which the MBOA (Multiband OFDM Alliance) recently became part, primarily supports MB-OFDM (multiband orthogonal frequency-division multiplexing); the UWB Forum supports DS (direct-sequence)-UWB, which some—but not Forum members—call impulse radio. Both groups are rumored to also support—or at least to be considering support of—a low-cost, lower speed, low-power version of impulse radio, whose key application may be a future version of ZigBee, the wireless-sensor standard.

Both groups' Web sites list large numbers of member companies, but, of the two, the WiMedia Alliance includes more merchant IC manufacturers, including the all-important Intel, and a technology that many—albeit, not UWB Forum members—call more elegant and more readily scalable to higher data rates. The Forum says, however, that its DS-UWB technology uses less power to achieve equivalent range—a key attribute in battery-powered-system applications—and is both inherently simpler, thus less costly, and closer to market readiness than is the Alliance's computationally intensive MB-OFDM. The Alliance disagrees on several counts. As Jason Ellis, senior manager of business development and marketing at Alliance member Staccato Communications, puts it, "Real MB-OFDM devices cost substantially less and use less power." However, a Forum spokesman insists that the DS-UWB parts perform more required UWB functions than do the MB-OFDM parts that Alliance members are using for comparison and that if comparable parts of both types existed, the DS-UWB parts would cost less and use less power.

Early adopters of UWB will, therefore, need to choose between the two major approaches, and, in choosing, they will have to decide whether they think the Alliance's lead in industry popularity trumps the Forum's lead in developing its technology. At one time, the UWB Forum had proposed a CSM (common signaling mode), which would allow MB-OFDM and DS-UWB systems to communicate with each other at low data rates. The WiMedia Alliance has declared the CSM idea dead, but Forum members say that CSM can't and won't die unless one of the two technologies disappears, because, without it, neither type of system can detect the presence of nearby systems of the other type. The result, they insist, will be interference that may prevent either type of system from working satisfactorily.

For a wireless technology, UWB can be blazingly fast. According to Alliance members, early MB-OFDM-based products will permit data rates as high as 480 Mbps at a 3m distance and 110 Mbps at 10m. In a few years, these folks say, speeds should reach 1 Gbps at perhaps as much as 20m. Unfortunately, when quoting these performance figures, the Alliance members sidestepped the issue of BER (bit-error rate) or PER (packet-error rate). Without that information, the data rates are not particularly meaningful.

Alliance members assert that the 10^–12 error rates of which Forum members speak represent overkill in the area of the UWB market that many in the Alliance consider the most lucrative—consumer multimedia applications. Nevertheless, Freescale Semiconductor, the Forum's principal merchant-IC-manufacturer member, was first to announce an agreement to supply UWB devices to a major consumer-electronics manufacturer (Haier). Freescale was also first to deliver fully operational UWB silicon (with a raw data rate of 114 Mbps at 20m) and expects late this year or early next to announce parts almost six times as fast. The company also questions whether 16-QAM (16-level quadrature-amplitude modulation), which the Alliance recently specified for data rates greater than 200 Mbps, can really achieve 10m range and still meet the FCC's –41.3-dBm/MHz radiated-power limit, because, says a Freescale spokesman, "16-QAM takes approximately 4 dB more power per bit to go the same distance as DS-UWB's bipolar- and quadrature phase-shift keying."

How the UWB market divides among several segments will determine the kinds of products chip and module manufacturers emphasize (see sidebar "UWB drives to commercialize"). A key UWB application is almost certain to be WUSB, a wireless version of the Universal Serial Bus, whose wired embodiments truly dominate connections between peripheral devices and PCs. Industry watchers seem to agree, however, that growth of USB for peripheral interconnection has slowed and that sales of USB components—ICs, for example—for home-multimedia applications now exceed sales of such components for use in PCs and peripherals. This trend is likely to accelerate in WUSB; although there is little probability of a rush to substitute printers that use WUSB for units having wired connections, observers expect WUSB and a UWB version of IEEE 1394 (FireWire) to rapidly supplant wired connections among components of large-screen high-definition-TV systems (Figure 1).

UWB, like any other technology that is new and unproven in commercial applications, has doubters and detractors. In this case, many detractors fear that UWB will interfere with the other services whose spectrum it shares. UWB raises barely a whisper of concern, however, compared with another currently trendy shared-spectrum communication technology, BPL (broadband over power lines—see sidebar "BPL: the uninvited guest").

In a move to pre-empt problems, UWB-system architects have removed from the range of frequencies that UWB occupies the area in the vicinity of 5 GHz, which is home to several of the newer forms of wireless networks, such as IEEE 802.11a, g, and n, whose narrowband signals might otherwise cause interference to UWB signals. In addition, to avoid interference with the GPS (global positioning system), the FCC not only restricts UWB to frequencies well above those that GPS uses, but also demands stringent testing of UWB products and systems for out-of-band emissions at or near GPS frequencies.

The feeling in the UWB community is that the FCC has gone overboard in restricting the field strength of UWB emissions, even though, in some other technologies, such as BPL, the commission appears to have yielded to political pressures and ignored—or at least downplayed—legitimate concerns about interference to established services.

Currently, however, limits that UWB developers regard as excessively strict on EMI (electromagnetic-interference) emissions from UWB transmissions make compliance measurements more difficult, expensive, and time-consuming than the developers feel they need to be and can sometimes even make meaningful measurements impossible. Moreover, the limits are often far below FCC limits on and actual emissions from portions of the equipment under test that are unrelated to UWB. These emissions—for example, from PC power supplies and processor ICs—are present whether or not UWB is in use and can so completely dwarf those from UWB that, when UWB is activated, its contribution is undetectable (Figure 2).

DS-UWB and MB-OFDM differ profoundly. DS-UWB is sometimes referred to as a zero-carrier system because some implementations do not use conventional modulation of a sinusoidal carrier. The bandwidth can exceed 25% of the signal's center frequency, however. For example, signals that occupy a frequency range of 3.1 to 4.9 GHz have a center frequency of 4 GHz and a bandwidth of 1.8 GHz, which is 45% of the center frequency. In this type of UWB, the signals can be on for a few nanoseconds or, in some cases, even less than 1 nsec and can be off for considerably longer periods.

MB-OFDM is not a zero-carrier system; it uses a large number of carriers. Because attempting to continuously use the full 1.8 GHz would enormously increase the design complexity and might increase the power requirements beyond acceptable limits, the WiMedia Alliance system divides the 3.1- to 4.9-GHz spectrum into three bands, each slightly wider than 500 MHz, and periodically switches the signal among the bands. Each band contains 128 OFDM carriers, each of which is modulated either by QPSK (quadrature-phase-shift keying), which transmits two bits per symbol, or at higher data rates, by 16-QAM, which transmits four bits per symbol. The OFDM carriers are not subcarriers, however; there is no main carrier.

The ensemble of modulated carriers occupies one of the 500-MHz-wide bands for 250 nsec. Then, after a short delay, it switches to one of the other two 500-MHz-wide bands for another 250 nsec. Then, after another short delay, it switches to the third 500-MHz-wide band for another 250 nsec. The process repeats approximately every microsecond. A key reason for occupying three 500-MHz-wide bands, each for less than one-third of the time, instead of continuously occupying just one 500-MHz band, is to hold down the average energy radiated in any one band—and, hence, across the entire 1.8 GHz—to comply with the FCC radiation limits. According to the UWB Forum, however, the band-switching approach raises by 6 dB the ratio of peak-to-average signal, preventing operation at the FCC's maximum permitted level, thereby reducing the range.

The reason for using only the lowest one-fourth of the allocated 3.1- to 10.6-GHz spectrum is to stay in a frequency range that is compatible with current, well-understood, cost-effective CMOS-IC-fabrication technology. As IC manufacturers develop faster processes, the Alliance expects UWB to occupy nearly all of the allotted 7.5-GHz-wide spectrum.

"In testing a new-product design that incorporates UWB, four major areas require attention: conformance to regulatory requirements, performance, interoperability, and fairness," says Yoram Solomon, mobile-connectivity-solutions director for strategic marketing and industry relations at Texas Instruments (see sidebar "Ensuring fair sharing of the UWB channel"). At this stage of UWB's evolution, you must regard all of the areas as challenging. Many of the required measurements are inherently difficult, and developers haven't yet had the opportunity to learn through their own and others' experience most of the lessons they need to learn about what to do and what to avoid (Figure 3).

Only those who have access to a properly equipped compliance laboratory and who have significant knowledge and experience in RF measurements and the use of sophisticated RF-measurement instruments should attempt to make the FCC-required compliance measurements. In the absence of those prerequisites, opportunities abound to spend weeks of precious time and huge sums of money on meaningless measurements. Also, because you usually won't find out until much later that the measurements were meaningless, you risk basing costly, yet inappropriate decisions on them.

Mike Violette, chief executive officer of Washington Laboratories Ltd, an FCC-listed test lab in Gaithersburg, MD, makes the following comments: "There are about 170 FCC-listed labs in the United States and half-again as many overseas." You can get the full list from the FCC Directory of Test Labs. "There is no particular designation for labs that are allowed to do UWB testing; any listed lab that has the right equipment and know-how can do it," says Violette.

"One critical aspect of doing the testing is that the necessary equipment isn't cheap," Violette says. "The biggest issue is the available RBW [resolution bandwidth]. One of the UWB tests requires measuring 'power in a 50-MHz band.' To properly do this test, you need a receiver or spectrum analyzer with a 50-MHz RBW. To my knowledge, only two units, one from Rhode and Schwarz (Picture) and one from Agilent, provide both the required resolution and the dynamic range. They are costly and somewhat finicky, too."

John McCorkle, chief technologist of Freescale Semiconductor's UWB Organization, says, "Even with the proper instruments, it is easy to take too few measurements or to allow too little averaging time for rms measurements. To keep from missing narrowband spikes within bands of interest, measurements need to be spaced at 0.4times RBW-frequency intervals. To obtain accurate rms readings, the detector should dwell at each frequency for 1 msec. If you sweep the frequency instead of changing it in steps, you must set the sweep speed to meet the 1-msec averaging-time requirement. Taken together, the frequency-interval and averaging-time requirements impose a not-insignificant minimum on the time required to perform a UWB emissions test."

Washington Labs' Violette adds, "The other critical aspect of the measurement is that achieving the necessary sensitivity in the GPS bands requires painstaking technique and a very-low-noise measurement path. Also, to get the signal level above the noise, you must place the measurement antenna unusually close to the unit under test. The limits are absolutely too conservative in the GPS bands—something like 30 dB below the general-emissions limits for other FCC-regulated devices, such as computers and other unintentional emitters."

At press time, Peter Cain, an expert on UWB testing at Agilent and the author of an application note on the subject (Reference 1), commented, "FCC testing isn't the only issue. Regulatory testing is always a subset of the RF testing that a new radio design requires. Regardless of the technology, antenna-based testing always requires attention to detail; the same principles apply to UWB as apply to other [radio] technologies.

"I agree, though, that the levels tested for UWB are low and that you must exercise care to isolate signals not related to the UWB transmitter itself. Building low-noise amplifiers into spectrum analyzers is one way instrument manufacturers can remove the uncertainty of users trying to improvise noise-reduction techniques. Still, how low the emissions have to be depends on more than the other users of the band and the regulatory requirements. Placing the UWB transmitter or receiver near any other radio device or digital device with 'rich' emissions also has an impact. You can, however, desensitize either radio's receiver if the noise floor or out-of-band rejection isn't good enough.

"Before you go to the expense of booking a test at any test house, you should use the wide capabilities of your spectrum analyzer to understand what the device [under test] is doing. The FCC may have chosen certain ways of doing things, but that shouldn't prevent you from working in advance on ways to more quickly identify problems. If you expect your UWB device to be used near any other radio, it is very important to know how the two will interact.

"To overcome long measurement times, you can make the narrow-bandwidth and wide-span measurements with FFT-based techniques instead of swept-frequency ones. Using a 1-kHz RBW automatically guarantees a 1-msec measurement time because the measurement-time window has to be at least several milliseconds. With regard to the 50-MHz RBW requirement, I note that Wisair has just successfully gotten a device through FCC certification without using a 50-MHz RBW filter."

James Wright, director of marketing at LeCroy's Protocol Solutions Group (formerly, CATC), says, "Protocol testing is [yet another] nontrivial area of UWB testing. WUSB will offer more variations and more operating modes than its wired counterparts and will require highly sophisticated protocol-test instruments." LeCroy's initial protocol-test offerings focus on Certified WUSB, a version of WUSB specified by the USB-IF (USB Implementers' Forum). Though the UWB Forum proposed a DS-UWB-based version of WUSB, the USB-IF based Certified WUSB V1.0 on MB-OFDM. Although this development didn't please the UWB Forum, the group still believes that DS-UWB will play an important role in WUSB's future.

Another piece of the UWB-testing puzzle is interoperability testing. The quickest way for developers of UWB products to determine whether their products exhibit interoperability problems is through "unplugfests"—the wireless equivalents of the "plugfests" that organizations responsible for wired protocols have used to great advantage. The events, which have good credibility with special-interest groups and manufacturers alike, bring together for hands-on trials developers of products that are supposed to communicate and work harmoniously with each other. When they detect problems, the developers try to find solutions, and it is not uncommon for competitors to assist one another. Although the WiMedia Alliance hasn't yet announced any unplugfest plans, some members expect that early in 2006 and maybe even sooner, the organization will hold its first such event. It almost certainly won't be the last.

Author Information

Contributing Technical Editor Dan Strassberg has covered test and measurement for EDN for 18 years. He holds a bachelor's degree in electrical engineering from Rensselaer Polytechnic Institute (Troy, NY) and a master's in electrical engineering from the Massachusetts Institute of Technology (Cambridge) and is a life member of the IEEE. Contact him at StrassbergEDN@att.net.