DALLAS, Texas -- The debate is heating up within the IEEE 802.15.3a task group, which is trying to prepare a draft proposal for ultrawideband (UWB) communications. Texas Instruments' hometown of Dallas recently hosted a meeting of the project that saw 23 different proposals presented on how to standardize the operation of UWB as a short-range wireless communications network.
The goal of the task group is to produce a standard that can operate at about 110-Mbit/s at 10-meters range. Achieving that does not look too difficult in terms of pure physics, but doing it in the optimum way for cost and power consumption today, and for evolution to higher data rates in the future, is more contentious. And IEEE projects have a reputation for being talking-shops that find it hard to reach closure.
Texas Instruments engineers and executives said they want this IEEE committee to move swiftly and, if necessary make compromises in the interests of commercial expediency, but they are also adamant that their own proposal, based on orthogonal frequency domain modulation (OFDM) and quadrature phase shift keying (QPSK), is the way forward for low-cost manufacturability and for future evolution,
The seductive charm of UWB is that by spreading data across a large swathe of the electromagnetic spectrum wireless communications could theoretically go to data rates approaching 1-Gbit/s over short distances. With signals that are so highly spread, the energy level in any given part of the spectrum is low enough not interfere with established services, proponents claim. The bad news is that even if UWB does not knock out established services it would manifest itself as an increase in background noise and for some opponents of UWB any detrimental impact on the electromagnetic spectrum is to be resisted.
"The FCC has recently re-affirmed the allocation of spectrum for UWB," said Steve Turner, UWB business development manager for Texas Instruments (see February 13, story). "And the Japanese government seems to be taking an aggressive position on UWB as do Korea and Singapore. It is the impression that Europe is being more cautious but they are looking at the technology intently," Turner added.
According to Turner now that the FCC has ruled that in the United States UWB can extend up to 7.5-GHz and should be divided into fifteen 500-MHz to 700-MHz sub-bands there is an urgency to elaborate a standard and bring products to market.
"Every company we've talked to agrees that this needs to happen fast. We need to try and find an early consensus around a merged proposal sometime around September, November 2003. That could allow products in the market in 2004. This is not a technology looking for a problem to solve. The market needs are out there," said Turner.
However, Turner's talk of consensus appears optimistic after news of consortia forming around Intel Corp. and around XtremeSpectrum Inc. emerged at the time of the meeting a couple of weeks ago (see March 11 story).
XtremeSpectrum already has a UWB chipset (see July 16, 2002, story) and Intel has the zeal of a new convert for all things related to communications.
Ironically despite coming under the title of ultrawideband many of the 802.15.3a proposals are keen to minimize the charges of RF pollution laid against UWB and to reduce the cost of development by taking their proposals narrower rather than broader.
So whereas FCC is recommending the use of up to 15 sub-bands none of the proposals has recommended using more than seven, according to Anuj Batra, a member of the technical staff at the DSP-Solutions R&D center in Dallas.
Early proposals for UWB in the 1990s, which originated with Time Domain Corp., were for a simple digital pulse technology (see September 29, 2000, story).
"If the pulse was early you recorded a zero, if it was late you recorded a one," said Batra. "A few years ago people improved on that by also modulating the phase of the pulse, which gives another 3-dB off the bat. But now within the last year and with the 7.5-GHz limit we've started to look at a more classical radio approach," Batra said.
TI is proposing the standard only use three sub-bands between 3.1-GHz and 4.8-GHz thereby avoiding interfering with the 5-GHz band where the 802.11a wireless LAN sits. Batra and his team are claiming speeds of 55-Mbit/s, 110-Mbit/s, 200-Mbit/s and even 480-Mbit/s in a more custom-tuned implementation.
"We think there would be a problem to seven sub-bands. You would have to design a wide-open LNA low-noise amplifier, a wide-open pre-select filter. As you do that the noise in the LNA tends to go up. We worked out that to go above 4.8-GHz would only gain 1-dB in signal margin."
Batra sees the extra sub-bands as providing room for growth and evolution as circuit implementation technologies improve.
For now within each of the three sub-bands Texas Instruments is proposing the use of an OFDM signal, itself a form a spread-spectrum transmission, based on QPSK modulation to give the maximum flexibility. This should allow data to shifted onto the best channels, and create and suffer from the least interference.
According to Batra the TI UWB scheme can be completed with CMOS components today. This in turn helps promote low costs and low power, and a quick time to market. It also promotes future circuit integration and the low-cost evolution of the standard.
"OFDM is spectrally efficient and robust to multi-path issues, 60-nanosecond delay spreads can be coped with. It is robust against frequency-selective fading," said Batra.
But does the proposal introduce complexity, something that Texas Instruments with their established position in DSP might welcome.
"We've come up with a digital heavy solution. But that means we can take advantage of Moore's Law and scale the solution over time. But there also has to be analog expertise and TI has that," said Turner.
Nor was Batra prepared to concede that the solution is overly-complex. "It may sound complex but in reality the implementation is pretty simple. We do require a 128-point FFT and inverse FFT at about 100-MHz clock frequency."
But below 200-Mbit/s the solution can be contrived to be "conjugate symmetric" which means only the I value of an I/Q quadrature and phase system needs to be calculated, Batra said.
"We don't believe the complexity issue is a problem. This is something we can do quickly," said Turner.
The next meeting of the IEEE 802.15.3a task group takes place in May in Singapore followed by a meeting in July in San Francisco.
