Ultra-Wideband for Multiple Access Communications

MULTIPLE ACCESS TECHNOLOGIES FOR B3G WIRELESS COMMUNICATIONS

Robert C. Qiu, Tennessee Tech University Huaping Liu, Oregon State University Xuemin (Sherman) Shen, University of Waterloo

ABSTRACT two laws, while narrowband communications has shifted to the last two laws. Ultra-wideband wireless communications Although often considered a recent break- techniques have many merits, including an through in wireless communications, UWB has extremely simple radio that inherently leads to actually experienced well over 40 years of tech- low-cost design, large processing gain for robust nological development. The physical cornerstone operations in the presence of narrowband inter- for understanding UWB pulse propagation was ference, covert operations, and fine time reso- established by Sommerfeld a century ago (1901) lution for accurate position sensing. However, when he attacked the diffraction of a time there are a number of challenges in UWB domain pulse by a perfectly conducting wedge. receiver design, such as capturing multipath In fact, one may reasonably argue that UWB energy, intersymbol interference especially in a actually had its origins in the spark gap transmis- non-line-of-sight environment, and the need for sion design of Marconi and Hertz in the late high-sampling-rate analog-to-digital converters. 1890s. In other words, the first wireless commu- In this article we provide a comprehensive nications system was based on UWB. Due to review of UWB multiple access and modulation technical limitations, narrowband communica- schemes, and their comparison with narrow- tions was preferred to UWB. Much like spread band radios. We also outline issues with UWB spectrum or code-division multiple access signal reception and detection, and explore var- (CDMA), UWB followed a path with early sys- ious suboptimal low-complexity receiving tems designed for military covert radar and com- schemes. munications. After accelerating development since 1994 when some of the research activities INTRODUCTION were unclassified, UWB picked up its momentum after the Federal Communications Commis- Modern communication theory was originated sion (FCC) Notice of Inquiry in 1998. Interest in from the attempts of communication engineers UWB was sparked when the FCC issued a who wanted to understand what they were doing Report and Order in February 2002 allowing its in the most general terms. The limits of digital commercial deployment with a given spectral wireless communications systems depend pri- mask requirement for both indoor and outdoor marily on four basic laws and their underlying applications. theories, attributed to Maxwell and Hertz, Shan- The main limiting factor of UWB wireless non, Moore, and Metcalfe. The first two laws systems is power spectral density rather than are laws of nature, while the last two are laws of bandwidth. Thus, the initially targeted applica- behavior. The order is in the sequence of their tion was short-range (< 10 m) high-data-rate (> discovery and importance. As the field of wire- 100 Mb/s) within IEEE 802.15.3a. Due to imple- less communications has matured, emphasis and mentation limitations, systems with moderate immediate relevance have shifted gradually range (100–300 m) and low data rates (less than downward in the list. Without an appreciation a few megabits per second) have received signifi- for Maxwell and Hertz’s theories, there would be cant attention since late 2003 within IEEE no controlled wireless propagation of electro- 802.15.4a. In addition, UWB applications should magnetic waves. Without an understanding of also include safety/health monitoring, personnel Shannon’s theories, efficient use of the spectrum security, logistics, industrial inventory control, through sophisticated signal processing could not industrial process control and maintenance, have been achieved. Ultra-wideband (UWB) is home sensing, control, and media delivery. In experiencing this shift, probably from the first this article our goal is to provide a comprehen-

sive review of UWB multiple access and modulation schemes, and their comparison with narrow- Relative time shift band radios. We will also outline issues in UWB ‘1’ signal reception and detection, and explore vari- Time reference ous suboptimal low-complexity receiving ‘0’ schemes. The rest of the article is organized as follows. We describe the basics of pulsed UWB modulation schemes. Multiple access approaches employing pulsed UWB technologies are pre- (a) sented. UWB receiver designs and challenges are discussed, followed by conclusions. Pulse repetition interval ‘1’ ‘0’ MODULATION SCHEMES Based on FCC regulation, pulsed UWB is a form of bandpass communications. The transmitter sends pulses with a bandwidth (10 dB) that is at least 500 MHz and edge frequencies Pulse width within 3.1–10.6 GHz [1]. (b) MULTIBAND DESIGN The major reasons for the standards body (IEEE ‘1’ ‘0’ 802.15.3a) to adopt a multiband scheme are: • Spectrum flexibility/agility: regulatory regimes may lack large contiguous spectrum allocation; spectrum agility may ease coexistence with existing services. • Energy collected per RAKE finger scales (c) with longer pulse widths used, which prefers fewer RAKE fingers. n Figure 1. UWB modulation options: a) binary PPM; b) binary PAM; c) on- • Reduced bandwidth after downconversion off keying. mixing reduces receiver power consumption and linearity requirements. • A fully digital solution for signal processing munication system. The multiband orthogonal is more feasible than a single-band solution frequency-division multiplexing (OFDM) system for the same occupied bandwidth. transmits information on each subband. This • Transmitter pulse shaping is made easier; technique has nice properties including the abili- longer pulses are easier to synthesize and ty to efficiently capture multipath energy with a less distorted by integrated circuit (IC) single RF chain. The drawback is that the trans- package and antenna properties. mitter is slightly more complex because it • It is capable of utilizing frequency-division requires an inverse fast Fourier transform multiple access (FDMA) for severe near-far (IFFT), and the peak-to-average ratio may be scenarios. higher than that of pulse-based multiband approaches. It seems to be very challenging for The Multiband Pulsed Scheme — The main both the pulse-based and OFDM-based multi- disadvantage of narrow time domain pulses is that band solutions to meet the target costs imposed building radio frequency (RF) and analog circuits by the market. as well as high-speed analog-to-digital converters (ADCs) to process signals of extremely wide MODULATION OPTIONS bandwidth is challenging, and usually results in In choosing modulation schemes for UWB sys- high power consumption [2]. Collection of suffi- tems, one must consider a number of aspects cient energy in dense multipath environments such as data rate, transceiver complexity, spec- requires a large number of RAKE fingers. The tral characteristics, robustness against narrow- pulsed multiband approach can eliminate the dis- band interference, intersymbol interference, and advantages associated with large front-end pro- error performance. For pulsed UWB systems, cessing bandwidth by dividing the spectrum into the widely used forms of modulation schemes several subbands. The advantage of this approach include pulse amplitude modulation (PAM), on- is that the information can be processed over off keying (OOK), and pulse position modula- much smaller bandwidth, thereby reducing the tion (PPM). To satisfy the FCC spectral mask, design complexity and power consumption, lower- passband pulses are used to transmit information ing cost, and improving spectrum flexibility and in pulsed UWB systems. Although these pass- worldwide compliance. However, it is difficult to band pulses could be obtained by modulating a collect significant multipath energy using a single baseband pulse using a sinusoidal carrier signal, RF chain. In addition, there are very stringent the term binary phase-shift keying (BPSK) is still frequency switching time requirements (< 100 ps) somewhat imprecise in the context of pulsed at both transmitter and receiver. UWB signaling as the carrier-modulated baseband pulse is usually treated as a single entity, The Multiband OFDM Scheme — Multipath the UWB pulse shape. energy collection is an important issue as it is a The three modulation schemes mentioned major factor that determines the range of a com- above are illustrated in Fig. 1. For a single-user

IEEE Communications Magazine • February 2005 81

system with binary PPM signaling, bit 1 is rep- minimize the probability of collisions due to Because of the resented by a pulse without any delay and bit 0 multiple access. In TH UWB, each frame is sub- by a pulse with delay relative to the time refer- divided into Nh chips of duration Tc. Each user random polarities of ence. A major factor governing the perfor- (indexed by k) is assigned a unique pseudo-ran- the information mance of this system, like any other PPM-based dom time shift pattern, {hk,n}, 0 ≤ hk,n < Nh, symbols, the PAM system, is the set of time shifts used to repre- called a TH sequence, which provides an addi- sent different symbols. The most commonly tional time shift to each pulse in the pulse train. scheme inherently used PPM scheme is the orthogonal signaling The nth pulse undergoes an additional shift of offers a smooth PSD scheme for which the UWB pulse shape is hk,nTc s, where chip duration Tc is also the dura- orthogonal to its time-shifted version. There tion of an addressable time delay bin [4]. The when averaged over also exists an optimal time shift for an M-ary addressable TH duration must be strictly less a number of symbol PPM scheme. The time shifts for both the than the frame time. For TH PPM, information intervals. In this orthogonal and optimal schemes depend on data are carried by the additional time shift τ. the choice of UWB pulse p(t). For binary PAM For TH PAM, information data are transmitted sense, PAM signaling, information bits modulate the pulse by changing the amplitude of the time-hopped signaling is polarity. For OOK signaling, information bit 1 pulse. With binary signaling, the transmitted TH is represented by the presence of a pulse, and PAM or TH PPM signal of the kth (k = 1, …, attractive. no pulse is sent for bit 0. K) user can be written in a general mathematical Binary PAM and PPM schemes have similar form as performance. The OOK scheme is less attractive  τ  than PAM or PPM because of its inferior error ∞ tnThT−−b kn, c − St() Ebp0  2 performance in the same environment. Howev- k = ∑ b kn,   (1) n=−∞  1  er, if receiver complexity is the main design con-  (1− bk,n ) cern, a simple energy detection scheme can be applied to OOK signals, resulting in a receiver where Eb is the transmitted energy per bit, n is of lowest achievable complexity. Power spectral the bit index, Tb is the bit interval, p(t) is the 0 1 density (PSD) of the modulated UWB signal UWB pulse, and bn, bn are information bits (may must satisfy the spectral masks specified by have been converted to non-return-to-zero 1 0 spectrum regulating agencies. In the United form). For TH-PAM, bk,n is set to 1 and bk,n States, the spectral mask for indoor applications ∈{–1, 1} carries information. For TH PPM, 0 1 specified by the FCC starts from 3.1 to 10.6 bk,n is set to 1 and bk,n ∈{–1, 1} carries informa- GHz. OOK and PPM signals have discrete spec- tion. The pseudo-random hopping sequence tral lines. These spectral lines could cause severe {hk,n} is used to identify the kth user. interference to existing narrowband radios, and various techniques such as random dithering DIRECT SEQUENCE FOR MULTIPLE ACCESS could be applied in PPM to lower these discrete Direct sequence can also be used for multiple spectral lines and smooth the spectrum. Because access in a PAM-UWB or PPM-UWB system. of the random polarities of the information In such a system, each symbol is represented symbols, the PAM scheme inherently offers by a series of pulses that are pulse-amplitude- smooth PSD when averaged over a number of modulated by a chip sequence. Input symbols symbol intervals. In this sense, PAM signaling is are modulated onto either the amplitude or attractive. the relative positions of each sequence of pulses. ULTIPLE CCESS CHEMES For binary signaling, the transmitted signal M A S for direct-sequence PAM or direct-sequence Multiple access communications employing PPM signals of the kth (k = 1, …, K) user can pulsed UWB technologies has drawn significant be written in a general mathematical form as research interest. Various multiple access  tnTiT−−  schemes and their performance have been ∞ Nc −1  bc reported in the literature [3, 4]. Time hopping 0     Stkbknki()= ∑ Ebp,,∑ a p τ 1 (TH) has been found to be a good multiple   −−1 b   n=−∞  i=0  2 ()kn,   access technique for pulsed UWB systems [4], and various modulation options for TH UWB (2) systems are discussed in terms of their spectral where Eb is the transmitted energy per bit characteristics and hardware complexities in [5]. (assuming the total energy of Nc pulses is nor- Direct sequence (DS) spreading is also an malized to unity), Tc is the chip duration, ak,i ∈ attractive method for multiple access in UWB {–1,1} is the ith chip of the kth user, Nc is the systems. Performance of a DS-UWB system number of chips used to represent one symbol, 0 1 with PAM was analyzed in [3]. Since pulsed and bk,n, bk,n∈{–1,1} are the information bits of UWB systems are inherently spread spectrum the kth user. The Nc-chip PN sequence, {ak,0, … systems, the use of spreading codes in DS- ak,(Nc–1)}, is used to identify the kth user. It must UWB systems is solely for accommodating mul- be ensured that the pulse duration Tp is less than tiple users. the chip duration Tc, and NcTc must be less than the symbol repetition interval. TIME HOPPING FOR MULTIPLE ACCESS For binary DS-PPM systems, information bit In pulsed UWB systems, the pulse duty cycle is 1 is represented by a frame of pulses without any very small. Thus, the transmitter is gated off for delay, and information bit 0 is represented by the bulk of a symbol period. Multiple access can the same frame of pulses but with a delay τ rela- be implemented by employing appropriately cho- tive to the time reference. Let K be the number 0 sen hopping sequences for different users to of users in the system. Thus, bk,n is set to 1 and

82 IEEE Communications Magazine • February 2005

the information bit of the kth user is carried by 1 L bk,n in such a system. For binary DS-PAM sig- The physics-based 1 hTurin(τ )()= ∑ A nδτ− τ n nals, bk,n is set to 1 and the information bit of 0 n=1 unified model forms the kth user is carried by bk,n. Similar to single-user PPM signaling, the is a special case of h(τ) defined above when a new basis for most commonly used DS-PPM scheme is the hn(τ) = δ(τ). Pulse distortion is not as severe future system and orthogonal signaling scheme for which the DS for indoor applications such as those targeted by spread UWB pulse is orthogonal to its time- IEEE 802.15.3a, but it could cause serious prob- signal processing shifted version. There also exists an optimal time lems for IEEE 802.15.4a that adopts a special models, as Turin’s shift for a DS-PPM scheme. Let the sequence of form of h(τ) [6]. In [6] pulse distortion is the model does for pulses used to represent bits 1 and 0 of the kth same for each path (i.e., hn(τ) = h0(τ) for all n). user be Sk,1(t) and Sk,0(t) = Sk,1(t – τ), where Note that in [6] the older notion of frequency narrowband systems. ∞ 2 Sk,1(t), Sk,0(t) have unit energy (i.e., ∫ –∞ Sk,j(t)dt dependency [7, earlier references therein] is The physical origins = 1, j = 0, 1). For the orthogonal signaling used to characterize pulse distortion. From a of pulse distortion scheme, τortho is chosen to be such that theoretical point of view, the notion of per-path ∞ ∫ –∞ Sk,1(t)Sk,1(t – τortho)dt = 0. The orthogonality pulse distortion provides a framework to unify due to diffraction of condition may be met for a number of choices of the deterministic physical models of time a UWB pulse by τ (0 < τ < Tb), and the minimum of those val- domain electromagnetics. The physics-based ues is usually chosen as the time shift. For opti- unified model forms a new basis for future sys- some simple mally spaced signaling, the time shift τopt is tem and signal processing models, as Turin’s structures were chosen in the range 0 < τopt < Tp such that model does for narrowband systems. The physi- cross-correlation between the symbols is mini- cal origins of pulse distortion due to diffraction well understood. mized. of a UWB pulse by some simple structures were well understood. Different from a narrowband system, a UWB RECEIVER UWB system must include antennas as pulse DESIGN AND CHALLENGES shaping filters. In addition, antennas act as different pulse shaping filters for different angles. CHANNEL MODELS Due to unpredictable arriving angles of multi- UWB represents a major shift in design consid- path, antennas “distort” or “shape” the trans- erations in terms of signal propagation. A con- mitted pulses differently for different paths, as ventional narrowband system typically operates experimentally observed. Thus, both antennas in less than 20 MHz bandwidth (GSM, wideband and propagation environments suggest channel CDMA [WCDMA], IEEE 802.11). However, model h(τ). One difficulty is to decouple the UWB systems operate using up to 7.5 GHz band- effect of highly dispersive antennas and highly width. Although current implementations pre- dispersive propagation environments such as clude direct use of such large bandwidth, one dense urban. There is an ongoing demand for must understand its significance. A radical dif- the solution of basic pulse propagation prob- ference of a UWB channel from a narrowband lems to permit propagation models to be based channel is the number of resolvable multipaths. on sound physical principles. This is called the With a 0.167 ns multipath resolution (about 6 physics-based modeling framework [7]. The GHz bandwidth), more than 30 significant paths coarse but rapid results produced by empirical exist in a typical indoor environment. Collecting models represent another extreme. With the all multipath energy for reliable detection repre- high resolution of UWB signals and small area sents a major challenge for UWB receivers. In of interest, there is an increasing trend toward addition, maximum excess delay values of greater the physics-based framework, which facilitates than 60–70 ns (root mean square, RMS, delay the understanding of the theoretical founda- of 20–25 ns) were commonly observed for indoor tion and ultimate limits imposed by electro- environments. For high-speed short-range appli- magnetic wave theory and communications cations such as 100 MHz (10 ns symbol repeti- theory. tion interval), the multipath delay will lead to at least six or seven overlapping symbols. This sym- INTERFERENCE ISSUES bol overlapping requires mitigation via (symbol- Interference is one of the major issues in the rate) equalization followed by the matched filter receiver design for a pulsed UWB multiple (RAKE receiver). access system. The first problem is intersymbol Another potential problem of UWB is per- interference (ISI) caused by multipath delay. path pulse distortion. Mathematically the gener- In order to illustrate the ISI problem, Fig. 2 alized channel model can be expressed by shows the simulated received signals when two pulses of 0.5 ns duration spaced 25 ns apart are L hAh()ττδττ=− ()*( ), transmitted and go through a multipath chan- ∑ n n nel. The channel RMS delay spread is 25 ns, n=1 and it is assumed there is a resolvable path in where L generalized paths are associated with each bin of one bit duration. For a channel amplitude An, delay τn, and per-path impulse with a given delay spread, a higher data rate response hn(τ) ⋅ hn(τ) represents an arbitrary will result in a lower pulse repetition interval, function that has finite energy. Symbol * causing a larger amount of ISI. A complex denotes convolution, and δ(x) is the Dirac equalizer is probably impractical in a real sys- Delta function. Statistical parameterization of tem. Thus, ISI, rather than the bandwidth, will hn(τ) is a challenging task. Turin’s model eventually limit the achievable data rates in a expressed as pulsed UWB system.

IEEE Communications Magazine • February 2005 83

Figure 4 shows the normalized auto-correla- 0.6 tion function of the impulse response of a channel Rhh(t)/Rhh(0). It is interesting to notice that the first peak has an energy above 50 percent of 0.4 the total energy. We can think of each pulse as a chip for a spread spectrum system, and the chan- 0.2 nel impulse response can be regarded as a pseudo-noise code where the width of multipath delay spread corresponds to the length of the 0 PN code. This observation intuitively suggests that it is simpler to base the detection on the d signal

ive auto-correlation Rhh(t) rather than the noisy h(t). -0.2 If the first tap is used (shown in Fig. 4), the col- Rece lected energy is above 50 percent of the total -0.4 energy. SUBOPTIMUM RECEIVER STRUCTURES -0.6 Let us simplify optimum receiver structures in two directions. First, the matched filter can be simplified in a suboptimum manner. The out- -0.8 0 10 20 30 40 50 60 70 80 put of the matched filter is the auto-correlation of y(t). Since the per-path pulse waveform Time (ns) has been included in the channel impulse, the resultant matched filter is sometimes called a n Figure 2. A simulated received signal for two transmited UWB pulses (0.5 ns generalized RAKE receiver structure, of which duration) spaced 25 ns apart. Altes’ receiver structure is a special case. When per-path pulse distortion is included, the suboptimum implementation of the In a multi-user scenario, multiple access matched filter is the RAKE receiver structure. interference (MAI) for either the DS-UWB or In the absence of per-path pulse distortion, TH-UWB system must be considered, just like the matched filter is identical to the RAKE in a conventional CDMA system. Since a series receiver of Price and Green (1958). The trans- of pulses are used to represent one symbol in a mitted reference scheme with an auto-correla- DS-UWB (PAM or PPM) system, many multi- tion receiver [9], the differential scheme with path components are received within the obser- energy detector [10], and time reversal method vation window of a particular symbol, causing [11] are special suboptimum forms to approxi- interpath interference (IPI). If a RAKE-type mate the auto-correlation of y(t). Second, sub- receiver is employed, each path at the input of optimum linear equalizers such as zero-forcing, the combiner is corrupted by its preceding and decision feedback, and adaptive equalization following paths, which eventually limits perfor- can be used to replace the optimum MLSE. mance. For TH multiple access, a larger num- The receiver structure of RAKE plus MMSE ber of users in the system causes a higher equalizer, as proposed in IEEE 802.15.3a, is a probability of collision. Also, the randomly simplified implementation of the optimum hopped chips placed at the latter portion of a receiver shown in Fig. 3. symbol interval may cause severe ISI to other It is found that per-path pulse distortion users. Finally, a system’s robustness to inten- affects the system performance through the tional jamming is also an important design output of the matched filter. Interpulse over- consideration [8]. lapping combined with per-path pulse distortion makes receiver design very challenging. OPTIMUM RECEIVER STRUCTURE With suitable manipulation, the structure of We incorporate the per-path impulse response Fig. 3 and its suboptimum forms are valid for into the optimal receiver when ISI is present. It multi-user detection. The structure of Fig. 3 has been shown for a narrowband system that forms a unified framework that connects the the optimum receiver structure is illustrated in physics-based time domain channel model [7] Fig. 3. with detection theory. If the number of overlap- The received signal r(t) = y(t) + n(t) first ping symbols is less than 10, the MLSE (in Fig. passes through the matched filter y*(–t) fol- 3) may even be feasible for state-of-the-art sig- lowed by a sampler at a sampling rate of 1/Ts. nal processing capability. The composite signal y(t) is expressed as y(t) = x(t) * h(t), where x(t) is the transmitted signal RECEIVER CHALLENGES and h(t) is the generalized channel model. The High-Sampling-Rate ADC — Most existing sampled sequence is further processed by a detection schemes require an ADC that operates maximum likelihood sequential estimator at a minimum of the Nyquist rate. For high-per- (MLSE) detector that was first studied by For- formance design, the sampling rate is in the ney (1972). This receiver structure is optimal in multi-gigahertz range for UWB signaling (mini- terms of minimizing the probability of transmis- mum bandwidth is 500 MHz). In addition to the sion error in detecting the information extremely high sampling frequency, the ADC sequence. If y(t) is of finite energy, the struc- must support a relatively high resolution (e.g., ture of Fig. 3 can immediately be applied to greater than 4-6 bits) to resolve signals from nar- UWB receiver design. rowband interference. Such a speed mandates

84 IEEE Communications Magazine • February 2005

the use of an interleaved flash ADC, which tends to be power hungry with a power scaling expo- nentially increasing with bit precision. Although Received achievable with today’s complementary metal signal Matched MLSE filter Sampler (Viterbi oxide semiconductor (CMOS) technology, such y*(-t) algorithm) an ADC must be avoided for low-power opera- Output tions. One technique to avoid using such a high- Clock data speed ADC is to implement the correlator in the r(t)=y(t)+n(t) t=kT analog domain. One technical challenge associated with the analog implementation method is n the difficulty in obtaining the receiver template Figure 3. Optimal receiver structure in the presence of ISI in an additive waveform and deriving the precise timing of white Gaussian noise channel. symbols and received paths. In the digital implementation, a bank of correlators that are delayed Autocorrelation channel relative to one another by a fraction of the pulse 1.2 duration can be applied to correlate with the received digitized signal, and the local peaks cor- CM3, channel 1, energy ratio = 50% responding to the possible received pulses are 1 located. With analog implementation, this is difficult to achieve because some sort of analog delay units are needed, and novel multipath 0.8 tracking methods are needed. Another approach is to represent the samples using fewer bits. One 0.6 bit (monobit) representation of samples, a traditional approach, finds new application in UWB to reduce ADC speed. 0.4

Multipath Capture — One of the advantages Normalized amplitude of pulsed UWB communication is its ability to 0.2 resolve individual multipath components. How- ever, the large number of resolvable paths in such a system makes it unrealistic to employ 0 the traditional RAKE receiver to capture a significant portion of the energy contained in the -0.2 received multipath components, and most exist- -100 -80 -60 -40 -20 0 20 40 60 80 100 ing receivers must resort to a partial or selec- Time (ns) tive RAKE for multipath combining. A partial or selective RAKE can practically combine up n to a few paths, which represent only a small Figure 4. Auto-correlation of a channel impulse response Rhh(t) = h(t) * percentage (e.g., less than 10 percent) of the h(–t) for a CM3 channel. total received signal energy in a non-line-of- sight (NLOS) environment. A RAKE with more than three to five paths will lead to an expo- which the received waveform is averaged. How- nential increase in complexity because multi- ever, the analog delay units needed to derive path acquisition, multipath tracking, and the receiver template could be costly in imple- channel estimation consume too much process- mentation. ing resources. The insufficient multipath energy In a time reversal scheme [11] the channel captured by the receiver results in poor system impulse response is used as a prefilter at the range and almost no tolerance to ISI caused by transmitter. After the encoded information bits multipath delay. Suboptimal schemes such as pass through the channel, the response of the the transmit reference [9] or differential scheme channel is the auto-correlation of the channel [10] can perform successful multipath energy impulse response as shown in Fig. 4. A pilot sig- capture and detection without requiring chan- nal can be used to sound the channel such that nel estimation. The primary drawback of these the transmitter has the channel impulse response receivers is the significant performance degra- available for precoding. This scheme is used to dation associated with employing noisy received shift the design complexity from the receiver to signals as the reference signals for data detec- the transmitter [11]. In the time reversal scheme, tion. The net result is that such a receiver could a signal is precoded such that it focuses in both perform much worse than a partial or selective time and space at a particular receiver. Due to RAKE receiver, which implements only a small temporal focusing, the received power is concen- number of fingers. The decision feedback auto- trated within a few taps, and the task of equaliz- correlation receiver [12] overcomes the draw- er design becomes much simpler than without backs of these receivers. In this scheme, the focusing. received signal waveform is delayed using analog delay units. The respective symbol decisions UTURE UTLOOK are applied to data-demodulate the outputs of F O these delay units, which are then summed up, Industrial standards such as IEEE 802.15.3a forming the receive template waveform. The (high data rate) and IEEE 802.15.4a (very low template waveform quality can be adjusted by data rate) have been introduced based on UWB changing the number of symbol intervals over technology. Department of Defense (DOD)

IEEE Communications Magazine • February 2005 85

UWB systems are different from commercial chunk of unlicensed spectrum. In the long run The promising systems in that mobility is a significant concern. UWB will change some basic thinking in the UWB has found a new application for lower- standard wireless textbooks. For one thing, the future of UWB for data-rate moderate-range wireless communica- concept of gigabit wireless communications is wireless multiple tions, illustrated by IEEE 802.15.4a and DOD enabled by UWB. One must admit that people access is optimistic, systems enabling joint communication and rang- still know little about UWB compared to its ing capabilities unique to UWB. Unlike the narrowband counterpart developed in the last simply because of indoor environment in 802.15.3a (wireless per- half century. the unprecedented sonal area network, WPAN), the new environments for sensors, IEEE 802.15.4a, and DOD REFERENCES huge chunk of unli- systems will be very different, ranging from [1] W. Zhuang, X. Shen, and Q. Bi, “Ultra-Wideband Wireless censed spectrum. dense foliage to dense urban obstructions. Pulse- communications,” Wireless Commun. and Mobile Comp., In the long run based UWB has gained momentum again for Special Issue on Ultra-Broadband Wireless Communications some new applications after a short setback for the Future, vol. 3, Sept. 2003, pp. 663–85. UWB will change [2] J. Balakrishnan, A. Batra, and A. Dabak, “A Multi-Band when OFDM-based UWB was popular in IEEE OFDM System for UWB Communication,” Proc. 2003 some basic thinking 802.15.3a. The mainstream papers in the litera- IEEE Conf. UWB Sys. and Tech., Nov. 2003. ture deal with pulse-based UWB systems. One [3] J. R. Foerster, “The Performance of a Direct-Sequence in the standard reason may be the former was not sufficiently Spread Spectrum Ultra-Wideband System in the Pres- wireless textbook. ence of Multipath, Narrowband Interference and Mul- understood compared to OFDM. It should not tiuser Interference,” Proc. 2002 IEEE Conf. UWB Sys. be regarded as a sign that industry should favor and Tech., May 2002, pp. 87–91. the former. It is beyond the scope of this article [4] M. Z. Win and R. Scholtz, “Ultra-Wide Bandwidth Time- to comment on the pros and cons of the two hopping Spread Spectrum Impulse Radio for Wireless Multiple-access Communications,” IEEE Trans. types of systems. Commun., vol. 48, Apr. 2000, pp. 676–91. Most existing work regarding the CDMA sys- [5] I. Guvenc and H. Arslan, “On the Modulation Options tem seems to still be valid. There exist some for UWB Systems,” Proc. IEEE MILCOM ’03, Oct. 2003, unique issues: pp. 892–97. [6] IEEE 802.15.4a, “Status of Models for UWB Propagation •Pulse distortion caused by antennas and Channel, Final Report,” http://www.ieee802.org/15/pub/ dispersive propagation environments is critical. TG4a.html, Sept. 2004. The challenge also comes from the fact that it [7] R. C. Qiu, “A Study of the Ultra-Wideband Wireless is very difficult to decouple the effect of anten- Propagation Channel and Optimum UWB Receiver Design,” IEEE JSAC, Special Issue on UWB Radio in Mul- nas with propagation environments. A typical tiple Access Wireless Communications, vol. 20, no. 9, pulse-based system usually does not have a Dec. 2002, pp. 1628–37. front-end filter like its narrowband counter- [8] B. Sadler and A. Swami, “On the Performance of part, so pulse distortion serves as a pulse shap- Episodic UWB and Direct-sequence Communication Sys- tem,” IEEE Trans. Wireless Commun., to appear. ing filtering effect that must be included in the [9] J. D. Choi and W. E. Stark, “Performance of Ultra-Wide- system model band Communications with Suboptimal Receiver in •The receiver structures need more evalua- Multipath Channels,” IEEE JSAC, vol. 20, no. 9, Dec. tion. A RAKE structure is often suggested. 2002, pp. 1754–66. [10] S. Roy et al., “Ultra-Wideband Radio Design: The Promise When pulse distortion occurs, the generalized of High-Speed, Short Range Wireless Connectivity,” Proc. RAKE structure is required for optimum recep- IEEE, vol. 92, no. 2, Feb. 2004, pp. 295–311. tion. Channel estimation and synchronization [11] S. M. Emami et al., “Predicted Time Reversal Perfor- required for RAKE are too difficult for short mance in Wireless Communications Using Channel Measurements,” to appear, IEEE Commun. Lett. UWB pulses, so some suboptimum alternatives [12] S. Zhao, H. Liu, and Z. Tian, “A Decision Feedback are investigated. To avoid these difficulties the Autocorrelation Receiver for Pulsed Ultra-Wideband Sys- TR philosophy seems to be very promising, espe- tems,” Proc. IEEE Rawcon ’04, Sept. 2004. cially for the DOD’s adverse environments and IEEE 802.15.4a. But the noisy template problem BIOGRAPHIES and weak capability to counteract intentional HUAPING LIU received his B.S. and M.S. degrees from Nan- interference are two basic problems. jing University of Posts and Telecommunications, P. R. •UWB antennas are important and often China, in 1987 and 1990, respectively, and his Ph.D. require co-design with other parts of the base- degree from New Jersey Institute of Technology, Newark, in 1997, all in electrical engineering. From July 1997 to band. Communications imposes stringent August 2001 he was with Lucent Technologies, New Jersey. requirements on antennas. Low-cost omni anten- Since September 2001 he has been an assistant professor nas are challenging. UWB multiple antennas are with the Department of Electrical and Computer Engineer- rarely understood. ing at Oregon State University. His research interests include design and performance analysis of wireless communication systems, and communication techniques for CONCLUSION multi-user time-varying environments with applications to cellular and indoor wireless communications. Current spe- In this article we provide a comprehensive cific interests include ultra-wideband techniques, detection and signal processing for MIMO systems, and spread spec- review of UWB multiple access and modula- trum communications. tion schemes and compare them with narrowband radios. We also outline issues with UWB ROBERT C. QIU is currently an associate professor in the signal reception and detection, and explore Department of Electrical and Computer Engineering, Center for Manufacturing Research, Tennessee Tech University. various suboptimal low-complexity receiving Before joining TTU, he spent eight years in industrial labo- schemes. Its focus is on balancing the treat- ratories including Bell Laboratories. His research interests ments of theoretical and practical designs. We include wireless communications and networking, ultra- also mix the needs of two major different appli- wideband radio propagation, signal processing, and system prototyping. He has published over 40 journal and confer- cations (IEEE 802.15.3a and 4a). The future of ence papers as well as 10 pending U.S. patents. He was UWB for wireless multiple access is optimistic, founder, CEO/president, and chief scientist of Wiscom simply because of the unprecedented huge Technologies Inc., a 3G WCDMA chipset designer and ven-

86 IEEE Communications Magazine • February 2005

dor that was sold to Intel in 2003. He is an Editor for IEEE Transactions on Vehicular Technology, and Wireless Com- munications and Mobile Computing (Wiley). He is also a guest editor of a special issue of IEEE JSAC on UWB, and editor of the book Ultra-Wideband Wireless Communica- tions (Wiley, 2005).

XUEMIN (SHERMAN) SHEN [SM] ([email protected]) received a B.Sc. (1982) degree from Dalian Maritime Uni- versity, China, and M.Sc. (1987) and Ph.D. (1990) degrees from Rutgers University, New Jersey, all in electrical engineering. From September 1990 to September 1993 he was first with Howard University, Washington D.C., and then the University of Alberta, Edmonton, Canada. Since Octo- ber 1993 he has been with the Department of Electrical and Computer Engineering, University of Waterloo, Cana- da, where he is a professor and associate chair for gradu- ate studies. His research focuses on mobility and resource management in interconnected wireless/wireline networks, UWB wireless communications systems, wireless security, and ad hoc and sensor networks. He is a coauthor of two books, and has published more than 150 papers and book chapters in wireless communications and networks, control, and filtering. He was Technical Co-Chair for the IEEE GLOBECOM ’03 Symposium on Next Generation Networks and Internet and ISPAN ’04, and is Special Track Chair of the 2005 IFIP Networking Conference. He serves as Associ- ate Editor for IEEE Transactions on Wireless Communica- tions; IEEE Transactions on Vehicular Technology; Computer Networks; Dynamics of Continuous, Discrete and Impulsive — Series B: Applications and Algorithms; Wireless Commu- nications and Mobile Computing (Wiley); and International Journal Computer and Applications. He has also served as Guest Editor for IEEE JSAC, IEEE Wireless Communications, and IEEE Communications Magazine. He received the Pre- mier's Research Excellence Award (PREA) from the Province of Ontario, Canada for demonstrated excellence of scientif- ic and academic contributions in 2003, and the Distin- guished Performance Award from the Faculty of Engineering, University of Waterloo, for outstanding contributions in teaching, scholarship, and service in 2002. He is a registered Professional Engineer of Ontario, Canada.

IEEE Communications Magazine • February 2005 87