1

A UWB Radio Network Using Multiple Delay Capture Enabled by Time Reversal
Nan Guo, Robert C. Qiu, and Brian M. Sadler

Abstract An uplink multiple-access method for a centralized ultra-wideband (UWB) radio network is studied. The multipleaccess method is based on a phenomenon of multiple delay capture and is enabled by time reversal technique. Multiple delay capture is a random multiple-access method that can be viewed as a time-of-arrival (TOA) division multiple-access. When received signals are of low duty cycle, the chance of collision is low, and the base station is able to communicate with different users simultaneously. At much higher data rate compared to the narrow band scenarios, the delay spread of a UWB channel is relatively large, and the multipath-distorted signal may not be of low duty cycle. To improve the multiple-access efciency, time reversal in conjunction with transmitter-side antenna array is considered to shorten the UWB channel impulse response (CIR). Time reversal takes advantage of rich scattering environments to achieve signal focusing, while the antenna array further enhances signal focusing. Our focus in this paper will be on evaluating the collision-free probabilities using both analysis and simulation. A lower bound is derived and it is actually a good estimate of the probability that all received signals are collisionfree. Finally, quantitative results are provided, showing that the normalized system throughput increases fast as the number of users increases. Index Terms UWB, delay capture, time reversal.

I. I NTRODUCTION Recently time reversal [1] has received attention in ultrawideband (UWB) radio community [2]-[8]. Time reversal can be viewed as a transmitter-side signal processing technique that takes advantage of rich multipath scattering to achieve signal focusing both temporally and spatially. Time reversal uses time-reversed channel impulse response (CIR) to set up a transmitter-side pre-lter, leading to an equivalent CIR with sharp prole. Although this pre-lter setting would not result in a delta-function-like signal at the receiver, its processing is much simpler than inverse ltering that aims at ideal channel inversing. A UWB channel is typically contains a large number of multipath components, and UWB with time reversal is indeed a good combination. Furthermore, the spatio-temporal focusing can be effectively enhanced using transmitter-side antenna array [3][4][7][8]. UWB time reversal enables a few unique techniques. Time reversals temporal focusing feature can reduce the impact of ISI, while its spatial focusing feature can not only enable downlink spatial-division multipleaccess [4], but also provide a certain level of security for a
Nan Guo and Robert C. Qiu are with the Center for Manufacturing Research, Tennessee Technological University (TTU), Cookeville, TN 38505, USA (e-mail: nguo@tntech.edu, rqiu@tntech.edu). R.C. Qiu is also an associate professor of the Department of Electrical and Computer Engineering, TTU. Brian M. Sadler is with US Army Research Laboratory, AMSRDARL-CI-CN, 2800 Powder Mill Rd., Adelphi, MD 20783. (e-mail: bsadler@arl.army.mil)

system with symbol duration much less than the multipath channel excess delay spread [6]. Low-complexity multiple-access is badly demanded for lowcost applications such as sensor networks. There have been many existing multiple-access techniques including a recently proposed rate-division multiple-access (RDMA) [9], a random multiple-access working for both downlink and uplink. Our in-ofce test has shown that the downlink spatial-division multiple-access works very well. This multiple-access utilizes time reversals spatial focusing characteristic to separate users, and similar to RDMA, it does not need additional radio resource such as orthogonal codes. Can a group of users transmit data to a base station at the same data rate without using any orthogonal function? In this paper we investigate an uplink random access method that takes advantage of time reversals temporal focusing characteristic. The proposed multiple-access method is based on a phenomenon called multiple delay capture [10]-[13]. This is a random multiple-access that can be viewed as time-of-arrival (TOA) division multiple-access, and its performance depends on the probability that signals collide (or overlap) at the receiver. To reduce the chance of collision, a multiple input single output (MISO) time reversal technique is considered to shorten the UWB CIR. With MISO time reversal, a low duty cycle transmitted waveform would trigger off a low duty cycle waveform at the receiver, which is in favor of multiple delay capture. Our focus in this paper is on analysis of the collision probability and throughputs. The rest of this paper is organized as follows. Section II describes the system conguration. The collision-free probabilities are studies in Section III, followed by some numerical results and remarks in Section IV. II. S YSTEM D ESCRIPTION Consider a cluster of K users that share a single base station. Without any coordination, these K users simultaneously send information via uplink to the base station, and at the base station all received single-user signals have comparable power levels. Assume the base station has delay capture ability to lock to the signals with different time of arrivals (TOAs) [10][13]. Also consider pulse based signals with pulse width Tp and pulse repetition interval T (> Tp ) for all users. Modulation scheme can be antipodal modulation or pulse amplitude modulation (PAM). A MISO time reversal conguration is conceptually illustrated in Fig.1, where there are M transmitter antenna elements targeting at the receiver antenna, cm (t) is the pre-lter connected with the antenna element m, and hm (t) is the CIR for the channel associated with an antenna pair of the m-th transmitter antenna element and the receiver antenna. The system is set up during a handshaking phase. Letting ptx (t) be

0.15

Tx
0.1

h1 (t)
Amplitude (Volt)

Main Lobe

wtx,1 (t) h2 (t) wtx,2 (t) hM (t)

0.05

Rx

wtx,M (t)
0.1 0 20 40 60 80 100 120 140

Fig. 1.

MISO conguration.

.
0.05

Time (ns)

Tm
Fig. 3. Rectangular signal model.

1 0 0 1 0 1 0 1 0 1 0 1 0 1 0 1

Fig. 2.

Received waveforms of MISO time reversal.

collision

the transmitted pulse waveform, ideally the received symbol waveform from antenna element m can be represented by [hm (t) hm (t)] [ptx (t) ptx (t)], implying the main lobe in the received waveform almost has the same prole as the autocorrelation waveform ptx (t) ptx (t). Indeed, the mainlobe prole is not greatly affected by undesired distortion. Shown in Fig.2 is a received pulse waveform measured in our ofce, where a 4-element linear antenna array is employed to form a 4 by 1 MISO time reversal conguration, and the pulse distortion introduced by the antennas has been taken into account. Because of time reversals temporal focusing characteristic, each received single-user waveform has a main lobe of width Tm . From multiple-access point of view, we model the received single-user waveform as a rectangular waveform with duty ratio Tm : T (see Fig.3), which will make our description and analysis easier. Because of asynchronous access fashion, the signals can collide with each other and a proper protocol and some error detection mechanism are necessary to maintain the network. Theoretically speaking, if some single-user signals observed at the base station are at least Tm seconds apart, these users are collision-free and can communicate with the base station simultaneously; while the collided signals would result in serious errors. Feasibility of this TOA-division multiple-access depends on a number of issues. Quantifying the collision-free probability is one of important task that can guides system design. III. C OLLISION -F REE P ROBABILITIES We do not have a formula for calculating the collision-free probabilities. The Monte-Carlo simulation and a lower-bound

will be adopted to quantify the collision-free probabilities. Dene the collision-free probability as Pcf (k | K ) = P r(k out of K received signals are collision f ree), 0kK (1)

where a collision means two rectangular waveforms overlap. It is obvious that Pcf (K 1 | K ) = 0. (2)

Pcf (K | K ) and Pcf (K 2 | K ) are two dominants among all of Pcf (k | K )s. From a set of simulation results given in Table I it can be veried that when Pcf (K | K ) is above 0.5 we have:
K 3

Pcf (K | K ) + Pcf (K 2 | K )
k=0

Pcf (k | K ) 0, (3)

k3 and Pcf (K | K ) + Pcf (K 2 | K ) 1.

(4)

In the following we will estimate the special probability Pcf (K | K ). Let rg be the received signal from user g , 1 g K . Dene an event CF g by
g 1

CF g =
i=1

{rg is collision f ree with ri }.

(5)

Table I. Simulation Results of Collision-Free Probabilities Pcf (k | K ) for Tm /T = 1/128 k=0 1 2 3 4 5 6 7 8 9 10 K=0 1 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 2 2.169 102 0 9.783 101 0 0 0 0 0 0 0 0 3 3.297 104 4.451 102 0 9.552 101 0 0 0 0 0 0 0 4 1.199 103 1.798 103 8.980 102 0 9.072 101 0 0 0 0 0 0 5 8.704 105 3.682 103 3.421 103 1.396 101 0 8.532 101 0 0 0 0 0 6 3.886 105 2.798 104 9.793 103 6.661 103 1.967 101 0 7.865 101 0 0 0 0 7 1.455 105 3.768 104 1.007 103 2.139 102 1.033 102 2.526 101 0 7.143 101 0 0 0 8 5.095 106 6.777 105 1.394 103 2.600 103 3.918 102 1.541 102 3.042 101 0 6.372 101 0 0 9 1.386 106 5.794 105 2.774 104 3.761 103 5.336 103 6.330 102 2.077 102 3.469 101 0 5.596 101 0 10 7.950 107 1.489 105 2.520 104 8.387 103 8.481 103 9.538 103 9.346 102 2.626 102 3.788 101 0 4.824 101

Dene another relevant event MCF g , called the mutually collision-free event, as MCF g = {any pair of rg , rg1 , , r1 are collision f ree}. (6) It can be veried that CF i MCF i1 , MCF i = CF 2 , i3 , i=2

Corollary: If the TOA of the g th user in the system is uniformly distributed on [0, T ) and g 3, then P r(CF g |MCF g1 ) is lower bounded by P r(CF g |MCF g1 ) 1 2(g 1) Tm . T (11)

(7)

Therefore, the event that all K received signals are collisionfree can be expressed as MCF K = CF K MCF K 1 = CF K CF K 1 MCF K 2 = = CF K CF K 1 CF K 2 CF 2 , (8) and the probability Pcf (K | K ) is given by Pcf (K | K ) = P r(MCF K ) = P r(CF K MCF K 1 ) = P r(CF K |MCF K 1 )P r(MCF K 1 ) = P r(CF K |MCF K 1 )P r(CF K 1 MCF K 2 ) = = P r(CF K |MCF K 1 )P r(CF K 1 |MCF K 2 ) P r(CF K 2 |MCF K 3 ) P r(CF 3 |MCF 2 )P r(CF 2 ). (9) Without loss of generality, in the following we assume all K TOAs are uniformly distributed over [0, T ). When there are only two users the collision-free probability is Tm T 2Tm =12 . (10) P r(CF 2 ) = T T Since we have not found any approach to calculate P r(CF g |MCF g1 ) exactly, we develop a lower bound to estimate it.

Proof: Let g be the TOA of the signal rg . MCF g1 stands for that any pair of 1 , 2 , 3 , , g1 are at least Tm apart from one-another. If we look at fg (x|MCF g1 ), the pdf of g conditioned on MCF g1 , it has to be a two-valued function on [0, T ), with g 1 zero-value regions of width Tm . Shown in Fig.4 is an example of such conditional pdf , where the g 1 zero-value regions, or notches, are because of the collision-free condition MCF g1 , and to satisfy MCF g1 , g is not allowed to be in theses regions. P r(CF g |MCF g1 ) is related to fg (x|CF g |MCF g1 ), the pdf of g conditioned on {CF g |MCF g1 }. Similarly, fg (x|{CF g |MCF g1 }) is a two-valued function on [0, T ), with some bigger zero-value regions. Note that g has to be at least Tm apart from the g 1 given TOAs to satisfy {CF g |MCF g1 }, implying that fg (x|{CF g |MCF g1 }) contains g 1 zero-value regions of width 2Tm (but some of then may overlap). Now let us consider two possible cases. Case 1: Among the g 1 given TOAs, any pair of adjacent TOAs are more than 2Tm apart from one-another. In this case, fg (x|{CF g |MCF g1 }) has g 1 non-overlapping notches of width 2Tm , corresponding to the g 1 given TOAs, meaning P r(CF g |MCF g1 ) = [T 2(g 1)Tm ]/T . Case 2: Among the g 1 given TOAs, at least there is one pair of TOAs separated at a distance larger than Tm but less than 2Tm . In this case, some the zero-value regions overlap, resulting in an increased probability; in other words, P r(CF g |MCF g1 ) > [T 2(g 1)Tm ]/T . Combining the cases 1 and 2, we have (11).

fg (x|M CFg1 ) x T

0
Fig. 4. Conditional pdf
1

//

0.9

K=4

0.8

Pcf (K | K )

0.7

0.6

K = 13

0.5

Simulation Lower bound

0.4

0.3 1

1.5

2.5

3.5 Tm /T

4.5

5.5 x 10

6
3

Fig. 5.

Behavior of Pcf (K | K ).

From (9)-(11), we obtain a lower bound for Pcf (K | K ):

K

The normalized per-user throughput should be /K . (14) can be upper-bounded as follows:

K

Pcf (K | K )
g =2

1 2(g 1)

Tm , K 2. T

(12) =

kPcf (k | K )
k=1 K 2

IV. N UMERICAL R ESULTS AND R EMARKS Shown in Fig.5 is Pcf (K | K ) vs. the parameter Tm /T based on both the theoretical lower bound and the MonteCarlo simulation. These results suggest that the lower bound is so tight that we can use the bound as a good estimate of Pcf (K | K ). An interesting phenomenon observed from Table I and Fig.5 is that Pcf (K | K ) decreases as K increases, while Pcf (K i | K ), i 2, increases with K . Assume all data packets retrieved from the collided signals are erroneous and have to be thrown out, thus the system throughput for K 2 is 1 T
K

= KPcf (K | K ) +
k=1

kPcf (k | K )
K 2

< KPcf (K | K ) + (K 2)
k=1

Pcf (k | K )

= KPcf (K | K ) + (K 2)[1 Pcf (K | K )] = (K 2) + 2Pcf (K | K ), K 2. (15)

This upper bound is tight if probabilities Pcf (k | K ), k < K 2, are sufciently small. Recall (12), can be approximated by
K

kPcf (k | K ) [bits/sec].
k=1

(13) (K 2) + 2

1 2(g 1)
g =2

We further dene , the normalized system throughput that is independent of T :

K

Tm . T

(16)

=
k=1

kPcf (k | K ).

(14)

Fig.6 shows the throughputs which are based on either the collision-free probabilities given in Table I, or the approximation in (16). The normalized system throughput indeed increases fast with K . Also, from these results it appears that

Tm/T = 1/128

Approximation in (16)

Normalized Throughputs

Normalized System Throughput

Based on Table I

Normalized PerUser Throughput (based on Table I)

0 1

10

Number of Users

Fig. 6.

Normalized throughput for Tm /T = 1/128.

the approximation is very accurate when K is smaller, but exhibits overestimated as K increases. We have investigated collision-free probabilities by assuming that all received signal have an effective width Tm in an observation window of length T . The collision-free probability Pcf (k | K ) only depends upon the duty ratio Tm : T and K , the total number of users in the system. In particular, Pcf (K | K ) can be well estimated by using the bound (12). Practically, with UWB MISO time reversal, Tm is a few of nanoseconds. To accommodate ten users and keep an acceptable per-user throughput (say, /K 0.9), the pulse repetition interval T should be in the order of several hundreds of nanosecond, converting to a bit rate of one to a few Mbps (assuming one pulse per bit). Low duty cycle pulsed signaling is the fundamental condition for the TOA-division multiple-access to function. The proposed method can be applied to covert pulsed communication systems such as the one investigated in [14]. Based on the above analysis, in a low-duty-cycle direct-sequence (DS) system, a group of users can share a common spreading code to communicate in an uncoordinated manner, which is a way to save precious resource of spreading codes. ACKNOWLEDGMENT This work was supported in part by the Army Research Laboratory through a Short Term Innovative Research (STIR) grant (W911NF-06-1-0349), and by the Army Research Ofce through a Defense University Research Instrumentation Program (DURIP) grant (W911NF-05-1-0111). R EFERENCES
[1] M. Fink, Time reversal of ultrasonic eldsPart I: Basic principles, IEEE Trans. Ultrason., Ferroelec. Frequency Control, vol.39, no.5, pp.555-566, Sept. 1992.

[2] T. Strohmer, M. Emami, J. Hansen, G. Pananicolaou and A. J. Paulraj, Application of time-reversal with MMSE equalizer to UWB commplications, in Proc. Globecom04, Dallas, TX, Dec. 2004, pp.3123-3127. [3] P. Kyristi, G.G. Pananicolaou, P. Eggers and A. Oprea, MISO time reversal and delay spread compression for FWA channels at 5 GHz, IEEE Ant. and Wireless Prop. Lett., vol.3, pp.96-99, 2004. [4] H. T. Nguyen, J.B. Andersen and G.F. Pedersen, The potential use of time reversal techniques in multiple element antenna systems, IEEE Com. Lett., vol.9, pp.40-42, Jan. 2005. [5] A.E. Akogun, R.C. Qiu and N. Guo, Demonstrating time reversal in ultra-wideband communications using time domain measurements, in Proc. 51st International Instrumentation Symposium, Knoxville, TN, May 8-12, 2005. [6] N. Guo, R.C. Qiu, and B.M. Sadler, An ultra-wideband autocorrelation demodulation scheme with low-complexity time Reversal enhancement, in Proc. IEEE MILCOM05, Atlantic City, NJ, Oct. 17-20, 2005. [7] R.C. Qiu, C. Zhou, N. Guo, and J.Q. Zhang, Time reversal with MISO for ultra-wideband communications: experimental results, IEEE Antenna and Wireless Propagation Letters, to appear, 2006. [8] N. Guo, J. Zhang and R. Qiu, UWB MISO Time Reversal With Energy Detector Receiver Over ISI Channels, submitted to The 2006 IEEE International Conference on Ultra-Wideband. [9] M. Weisenhorn and W. Hirt, Uncoordinated rate-division multiple-access scheme for pulsed UWB signals, IEEE Trans. on Vehicular Tech., vol.54, No.5, pp.1646-1662, Sept. 2005. [10] D. Davis and S.A. Gronemeyer, Performance of slotted ALOHA random access with delay capture and randomized time of arrival, IEEE Trans. Commun., vol. COM-28, pp. 703-710, May 1980. [11] M. Soroushnejad and E. Geraniotis, Probability of capture and rejection of primary multiple-access interference in spread-spectrum networks, IEEE Trans. Commun., vol. COM-39, pp.986-994, June 1991. [12] D. I. Kim, Multiple-capture performance of DS-SS packet radio systems with common spreading code, IEE Proceedings-I, vol.140, No.6, pp.471-, Dec. 1993. [13] K. Cheun, Optimum arrival-time distribution for delay capture in spread-spectrum packet radio networks, IEEE Trans. on Vehicular Tech., vol.46, No.4, pp.981-991, Nov. 1997. [14] B.M. Sadler, A. Swami, On the performance of episodic UWB and direct-sequence communication systems, IEEE Transactions on Wireless Communications, vol.3, pp.2246-2255, Nov. 2004.