You are on page 1of 5

Low DataRate UltraWideband Impulse Radio Communications for Wireless Medical Body Area Networks

HighlyReliable Noncoherent System Design


Igor Dotli c
Dependable Wireless Laboratory, Wireless Network Research Institute National Institute for Information and Communications Technology 3-4, Hikarino-oka, Yokosuka, 239-0847, Japan

Ryu Miura
Dependable Wireless Laboratory, Wireless Network Research Institute National Institute for Information and Communications Technology 3-4, Hikarino-oka, Yokosuka, 239-0847, Japan

dotlic@nict.go.jp ABSTRACT

ryu@nict.go.jp

between onbody and obody data hubs. Current standardization eorts in IEEE 802.15.6 group for BAN [1] resulted in a standard draft [10] that, among other Physical Layers (PHYs), include UltraWideband (UWB) PHYs; namely UWB ImpulseRadio (UWBIR) and Frequency Modulated UWB (FMUWB) PHYs. FMUWB [6] features good interference resistivity at very low data rates. Although ecient in interference suppression, FMUWB has a drawback of very low aggregate data rates in multiuser environments of less than 250 kbps per over 500 MHz wide UWB channel. Indeed, such low channel capacity imposes signicant limitations on the application of FMUWB systems. On the other hand, UWBIR holds a promise of personal radio with high interference and multipath fading resistivity and low power consumption. Interference and multipath fading resistivity are possible due to the high processing gain Categories and Subject Descriptors of UWBIR. Furthermore, due to impulse nature of UWB C.2 [COMPUTER-COMMUNICATION NETWORKS]: IR, low power consumption can be achieved through duty cycling of the RF frontend which is, in contemporary perGeneralData communications sonal radio systems, typically responsible for a large portion of overall transceiver power consumption. General Terms Although UWB band is unlicensed, these features of UWB IR make it a good candidate for low datarate medical BAN. Design, Performance, Reliability, Standardization Nevertheless, the aforementioned UWBIR features of high interference resistivity and low power consumption are hard Keywords to achieve simultaneously. On one hand, fully coherent systems like [5, 9] are highly interferenceresistant, but do not Impulse Radio, UltraWideband (UWB), Chirp, Pulse Comfully dutycycle transceiver RF frontend, have high analog pression, Interference suppression, Body Area Networks (BAN) timing estimation accuracy requirements and/or high digital backend complexity which addup to their high power 1. INTRODUCTION consumption. On the other hand, systems based on low complexity EnergyDetection (ED) like [13,15] feature low Onbody medical Body Area Network (BAN) is an emergpower consumption, but are highly sensitive to interference. ing technology of communication between sensor nodes loUWBIR systems with dierential detection like [8] reprecated on the human body and an onbody data hub, and sent a good compromise between low power consumption of ED based systems and high interferenceresistivity of coherent systems; symbol dierentiation is typically done without symbol signature estimation, which we termed Samplewise c 2011 Association for Computing Machinery. ACM acknowledges that DPSK (SDSPK) [2, 4]. Like ED systems, SDPSK systhis contribution was authored or co-authored by an employee, contractor tems can exploit a large portion of multipath energy with or afliate of the national government of Japan. As such, the government of low timingprecision requirements while attaining better inJapan retains a nonexclusive, royalty-free right to publish or reproduce this terference resistivity than ED. As analog delay lines for dearticle, or to allow others to do so, for Government purposes only. lays with order of magnitude of UWBIR symbol durations ISABEL 11 Barcelona, Catalonia, Spain In this paper we present receiver design considerations and performance analysis for the physical layer of low data-rate non-coherent UltraWideband Impulse Radio (UWBIR) communication systems. Such UWBIR systems are of particular interest in wireless medical Body Area Networks (BAN); in medical BAN the amount of trac is relatively low, however, reliability of communication under body-shadowing and interference conditions is of utmost importance. Two non-coherent modulation/detection schemes are treated analytically and numerically through noise and interference performance studies. Particular attention is given to the performance enhancement achieved by compression of the received signal by methods previously introduced by the authors.
Copyright 2011 ACM 978-1-4503-0913-4/11/10 ...$10.00.

are (at the current state of technology) hard to implement, SDPSK detection is foremost performed in the digital domain. Digital complexity of the SDPSK is proportional to the dimension of the signal in the detection, which typically increases with the symbolrate decrease. Furthermore, performance of both aforementioned noncoherent detection schemes ED and SDPSK deteriorates with the increase of the dimension of the signal in the detection. For the sake of lowering digital complexity of dierential detection receivers with classic DPSK detection, in [3] we proposed chirp pulse compression, which is compatible with contemporary analog integrated circuit technology, i.e. it does not need separate discrete analog components. This technique has potential to also improve performance of the aforementioned noncoherent detection techniques through lowering the dimension of the signal in detection [2, 4]. This is especially important in medical BAN, where at very low data rates received signals dimensions are high. For this reason, based on our proposal, IEEE 802.15.6 standard draft [10] included chirp pulse waveform. Thus, for both PPM and DPSK modulations, IEEE 802.15.6 UWBIR system designer has an option to implement chirp pulse compression in the receiver. Although BAN standard draft [10] prescribes mandatory (lowest) raw data rate of 490 kbps for UWBIR PHY, for some lifecritical applications like insulin pump control [16] we may foresee using lower data rates in order to increase the systems reliability. This work will investigate performance in the presence of noise and interference of aforementioned noncoherent detection schemes with and without chirp pulse compression in order to determine their usability in these lifecritical applications.

with carrier frequency f0T , duration Tw and frequency sweep of fc ; its analytical signal representation is c(t) = exp j2 f0T t + Kc 2 t 2 w(t, Tw /2, Tw /2). (1)

Function w(t, t1 , t2 ) is the rectangular window function with limits t1 and t2 and Kc is a chirping slope dened as fc . (2) Tw In this work we will use fc = 520 MHz, as prescribed by the IEEE 802.15.6 standard draft [10]. Signal c(t) (1) is DBPSK modulated like in [2,3] or binary PPM modulated [18] with the time shift T . IEEE 802.15.6 UWBIR PHY hopping sequences are used [10] for either of the modulation schemes; the guard interval inherent to hopping sequences used [10] assures that there is no inter symbolinterference (ISI) as well as no interference among PPM positions at the receiver. Modulated waveform of the kth symbol will be denoted qk (t). Kc =

2.3

Receiver architectures

2. 2.1

PRELIMINARIES UWBIR signaling strategies

Two UWBIR signaling strategies considered are shown in Fig. 1. The upper part of the gure represents the classic UWBIR signaling scheme of emitting a number of isolated shortpulse chips per symbol [17]. The lower part of the gure represents signaling scheme adopted by IEEE 802.15.4a UWBIR standard [7] and retained in the IEEE 802.15.6 UWBIR standard draft [10]; it consists of emitting a single, uninterrupted package of energy as symbol. In [7] this package consists of a shortpulse burst, while in [10] it can be either a shortpulse burst or single long chirp pulse depicted in Fig. 1. As Fig. 1 shows, the signaling scheme of [7, 10] has an advantage of involving only a single ono cycle per symbol, which helps lower power consumption through reducing energy that the transceiver consumes during turning on and o. For this reason, and reasons of being compatible as much as possible with [7, 10], we will also use it here. The ratio of the width of the symbol waveform (Tw ) to the symbol time (Tsym ) is called duty cycle: DC = Tw /Tsym . Both [7] and [10] have a constant DC = 1/32 for all data rates described which we will also adopt here. Constant DC is very benecial for transmitter implementation, since it maintains constant radiated pulse power regardless of the data rate, i.e. Tsym used.

2.2

The chirp pulse waveform

For a transmitted waveform we will use a linear chirp pulse

Modulated waveform for the kth symbol passes through the channel with impulse response hc (t) and RMS delay spread to produce signal at the input of the receiver: rk (t) = qk (t) hc (t), where represents convolution. Receiver architectures shown in Figs. 2(a) and 2(b) represent noncoherent heterodynetype receivers without and with chirp pulse compression respectively. While receiver without pulse compression, shown in Fig. 2(a) can receive practically any transmitted waveform with a given Tw equally well, chirp receiver shown in Fig. 2(b) fullls its purpose only if the transmitted waveform is a chirp pulse (1). Both receivers receive a part of the rk (t) signal of duration Tw and translate it to the intermediate frequency (IF): fIF = f0T f0R to produce signal vk (t). Here, f0R is carrier frequency of the pulse generator in the receiver. After that receivers lter signal vk (t) with the bandpass lter around fIF : hBP (t) to produce signal sk (t). Since sk (t) generally does not contain the entire energy of rk (t), receiver will have a SNR loss (Eb /N0 )r . (3) = (Eb /N0 )s The dierence between the chirp receiver and the receiver without pulse compression is found in the mixing and ltering part of the receiver. The receiver without pulse compression mixes the received signal with an unmodulated carrier pulse at f0R , in order to merely translate it in frequency to the IF. On the other side, chirp receiver mixes the received signal with chirp pulse of the form (1) with f0T f0R , i.e. chirp pulse with the carrier frequency f0R . This step in the chirp receiver has an eect of lowering the bandwidth of the signal vk (t) compared to the rk (t), i.e. compressing the signal in frequency. Principles of this technique are depicted in Fig. 3, while for a detailed mathematical analysis we direct interested readers to [3]. As Fig. 3 shows, in the chirp receiver under assumption Tw , bandwidth of the vk (t) signal is Bv Kc and thus its timebandwidth product is T Bv = Bv Tw fc . For this reason, in the chirp receiver bandwidth of the band pass lter hBP (t) BBP can be set to BBP Bv and still achieve low [3]. Hence, T Bs T Bv fc . In comparison, it is trivial to show that in the receiver without pulse

on time

o time

Tsym

Tw

Figure 1: Signaling strategies. Multiple shortpulse chips per symbol (upper). Single continuous waveform as symbol (lower).

rk (t)

vk (t)

hBP (t)

sk (t)

Detection

rk (t)

vk (t)

hBP (t)

sk (t)

Detection

f0R

Tone pulse generator

f0R

Chirp pulse generator

(a) Receiver without pulse compresion.

(b) Chirp receiver.

Figure 2: Heterodyne receiver architectures considered.


f (t) fc Tw Transmitted signal t f (t) fc Tw Received signal t f (t)
fc Tw

Tw Tw t

Signal after mixing

Figure 3: Principles of the chirp pulse compression. compression BBP Bv = fc needs to be set to achieve low and thus T Bs T Bv fc Tw ; for high Tw this can be signicantly higher than in the case of the chirp receiver. where t = (Eb /N0 )s = (Eb /N0 )r /, Qm (, ) being a generalized Marcum Qfunction of mth order, while b = 1 for EDBPPM and b = 2 for SDBPSK.

2.4

Detection methods 3. NUMERICAL ANALYSIS


In order to test the performance of the aforementioned detection schemes EDBPPM and SDBPSK with and without chirp pulse compression, we performed a simulation study of performance in noise and interference. All simulations were done on IEEE 802.15.6 channel model CM3 [19] made for the scenario of communication between BAN nodes and an onbody BAN hub. The data rate chosen was 122 kbps, four times lower than the mandatory data rate of IEEE 802.15.6., i.e. Tsym = 8192 ns and Tw = Tsym DC = 256 ns. T = 512 ns was chosen for BPPM modulation. As described above, dimensions of the vk (t) signals in receivers with and without pulse compression is dierent, thus M = 128 in the receiver without pulse compression and M = 16 in the chirp receiver were set. All simulation were performed on 5000 scenarios with 10000 bits transfered at each scenario. Receiver timing information is assumed to be perfect. The chirp pulse is used as the transmitted waveform for the BAN of interest. First test performed is one shown in Fig. 4 of performance in noise of the aforementioned modulation/detection pairs together with classic DBPSK detection, i.e. DBPSK detection with symbolsignature estimation [3]. DBPSK detection performance is not a function of M but only of [3], thus DBPSK detection curves in Fig.

For two considered modulations: BPPM and DBPSK we will analyze corresponding detection methods that do not include symbol signature estimation; for binary PPM it will be ED (EDBPPM), while for DBPSK it will be SDPSK (SDBPSK). Since sk (t) is limited in frequency and time, by using the sampling theorem [12] it can be represented by a vector of M = T Bs complex samples sk [3]. For S-DBPSK detection, decision variable dk is computed as [2, 4]: dk = sH sk , k1 while for the EDBPPM dk can be represented as dk = s1 k s1 k s0 k
2

(4)

s0 k

(5)

where and represent sk vectors at BPPM positions 1 and 0 respectively. It is virtually unknown in UWBIR literature that the problem of deriving probability of error in noise for decision variables of forms (4) and (5) has been solved in [11] and rewritten in the closed form in [14]. For (4) and (5) probability of error can be expressed as Pe = 2M m 1 1 + 2M 1 2 2 M m m=1 [Qm (0, b t ) Qm (b t , 0)],
M

(6)

100 101 102 103 Pe 104 105 106 107 0 Pe

100 101 102 103 104 105 106 5 10 15 (Eb /N0 )r (dB) 20 107

35

30

25 SIR (dB)

20

15

Figure 4: Performance in noise of detection schemes considered. Legend: no comp. & EDBPPM: simulation ( ), theory ( ); chirp comp. & EDBPPM: simulation ( ), theory ( ); no comp. & SDBPSK: simulation ( ), theory ); chirp comp. & SDBPSK: simulation ( ), theory ( ); no comp. & DBPSK: simulation ( ), theory ( ); ( chirp comp. & DBPSK: simulation ( ), theory ( ).

Figure 5: FMUWB interference performance of detection schemes considered. Legend: no comp. & ED BPPM ( ); chirp comp. & EDBPPM ( ); no comp. ); chirp comp. & SDBPSK ( ); no & SDBPSK ( comp. & DBPSK ( ); chirp comp. & DBPSK ( ). and have the same power at the receiver as the BAN signal of interest; Eb /N0 = 30 dB was set. Since it only detects total energy in a given time window, EDBPPM without pulse compression performs by far the worst and the same for both interfering waveform types. Chirp pulse compression introduces a signicant performance gain in EDBPPM, which becomes very large if the interfering waveform type is a shortpulse burst. Indeed, as Fig. 6 shows, chirp pulse compression signicantly reduces the eect of shortpulse burst interference to the detection for both EDBPPM and SDBPSK. In contrast with EDBPPM, chirp pulse compression does not improve performance of the SDBPSK, if the interfering waveform is a chirp pulse. Overall, SDBPSK performed signicantly better than EDBPPM in the multiBAN environment and allowed for a much larger number of colocated BANs.

4 show roughly 1 dB higher of the chirp receiver. Nevertheless, the chirp receiver performs somewhat better with other two detection techniques plotted: ED-BPPM and S DBPSK. Indeed, since performance of both ED-BPPM and SDBPSK deteriorates with increase of M (6), lower M of the chirp receiver more than compensates for its higher compared to the receiver without pulse compression. Theoretical curves plotted in Fig. 4 from (6) show an excellent match of theory and simulation. The second performance evaluation performed was one in the presence of FMUWB interference [6]. Since FMUWB is another candidate for medical BAN [10], it is likely that FMUWB BANs will be colocated with UWBIR BANs. Thus, it is important to evaluate the impact of FMUWB systems to UWBIR. Results shown in Fig. 5 show about the same order of performance among modulation/detection pairs as in the case of noise (Fig. 4). The only notable dierence is that the SDPSK detection without pulse compression performs signicantly worse than the SDBPSK detection with chirp pulse compression. The reason for this might be that the smaller dimension of detection in the chirp pulse compression signicantly improves performance under FM UWB interference. The third and probably the most important interference test done is the one shown in Fig. 6 of BAN performance in multiBAN UWBIR environments. Indeed, this scenario is very likely in healthcare facilities since it is reasonable to assume that BAN devices of dierent patients will be similar or even the same. We considered interfering from BANs that use the same modulation and the same signaling parameters (Tsym and Tw ) as the BAN of interest with either chirp pulse or shortpulse burst [10] waveform transmitted

4.

CONCLUSIONS

In this work we investigated the performance of two non coherent modulation/detection pairs for UWBIR with and without chirp pulse compression. Investigation was made at a data rate four times lower than the mandatory data rate of current IEEE 802.15.6 UWBIR standard draft [10] in an attempt to improve interference and shadowing resistance and thus increase reliability. While performance of EDBPPM is worse than SDBPSK in noise and FM UWB interference, it is still relatively comparable. Hence, if only these two performances are considered, the choice of EDBPPM could be welljustied by a tradeo of a somewhat poorer performance for a lowercomplexity analog detection implementation. Nevertheless, EDBPPM is unable to satisfy the requirement of reliable BAN communication in the multiBAN scenarios which is expected to be common in health care facilities. On the other hand, SDBPSK

100 101 102 103 Pe 104 105 106 107

10 20 30 Number of interfering BANs

40

Figure 6: MultiBAN interference performance of detection schemes considered. Legend: no comp. & EDBPPM: chirp interf. ( ), burst interf. ( ); chirp comp. & EDBPPM: chirp interf. ( ), burst interf. ); no comp. & SDBPSK: chirp interf. ( ), burst in( terf. ( ); chirp comp. & SDBPSK: chirp interf. ( ), burst interf. ( ). shows that it can endure a much larger number of colocated BANs. Chirp pulse compression produces performance enhancements in all numerical tests performed. Moreover, in digital detection implementations considered here it reduced the number of complex samples from M = 128 to M = 16. From all written above we conclude that chirp pulse compression in low datarate medical BAN is a very benecial technique that has an eect of both complexity reduction and performance enhancement; hence, we recommend its usage in highly reliable medical UWBIR BAN devices that are expected to use lower datarates than IEEE 802.15.6 UWBIR PHY.

5.

REFERENCES

[1] IEEE 802.15 WPAN Task Group 6 Body Area Networks (BAN). [2] I. Dotli and R. Kohno. Performance analysis of c Impulse Radio UltraWideband dierential detection schemes for Body Area Networks. In IEEE 21st International Symposium on Personal, Indoor and Mobile Radio Communications Workshops (PIMRC Workshops), pages 72 77, September 2010. [3] I. Dotli and R. Kohno. Low Complexity Chirp Pulsed c Ultra-Wideband System with Near-Optimum Multipath Performance. IEEE Transactions on Wireless Communications, 10(1):208 218, January 2011. [4] I. Dotli and R. Kohno. Preamble Structure and c Synchronization for IEEE 802.15.6 ImpulseU-Radio Ultra-UWideband Physical Layer. In 5th International Symposium on Medical Information and Communication Technology, ISMICT, March 2011.

[5] R. Fujiwara, A. Maeki, K. Mizugai, G. Ono, T. Nakagawa, T. Norimatsu, and M. Kokubo. 0.7-GHz-Bandwidth DS-UWB-IR System for Low-Power Wireless Communications. IEICE Transactions on Communications, E91-B(2):518526, Feb. 2008. [6] J. F. M. Gerrits, M. H. L. Kouwenhoven, P. R. van der Meer, J. R. Farserotu, and J. R. Long. Principles and limitations of ultrawideband FM communications systems. EURASIP J. Appl. Signal Process., 2005(1):382396, 2005. [7] P. W. Kinney, V. Brethour, J. Bain, P. Houghton, J. Lampe, V. Brethour, Z. Sahinoglu, P. Orlik, I. Lakkis, R. Hach, K.-K. Lee, M. Welborn, J. Bain, B. A. Rolfe, M. Welborn, C. Gentile, and M. McLaughlin. IEEE Standard for PART 15.4: Wireless MAC and PHY Specications for Low-Rate Wireless Personal Area Networks (LR-WPANs): Amendment 1: Add Alternate PHY. [8] D. Lachartre, B. Denis, D. Morche, L. Ouvry, M. Pezzin, B. Piaget, J. Prouvee, and P. Vincent. A 1.1nJ/b 802.15.4a compliant fully integrated UWB transceiver in 0.13 m CMOS. In IEEE International Solid-State Circuits Conference - Digest of Technical Papers, ISSCC 09, pages 312313a, 2009. [9] F. Lee and A. Chandrakasan. A BiCMOS ultra-wideband 3.1U-10.6 GHz front-end. IEEE Journal of Solid-State Circuits, 41(8):17841791, Aug. 2006. [10] D. Lewis. IEEE P802.15.6 D04 Draft Trial-Use Standard for Body Area Network. IEEE 802.15 WPAN, June 2011. [11] J. Proakis. On the Probability of Error for Multichannel Reception of Binary Signals. IEEE Transactions on Communication Technology, 16(1):68 71, february 1968. [12] C. Shannon. Communication in the presence of noise. Proceedings of the IEEE, 72(9):11921201, Sept. 1984. [13] D. Simi, A. Jordan, R. Tao, N. Gungl, J. Simi, c c M. Lang, L. V. Ngo, and V. Brankovi. Impulse UWB c radio system architecture for body area networks. Mobile and Wireless Communications Summit, 2007. 16th IST, pages 15, July 2007. [14] M. K. Simon and M.-S. Alouini. Digital Communication over Fading Channels: A Unied Approach to Performance Analysis. John Wiley, 2000. [15] L. Stoica, A. Rabbachin, and I. Oppermann. A low-complexity noncoherent IR-UWB transceiver architecture with TOA estimation. Microwave Theory and Techniques, 54(4):16371646, June 2006. [16] D. M. Wilson. IEEE Body Area Network Diabetes July 11. IEEE 802.15 WPAN, (IEEE P802.15-11-0564-00-0006), July 2011. [17] M. Win and R. Scholtz. Impulse radio: how it works. IEEE Communications Letters, 2(2):3638, Feb 1998. [18] K. Witrisal, G. Leus, G. Janssen, M. Pausini, F. Troesch, T. Zasowski, and J. Romme. Noncoherent UltraWideband Systems. IEEE Signal Processing Magazine, 26(4):48 66, July 2009. [19] K. Y. Yazdandoost. IEEE Body Area Network Channel Model. IEEE 802.15 WPAN, (IEEE P802.15-08-0780-12-0006), November 2010.

You might also like