Experimental Impulse Radio IEEE 802.15.4a UWB Based Wireless … · 2018-12-19 · This paper presents a thorough ap-plied examination of prototype IEEE 802.15.4a impulse UWB transceiver

ISSC 2011, Trinity College Dublin, June 23–24

Experimental Impulse Radio IEEE 802.15.4a UWBBased Wireless Sensor Localization Technology:

Characterization, Reliability and Ranging

Tingcong Ye∗, Michael Walsh, Peter Haigh, John Barton, Alan Mathewson,Brendan O’Flynn

Tyndall National InstituteUniversity College Ireland, UCC, Cork

E-mail: ∗[email protected]

Abstract — Ultra Wide Band (UWB) wireless transmission has recently been the ob-ject of considerable attention in the field of next generation location aware wirelesssensor networks (WSNs). This is due to its fine time resolution, energy efficiency androbustness to interference in harsh environments. This paper presents a thorough ap-plied examination of prototype IEEE 802.15.4a impulse UWB transceiver technologyto quantify the effect of line of sight (LOS) and non line of sight (NLOS) ranging inreal indoor and outdoor environments. The results included draw on an extensive arrayof experiments that fully characterize the 802.15.4a UWB transceiver technology, itsreliability and ranging capabilities for the first time. The goal of this work is to validatethe technology as a dependable wireless communications mechanism for the subset ofsensor network localization applications where reliability and precision positioning arekey concerns.

Keywords — Ultra Wide Band, Two Way Ranging, Reliability, LOS, NLOS

I Introduction

Ultra Wide Band 802.15.4a transceiver technol-ogy is emerging as an ideal fit for the require-ments of the next generation wireless sensor net-work [1]. IEEE has recognized the need to stan-dardize UWB technology for use in personal areanetworks (PANs) and has established the IEEE802.15.4a standard specifying a new UWB phys-ical layer for WSNs [2]. For some WSN appli-cations such as tracking missing persons after anearthquake, sensing data without knowing its sen-sor location is meaningless. Thusly, real-time lo-calization is becoming a more important concept inWSNs. Accurate range-based localization dependson a precise ranging measurement of the wirelesssensor systems. Some classical ranging techniquesinclude received signal strength (RSS), angle of ar-rival (AOA) and time of arrival (TOA) which arerelated to energy, direction and the timing of thewireless signals traveling between two sensor nodesrespectively. However, UWB techniques employ-

ing RSS methods can not obtain accurate rang-ing estimates due to its strong dependance on thechannel parameters, which makes the received en-ergy more sensitive to distance changes in NLOSareas. AOA methods can facilitate accurate rang-ing when the UWB signal bandwidth is increased,but it needs multiple antennae that makes systemlarger and costly. TOA parameter based methodsprovide more accurate range estimates but lowercost compared to the RSS and AOA. The IEEE802.15.4a standard based on an impulse UWB sig-nal supports a TOA ranging mechanism [3]. Em-ploying prototype fully IEEE 802.15.4a compli-ant transceiver technology, the world’s first IEEE802.15.4a UWB wireless packet was transmittedand successfully coherently received in real timein March 2009 [4]. This impulse UWB prototypetransceiver technology can easily be placed in thenext generation Wireless WSN category.

The goal of this paper is to examine the charac-terization, reliability and ranging precision of an

impulse UWB based transceiver for both indoorand outdoor environments. A two way ranging al-gorithm based on TOA employed as part of thiswork is described in detail. A theoretical analysisof impulse UWB radio for wireless communicationand ranging is provided employing the ShannonHartley theorem [5] and Cramer-Rao lower bound(CRLB) [6] method.

To fully test the reliability of the UWB rangingsystem, a distance measuring experiment is firstlyimplemented in an indoor environment to inves-tigate the case of LOS with multi-path, and alsothe case of NLOS with different materials, for in-stance, chair, counter, door and walls; secondly,the outdoor open field case is evaluated with re-flections are presented at the receiver. RelevantUWB signal characterization, pulse response andtransmitting power are measured. A realistic twoway ranging model of the system in operation isdescribed. Finally some conclusion are included.

II Theoretical analysis of UWB ranging

A classical time of arrival based method calledtwo-way ranging (TWR) was originally proposedin [3]. The practical ranging demonstration is de-scribed in Fig.1. The leader observes a round trip

Fig. 1: TW-Ranging Method

time LRT = TRR − TSB and a turn around timeLTA = TSF − TRR, where TSB , TRR and TSF arethe leader send-time, receive-time and future send-time respectively. The follower observes a roundtrip time FRT = TRF − TSR and a turn aroundtime FTA = TSR−TRB , where TRB , TSR and TRF

are the follower receive-time, send-time and futurereceive-time respectively. The value of transmis-sion time T is computed at both leader (Tl) andfollower (Tf ):

2Tl = (TRR − TSB)− (TSR − TRB) (1)

2Tf = (TRF − TSR)− (TSF − TRR) (2)

The follower or leader can then combine these tworesultant round trip times (by averaging) to re-move by effects of clock differences. The result isthen divided by 2 to get one way trip time.

T =2Tl + 2Tf

2× 2(3)

Thus the distance between the two prototypes is:

d = T × C (4)

Here C is the speed of light.When implementing this ranging method in a

practical operation, reliable ranging depends onthe system being able to accurately determine thetransmitting and receiving times of the signal mes-sages at the antenna. The IEEE 802.15.4a specifi-cally defines the time stamps reflecting the instanttime of the first ultra wide band pulse of the firstbit of the physical layer header (PHR)of a rang-ing frame. Moreover, there are a number of chal-lenges that remain before accurate ranging can beachieved in a harsh environment including multi-path, radio interference, and the NLOS propaga-tions. Thus, the TWR requires that the radio fre-quency (RF) signal should have a good channel ca-pacity and be robust to interference to enable thedetection of the first path of the received signal.

For any given RF radio, Shannon’s theory [5] ex-amines the characterization of the signal channel.

C = B × log2(1 + SNR) (5)

In equation (5), C is the maximum channel ca-pacity, B is the channel bandwidth, and SNR isthe signal to noise ratio. This equation indicatesthat for high frequency bands, large channel capac-ity is available despite a reduction in transmissionpower. An UWB signal is defined by the FCC tohave either a signal bandwidth exceeding 500MHzor a fractional bandwidth exceeding 0.2. If bothUWB (500MHz) and narrow band signal such asIEEE 802.11n [7] (40MHz) have the same SNR,the channel capacity of UWB is approximately 12times larger than IEEE 802.11n.

UWB signal is broadly categorized into impulseUWB and multi-carrier UWB. The UWB pulsecan be easily generated from a Gaussian pulse andits derivatives. A Gaussian pulse in the time do-main is described in [8] as:

P (t) = ±√

2α

e−2πt2

α2 (6)

Here

α2 = 4πσ2 (7)

where α is the pulse form factor and σ2 is thevariance. The Gaussian pulse and its first 15

derivatives are shown at Fig.2. These pulses arenot directly used for practical applications becauseof the need to meet the spectrum as mandated bythe FCC. The first pulse is a Gaussian monocycle.

Fig. 2: The Gaussian pulse and its derivatives

From equation 6, it is clear that the bigger theα value is, the narrower the pulse width, but thewider the frequency bandwidth. When the orderof the Gaussian pulse derivative increases, its peakfrequency increases as indicated in Fig.3.

Fig. 3: 15 order derivative verus pulse Peak Frequencyand α

It is most likely that signals formed by pulseswith duration on the order of fractions of nanosec-onds will be UWB signals. This paper is specificto impulse UWB.

The impulse UWB signal can be modulated byTime-Hopping (TH) modulation and Direct Se-quence (DS) modulation. The DS modulation isselected to generate a UWB signal whose trans-mitting signal is indicated as follows:

S(t) =+∞∑

j=−∞djP(t− jTs) (8)

Here, j is integer, dj is the pseudo random se-quence, Ts is the pulse duration, P(t) is signalpulse. A simulation of a DS UWB signal is shownin Fig.4, pulse repetition period 2e−9s, 10 pulsesper bit, periodicity of the DS code is 1 0, and

Fig. 4: Direct Sequence UWB Signal Simulation

pulse duration is 0.5e−9s. From equation 8 andFig.4, the UWB system architecture is simplerwhen compared to the Direct Sequence SpreadSpectrum (DS-SS) system with Binary Phase ShiftKeying (BPSK) carrier modulation. The discreteUWB signal is narrow in width and can be dis-tinguished from multi-path reflections, making ac-curate pulse timing more strength when comparedwith conventional RF radios for indoor environ-ments.

For ranging applications, the system capturesthe transmitting time employing the time of arrivalmethod. The Cramer-Rao lower bound states thatthe variance of any unbiased estimator would notbe lower than the inverse of the Fisher informa-tion. Where the Fisher information is defined as away of measuring the amount of information thatan observable random variable X carries about anunknown parameter θ upon which the probabilityof X depends. The CRLB for impulse UWB signaltime delay estimate is given by Kay [6] as follows:

V ar(τ) ≥ 1SNR× β2

(9)

In the equation 9, τ is the transmission time, SNRis the system signal to noise ratio, and β is theeffective bandwidth. Therefore, the wider the sig-nal bandwidth or the better the SNR, the moreaccurate the ranging precision.

An Impulse UWB (500MHz) is therefore apromising technology for location aware sensornetworks, due to its high quality communicationand accurate ranging. Owing primarily to the finetime resolution on the order of sub-nanosecondpulses, accuracy of a few centimeters in distancemeasurement can be obtained.

III Experimental Activities

The goal of this experiment is to validate the char-acterization, reliability and ranging of the UWBtransceiver technology in indoor and outdoor en-vironments. The measurements were made usingtwo FCC-compliant UWB radios obtained fromDecawave, a company based in Dublin, Ireland[9].The primary focus is to characterize the electricalproperties of these UWB transceivers. The two

radios, a leader and a follower, each of which is15cm wide and 30cm high, generate impulse UWBsignals resulting in a signal with a bandwidth of500MHz and a center frequency of 4GHz.

In the system configuration, a Pulse RepetitionFrequency (PRF) of 16MHz, a preamble length of1024, preamble code 4 and data rate 850kbit/s areselected for both transceivers. The system delaywhich is the key factor for ranging measurement istested and determined to be 277.850ns. However,the system delay may need to be tuned if operatingmodes are changed, for instance, the delay needs tobe changed down when increasing the PRF value.Large system delays will lead to the time of arrivalbeing a negative value. The prototype has no Au-tomatic Gain Control (AGC) unit, thus setting ofthe transmission attenuation value is needed whichis 13.5dB to ensure the transmitting power is com-pliant with the FCC emission limits.

The UWB radio signal spectrum captured byan agilent spectrum analyzer is shown as Fig.5.The effective bandwidth is about 500MHz, from alower frequency of 3.75GHz to an upper frequencyof 4.25GHz, is observed with 1MHz resolutionbandwidth. The Power Spectrum Density(PSD) isapproximately -41.3dBm/MHz which meets withFCC emission limits in Fig.6. The band poweris measured as -13.31dBm at the 4.22GHz point,from which, the antenna gain is calculated, result-ing in a value of 3.1 and the transmitting poweris about 0.047mW. The TWR ranging algorithm

Fig. 5: UWB Signal Spectrum

(equation (1),(2),(3),(4)) is embedded in the sys-tem IC for each of the prototype boards. Suc-cessful ranging relies on the system being able toaccurately determine the transmitting and receiv-ing times of the signals as they leave one antennaand arrive at the other antenna. Due to the dig-ital hardware that makes the generation or re-ception of the ranging marker which is the firstbit of the PHR, the adjustments are needed toadd the transmitting system-to-antenna delay tothe transmitting time-stamp and subtract the re-

Fig. 6: Indoor Emission Mask

ceiving antenna-to-system delay from the receiv-ing time-stamp. The first path arrival time is de-termined from the channel impulse response dataread from the register and the time delta thus de-termined is added. The time delta is the distancebetween the time stamp window index, as indi-cated by the first byte of channel response read,and the software determined leading path whichis the sample index or point between two indiceswhere the software determines the first arrivingsignal to be. The system delay as previously spec-ified is added. Compensation for the clock offset isachieved by adding the relevant offset. Thus, therealistic two way ranging model based on the orig-inal TW ranging protocol[3] based on equation 1and equation 2 should be as follows:

2T = TLRTT − TF

TAT (10)

Where

TFTAT = TDELAY −AR

F + LpRF + TF

off + ATF (11)

TLRTT = AT

L + TFTAT −AR

L + LpRL + TL

off (12)

In equation(10), TLRTT is the round-trip-time of

the leader, TFTAT is the turn around time of the

follower. In equation(11) and (12), the TDELAY

is the specified time delay, ARF and AR

L are thereceiving antenna-to-system delay at follower andleader respectively; while, AT

F and ATL are the

transmitting system-to-antenna delay at followerand leader respectively; LpR

F and LpRL are the first

path detection delay of the follower and the leader;TF

off and TLoff are the relevant time offset at both

follower and leader respectively. Thus, the totalranging time is the summation of ranging countand clock offset.

When the system is running well, the channelresponse can be captured from a ranging GUI as

shown in Fig.7, in which the red line (c) is theplot of the real pulse response, the green line (d)is the imaginary response and the blue line (b)is the computed magnitude values. The verticalcyan line (a) is the leading path which is used tofind the first in time arriving receive signal. Thesystem SNR (44.4dB), first path SNR (38.4dB),and the real time distance (1.31m) is tested andrecorded. This distance is firstly measured in anoffice with no obstructions and subsequently lessreflections. Compared with the true value 1.30mmeasured physically using a tape, the measured in-stant and average distances are very accurate withonly 1cm error. The round trip time value L, R, ofboth local and remote prototypes and their aver-age value C are calculated in device time units [3].The device time units follow the definition of timecontained in the IEEE Std 802.15.4a, which statesthe LSB of a time value represents 1/128 of a chiptime at the mandatory chipping rate of 499.2MHz.After dividing C by 128 and multiplying by thespeed of light, the distance is finally calculated.

Fig. 7: Received UWB Signal Response

IV experimental ranging results

To validate the reliability of the UWB transceiverranging in cluttered indoor and outdoor environ-ments, according to the different channel models, 3scenarios are defined and implemented in some of-fices, hallways, and yards in the Tyndall NationalInstitute as shown in Fig.8. The two motes aremounted on the floor, One is located at the fixedorigin, while the other moves to perform point-to-point measurement. The distance between the tworadios is varied from 1m up to 45m to capture a va-riety of operating conditions. In each measurementlocation, the values such as the pulse response, in-stant distance estimates, SNR, as well as the actualdistance from a tape are recorded.

a) LOS Test

The line of sight ranging experiments are con-ducted in a library, an indoor hallway, an outdoorhallway and in the open field such as Fig.8(a,b)where is no obstructions between the two motes.This experiment is to validate the performance ofthe UWB ranging system in an environment withsignificant multi-path propagation, thermal noise

and narrowband interference such as WIFI signals.Testing Points are placed randomly but are re-stricted to the tape measured points.

Fig. 8: LOS, Soft-NLOS, Hard-NLOS Ranging Campaigns

Fig. 9: Ranging Results of LOS Area Measurement

Fig.9 shows the average LOS ranging results inthe library which has some counters and chairsaround the motes, the outdoor hallway in whichthe two boards are located between the outsidewalls and bicycle shed, the indoor hallway withsome doors, WIFI sites and walls on the both sidesand in the open field where there are no obstruc-tions, and the only sources of signal reflection arethe operators, the equipment and the ground. Theaverage ranging errors of these tested points areless than ±20cm regardless of the system error.The errors measured in the open field are less than10cm, which is more accurate than the results ofother conditions with more reflections and sourcesof interference in between. The outdoor hallwayresult is less than the positive value measured viaa tape, that is likely because the PHY header checkbits of some frames were in error.

b) Soft-NLOS Test

This experiment aims to validate the reliability ofthe UWB ranging system through a Soft-NLOSchannel which occurs when the LOS path is ob-structed by materials with relatively low attenua-tion or by a combination of these materials suchas glass, chairs, counters and doors. A 3m test-ing point is selected for each experimental clusterand different obstructions are deployed betweenthe two motes as in Fig.8 (c). Some obstructions

are combined as this would likely be the case in ac-tual conditions. All the results of clusters in Fig.10show that the error from different obstructionsvaries from 6cm (Glass) to 29cm (Door). Theseexperimental conditions represent the most com-mon channel model over the distances of interestin most European offices, the average results of lessthan 30cm accuracy are acceptable to validate thatthe UWB signal can be utilized in indoor rangingmeasurements.

Fig. 10: Ranging Results of Soft-NLOS Area at 3m Point

c) Hard-NLOS Test

This experiment (Fig.8 (d)) attempts to validatethe capability of the UWB ranging system in ahard-NLOS channel with multiple concrete wallsor multi-obstructions in the environment. Resultsin Fig.11 indicate that hard obstructions severelyattenuate the UWB signal propagation and gener-ate large positive bias in the range estimates. Thedistance error varies widely from 26cm for one wallto 87cm for 4 walls. While at the 38m point, the re-ceiver is unable to receive the signal. It is clear thatlocalization in this hard NLOS rooms can not ob-tain high precision by only using TWR algorithm.Some method and algorithms [3] are proposed toimprove or solve the hard NLOS ranging problems,but ranging in an NLOS area is still a challengingproblem for indoor locations.

Fig. 11: Ranging Results of Hard-NLOS Area

V Conclusion

In this paper, we theoretically analyzed and realis-tically validated the reliability of an impulse UWB

transceiver based point-to-point ranging systemusing a two way ranging algorithm in both indoorand outdoor environments. In theory, the UWBsignal is very resistant to multi-path and reflec-tions, the CRLB method proves that the UWBsignal has a high precision on the order of severalcentimeters. In practical operation, the two UWBtransceivers were evaluated in this study satisfyingthe FCC limits. A realistic two way ranging modelwas generated with antenna-to-system delay, firstpath detection delay and time offset and imple-mented in the ranging system. Results recordedof LOS, Soft-NLOS and Hard-NLOS ranging ex-periments show that UWB transceivers are capa-ble of capturing accurate transmission time be-tween two radios which can be used in turn tocompute the real distance. With features such aslarge channel capacity, robustness to interferenceand multi-path, energy efficiency and fine resolu-tion, UWB transceiver technology is a dependablewireless communications mechanism for WSN lo-calization applications.

References

[1] M.G. diBenedetto, T. Kaiser, A.F. Molish,I. Oppermann, C. Plitano, and D. Porcino(eds.). “UWB Communications Systems: AComprehensive Overview”. EURASIP Publish-ing,2005.

[2] IEEE Working Group 802.15.4a. “Draft speci-fications for IEEE 802.15.4a standard”.

[3] IEEE Std 802.15.4aTM -2007, IEEE Standardfor information Technology

[4] http://www.idtechex.com/events/presentations/scensor_an_ieee802_15_4a_uwb_compliant_chip_ripe_for_energy_harvesing_001824.asp

[5] Kazimierz Siwiak and Debra McKeown.“Ultra-Wideband Radio Technology”. JohnWiley and Sons,2006.

[6] S.M. Kay. “Fundamentals of Statistical Sig-nal Processing Estimation Theory”. EnglewoodCliffs, New Jersey:Prentice-Hall,1993.

[7] IEEE Std 802.11n-2009,IEEE Standard forinformation Technology–Telecommunicationsand information exchange between systems

[8] Maria-Gabriella Di Benedetto and Branimir R.Vojcic. “Ultra Wide Band Wireless Communi-cations : A Tutorial”. Journal of communica-tions and networks, Vol.5, No.4, , December2003.

[9] http://www.decawave.com/index.html