1A Robust Ultra Broadband WirelessCommunication System Using SAW Chirped DelayLinesAndreas Springer, Mario Huemer, Leo Reindl, Clemens C. W. Ruppel, Alfred Pohl, Franz Seifert,Wolfgang Gugler and Robert WeigelAbstract|Design and performance of a low-cost wirelesscommunication system for indoor and industrial environ-ments are presented. The system is based on chirp signaltransmission to achieve a robust communication link. Forthe chirp expansion and compression SAW chirped delaylines fabricated from on LiTaO3 - X112rotY are used. Cen-ter frequency, bandwidth, and chirp rate are 348.8 MHz, 80MHz, and �40MHz=�s, respectively. An optimized square-root weighting was chosen to reduce the sidelobes of thecompressed pulse to -42 dB compared to the correlationpeak. The chirp �lters have been deployed in a hardwaredemonstrator for data rates of up to 5 MBit/s. Limitingfactors for the data rate according to simulations and mea-surements are mainly intersymbol interferences due to thetime-overlapping of consecutive symbols and to a lower ex-tent the multipath propagation.Keywords|spread spectrum communication, surface acou-stic waves, wireless communcations, chirp signals.I. IntroductionThe wireless communications market undergoes a tre-mendous growth. This is mainly due to the rapidly expan-ding area of personal communication services (PCS) likecellular or cordless phone systems, paging and wireless lo-cal area networks (WLAN's). Some applications like cel-lular phone systems are already widely accepted althoughmarket penetration is still far from saturation, while others,e.g. WLAN's, are just starting to become a common tech-nology [1]. Driven by the rapid development of RF tech-nologies for PCS, wireless communication systems for in-dustrial environments are now facing strong interest. Here,major applications are exible and mobile data transmissi-on links between sensors, actuators, autonomous vehicles,robots, and controller units. Due to the hostile electroma-gnetic environment which includes severe electromagneticemissions from other devices as well as heavy distortionsdue to multipath propagation the robustness of the com-munication link is extremely important.The spread spectrum technology is especially well suitedto provide such a robust data transmission even in very noi-sy radio environments [2]. Originally developed for militarypurposes, spread spectrum communication is nowadays apopular technology in many areas of mobile communicati-A. Springer, M. Huemer, W. Gugler and R. Weigel are with theUniversity of Linz, Institute for Communications and InformationEngineering, A-4040 Linz, Austria.L. Reindl and C. C. W. Ruppel are with Siemens AG Munich,Research and Development Center, D-81739 Munich, Germany.A. Pohl and F. Seifert are with the University of Technology Vienna,Applied Electronics Laboratory, A-1040 Vienna, Austria.

ons. It is established in the IS-95 standard for cellular pho-nes as well as in the IEEE 802.11 standard for WLAN's [1],and very recently it has been chosen to be part of the thirdgeneration cellular phone system standard UMTS. Key fea-tures of the spread spectrum technology are high immunityagainst multipath phenomena, high spectral e�ciency, lowpower and low cost. The critical operations in spread spec-trum systems are the spreading and despreading functionsin the transmitter and receiver, respectively. The commonsystem concepts (direct-sequence (DS) and frequency hop-ping (FH)) require rather sophisticated circuit designs andsystem realizations to accomplish the despreading in thereceiver. In particular, the synchronization of the desprea-ding code sequence is a di�cult task. With the well knownFM chirp signals and the associated technique of pulse com-pression with its high processing gain, which is widely usedin radar systems [3], another kind of spread spectrum com-munication systems can be realized [4]. In such a systemthe spectrum spreading is used for combating the multipathdistortions whereas code division multiple access (CDMA)can only be realized if additional coding is introduced [5].The generation of the transmitted chirp signals and thecorrelation process in the receiver are vital to the chirpsystem. Both functions are easily accomplished by usingsurface acoustic wave (SAW) chirped delay lines [3], whichare well suited for use as expansion and compression �lterin the proposed communication system due to their highdegree of exibility in designing desired transfer functions.SAW chirped delay lines can be realized at small size andlow cost as is the case with the well established SAW �ltersand resonators in wireless communication products [6], [7].Our system which is designed for unlicensed use in indoorand industrial environments in the ISM band at 2.45 GHzuses binary orthogonal keying (BOK) as modulation sche-me by making use of the quasi-orthogonality of up- anddown chirp signals which leads to a very robust communi-cation link. In the following, we give a short description ofthe proposed system and its parameters, design and fabri-cation of the SAW chirped delay lines are described, radiochannel simulation and measurement, an issue of great rele-vance to every wireless communication system, is covered,and results of the system simulation as well as measure-ment results achieved with a hardware demonstrator aregiven.

2II. System OverviewA. Basic Chirp TheoryA chirp waveform can be written ass(t) = a(t) cos [�(t)] (1)where �(t) is the time domain phase, and a(t) is the timedomain envelope which is zero outside a time interval oflength T . The instantaneous frequency is de�ned asfM (t) = 12� d�(t)dt : (2)The rate of change of the instantaneous frequency�(t) = dfM (t)dt = 12� d2�(t)dt2 (3)is denoted the chirp rate. Waveforms with �(t) > 0 arecalled up-chirps while those with �(t) < 0 are called down-chirps. For a linear chirp �(t) is constant, and hence fM (t)is a linear function of t; and �(t) is a quadratic function.If we take the waveform s(t) to be centered at t = 0 it canbe written ass(t) = a(t) cos �2�f0t+ ��t2 + '0� (4)where fc is the center frequency, '0 is the phase constantand a(t) = 0 for jtj > T=2. It is convenient to de�ne thechirp bandwidth B as the range of instantaneous frequen-cies, so that B = j�jT: (5)The 3 dB bandwidth of the spectrum is dependent on theform of a(t), and can be quite di�erent from B.If a chirp waveform is fed into its matched �lter, whoseimpulse response is also a chirp waveform but with its fre-quency varying in the opposite direction, then the outputsignal typically has a narrow RF peak at the chirp centerfrequency. Generally the width of the output peak is muchless than the length T of the input waveform and hencethe process is called pulse compression, and the matched�lter is commonly named compressor. If we regard chirpwaveforms with at envelopes and if we take the matched�lter also to be centered at t = 0, i. e.h(t) � s(�t); (6)then we can �nd an analytical expression for the outputwaveform g(t) of the matched �lter. We haveg(t) = h(t) � s(t) = 'ss(t) (7)where 'ss(t) is the autocorrelation function of s(t). It canbe shown [8] that 'ss(t) is given by'ss(t) = pBT sinn�Bt�1� jtjT �o�Bt cos(2�f0t) (8)for �T < t < T . This is illustrated in Fig. 1 for the para-meters T = 2 �s, B = 80 MHz and fc = 348:8 MHz which

−40 −30 −20 −10 0 10 20 30 40−1

−0.8

−0.6

−0.4

−0.2

0

0.2

0.4

0.6

0.8

1

Time [ns]

Rel

ativ

e V

olta

ge

Fig. 1. Autocorrelation function for a linear chirp signal with theparameters T = 2 �s, B = 80 MHz and fc = 348:8 MHz.are used in our proposed system. The envelope has its ma-ximum at t = 0, and its �rst zeros at t � �1=B. It is conve-nient to specify the pulse width as 1=B. This gives a valueof 12:5 ns for the given example. The ratio of the inputand output pulse widths is given by the time-bandwidthproduct TB which is known as compression ratio or pro-cessing gain. Another important parameter is the sideloberejection, which here is about 14 dB. A common method ofreducing the sidelobes is to apply amplitude weigthing ofthe chirp signal which is described in more detail in sectionIII.B. Binary Orthogonal KeyingFig. 2 shows the principle of the binary orthogonal key-ing (BOK) modulation scheme in the time-frequency plane.If we send a 'high', an IF pulse at the chirp center frequen-Fig. 2. Schematic of a SAW BOK wireless LAN.cy of 348:8 MHz stimulates the up-chirp �lter, if we send a'low', the down-chirp �lter is stimulated. With a chirp si-gnal duration of 2�s the highest achievable data rate wouldbe only 500 kBit/s which is too low for today's systems. Toincrease the data rate the chirp signals have to overlap intime as is shown in Fig. 2. By that measure, the data rateof our system is 2 MBit/s. The sum of both �lter outputsis launched at a transmitter frequency of 2:45 GHz. In thereceiver the signal, which is disturbed by the frequency se-lective radio channel and additive noise, is mixed down and

3fed into the SAW compressor �lters. The up-chirp �lter ismatched to the 'low' signal, the down-chirp �lter to the'high' signal. In both paths non-coherent envelope detec-tion with hard decision and an adaptive threshold controlfollows. The schematic of the system is presented in Fig. 4in section VI together with a more detailed description ofthe realized system.The e�ect of overlapping and the fact that up- and down-chirps are not exactly orthogonal leads to disturbing cross-correlations in both �lter outputs. This limits the achie-vable data rate because the cross-correlation peaks growtogether with the overlapping of consecutive chirp signalsin time. The ultimative limit for the data rate is given bythe time spreading of the correlation peaks caused by themultipath channel.III. SAW Chirped Delay LinesUp- and down-chirp �lters have been designed and fa-bricated from LiTaO3-X112rotY substrate using standardoptical lithography technique. As metalization a layer of 72nm aluminum has been deployed. The �lters have a centerfrequency and bandwidth of 348.8 MHz and 80 MHz, re-spectively. According to [9] two apodized transducers wereused and a split �nger arrangement was chosen to avoidinternal re ections. The chip size is 11:8�2:0mm2. An op-timized square-root weighting of the magnitude of the �ltertransfer function was employed to reduce the sidelobes ofthe compressed pulse to -42 dB compared to the correlati-on peak. measurement results of the transfer function and

300 310 320 330 340 350 360 370 380 390 400−90

−80

−70

−60

−50

−40

Frequency [MHz]

Mag

nitu

de S

11 [d

B]

Gro

up D

elay

[s]

300 310 320 330 340 350 360 370 380 390 4000

0.5

1

1.5

2

2.5x 10

−6

Fig. 3. Measured transfer function and group delay of an optimizedsquare-root weighted down-chirp �lter.the group delay of a down-chirp �lter are shown in Fig. 3.The chirp rate of the �lters is about �40 MHz/�s whichresults in a dispersion time of about 2�s. This correspondsto a time-bandwidth product of 22 dB.IV. Radio Channel Measurements andSimulationsTo improve the performance of a communication systemin a timely, cost e�ective and e�ort free manner it is ne-

cessary to use computer aided analysis. The computatio-nal model for the indoor radio channel is crucial to thereliability of the simulation results for our proposed sy-stem. To collect data concerning the properties of the in-door radio channel wideband measurements [10], [11] ofsuch channels at 2:45 GHz have been conducted for line-of-sight (LOS) and non-line-of-sight (NLOS) scenarios in alaboratory/o�ce environment in the fourth oor of a sixstory building. From the measured impulse response pro�-les, mean delay spread, coherence bandwidth and path lossexponents have been derived. Comparison with simulationresults extracted from the SIRCIM simulator [12] - a toolfor modeling indoor radio channels - has also been made.A. Parameters of the Indoor Radio ChannelThe baseband complex channel impulse response of theindoor radio channel is modeled as [13]h(t) = N�1Xk=0 akej�k�(t� �k): (9)Here k is the path index, ak the path gain, �k the phaseshift, and �k the time delay of the kth path. The absolutedelay of the channel is not important, so �0 is set to zero.The rms delay spread which is a measure for the time-dispersion of a transmitted signal is de�ned as�rms =vuutR1�1 (t� �m)2 jh(t)j2 dtR1�1 jh(t)j2 dt ; (10)where �m is the mean excess delay given by�m = R1�1 t jh(t)j2 dtR1�1 jh(t)j2 dt : (11)We use a simple path loss model of the form dn where d isthe distance between the transmitter and the receiver, andn indicates the path loss exponent. The mean path loss indB is described as [14]PL(d) = PL(d0) + 10n log10 � dd0� ; (12)where free space path loss is assumed between the trans-mitter and a reference distance d0, so that PL(d0) is givenby PL(d0) = 20 log10�4�d� ��GT �GR; (13)where � is the wavelength, and GT and GR are the antennagains in dB.B. Measurement SetupThe measurement technique applied here is a widebandcoherent frequency response measurement, where magni-tude and phase shift at 1601 equidistant frequency pointsin the selected frequency band (2:4 GHz - 2:5 GHz) arerecorded. The measurement setup is shown in Fig. 4. The

4Fig. 4. Setup for radio channel measurement.system consists of a HP 8753 network analyzer, two broad-band ground plane antennas with a gain of 2 dBi, and alow noise ampli�er.Before the measurements are carried out, the system hasto be calibrated at the frequency band of interest in orderto compensate for the in uence of phase and amplitudevariations imposed by the cables, ampli�ers and the othermeasurement equipment.C. Measurement EnvironmentThe measurements have been performed in a laborato-ry/o�ce environment in the fourth oor of a six story buil-ding. We used three types of location (S1: laboratory room,S2: o�ce room, S3: hallway) for our measurements as isshown in Fig. 5. The tests have been done at 19 discretelocations separated by �=4 along a 4:5 � track at each ofthe three positions marked with S1, S2, and S3 as well asat di�erent positions in all rooms to achieve enough da-ta for a delay spread and a path loss estimation. Duringthe measurements we kept the channel as time-invariant aspossible.D. Measurement and Simulation ResultsMeasured impulse response results have been compa-red with results produced with the SIRCIM simulator [15].SIRCIM is a microwave multipath indoor radio channel si-mulator designed from propagation measurements made inover ten di�erent buildings. The simulator allows the choiceof the carrier frequency, the building type, the topography,the velocity of the mobile receiver, and the transmitter-receiver separation. After the parameters are chosen thesimulator produces multipath power delay pro�les at �=4separations along 4:5 � tracks.Fig. 6 shows the frequency response of the channel forthe transmitter position S1 in the laboratory room. Thepath loss has been computed as the average power lossover the measured frequency band. The path loss exponenthas been determined separately for LOS and NLOS topo-graphies. In both cases the path loss PL(d) in dB has beendrawn as a function of the logarithm of d=d0. From theresults the path loss exponent n has been determined by�tting a regression line through the measured points. TableI compares experimental and computed results.Fig. 7 compares measured and simulated impulse respon-se pro�les for a transmitter receiver distance of 5 m andLOS topography (S1). Table II shows measured and simu-

Fig. 5. Environment for radio channel measurement.TABLE IComparison of measured and simulated path loss exponents.Topography n measured n simulatedLOS 2:1 2:3NLOS 2:7 2:9lated estimates for the rms delay spread. Measurementsand simulations indicate that the mean delay spread is si-gni�cantly larger in NLOS topographies. Summarizing, wecan say that experimental and computed data coincide wellso that the SIRCIM simulator can be used as a basis forrealistic simulations of indoor wireless systems.V. Simulation ResultsSimulations have been carried out to evaluate the perfor-mance of the proposed chirp spread spectrum system. TheMATLAB software has been used for the simulation andall computations have been carried out in baseband. Boththe chirp �lters and the radio channel model have beenimplemented as discrete frequency transfer functions. Thee�ect of the multi-path fading radio channel has been mo-

5

2.4 2.41 2.42 2.43 2.44 2.45 2.46 2.47 2.48 2.49 2.5−65

−60

−55

−50

−45

−40

Frequency [GHz]

Pat

h Lo

ss [d

B]

Fig. 6. Frequency response of the radio channel for transmitter po-sition S1 in the laboratory room.

0 50 100 1500

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Time [ns]

Rel

ativ

e P

ower

Fig. 7. Comparison between measured and simulated indoor channelimpulse responses for a transmitter receiver separation of 5 m andLOS topography.deled with the simulation software SIRCIM as is describedin section IV. In the receiver an envelope detector followedby a decision device with a �xed threshold has been imple-mented and ideal symbol timing and clock recovery havebeen assumed.For the ideal case of an additive white Gaussian noi-se (AWGN) channel, the e�ect of the time overlapping ofconsecutive chirp signals is plotted in Fig. 8 in terms of biterror rate (BER) vs. the signal-to-noise-ratio Eb=N0 whichis the ratio of the average bit energy to the noise powerdensity. The solid line represents the theoretical result forfully orthogonal signals [16]. The result for non-overlapping2�s long chirp signals (at a data rate of 500 kBit/s, -�- cur-ve) almost approaches the theoretical limit. Only for highvalues of Eb=N0 the chirp system is slightly worse due tothe fact that up- and down chirps are not exactly orthogo-

TABLE IIComparison of measured and simulated estimates for the rmsdelay spread.Location T-R sepa-ration [m] �rms [ns]measured �rms [ns]simulatedS1 (LOS) 5 12.8 14.1S2 (NLOS) 12 37.9 36.6S3 (NLOS) 22 48.2 34.7

0 2 4 6 8 10 12 14−5.5

−5

−4.5

−4

−3.5

−3

−2.5

−2

−1.5

−1

−0.5

0

Eb/N0 in [dB]

Log(

BE

R)

Fig. 8. BER vs. Eb=N0 for orthogonal signals. Theoretical limit after[16] (solid line), non-overlapping 2�s long chirp signals (-�-), andoverlapping 2 �s long chirp signals with a data rate of 2 MBit/s(-�-).nal. So if an up-chirp signal is compressed in an up-chirp�lter (which is the matched �lter for a down-chirp signal)a small cross-correlation peak appears at the �lter output.This can be seen in Fig. 9 where the output of the recei-ver chirp �lters after envelope detection is depicted. Theoccouring intersymbol interference (ISI) between up- anddown-chirp grows as the overlapping of consecutive chirpsincreases which is demonstrated in Fig. 8 for a data rateof 4 MBit/s with 2 �s long chirp signals (-�- curve). Herethe deviation from the theoretical limit starts already atvalues of Eb=N0 as low as 4 dB.The ISI between up- and down-chirp mainly limits theactual data rate. The time dispersion due to the multipathfading channel causes distortions in the receiver signal, too.As shown in Fig. 9 these multipath distortions are repre-sented by multiple pulses in the compressed signal. Delayspreads in an indoor environment at UHF frequencies aretypical below 50 ns. Therefore, these distortions do notcontribute signi�cantly to the BER compared to the cross-correlation between up- and down-chirp. As can be seen inFig. 9 the distortions due to ISI are increasing if severalconsecutive chirps of one kind are transmitted. This casemust be prevented by using proper source coding.

6

Fig. 9. Simulated output of the receiver chirp �lters after envelopedetection. VI. Experimental ResultsA hardware demonstrator has been set up to prove thefeasibility of the proposed system as well as to demonstra-te its robustness for wireless indoor data communications.The schematic of the demonstrator is depicted in Fig. 10.After combining the two quasi-orthogonal chirp-signals the

Fig. 10. Schematic of the hardware demonstrator.IF-signal is up-converted to the 2.45 GHz band. The trans-mitted power was +13 dBm. We used time-duplex opera-tion for the system. A Tx/Rx switch isolates the transmit-ter from the receiver path. A low noise ampli�er combinedwith a bandpass �lter in the receiver path is followed by amixer and an automatic gain control ampli�er. The down-converted signal is fed into the matched dispersive SAWdelay lines, by which the signal is compressed in time, thus

enhancing the amplitude for sampling. After a simple butreliable envelope detector the signal is processed by an ad-aptive threshold decision device followed by the interfacelogic as well as automatic frequency and automatic gaincontrol circuits. The output of the receiver �lter after en-velope detection for a data rate of about 3.2 MBit/s isshown in Fig. 11. There is a good qualitative agreement

Fig. 11. Measured output of the receiver �lter after envelope detec-tion.between the measured and the simulated output signalsof the chirp-�lters in the receiver after envelope detecti-on. The slightly higher noise in the measured result canbe attributed to the transmitter and receiver electronicswhich has been modeled ideally for the simulation. Bothsimulations and measurements indicate that for data ratesof a few MBit/s and for typical delay spreads in indoorenvironments the simple threshold device in the receivercan be used. For higher data rates more elaborate receiverstructures like time-windowing receivers, Rake receivers orhigher-order modulation receivers have to be used and theweighting of the chirp �lters should be optimized regardingpulse energy, sidelobe suppression, and cross-correlation.VII. ConclusionWe have presented performance results of a low-cost wi-reless communication system based on SAW chirped delaylines fabricated on LiTaO3-X112rotY. The system uses bi-nary orthogonal keying as a modulation scheme to providea both robust and low-cost solution. Simulations show thatour system almost attains the theoretical BER limit forthe AWGN case. For use in indoor and industrial environ-ments with their strong multipath fading distortions as wellas severe electromagnetic emissions from other devices ourresults from simulations and measurements made with ahardware demonstrator indicate that the system performswell up to bit rates of 5 MBit/s which easily covers manycommercial applications such as e. g., wireless Internet ac-cess. The achievable data rate is limited by the chirp signalduration and the chirp signal overlapping, respectively. Forhigher data rates more re�ned transmission techniques andmore complex receiver structures have to be implemented.

7AcknowledgmentThe authors would like to acknowledge the many fruitfuldiscussions with R. B�achtiger, M. Loder, and H. K�upfer, allof Siemens Schweiz AG, Z�urich, Switzerland.References[1] B. P. Crow, I. Widjaja, J. G. Kim, P. T. Sakai, "IEEE 802.11 Wi-reless Local Area Networks", IEEE Communications Magazine,vol. 35, no. 9, September 1997, pp. 116-126.[2] R. C. Dixon, Spread Spectrum Systems with Commercial App-lications, New York: Wiley, 1994.[3] D. P. Morgan, Surface-Wave Devices for Signal Processing, 2ndEdition, Elsevier, Amsterdam, 1991.[4] W. Hirt, S. Pasupathy, "Continuous Phase Chirp (CPC) Signalsfor Binary Data Communication{Part I: Coherent Detection andPart II: Noncoherent Detection", IEEE Trans. on Communica-tions, vol. COM-29, pp. 836-858, June 1981.[5] M. Kowatsch, J. T. La�erl, "A Spread-Spectrum Concept Com-bining Chirp Modulation and Pseudonoise Coding", IEEETrans. on Communications, vol. COM-31, no. 10, pp. 1133-1142, Oct. 1983.[6] D. Penunuri, "Recent Progress in SAW Filters at GHz Frequen-cies", IEEE MTT-S International Microwave Symposium Digest1997, vol. I, pp. 169-172.[7] M. Hikita, N. Shibagaki, A. Isobe, K. Asai, K. Sakiyama, "Re-cent and Future RF SAW Technology for Mobile Communicati-ons", IEEE MTT-S International Microwave Symposium Digest1997, vol. I, pp. 173-176.[8] Ch. E. Cook, M. Bernfeld Radar Signals - an introduction totheory and application, Academic Press, New York, 1967.[9] C. C. W. Ruppel, L. Reindl, K. Wagner, "Optimum Design ofLow Time-Bandwidth Product SAW Filters", Proc. IEEE Ul-trasonics Symp. 1994, pp. 61-65.[10] M. Huemer, A. Pohl, W. Gugler, A. Springer, R. Weigel, F. Sei-fert "Design and Veri�cation of a SAWBased Chirp Spread Spec-trum System", IEEE MTT-S International Microwave Sympo-sium Digest 1998, vol. I, pp. 189-192.[11] N. Papadakis, A. Hatziefremidis, A. Tserolas, P. Constantinou,"Wideband Propagation Measurements and Modeling in IndoorEnvironment", International Journal of Wireless InformationNetworks, vol. 4, no. 2, pp. 101-111, 1997.[12] T. S. Rappaport, S. Y. Seidel and K. Takamizawa, "StatisticalChannel Impulse Response Models for Factory and Open PlanBuilding Radio Communication System Design", IEEE Trans.on Communications, vol. 39, pp. 794-807, May 1991.[13] H. Hashemi, "The Indoor Radio Propagation Channel", Procee-dings of the IEEE, vol. 81, no. 7, pp. 943-968, July 1993.[14] T. S. Rappaport, Wireless Communications: Principles andPractice, Prentice Hall, 1996.[15] D. M. Krizman, B. J. Ellison, T. S. Rappaport, SIRCIM Plus:Simulation of Indoor Radio Channel Impulse Response Modelswith Impulse Noise, Rev 1.0, December 1996.[16] J. G. Proakis, Digital Communications, 3rd Edition, New York:McGraw-Hill, 1995.