Spread Spectrum Communications Using Chirp Signals

A. Springer', W. Guglerl, M. Huemerl, L. Reind12, C.C.W. Ruppe13, R. Weigel'

'University of Linz, Institute for Communications and Information Engineering, Austria
Email: springer@mechatronik.uni-linz.ac.at,
2University of Clausthal, Institute for Electrical Information Technology, Germany.
3Siemens AG, Corporate Technology, ZT KM 1, Munich, Germany.

Ahtmct- We report on the use of broadband chirp s i p acoustic wave (SAW) chirped delay lines [5]. These de-
nals for spread spectrum systems in indoor applications. vices can be realized at small size and low cost and due to
The presented system concepts make use of chirp transmis-
sion and pulse compression. Different modulation schemes the analog correlation process the complex synchronization
for chirp signals resulting in different system performance circuits can be economized.
and complexity are compared in terms of bit error rate for After an introduction into the theory of chirp signals
the AWGN channel and for frequency selective indoor radio
channels. We present simulations and measurement results
we describe different incoherent and coherent modulation
from demonstrator systems which use surface acoustic wave schemes for chirp spread spectrum systems. Simulation
(SAW) devices for the generation and matched Altering of results and first measurements obtained with a hardware
the chirp signals. RF and IF frequency and transmission demonstrator are given.
bandwidth of the presented systems are 2.4 GHz, 348.8 MHa,
and 80 MHz, respectively. Due to the processing gain of 16
- -
d B made possible by the use of SAW devices and the large 11. CHIRPTHEORY
transmission bandwidth the system is insensitive against fre- A chirp waveform [3] can be written as
quency selective fading, CW interference and noise.
s ( t ) = a(t)cos [ O ( t ) ] (1)
I. INTRODUCTION where O(t) is the phase, and a(t) is the envelope of the
Indoor wireless communications has gained increasing chirp signal which is zero outside a time interval of length
attention over some time and its market share is expect- T. The instantaneous frequency is defined as
ing to grow rapidly in the next years due t o advantages 1 dO
over cable networks, such as mobility of users, elimination fM(t) = --.
2n dt
of cabling and flexibility. Typical applications are cordless
phone systems and wireless local area networks (WLAN's) The chirp rate is defined by
for home and office applications and flexible and mobile dfM 1 dLO
data transmission links between sensors, actuators, robots, p(t) = - = -- (3)
dt 27r dt2
and controller units in industrial environments. Due t o
the hostile electromagnetic environment, which includes se- and represents the rate of change of the instantanous fre-
vere electromagnetic emissions from other devices as well quency. We call waveforms with p(t) > 0 upchirps an
as heavy distortions due to multipath propagation [ l ]the those with p ( t ) < 0 down-chirps. For a linear chirp p ( t )
robustness of the communication link is an extremely im- is constant, and hence f ~ ( t is) a linear function of t, and
portant feature of a wireless communication system. @(t) is a quadratic function. If we take the waveform t o be
The spread spectrum technology is especially well suited centered at t = 0 it can be written as
to provide such a robust data transmission even in very s ( t ) = a(t)cos [27rfot + 7rpP + VO] (4)
noisy radio environments [2]. The critical operations in
spread spectrum systems are the spreading and despread- where fc is the center frequency and a(t) = 0 for It1 > T / 2 .
ing functions in the transmitter and receiver. In common It is convenient t o define the bandwidth B as the range of
system concepts (direct-sequence and frequency hopping) the instantanous frequency, so that
the synchronization of the despreading code is a difIicult
task which needs high computational effort. With the well B = 1p1T. (5)
known FM chirp signals and the associated technique of The impulse response of a matched filter for a linear chirp
pulse compression with its high processing gain, which is signal is again a h e a r chirp signal but with a chirp rate of
widely used in radar systems [3], another kind of spread opposite sign. If a chirp waveform is fed into its matched
spectrum technique can be realized [5],[6],[7]. In this sys- filter the output signal typically has a narrow IF peak at the
tem concept, the spreading is used solely for combating chirp center frequency. If we consider chirp waveforms with
multipath distortions, whereas code-division multiple ac- flat time domain envelopes and take the matched filter to
ces (CDMA) can only be realized if additional coding is be centered at t = 0 an analytical expression for the output
introduced [4]. waveform g ( t ) of the matched filter can be given. We have
The implementation of the spreading and despreading
with chirp signals is easily accomplished by using surface g(t) = W ) * 4 = cpss(t) (6)

(c) 2000 IEEE

0-7803-6323-W00/$10.00 166
 where pes(t)is the autocorrelation function of s ( t ) . It can the systems data rate are the fived time shift At and multi-
be shown [9] that ve8(t)is given by path fading. Since there are no crosscorrelations the exact
BER given in eq. (7) is reached in the AWGN case.
V. n/4 DQPSK MODULATION OF CHIRPSIGNALS
d t )= m 7rBt Traditionally, the phase shift keying (PSK) modulation
modifies the phase of the carrier signal. For the coherent
for -T < t < T. The envelope has its maximum at t = 0, demodulation of any M-ary PSK signal the absolute phase
and its first zeros at t M f l / B . It is convenient to specify of the signal has to be known. To overcome the problems
the pulse width as 1/B. The ratio of the input and out- associated with the necessary phase synchronisation the
put pulse widths is therefore given by the time-bandwidth technique of phase comparison detection (DQPSK) is used.
product T B which is known as compression ratio or pro- This also solves the problems of carrier recovery. In the
cessing gain. Another important parameter is the sidelobe presented system the well known n/4 DQPSK modulation
rejection, which is about 13 dB for chirp signals with rect- scheme is used. Different from traditionally n/4 DQPSK
angular time domain envelope a(t). A common method of systems we do not modulate the carrier. Instead, the IF-
reducing the sidelobes is to apply amplitude weigthing t o pulse which is stimulating the SAW-chirp filter is phase
the chirp signals. modulated.
The schematic of the chirp n/4 DQPSK system is de-
111. BINARY
ORTHOGONAL KEYINGUSING UP- AND
picted in Fig. 3. The input data are split in an inphase-
DOWN-CHIRP SIGNALS
and quadraturephase component (I and Q component) and
As u p and down-chirp signals are almost orthogonal .fed into the modulator. Another input of the modulator is
they can be used for a binary orthogonal keying (BOK) a short IF pulse at the chirp center frequency. The output
modulation scheme as is shown in Fig. 1a) in the time fre- of the modulator, which is the n/4 DQPSK modulated IF
quency plane. If a ’1’-bit is sent, an IF pulse at the chirp pulse, stimulates the transmit chirp (expander) filter re-
center frequency stimulates an upchirp filter, if a ’0’-bit sulting in the desired phase modulated chirp signal. The
is sent, a down-chirp filter is stimulated. To increase the received signal which is disturbed by the frequency selective
data rate beyond the limit imposed by the chirp duration radio channel and additive white noise is fed into the SAW
the chirp signals have to overlap in time. In the receiver compressor filter followed by the n/4 DQPSK demodulator
the signal is fed into the SAW compressor Glters. The u p which is s h o w in Fig. 4.
chirp filter is matched to the ’0’-bit signal, the down-chirp If we assume a signal
filter to the ’1’-bit signal. Due to the overlapping and the
fact that up- and down-chirps are not exactly orthogonal 21 ( t ) = A COS (Ut - 91 f (Pch)
disturbing crosscorrelation functions in both filter outputs
occure. Together with the time spreading of the correla- at the input of the demodulator, the signal after the delay
tion peaks caused by the multipath channel this limits the T s y m and a p h m shift of = wTsymbecomes
achievable data rate. For the same reasons the theoretical
BER of a system using two orthogonal signals 2 2 ( t )= Aces (U (t Tsym) - 9 2 + 0 f v c h ) 1
where 91, 9 2 , (Pch and Tsymare the phase of the present
symbol, the phase of the delayed symbol, the phase shift
due to the channel, and the symbol duration, respectively.
We also assume that the phase of the channel v c h is con-
can not be reached exactly (Fig. 2). stant during a symbol period since the indoor radio channel
does not change significantlyduring this time intervall(200
OW~HOGONAL
IV. BINARY KEYINGUSINGPULSE nsec at a data rate of 10 Mbps). At the output of the de-
POSITION
MODULATION modulator the I- and &-data after low pass filtering and
When applying pulse position modulation (PPM) to sampling axe
chirp spread spectrum transmission only one chirp wave- I
form is needed for transmission and therefore disturbances vr = SA2cos (vi - 9 2 )
due to crosscorrelations do not occur. Furthermore we 1
economize one SAW-filer in the transmitter and one in the VQ = sA2sin(vl - cp2)
receiver. The schematic of this system is depicted in Fig.
1 b). We do not send the data in equidistant time steps. which represents the known constellation in the I-& dia-
If we send a ’1’-bit the chirp signal is sent at a k e d time gram of any 7r/4 DQPSK system. Due to the differential
At before and if we send a ’0’-bit the signal is sent at the detection the phase shift due to the channel is canceled. In
same k e d time after the reference clock. An increase in case of the 7r/4 DQPSK chirp system the data rate is only
the data-rate by overlapping the chirp signals is possible limited by the time duration of the autocorrelation pulse
without any disturbances. Therefor the limiting &tors for and the time spreading due to multipath fading.

0-7803-6323-X/00/$10.00(c) 2000 IEEE 167

 VI. SYSTEMSIMULATION GHz. Regarding measurement results of the BOK demon-
System simulations have been carried out to evaluate the strator we refer to [5]. Fig. 7 shows measurements of the
performance of the proposed modulation concepts. ADS PPM chirp system at a data rate of 1Mbps. The plot above
from HP-EEsof has been used for the simulations and all shows the output of the chirp compressor filter without any
computations have been carried out in complex baseband. multipath fading. The distance between transmitter and
Measured data for the transfer function of the SAW chirp receiver in the laboratory was 2 m and the line-of-sight
filter have been used in the simulations. For the hardware (LOS)component predominated. The different time inter-
demonstrator of the PPM system up- and down-chirp filters valls between the autocorrelation peaks according to the
pulse position modulation can clearly be seen. In the lower
have been designed and fabricated from LiTaO~-Xll2rotY
substrate using standard optical lithography technique. An plot the output of the chirp compressor filter at a distance
optimized weighting of the magnitude of the filter transfer- of about 15 meters with two reinforced concrete walls in
between transmitter and receiver is depicted. At least two
function was employed to obtain an improved sidelobe re-
jection while keeping the compressed pulse narrow. Fig. 5 propagation paths can be distinguished. The limiting fac-
depicts the transfer function and the group delay of an tor for this kind of modulation is the time shift according
up-chirp filter. The center frequency, bandwidth, and time to the PPM and time spreading due to multipath fadiig.
duration of the chirp signals used in the system simulations VIII. CONCLUSION
as well as in the hardware demonstrator are 348.8 MHz, 80
MHz, and 500 nsec, respectively. The chirp rate of the We presented chirp spread spectrum systems for robust
filters is about f40 MHz/ps which results in a dispersion wireless communications in indoor environments. The sys-
time of about 0 . 5 ~ This
. corresponds to a time-bandwidth tems take advance of the low cost, small size and power
product of 16 dB. efficient SAW technology. SAW devices are used for chirp
As is shown in [8]the exact average bit error rate (BER) generation in the transmitter and for pulse compression in
of a n/4 DQPSK system under AWGN conditions is given the receiver. Due to the analog correlation process the usu-
ally complex synchronization task can be greatly simplified.
by
Different modulation schemes were discussed and simula-
tion results as well as measurements of the data through-
put were presented. Simulations and measurements indi-
cate that the proposed systems overcome the disturbances
due to frequency selective fading and CW interferers by
the transmission of ultra broadband signals. Therefore the
presented chirp spread s p e c t d technique is a well suited
solution for very robust wireless communication systems in
where Q is the Marcum &-function. If we assume a data highly distorted environments.
rate of 10 Mbps in our system we get an autocorrelation
pulse every 200 11s. The duration of the autocorrelation REFERENCES
pulse is about 20 ns. In Fig. 6 the theoretical BER of the H. Hashemi, The Indoor Radio Propagation Channel, Proceed-
traditional r / 4 DQPSK system and the BER of our system ings of the IEEE, Vo1.81, "0.7, (1993).
with a data rate of 10 Mbps and ideal data clock recovery R..P. Dixon, Spmad Spectmm Systems with Commercial Appli-
cations, New York: Wiley, (1994).
is plotted. By increasing the data rate up t o 67 Mbps an D.P. Morgan, Surface Wave Devices for Signal Pmcessing, El-
autocorrelation pulse occurs every 30 ns. As can be seen d e r : Amsterdam, (1985).
M. Kowateeh, J. T. Lafferl, A Sprecld Spectrum Concept Com-
the simulated BER coincides well with the theoretical BER bining Chirp Modulation and Pseudonoise Coding, IEEE Tkans-
of a x/4 DQPSK system. If data clock recovery by means actions on Communications, Vo1.29, No.6, (1981).
of a PLL is applied in the simulation, the results from Fig. 6 A. Springer, M. Huemer, L. Reindl, C.C.W. Ruppel, A. Pohl, F.
Seifert, W. Gugler, R. Weigel, A Robust Ultm Brwdband Wire-
show a loss of about 1dB. leas Communication System Using SAW Chirped Delay Lines,
In a further step we considered an indoor radio channel IEEE 'Itansactioneon Microwave Theorie and Techniques,Vo1.46,
with non lime-of-sight (NLOS) condition and intersymbol N0.12, (1998).
M. Huemer, A. Pohl, W. Gugler, A. Springer, R. Weigel, F.
interference (ISI) according to the model used in the IEEE Seifert, Design and Performance of a SAW Based Chirp Sprecrd
802.11 wireless LAN standardization group. For improved Spectrum System, Proc. 1998 IEEE MTT-S International Mi-
crowave. Symposium, (1998).
detection both outputs of the demodulator are integrated W. Gugler, A. Springer, R. Weigel, H.P. Kpfer, Simulation of a
over the duration of one symbol and sampled according to SAW-Baaed WLAN Using Chirp-Pi/A DQPSK Modulation, Pro-
the data clock recovery. Thus we were able to sum up the ceedings 1998 IEEE Ultrasonica Symposium, (1999).
L. Miller, J. Lee, BER Ezpmsions for DifferentiallyDetected m / 4
energy of all radio channel paths during one symbol peri- DQPSK Modulation, IEEE 'Itansactions on Communications,
ode. The BER including the indoor radio channel, the data VO1.46, (1998).
clock recovery and the integration can be seen in Fig. 6. -
Ch. E. Cook, M. Bernfeld, Radar Signals an introduction to
theory and opplication, Academic Press, New York, (1967)
VII. MEASUREMENTS
Demonstrators for the BOK and the PPM chirp modula-
tion schemes were built operating in the ISM band at 2.45

(c) 2000 IEEE

0-7803-6323-X/00/$10.00 168
 -.. 1 ...up chirp
_ - - _ _V...dowochirp

Figure 1. Binary orthogonal keying schemes.

Figure 2. BER of a BOK chirp system using two orthogonal siganls.

chirp filter

n / 4 DQPSK
modulator
IF-pulse

channel
data out
4

j
dat y&
tlfo:
combine

n / 4 DQPSK
demodulator
chirp filter

Figure 3. Schematic of the chirp s/4 DQPSK system.

(c) 2000 IEEE

0-7803-6323-X/00/$10.00 169
 phase shift
I-data
90" b

IF-signal I I
I
I
I ,
I

delay phase shift

Figure 4. Chirp a/4 DQPSK demodulator.

>:
c
e

Figure 5. Measured transfer function and group delay of an optimized weighted chirp filter.

Eb/NO in dB
I3 2 4 6 8 in 12

-1

-8 -2
m
-F
v

-3
+ 67 Mbps, no data-clock recovery
0 10 Mbps, data-clock recovery
-4
- 10 Mbps, indoor channel, integrate
10 Mbps, data-clock recovery
-5 indoor channel. inteerate

Figure 6. BER of the chirp n/4 - DQPSK system.

i :ULJ
/:
.
. .
. .... .
.
I,.
b-wwL+
.
.
. .
......,.'......' .,....
.
..

Figure 7. Measurements of a chirp PPM system.

0-7803-6323-X/00/$10.00 (c) 2000 IEEE 170