Professional Documents
Culture Documents
of Galileo Signals
Adina Burian, Elena Simona Lohan, Markku Renfors
Institute of Communications Engineering, Tampere University of Technology
P.O.Box 553, FIN-33101, Finland
adina.burian@tut.fi, elena-simona.lohan@tut.fi, markku.renfors@tut.fi
Abstract— The Binary Offset Carrier (BOC) modulation which is multiplied by a rectangular sub-carrier (e.g., with +1 and -1
has been proposed for future Galileo and GPS M-code signals, values) at sub-carrier frequency. The modulation parameters
provides a higher spectral separation from BPSK-modulated are equal to m = fsc /fref and n = fc /fref where fsc is
signals, such as GPS C/A code. The absolute value of the auto-
correlation function of a BOC signal has a narrower main lobe, the sub-carrier frequency, fc is the code rate and fref =1.023
which may increase the resolution of delay estimates, but also MHz is the reference frequency. Thus, the power spectrum
presents deep fades, which may lead to a higher number of timing is split into two symmetrical components around the carrier
hypotheses to acquire the signal. In order to get rid of these frequency. The common baseline for OS structure (agreed by
ambiguities, several approaches have been proposed in literature, US and European negotiation in June 2004) employs the sine
which provide an unambiguous BPSK-like shape of correlation
function. In this paper we analyze, compare and develop further BOC(1,1) modulation, which uses a 1.023 MHz square-wave
two BPSK-like methods which allow to acquire a BOC-signal sub-carrier modulated by spreading code chips at a chip rate
unambiguously. The focus is on hybrid search, where several fc =1.023 MHz [8]. For PRS services, both sine and cosine
time-frequency bins are searched in parallel. BOC(15,2.5) have been proposed [8]. Other BOC modulations
We introduce here a modified version of a BPSK-like method have also been considered, such as BOC(15,10), BOC(10,5)
which decreases the receiver complexity and is valid for both
odd and even BOC orders. We analyze both single-side band or BOC(10,4) [5], [6].
(SSB) processing (i.e., only one band is used) and dual-side band In order to eliminate the ambiguities of the ACF envelope,
(DSB) processing (i.e., upper and lower bands are combined different methods have been approached in literature [3], [4],
non-coherently). While eliminating the ambiguities in auto- [5], [6], [7]. The main idea behind these ”BPSK-like” methods
correlation function, both SSB and DSB processing present some (generic name proposed in [5]) is that the BOC-modulated
performance degradation, induced by the band selection and
non-coherent processing. The analysis is done in the presence signal can be obtained as the sum of two BPSK-modulated
of multipath fading channels. As a benchmark, we keep also the signals, located at positive and negative sub-carrier frequen-
ambiguous BOC processing. We consider parameters specified in cies. The effect of sub-carrier modulation can be removed by
the proposals for Galileo system Open Service (OS), respectively using a pair of single-sideband correlators. We may have a
Publicly Regulated Service (PRS). single-side band (SSB) receiver, where either the negative or
the positive of the sidebands correlators is used, or a dual-side
I. I NTRODUCTION
band (DSB) receiver, when both bands are combined non-
The BOC modulation has been selected for both Galileo coherently. Due to filtering and correlation losses, the BPSK-
and modernized GPS signals [1]. The main target of BOC like methods bring some degradation in the signal level.
modulation has been to provide a better spectral separation In [3], [5] it is asserted that a SSB BPSK-like method
with existing BPSK-modulated GPS signals, while allowing induces at least 3 dB degradation in SNR. If DSB processing
optimal usage of the available bandwidth for different GNSS is used, this loss can be partially compensated, excepting
signals [1], [2]. Also, BOC signals present a narrower main anyway the squaring losses in non-coherent integration [5].
lobe of the absolute value of their autocorrelation function Compared to coherent processing of both sidebands, non-
(ACF), which enhances the tracking accuracy. On the other coherent processing looses about 0.5 dB of SNR.
hand, the additional fades which appear in ACF within the The ’BPSK-like’ methods proposed so far in the literature
two-chip intervals may induce a missed detection due to a fall in of the following two categories: either the main lobes
zero (or very low) sampling point and may lead to a longer of the signal and of the reference BOC-modulated PRN code
acquisition time. In consequence, the necessary step to search are selected (via filtering) and then correlated [3], [4], [7],
a given time-uncertainty window, ∆tbin , should be small or both the main lobes and secondary lobes between of the
enough in order to find the main lobe of ACF. Therefore, the received signal are kept and the reference code is based on
computational complexity in acquisition process is increased, the BPSK-modulated PRN code [5], [6]. We will refer to
the computational load being inversely proportional to the the former method as Fishman & al. method [3], [4], [7],
square of step time bin ∆tbin , as reported in [3]. and to the later method as the Martin & al. method [5], [6]
A BOC(m, n) signal (m and n are not necessarily integers) (from the authors’ names). The Martin & al. method, if the
is created by a square sub-carrier modulation, where the signal reference code is properly selected, has the advantage of lower
5211
1-4244-0355-3/06/$20.00 (c) 2006 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE ICC 2006 proceedings.
Received BOC-modulated signal
PSD
f
fcarrier -fsc fcarrier fcarrier +f sc
single
Similar processing as in right sideband,
SB
but uses Lower Side Band Filters
statistic
Fig. 1. Method 1 (Fishman & al.). Selecting the main lobes of the BOC-modulated received signal and reference PRN sequence and correlate them.
complexity, as we will discuss in our paper. On the other as expected, for smaller time steps, the performances of BPSK-
hand, its performance is dependent on the BOC-modulation like methods may be worse than that of the ambiguous BOC
order NBOC (defined here as NBOC = 2fsc /fc ) and may be approach. Therefore, as suggested in [5], [6] the ’BPSK-like’
slightly lower than the performance of Fishman & al. method. methods can be used in initial energy search, while sweeping
The purpose of this paper is to analyze these non-ambiguous the uncertainty time-domain with a higher time step ∆tbin .
BPSK-like approaches, which differ through the manner the Once the energy is found with enough confidence, the receiver
received signal and reference code are selected and processed: may return to the BOC processing and use a finer time step.
the first method is based on Fishman & al. approach, where
II. BPSK- LIKE METHODS
two filters are used and only the upper/lower main lobes are
selected, both for the incoming signal and the reference code. A. Method I (Fishman & al. method)
The second method is a modified version of the Martin & In this method, the receiver selects only the main lobes of
al. method, where only one filter is used, which includes the BOC-modulated received signal and the reference code.
both the main lobes and secondary lobes between them. Our The block diagram of this approach, illustrated in Fig. 1, it
proposed method differs from the Martin & al. method in two has been proposed in [3] and [4], and later analyzed in [7].
aspects: first, the reference sequence here is the pseudorandom The main lobe of one of the sidebands (upper or lower) of
(PRN) code kept at sub-sample level, and not the shifted PRN BOC-modulated received signal is selected (via filtering) and
code; second, our method is valid for both even and odd it is correlated with a filtered PRN BOC-modulated reference
BOC modulation orders, not only for even BOC modulation code, having the tentative delay τ and the tentative Doppler
orders which was the case in Martin & al. method. Our frequency fD . The BOC-modulated reference sequence is
proposed method has the advantage of a lower complexity obtained in a similar manner with the received signal, filtering
implementation than Fishman & al. method, as explained in out the main lobe. Hence, the SSB Fishman & al. approach
[9]. needs three sideband selection filters (one for the real code
In order to evaluate the performance of our method, we and two for the in-phase and quadrature components of the
have selected for comparison the Fishman & al. approach received signal). The Integrate & Dump (I&D) block performs
(and not Martin & al.), because this method is patented and the coherent integration on Nc ms. Further non-coherent
may bring an improvement in results. Also, we have verified integration on Nnc blocks may be employed to reduce the
through simulations that Martin & al. technique provides very noise.
similar performance to our method. We compare both SSB The correlation function between the received signal and
and DSB cases of each method. As a benchmark, we keep reference code, on each sideband, will be unambiguous and
also the ambiguous BOC approach, where all spectrum (of will resemble the ACF of a BPSK-modulated signal. However,
signal and reference code) is used and the reference code is the shape of resulting ACF is not exactly the one of a BPSK-
the BOC-modulated PRN sequence. modulated signal, since there are information losses due to
The performance of the studied methods is tested for a hy- selection of main lobes. This method has rather high complex-
brid search acquisition in multipath-fading channels. We will ity due to the fact that six complex filters are needed for DSB
show that, for a step of the time bin equal to half chip, there processing [4]. A less complex version of this implementation
is always an improvement in performance when DSB method will be to use just one of sidebands (SSB approach), but in
is used, but the SSB processing does not always improve the order to compensate for SNR degradation, a longer dwell time
performance compared to ambiguous-BOC processing. Also, is needed.
5212
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE ICC 2006 proceedings.
Filtered BOC-modulated signal (i.e. BOC(10,5))
PSD
Interference from GPS C/A
f
fcarrier -fsc f carrier fcarrier +f sc
single
Similar processing as in right sideband, but shifted at SB
the center frequency of band with exp(+j . 2. pi. fSC . a. t) statistic
Fig. 2. Method 2 (proposed; modified Martin & al.). Select the main lobes of received signal and the secondary lobes between them, shift them close to
the carrier frequency and correlate with a BPSK PRN reference code.
5213
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE ICC 2006 proceedings.
SinBOC(1,1), DSB
where ck,n is the k-th chip value corresponding to the n-th 1
data symbol, SF is the spreading factor or the code epoch
length (e.g., for GPS C/A signal, SF = 1023), p(·) is a train 0.8
of rectangular pulses, and Ts is the sampling interval (Ts =
1/(Ns fc )). 0.6
ACF
Depending on the method, at the receiver part either only
0.4
the main lobes of the incoming BOC signal are filtered (using
two complex filters per each band, upper or lower) or the 0.2
whole bandwidth of signal containing the main lobes and the
secondary lobes between them is selected (by using only one 0
−1.5 −1 −0.5 0 0.5 1 1.5
filter per band, since the reference signal remains unfiltered). Delay error [chips]
For the second method, after filtering, the signal is shifted SinBOC(10,4), SSB
1
down or up around the center frequency by multiplication
with exp(−j2πfsc at) and exp(+j2πfsc at), for upper and 0.8
the lower side-bands, respectively. The shifting parameter a
depends on the order of BOC modulation used and has been 0.6
ACF
found to be equal to:
0.4
1 if NBOC even, Sin- and CosBOC
NBOC −1
a= if NBOC odd, SinBOC (3)
NN BOC
BOC +1
0.2
5214
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE ICC 2006 proceedings.
computed as: test statistic was built over time-frequency windows of size 12
chips × 1 kHz (i.e. hybrid search [11]). The maximum fre-
RM SE[chips] = E(|τ − τ|) (6)
quency incertitude range in the searching space is considered
where E(·) is the expectation operator, τ is the true first 9 kHz.
path delay, τ is the estimated delay of the first path, in
Rayleigh fading channel, 2 paths, SinBOC(15,10), Nc=10, Nnc=4
case of acquisition. The RMSE in meters was computed as 300
RM SE[m] =RM SE[chips] ∗ c, where c is the speed of light. ambigous BOC
Fishman met. SSB
250 Fishman met. DSB
Nakagami fading channel, 2 paths, SinBOC(1,1), Nc=30, Nnc=3 Proposed met. SSB
Proposed met. DSB
500 200
RMSE [meters]
ambigous BOC
450 Fishman met. SSB
Fishman met. DSB 150
400
Proposed met. SSB
350 Proposed met. DSB
100
RMSE [meters]
300
250 50
200
150 0
30 31 32 33 34 35 36 37 38
CNR [dBHz]
100
50
120 xmax =2 chips, Nc =20 ms, Nnc =2 blocks and mobile speed 45
100 km/h. Similar curves (both RM SE and P d) were obtained
80 for other channel profiles and other BOC modulation orders
60 as well, but are not reproduced here due to lack of space.
40 From both figures it can be observed that the best results are
20 provided by the Dual Side Band processing, with an improve-
0 ment of about 2 dB when comparing with SSB processing.
30 31 32 33 34 35 36 37 38
CNR [dBHz] Also, it can be observed that for DSB processing, our method
approximates the performance of Fishman & al. method and
there is a small degradation in the SSB case, for the proposed
Fig. 6. RM SE vs CN R for a Rayleigh fading channel, SinBOC(10,5)
(even BOC order), ∆tbin =0.5 chips. method. This degradation can be compensated by employing
a longer integration time and it may be worthy considering
The detection probability Pd is the probability to have the the lower computational complexity of proposed method. We
decision variable higher than a decision threshold γ, provided also note that the ambiguous BOC may be outperformed by
that we are in correct bin. The false alarm probability Pf a is SSB processing.
the probability that decision variable is higher than γ, if we are Comparing with even BOC orders, for odd orders (i.e.
in an incorrect bin. In order to have a fair comparison between SinBOC(15,10)) the performance of proposed method is de-
different decision variables, the false alarm probability Pf a teriorating, as can be observed from Fig. 7. This degradation
was set from beginning to a fixed value (here, Pf a = 10−2 ) is expected, observing the small fades from ACF envelope for
and the corresponding threshold γ was computed in order to an odd BOC modulation order (Fig. 4, the lower plot). The
meet this fixed Pf a . The Pd was computed based on estimated channel is 2-paths Rayleigh distributed, with xmax =1 chip,
threshold γ. Nc =10 ms, Nnc =4 blocks and the mobile speed 25 km/h.
The studied methods were tested for different fading channel Fig. 8 presents the simulation results for a higher BOC
profiles with multipath propagation. The oversampling factor modulation order, namely SinBOC(15,2.5), proposed for PRS
was set to Ns =4 and the step time bin ∆tbin =0.5 chips. The Galileo signals. We expect to get similar performance curves
5215
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE ICC 2006 proceedings.
Fading Rician channel, 2 paths, SinBOC(15,2.5), Nc=20, Nnc=4
1 ∆tbin =0.5 chips, in order to speed the acquisition process and
0.9 ambigous BOC to reduce the implementation complexity.
Fishman met. SSB
0.8 Fishman met. DSB V. C ONCLUSIONS
Proposed met. SSB
0.7
Detection probability
Proposed met. DSB In this paper we have proposed a modified low complexity
0.6
’BPSK-like’ method in the context of BOC-modulated signals
0.5
for hybrid search acquisition and we have compared it with
0.4
traditional sideband processing method of Fishman & al. Our
0.3
method has the advantage of a lower computational complexity
0.2 at the cost of a very small CNR degradation, comparing to
0.1 Fishman & al. BPSK-like method, due to the fact that the
0
30 31 32 33 34 35 36 37 38
number of filters is reduced and the reference code sequence
CNR [dBHz] is kept as a sequence of ±1s, which allows an efficient
correlation implementation, as reported in [10].
Fig. 8. Simulation results for a Rician fading channel, SinBOC(15,2.5), Compared to ambiguous BOC processing, there is a sig-
∆tbin =0.5 chips: Pd vs CN R nificant gain if dual sideband processing is used in order to
remove the ambiguities of correlation function, since higher
Rayleigh fading channel, 3 paths, SinBOC(1,1), Nc=20, Nnc=4 steps in searching the timing hypotheses can be used. Also,
1
we have showed that single sideband processing does not
0.9 ambigous BOC
Fishman met. SSB always outperform the ambiguous-BOC approach due to the
0.8 Fishman met. DSB correlation losses. The results were checked via simulations
0.7 Proposed met. SSB
Detection probability
0.3
R EFERENCES
0.2
[1] J.W. Betz, “The Offset Carrier Modulation for GPS moderniza-
tion”, in Proc. of ION Technical meeting, 1999, pp. 639–648.
0.1 [2] J.W. Betz, ”Design and Performance of Code Tracking for the
GPS M Code Signal”, MITRE Technical Paper, Sep. 2000 .
0 [3] P. Fishman and J.W. Betz, “Predicting performance of direct
30 31 32 33 34 35 36 37 38
CNR [dBHz] acquisition for the M-code signal”, in Proc. of ION NMT, 2000,
pp. 574–582.
[4] J. Betz and P. Capozza, “System for direct acquisition
of received signals”, US Patent Application Publication
Fig. 9. Simulation results for a fading Rayleigh channel, small step 2004/0071200 A1, Apr 2004.
∆tbin =0.125 chips: Pd vs CN R [5] N. Martin, V. Leblond, G. Guillotel, V. Heiries, “BOC(x,y)
signal acquisition techniques and performances”, in Proc. of ION
GPS/GNSS 2003, 9-12 September 2003, Portland, pp. 188–198.
[6] V. Heiries, D. Roviras, L. Ries, V. Calmettes, ”Analysis of Non
for CosBOC(15,2.5), since the ACF for Sin- and Cos- Ambiguous BOC Signal Acquisition performance”, ION GNSS
17th International Technical Meeting of the Satellite Division,
BOC(15,2.5) are very similar. The channel is a 2-paths Rician 21-24 Sept. 2004, pp. 2611–2622.
distributed fading channel with mobile speed 45 km/h, Nc =20 [7] S. Fischer, A. Guerin, S. Berberich, ”Acquisition concepts for
Galileo BOC(2,2) signals in consideration of hardware limita-
ms and Nnc =5 blocks. The gap between performances of tion”, Proc. of IEEE Vehicular Technology Conference, vol. 5,
DSB and SSB processing is maintained also here. Also, it pp. 2852–2856, May 2004.
[8] G. Hein, M. Irsigler, J.A. Rodriguez, T. Pany, ”Performance
can be observed that the ambiguous BOC performs worse that of Galileo L1 signals candidates”, CDROM Proc. of European
SSB processing due to an increased number of ambiguities. Navigation Conference GNSS, May 2004.
[9] A. Burian, E.S. Lohan, V. Lehtinen, M. Renfors, “Complexity
The performances of Fishman & al. method and that of our Considerations for Unambiguous Acquisition of Galileo Sig-
proposed method are roughly similar, for both SSB and DSB nals”, accepted for presentation at 3rd Workshop on Positioning,
Navigation and Communication 2006 (WPNC 2006), Hannover,
processing. Germany, March 2006.
The results in case of a smaller step time bin 1/(Ns NBOC ) [10] A. Lakhzouri, E. S. Lohan, and M. Renfors, “Reduced-
Complexity Time-Domain Correlation for Acquisition and
chips are presented in Fig. 9 for a 3-paths fading Rayleigh Tracking of BOC-Modulated Signals”, in CDROM Proc. of
channel and SinBOC(1,1) modulation. Here the oversampling 1st ESA Workshop on Satellite Navigation User Equipment
Technologies (NAVITEC), Dec 2004.
factor Ns =4, xmax =2 chips, Nc =20 ms, Nnc =4 blocks. As [11] E.S. Lohan, A. Lakhzouri, and M. Renfors, ”Selection of
it can be observed, for a smaller step time bin ∆tbin =0.125 multiple-dwell hybrid-search strategy for acquisition of Galileo
chips, the ambiguous BOC method provides superior per- signals in fading channels”, Proc. of PIMRC 2004, vol. 4,
pp.2352-2356, Barcelona, Spain, Sep. 2004.
formance, compared with both DSB and SSB processing.
Also, the SSB processing of Fishman & al. method gives an
improvement of about 0.3 dB compared with the SSB of pro-
posed method. Therefore, the non-ambiguous BOC processing
makes sense to be used only with high enough steps, namely
5216
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE ICC 2006 proceedings.