Int. J. Communications, Network and System Sciences, 2010, 3, 730-736
doi:10.4236/ijcns.2010.39097 Published Online September 2010 (http://www.SciRP.org/journal/ijcns)
S tatistical Approach to Mitigating 3G Interference to GPS
in 3G Handset
Taher AlSharabati, Yinchao Chen
University of Sout h C arol i n a C ol u mbia, South Caroli n a, USA
Received June 10, 2010; revised July 18, 2010; accepted August 22, 2010
Abstract
In this paper, we show how statistical decision theory can be used to solve real life product problems. Global
Positioning System (GPS) performance in a mobile handset degrades whenever it is simultaneously used
with a 3G data or voice service. This degradation is due to the 3G transmitter interference. Mitigation meth-
ods to interference in GPS have been proposed. However, most of these methods depend on hardware and
signal processing or just hardware solutions. In some cases, it maybe difficult to implement these hardware meth-
ods, especially in mobile handsets due to its small size, printed circuit board (PCB) layout issues and the
added cost. In this paper, a novel signal processing statistical algorithm approach is proposed to mitigate the
3G interference to GPS. This statistical approach utilizes the knowledge of the statistical characteristics
and distributions of both the GPS signal and noise. Then the method utilizes the probabilities to make a sta-
tistical decision to remove the effect of noise. This method does not require room on the PCB of the mobile
handset and therefore no layout challenges arise. In addition, cost is minimized and the product turn in cycle
is shortened. This paper offers theoretical as well as practical insight to the GPS operation during 3G call inside
the mobile phone.
Keywords: Detection and Estimation, Global Positioning System (GPS), Satellite Acquisition, Wide Band
Code Division Multiple Access (WCDMA)
1. Introduction
Global Positioning System (GPS) plays a major role in
our daily lives. Its applications extend from naval, aero-
space, first time emergency responders and location
based-finding and directions in cell phones to name a few.
The existence of GPS in cell phones presents its own
challenges since it co-exists with other services such as
3G WCDMA and GSM carriers [1].
The spatial distance or proximity of these services and
subsystems and their components inside the cell phone is
very small. Especially since the mobile handset is being
packed with as many services as possible with the smallest
form factor. This distance is small enough to the point
where it does not present enough attenuation to the 3G
transmitted Radio Frequency (RF) signals toward other
receivers inside the handset. This, in turn, will present
itself as interference to the receivers of the services that
exist in the handset especially the GPS receiver. The effect
of the distance on the attenuation or path loss of the
transmitted signal strength is governed by the equation:
2
32.4 10log()
A
rf (1)
where A is the attenuation in dB, f is the frequency of the
transmitted signal and r is the distance between the
transmitter antenna and the GPS receiver.
The spectrum separation or proximity between the
frequency bands of the mobile handset wireless services
is another source of interference. The wireless services
are allocated frequency bands for their transmitters and
their receivers. The closer the frequency of a transmitter
to the frequency of a receiver, the more of interference
the frequency will present. This is due to the non ideal
characteristics of the band pass filter of the transmitter
[2]. As a matter of fact, some filter designs exhibit “fly
back”, which is worse insertion loss characteristics espe-
cially towards the band edges of the filter. These band
edges could be the operating frequencies of another ser-
vice like the GPS. For example, the GPS receive fre-
quency (called L1 band) is 1575.42 MHz, while the DCS
band lowest transmit frequency (or channel) in GSM is

T. ALSHARABATI ET AL. 731
1710 MHz and that of WCDMA band IV is 1712.4 MHz.
The frequency separation is less than 135 MHz.
It should be mentioned that architecture and system
engineers do a thorough interference lineup analysis and
calculations to assess the effects of different scenarios of
the effects of interference of each band to itself and others
with the given components of the subsystems before or
on the onset of any transceiver chipset deployment. The
effects of interference, if not mitigated, on receivers and
the GPS receiver in particular are severe. These effects
could range from degradation in user position accuracy
or loss of satellite acquisition. This situation, when it
happens to a receiver in a product like mobile phones, is
called “de sense” i.e. the receiver is “de sensed”. In [1],
the GPS receiver was de sensed by 5.5 dB. Shown in
Figure 1 is a bar graph of the GPS satellites where eight
of them have been acquired based on an acquisition metric
of 2.5, where the horizontal axis is the satellite number
and the vertical axis is the metric value, (Permission
from [3]) when there is no 3G interference added to the
data collected in [3]. Any satellite with an acquisition
metric of 2.5 and above is considered acquired.
2. Analysis
In the following we present an analysis and examples for
calculating two of the important parameters of the GPS
receiver, namely; the average signal strength and the root
mean squared (RMS) code tracking error.
2.1. Calculating the Average Signal Strength
The average carrier strength ()
av
C (or
s
C) of the ac-
quired satellites can be calculated based on Equation (9)
in [4]. Equation (9) is repeated here for clarity purposes
for the case1Db

and assuming that 20.0332  
(2

is the normalized GPS code tracking error variance
in units of sec2 [5], D is the two sided GPS correlator
spacing and
rc
bT

 (2)
b is the normalized bandwidth, assuming6
r
M
Hz

,c T
is the chip period ):
From this equation, the carrier to noise power ratio
s
o
C
Nis:
dB 21.67
s
o
CdB
N.
Assuming system noise figure (NF) of 2dB and ther-
mal noise power (No) of -174dB/Hz. From:
oso
CNNFCN
av

 (3)
av
C~ = –150 dBm. This is the average signal strength or
carrier power of one of the acquired satellites. Figure 2
shows a plot of the average carrier power of the acquired
satellites of Figure 1. The horizontal axis is the satellite
number, while the vertical axis is the average carrier
signal strength in dBm. Note how the values correspond
to the metric values in Figure 1.
2.2. Calculating the RMS Code Error
One of GPS measurement accuracy parameters is the
root mean squared (RMS) code tracking error in units of
meters. Let’s denote the RMS code tracking error as

then [5],
*c


 (4)
where σ is the standard deviation in units of chips [6] and
c is the speed of light. Each chip period is approxi-
mately1
s

. For example, for the above case where
20.0332
 ,

54.66 m. This is the code tracking
error. Figure 3 displays the RMS code tracking error (in
meters) of the acquired satellites. It is evident from this
plot and that of Figure 2 how the average signal power
affects this parameter. The stronger the power is; the less
the error is and therefore the accurate the user position is.
3. Procedure
3.1. GPS Data and 3G Noise Generation
Being inside the phone, the GPS receiver suffers the
most interference from the 3G transmitters inside the
User Equipment (UE) while they are transmitting. From
this point on, the 3G transmitted signals will be referred
to as 3G noise because they present themselves as noise
to the GPS receiver. The GPS data used in this paper are
data collected in [3], which is data output of the ADC in
that platform. The 3G noise will be simulated on the ba-
seband level by using the WCDMA Quadrature Phase
Shift Keying (QPSK) modulation scheme which is:
2
2(10.5) 11 2
() ()1, 1 Db (9) in [4]
1
2(2)
LL
ss
oo
BBTbD
TC C
bb
cTD
NN














T. ALSHARABATI ET AL. 732
23456789101112131415161718192021222324252627
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15 Acquisition Results
A cq u ired S a tel lite 's N u m b er
Acquisition Me tric
Not ac quired sign als
Acq uired signals
Figure 1. Number of acquired satellites when 3G noise is
not present.
34567891011 12 13 14 15 16 17 18 19 20 21 22 23 242526
-151
-150.5
-150
-149.5
-149
-148.5
-148
-147.5
-147
-146.5
-146
-145.5
-145 Averag e Carrier Power of Acquired Satellites
Acqu ired Satellites' Num b ers
Average Carrier Power Cave (dBm )
Figure 2. Plot of the average power of the acquired satel-
lites.
3 4 5 67 8 910 111213 14 15 16 17 18 19 20 21 22 23 24 25 26
20
25
30
35
40
45
50
55
60
65 RMS Cod e Tracking Error of Acqu ired Satellites
Acquired Satellite s' Num b ers
Code Tracking Error E (m)
Figure 3. Plot of RMS code tracking error of the acquired
satellites.
3()(()cos( )()sin( ))
44
2
(), ()1
IQ
GIFIF
IQ
Vrms
V tdtwtdtwt
dtd t




(5)
where (), () IQdtd t are the in phase and quadrature phase
baseband components of the data streams (which are
random and normally distributed with zero mean and
variance of one) and
I
F
w is the intermediate frequency
of the down converted signal. The 3G noise is sampled at
the rate of the sampling frequency of the ADC in [3].
Then the 3G and GPS data are added together to form the
GPS plus interference signal.
3.2. Determining 3G Noise Level
This task will take into consideration the total system
architecture components of both the 3G transmit (3G TX)
section and GPS receive section. Figure 4 shows a block
diagram of a typical GPS-3G coexistence scenario [1].
The GPS section could be the hardware platform in [3].
Going from right to left; the heart of the GPS (GPS RX)
receiver is the GPS Radio Frequency Integrated Chip
(RFIC). Most of GPS RFICs have on board cross func-
tional sections that span from RF to baseband. The SiGe
4120L [8] is one of SiGe’s GPS line of products. It has a
low noise amplifier (LNA) on board and it outputs serial
IQ data. So, its functionality span from RF to bits. A
bandpass filter is used to filter out any signal outside the
designed bandwidth of the GPS RX. At the forefront is
the GPS antenna and antenna matching circuitry to bring
the antenna impedance to a 50-ohm system. On the in-
terfering 3G TX side, the TX-RX 3G antenna accom-
modates both the WCDMA and GSM carrier bands.
Then, the WCDMA/GSM Power Amplifier (PA) along
with the duplexer.
Based on typical values of 3GPP WCDMA transmitter
noise into GPS, values of 3G noise into GPS [9] (–140
dBm/Hz), duplexer insertion loss and spatial isolation, a
value of rms
V (3G noise) in Equation (5) was calculated.
This value came to be ~0.28 mV. In these calculations
(for illustration purposes only), the spatial isolation is
assumed to be 15 dB, the total receive line up gain is
~100 dB in 50 ohm system. First the power (PdB) of the
interferer was calculated and then it was converted to
voltage based on a 50 ohm system:
dB noiseintoGPS+Lineup GaindB+
IsolationdB =55 dB
P
 (6)
30
dBw
p pdB

 (7)
2
10log(/50)
dBw rms
PV (8)
0.28
rms
VmV

T. ALSHARABATI ET AL. 733
Figure 4. Typical co-existence 3G-GPS scenario.
4. Approach to the Algorithm
Previous published work has focused on suggesting
either a combination of signal processing and hardware
or just hardware solutions to mitigating the interference.
In [7], antenna null steering was suggested to minimize
Wide Band (WB) RF interference (RFI). While this me-
thod works in products other than cell phones, it is diffi-
cult to implement in the cell phone because it requires
complex implementation of antenna design, RF front end
electronics, feedback and signal processing. This method
can not be implemented in the cell phone due to its com-
plexity and immaturity for mobile phone environment.
Other methods are proposed to prevent pulsed RFI [7].
While the methods just mentioned aimed at preventing
interference from occurring, they do not mitigate the
effects of interference after it happens. These methods do
not offset the effects of errors generated once the interfer-
ence is mixed along the line up and down converted to
baseband with the desired GPS signal. Instead, a statisti-
cal method is proposed to mitigate these errors.
The GPS data output from the ADC is Gaussian dis-
tributed with zero mean and therefore their probability
distribution function is normal and is in the form of:
2
2
1
() 2
x
px e



 (9)
Figure 5 shows that distribution and weights in bar
graph of the values and therefore it gives the probability
of each value output from the Analog to Digital Con-
verter (ADC). The red trace is the Normal distribution fit
of the data. The horizontal axis are the values output
from the ADC, the vertical axis are their weights. This
figure does not literally give the probability values, but it
shows the weights of these values.
5. Algorithm and Results
5.1. Mitigation Algorithm Development
The receiver tries to make a decision or best guess about
a symbol si given that r was received. In other words, the
receiver tries to decide the probability of s given that r
was received; p(s|r) based on the test statistic:
(| )()
ii
Tprsps (10)
Since the data follow the Gaussian distribution, then
for N = 2 [10], where N is the number of dimensions of
the signal space:
2
2
()
2
(| )2
i
rs
ie
pr s





 (11)
and (10) becomes;
2
2
()
2
(| )()()
2
i
rs
ii i
e
pr spsps




 (12)
()
i
ps is the probability of the noise free quantized
output GPS signal from the ADC and

is the standard
deviation of the 3G noise. The novelty of this approach is
as follows: since the Coarse Acquisition (C/A) code of
the satellites is comprised of 1’s and –1’s (–1 for 0), the
prototype messages can be reduced to two levels, namely;
1 and –1 and therefore calculating the probabilities in
Figure 5 reduces to calculating the probabilities of the
negative values of the signal and the probabilities of the
positive values of the signal. We just need to recover the
right phase of the signal. The decision reduces down to
deciding whether a 1 or a –1 is being sent. In this case
(12) generates two values:
2
2
()
2
(1 |)()(0)
2
i
rs
ii i
e
ps psps





 (13)
and
2
2
()
2
(1|)()(0)
2
i
rs
ii i
e
psps ps





 (14)
The decision goes to the one with the higher statistic.
5.2. Results
To determine user position, the GPS receiver has to ac-
quire at least four satellites. Figure 6 shows the results of
acquisition of GPS data plus 3G noise when no detection
algorithm is used [2], namely PRN#: 15, 18, 21. From
the results, only three satellites were acquired, not
enough to determine user position. While after using the
detection algorithm, five satellites (Figure 7) are ac-
quired, namely PRN#: 6, 15, 18, 21, and 22. This will
WCDMA/
GSM
Antenna
Internal
LNA
GPS
RFIC
Isolation (I)
TX
BPF
WCDMA
/GSM PA Duplexer
An-
tenna
Match
3G
Antenna
3G TX GPS RX
GPS
Antenna

T. ALSHARABATI ET AL. 734
help the GPS receiver in calculating the user position
even if the user is using the mobile handset in either a
data call or a voice call.
Based on Equations (2), (3) and (10) in [4], values of
the average carrier power (av
C) were approximated and a
comparison plot for three cases was obtained. The first
case corresponds to the acquired satellites in Figure 1
where the results are for the raw data where no 3G noise
was added to the data and therefore no detection algo-
rithm was used. The second case corresponds to the ac-
quired satellites in Figure 6 where 3G noise was added to
the data and no detection and estimation algorithm was
used. The third case corresponds to the acquired satellites
in Figure 7 where 3G noise was added to the data and the
detection and estimation algorithm was used. Figure 8
shows such a comparison. It is evident from this plot how
the proposed detection algorithm and estimation not only
recovers the satellites lost due to noise, but it improves the
average detected power of those satellites.
By using (4), a comparison was made for the code
-0.5 -0.4 -0.3-0.2 -0.100.1 0.2 0.3 0.4
0
5
10
15
20
25
30
Discretize d GP S Data Ou tpu t of ADC
Density
Probability Density Distribution of GPS D ata
GPS data
G PS Da ta F it
Figure 5. Distribution of GPS data from the ADC in [3].
Figure 6. Results of acquisition when no detect ion use d.
GPS data
GPS data Fit
012345678910 1112131415 16 171819 202122232425 26 272829 30 31 32 33
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
A cquisit io n result s
Sa tellite nu mbe r
Acquisition Metr ic
Not acquired signals
Acquired signals

T. ALSHARABATI ET AL. 735
01234567891011 1213141516171819 20 21 2223 24 252627 2829 303132 33
0
1
2
3
4
5
6
7
8
A cqu is itio n R es u l ts wt ih N ew D e te ctio n Al gorit hm
Satellites' number
Ac quisitio n Me t ric
Not acquired signals
Acquired signals
Figure 7. Results of acquisition when using detection algorithm.
3 45 6 7 8 91011 12 13 14 15 16 17 1819 2021 222324 25 26
-151
-150.5
-150
-149.5
-149
-148.5
-148
-147.5
-147
-146.5
-146
-145.5
-145
Comparison of Average C arrier Power of Acqu ired Satellites
Acq uired Satellites' Nu mb ers
Av e rag e Car rier P ow er Cave (dB m )
3G Noise Added-No D etec tion
No 3G Noise-No D etection
3G Noise Added-Detection Used
Figure 8. Comparison of average carrier power of acquired satellites when (1) No 3G noise added. (2) 3G noise added-no de -
tection algorithm used. (3) 3G noise added- detec t ion algor ithm use d.
3G Noise Added-No Detection
No 3G Noise-No Detection
3G Noise Added-Detection Used

T. ALSHARABATI ET AL. 736
345678910 11 12 13 141516 17 181920 21 22 2324 25 26
20
25
30
35
40
45
50
55
60
65
Comparison of RMS Code Tracking Error of Acquired Satellites
Ac qu ired Satellites' Nu m bers
Code Tracking Error E (m)
3G Noise Added-No Detec tion
No 3G Noise- No D etection
3G Noise Add ed-De tection Used
Figure 9. Comparison of RMS code tracking error of acquired satellites when (1) No 3G noise added. (2) 3G noise added-no
detection algorithm used. (3) 3G noise adde d- dete c tion algorithm used.
tracking error for the same three cases as shown in Fig-
ure 9. The standard deviation was calculated for each
tracked satellite. It is seen how the proposed algorithm
reduces the error by showing less numbers in meters.
This means that the user position is more accurate for the
satellites acquired with noise and detection used than
under noise without detection algorithm. It should be
mentioned that the situation in Figure 6 is very optimistic.
In many cases, the GPS receiver loses track of all satel-
lites and hence no GPS fix can be obtained. The sug-
gested algorithm guarantees to recover most if not all the
satellites that are tracked as if the 3G noise is not present.
6. Conclusions
A cross functional effort has been displayed in this paper.
The effort spanned from presenting architectural analysis
and calculations for a 3G-GPS RF TX-RX system fo-
cusing on the interference related aspects and sources, to
estimating the average power of the received signal, to
laying out the algorithm analysis and derivation of the
mitigation scheme. We showed how the mitigation
scheme can recover the acquisition of the satellites that
were lost due to the interference. Also, the average car-
rier power and RMS code tracking error of the recovered
satellites are improved and therefore the user position
accuracy is improved even though the mobile handset is
in a 3G data or voice service when using the GPS service.
This algorithm could be implemented at the phasing
stage of a mobile handset development and manufacturing.
7. References
[1] T. Sharabati and Y. Chen, “Mitigating 3G Carrier Inter-
ference to GPS Due to Co-Existence in 3G Handset,”
IEEE NAECON, 2009, pp. 86-91.
[2] T. Sharabati, “Proposal for Mitigating 3G Interferrence to
GPS,” University of South Carolina, South Carolina, 2009.
[3] K. Borre, D. M. Akos, N. Bertelsen, P. Rinder and S. H.
Jensen, “A Software Defined GPS and Galileo Receiver
Birkhauser,” Birkhauser, Boston, 2007.
[4] W. B. John and R. K. Kevin, “Extended Theory of
Early-Late Code Tracking for A Band limited GPS
Receiver,” Navigation, Journal of the Institute of
Navigation, Vol. 47, No. 3, 2000, pp. 211-226.
[5] W. B. John, “Effect of Narrowband Interference on GPS
Code Tracking Accuracy,” ION NTM, 26-28 January
2000, pp. 16-27.
[6] B. Michael and A. J. V. Dierendonck, “GPS Receiver
Architecture and Measurements,” Proceedins of the IEEE,
Vol. 87, No. 1, 1999, pp. 48-64.
[7] P. Ward, “Interference Heads Up,” GPS World, Vol. 19,
No. 6, June 2008, pp. 64-73.
[8] S. Semiconductor, “SE4120L GNSS Receiver IC,” June
2008.
[9] A. Technologies, “UTMS1700/2100(1710-1755MHz) and
UMTS1700 (1750-1785MHz),” Avago Technologies, AV02-
1907EN, 2009.
[10] W. Couch Leon II, “Digital and Analog Communication
Systems,” 3rd edition, Collier Macmillan, New York, 1990.
3G Noise Added- Detection Used
N
o 3G Noise-No Detection
3G Noise Added-No Detection