You are on page 1of 5

Upper Bound on C/A and L1C Code Spectral

Separation Coefficients
Srini H. Raghavan and Thomas D. Powell
The Aerospace Corporation
P.O. Box 92957
Los Angeles, CA 90009-2957
Srini.h.raghavan@aero.org
further complicated because of the Doppler frequency shifts
introduced due to the relative motion between the GPS
satellite and the GPS receiver. The net result of lack of
orthogonality is what is known as CDMA noise that
degrades the GPS receiver performance. CDMA noise gets
worse with the increased number of C/A code transmitters
in the system. The C/A code CDMA noise affects primarily
the C/A code receivers with all its severity, but it is still
important to characterize C/A code CDMA noise accurately
in radio frequency compatibility (RFC) studies done within
the Radio Navigation Satellite Systems (RNSS) frequency
band using other codes and signals in the same frequency
band.

AbstractIt is well known that the available effective


carrier-to-noise density ratio (C/N0)effective to a GPS receiver
is degraded to a greater degree due to spectral line effects
than when spectral line effects are ignored for the Global
Positioning System (GPS) coarse acquisition (C/A) codes.
12
(C/N0)effective is an important parameter used to characterize
the GPS receiver performance in terms of code acquisition,
carrier loss of lock, and data bit error rate. In todays signal
environment, which includes a growing number of Global
Navigation Satellite Systems (GNSSs) in the Radio
Navigation Satellite System (RNSS) band, there is a need to
account for all the interference sources accurately to make
sure that the intersystem and intrasystem Radio Frequency
Compatibility (RFC) is achieved. Towards this goal many
studies in the past considered quasi-analysis and simulation
methods to account for the spectral line effects on the GPS
C/A code receivers. The spectral separation coefficient
(SSC) provides a measure of the amount of interference that
can be expected from one signal to the other. An upper
bound on C/A code SSC was presented in last years IEEE
Aeroconference [1]. In this paper we make refinements to
the upper bound equations for the C/A codes presented in
[1] and also extend the bounds to codes other than the C/A
codes. Computational advantage is still retained with the
refinement to the equations.

Modernized GPS has a new civilian signal called L1C.


There are two parts to the L1C signalsone without data,
called the pilot signal and denoted as L1Cp, and the second
with data, denoted as L1Cd. Both L1Cp and L1Cd use
spreading codes based on Weil sequences [2]. The code
period for L1C codes is 10 ms (1 ms for C/A), the code
length is 10,230 chips (1023 chips for C/A), and the
chipping rate is 1.023 Mcps. An overlay code of length 18
seconds that is 1800 bits at a rate of 100 bps is used on
L1Cd. Since the chipping rate is the same as the C/A codes
but the code length is 10 times longer, the spectral spacing
is reduced to 100 Hz (10 times more lines in the power
spectrum) and also the line amplitudes are reduced by a
factor of 10. Cross-correlation properties of L1C codes are
better than the cross-correlation properties of the C/A codes,
resulting in reduced CDMA noise. But as far as the SSC is
concerned, it is lower-bounded by strictly orthogonal codes
for a given chipping rate. Since the chipping rate is 1.023
Mcps, SSC for the L1C codes is underbounded at 61.8
dB/Hz. For the L1C SSC upper-bound calculations, we
apply the equations developed for the C/A codes. In this
paper, only L1Cd codes are considered, and we simply call
them L1C codes from here onwards.

TABLE OF CONTENTS
1. INTRODUCTION .................................................................1
2. EFFECTIVE CARRIER POWER TO NOISE DENSITY RATIO
(C/N0) EFFECTIVE ................................................................1
3. MAXIMUM SSC: AN UPPER BOUND ................................2
4. CONCLUSIONS ..................................................................5
REFERENCES ........................................................................5

1. INTRODUCTION
GPS C/A spreading codes are implemented using Gold
codes. Gold codes have some very desirable properties, such
as a large number of codes available for code division
multiple access (CDMA) applications. Their correlation
properties are also well defined. On the down side the
spread-spectrum code length used in C/A codes is short (5%
of the bit period), resulting in spectral lines spaced 1 kHz
apart. Also because the Gold codes are not strictly
orthogonal, the spectral line magnitudes vary over the
spectral lines produced by strictly orthogonal codes. This is
1
2

2. EFFECTIVE CARRIER POWER


DENSITY RATIO (C/N0) EFFECTIVE

TO

NOISE

(C/N0)effective is the most important metric that is used in


determining the RFC with all other RNSS systems.
(C/N0)effective can be written as follows:

C
=

N 0 Effective

978-1-4244-7351-9/11/$26.00 2011 IEEE


IEEEAC paper#1154, Version 1, Updated 2011:01:06

CEffective
( N 0 )Effective + ( I0 )Effective

(1)

CEffective, (N0)Effective, and (I0)Effective are the received desired


signal, thermal noise psd, and the interference psd after
accounting for all the applicable losses from the transmitter
to the receiver output. The effective interference power
spectral density from all the satellites within an RNSS
system for a receiver at a given location, indexed by i, can
be written in terms of the spectral separation coefficient
(SSC), transmit power, transmit/receive antenna gain, path
loss, and processing loss as follows:

I i (t)=

M Si (t)

m=1

Pm
T
R
G i,m (t)G i,m (t) i,m L

N 0 Effective

Min

(2)

i = receiver index
t = time for which the aggregate interference
power is being calculated

M iS (t ) = number of satellites in view at the i-th

SSC by definition is a frequency domain property of spreadspectrum signals that helps to visualize the spectrum overlap
between two signals resulting in interference. Let us first
examine the definition of the SSC. For two spread-spectrum
signals s1(t) and s2(t) with power spectral densities
represented by the functions S1(f) and S2(f), SSC denoted by
is defined as follows:

m = index of the summation over satellites in


view and m = 0 for the desired signals
satellite index

G iT,m = transmit-antenna gain (relative to isotropic)


between the m-th satellite to the i-th receiver
location

G iR,m = receive-antenna gain (relative to isotropic)

S ( f conv ) =

between the i-th receiver location and the


m-th satellite

conv

f )S2 ( f ) df

(6)

(7)

As can be seen from Equation (7), is defined to be the


peak value of the convolution signal, which is within the
pass band of the subsequent filter in the correlation process.
Furthermore, this also signifies the implicit assumption in
the definition of that the convolution spectral density is
flat within the correlator bandwidth so that the interference
can be treated as receiver thermal noise increase with the
value equal to the peak value. For random codes, psd S1(f)
and S2(f) are given by a sinc2 function.

L = (dimensionless) processing loss


Minimum (C/N0)effective available as defined above occurs
when the interference psd is maximum. The maximum
interference psd can be written from Equation (2) as

G agg max Pmax

S1(f) = S2(f) = Tcsinc2(fTc)

(3)

(8)

Equation (6) is plotted as a function of fconv in Figure 1


below, and the SSC value given by Equation (7) for Tc =
1/(1.023*106) seconds is marked by a dot at the peak.

where Gagg is given by


T
R
G i,m (t)G i,m (t) i,m Pm

Pmax

= S ( 0 ) = S1 ( f )S2 ( f ) df

Pm = transmitted power (W) of the m-th satellite

G agg

S (f

= SSC (1/Hz)

M Si (t)
= max
m=1

i,m = (dimensionless) path loss from the m-th


satellite to the i-th receiver location

L min

Max ( N 0 )Effective + Max ( I 0 )Effective

3. MAXIMUM SSC: AN UPPER BOUND

receivers location at time t

max[I i (t)] =

(5)

The minimum desired signal, Min[CEffective], includes all the


relevant processing losses occurring in the entire transmitreceive link. It should also be noted that in Equation (5), a
lower bound on the Min[(C/N0)Effective] is found by treating
the three terms on the right-hand side of the equation
independent of each other and hence will always result in a
minimum (C/N0)Effective that is equal to or less than the value
computed by considering desired signal, receiver noise, and
the interference at any location simultaneously. We use
Hlders inequality to derive an upper bound on maximum
SSC (max) for both the C/A and L1C codes.

where

Min [ C Effective ]


(4)

The minimum (C/N0)effective can be further bound on the


lower side as

cai,n = i,n cn

(9)

where cn is n-th sample value of the sinc2 function sampled


1 kHz apart and i,n is the ratio of the corresponding C/A and
the sinc2 (random code lines) sample psds. For any two
C/A codes, CA-SSC can be written as

CA-SSCi,j =

ca

i,n

ca j,n 1000

(10)

where cas are the line amplitudes as shown in Figure 2.


This can be written in terms of random code coefficients by
using Equation (9) as
Figure 1 SSC 1.023 Mcps for random codes

CA-SSCi,j =

It can be seen from Figure 1 for random codes that SSC is


maximum at zero relative Doppler. This is not so for
nonrandom codes such as the C/A codes for which the psd
does not follow the sinc2 function. Depending upon the line
amplitude variation from the sinc2 function, the SSC could
be higher sometimes and lower some other times. We can
write the discrete psd for the i-th C/A code (spectral lines
replaced by step function to match a sinc2 function of the
same code chipping rate approximately) as Si(f) = (ai,k/1000)
for all frequencies in the range (2i1)500 to (2i+1)500 Hz,
for all integer values of i as shown in Figure 2. Similarly,
for L1C codes, we can write Si(f) = (ai,k/100) for all
frequencies in the range (2i1)50 to (2i+1)50 Hz, for all
integer values of i as shown in Figure 3.

c ) ( j,n c n ) 1000

(11)

i,n n

where i,n

ca i,n
cn

. By rearranging the terms in Equation

(11) we get

CA-SSCi,j = 1000 (c n c n ) ( i,n j,n )

(12)

Using Hlders inequality we can write the CA-SSC bound


from Equation (12) as
1

2
2
CA-SSCi,j 1000 (c n c n )2 (i,n j,n )2 (13)
n
n

Power Spectral Density

20

Considering the main lobe only and setting s equal to 1


and cn = TcSinc2(1000nTc) for random codes, we get the
SSC from Equation (13) as

40

60

200

400
600
Frequency in kHz

800

Figure 2 Random and C/A psd (1-sided)

Power Spectral Density

2
2
SSC-RND 1000 2 (c n c n ) 2 1
n
n

1000

30

2
1000 2 (c n c n ) 2 1023
n

40

-61.69 dB/Hz
The theoretical value of SSC-RND is 61.8 dB/Hz, and
hence the upper bound from Equation (13) seems quite tight
for the random codes, as expected. Equations (9) to (14) are
the same as in reference [1]. We include a fudge factor in
Equation 13 (see Equation (15)) to take into account the
sinc2 shape of the psds to deemphasize the effects of s on
SSC from the spectral lines away from the center frequency.
That is to say the real effect of the second term in Equation
(13) may diminish as we move away from the center
frequency.

50

60
0

2000

4000
6000
Frequency in kHz

8000

(14)

1 .10

Figure 3 Random and L1Cd code psd (1-sided)


From Figure 2 we can relate the C/A code psd and the sinc2
function by the following equation:
For the i-th C/A code the n-th line in the spectra can is
3

100

(15)
% Time Less than X-axis

CA-SSCi,j 1000 (c n cn ) 2 ( i,n j,n sinc21000nTc ) 2

Now for the C/A codes we first consider 36 C/A codes as


listed in GPS IS-200D [3]. Computing the pair-wise CASSC using Equation (15) and selecting the maximum from
them, we get the upper bound for C/A codes to be equal to
56.4 dB/Hz. Using Equation (13) this value is 56.3
dB/Hz, less than a tenth of a dB due to the fudge factor.

2
2
100 (c n c n ) 2 ( i,n j,n sinc 2100nTc ) 2
n
n

% Time Less than X-axis

100
75
62.5
50
37.5
25
12.5
59
58.5
58
57.5
57
CA-SSC : Upper Bound, dB/Hz

59.5

59
58.5
58
57.5
57
L1C-SSC : Upper Bound, dB/Hz

56.5

56

The CA-SSC upper bound is 56.3 dB/Hz. The L1C-SSC


upper bound is 58.7 dB/Hz. Using Hlders inequality, we
were able to separate the random and nonrandom part of the
C/A and L1C codes and apply a correction based on a
nonrandom part of the codes to the random code SSC. To
explain this further, we define line amplitude ratio (LAR) as
the ratio of the C/A or L1C code line amplitude to the
equivalent sinc2 line amplitude. A histogram of LARs for
both C/A and L1C is shown in Figure 6. From Figure 6 we
can clearly see the improvement expected using L1C codes
over C/A codes. For random codes LAR = 1 with
probability equal to 1.

87.5

59.5

60

Table 2. L1C-SSC Bound for 63 L1C Codes


Max L1C-SSC in dB/Hz
Percentile
less than or equal to
100
56.6
100
57.3
100
58.0
100
-58.6
35.913
59.0

(16)

Using the maximum CA-SSC for the 630 code pairs we can
construct the distribution function, which is plotted in
Figure 4.

60

25
12.5

If we consider the positive half of the main lobe for C/A


codes n = 01023 in Equation (15) and L1C codes n =
0.10,230 in Equation (16).

50
37.5

Figure 5 Upper bound on L1C-SSC using 63 codes

75
62.5

For L1C codes, Equation (15) can be written as

L1C-SSCi,j

87.5

56.5

56

0.01

Figure 4 Upper bound on C/A-SSC


using 36 C/A codes

Normalized Histogram

0.008

Table 1. CA-SSC Bound for 36 C/A Codes


Max CA-SSC in dB/Hz
Percentile
less than or equal to
99.991
56.6
99.795
57.3
96.482
58.0
76.956
58.6
44.475
59.0

0.006

0.004

0.002

0.5

1.5

2.5

3.5

4.5

.Line Amplitude Ratio

Figure 6 Histogram of line amplitude ratios C/A (red)


and L1C) (blue)

A similar bound on max L1C-SSC can be found using all


the 63 codes listed in the IS-800 Draft [2], for which the
results are presented in Figure 5 and Table 2.

BIOGRAPHY

4. CONCLUSIONS
Accounting for all interference sources is an important
aspect of RFC in the Radio Navigation Satellite System
band, which is crucial for all the GNSS receivers
performing at their best without undue interference among
the satellite navigation systems. Because of the nature of
C/A code behavior it is very difficult to estimate the C/A
code CDMA noise accurately. A previous paper on this
issue by the authors of this paper utilized a different
approach to consider the SSC in the frequency domain; the
results obtained using this approach provided an upper
bound to the maximum CA-SSC. In this paper, equations
developed to compute the upper bound are refined and
applied to both the C/A codes and the L1C codes. Since for
a given chipping rate SSC is bounded by random codes, the
improvement achievable using codes other than C/A codes
is also limited. However, in this paper we have illustrated
that the SSC upper bound moves closer to the limit by
changing the code length from 1023 to 10,230 bits. It is not
clear that any of the improvement seen is due to the code
properties, for example, Gold versus Weil sequences. The
upper bound obtained using this method is adequately
conservative and is not dependent on the receiver coherent
integration time or the relative Doppler shifts. In that sense
the computation of the bound is simplified. It is interesting
to see the bound when applied to other spreading codes used
in GPS.

Dr. Srini Raghavan has over 30 years


of experience in the design, simulation
and
analysis
of
satellite
communication
systems,
spread
spectrum
systems,
and
signal
processing. Currently, he is supporting
several communication systems and Global Positioning
System (GPS)-related activities at The Aerospace
Corporation, where he is a Senior Engineering
Specialist. He also has taught a number of in-house
courses on Digital communications and Spread
Spectrum Systems. He holds a Ph.D. in Electrical
Engineering from the University of Missouri-Rolla, and
is a Senior Member of IEEE and a member of ION, Eta
Kappa Nu and Sigma Xi.
Thomas Powell is a Systems Director in
the Navigation Division of The Aerospace
Corporation, managing the Aerospace
support to the Engineering and
Technology Branch of the GPS
Directorate. He has supported the GPS
program for over 10 years at Los Angeles Air Force Base
and in Washington, DC, including work on military user
equipment, constellation sustainment, and spectrum
management. He holds a Ph.D. in Aerospace Engineering
from UCLA.

REFERENCES
[1] S. Raghavan, Jason Hsu, and Thomas Powell, Upper
Bound on C/A Code Spectral Separation
Coefficient, 2010 IEEE Aeroconference.
[2] Draft IS-GPS-800, Navstar GPS Space Segment/User
Segment L1C Interfaces
[3] Interface Specification IS-GPS-200, Revision D,
Navstar GPS Space Segment/Navigation User
Interfaces, Navstar GPS Joint Program Office.

You might also like