Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Investigating the Impact of Ionospheric

Scintillation using a GPS Software Receiver


S. Skone, G. Lachapelle, D. Yao, W. Yu and R. Watson
Department of Geomatics Engineering
University of Calgary

BIOGRAPHIES various latitudes in the auroral and sub-auroral regions.


These measurements are used to characterize high-latitude
Susan Skone, Ph.D., is an Associate Professor in scintillation effects, and develop models for assessing
Geomatics Engineering at the University of Calgary. She GPS receiver performance in the presence of such effects.
has a background in space physics and conducts research
on ionospheric and tropospheric effects. She has The focus of this paper is a study of the effects of
developed licensed software for atmospheric research, ionospheric scintillation on GPS signals, based on a GPS
and is co-chair of an international study group focused on software receiver developed at University of Calgary.
remote sensing of the atmosphere using GPS. She is This study consists of several components: simulating
currently chair of the Canadian Navigation Society. ionospheric scintillation effects on L1 using an
intermediate frequency (IF) GPS software signal
Dr. Gérard Lachapelle holds a CRC/iCORE chair in simulator, investigating phase lock loop (PLL)
Wireless Location. He has been involved with GPS performance under scintillation conditions using a
developments and applications since 1980 and has software receiver, and developing improved tracking loop
authored/co-authored numerous related publications and models to minimize phase errors and loss of signal lock
software. More information is available on during scintillation events. Results indicate that PLL
http://plan.geomatics.ucalgary.ca performance is degraded for moderate to severe
scintillations, with loss of lock occurring for narrow
Wei Yu received his B. Sc in Electrical Engineering at bandwidths. By employing a fast adaptive bandwidth
Dalian University of Technology (China). Now he is a approach, reliable signal tracking can be achieved.
M.Sc candidate in Geomatics Engineering at University
of Calgary. INTRODUCTION

Robert Watson is a Research Associate with the PLAN The electromagnetic radiation, solar wind and
Research Group. He completed his MSc in Geomatics interplanetary magnetic field from the Sun have an impact
Engineering in 2005, following a BSc in Electrical on the Earth’s environment and can cause geomagnetic
Engineering in 2002, both at the University of Calgary. storms. Intense and very intense storms may be associated
His research interests include signal processing for indoor with high-latitude ionospheric scintillations, which can in
and high-sensitivity GPS, and modernized GPS tracking. turn adversely affect communication and navigation
systems. Ionospheric scintillation is observed as rapid and
ABSTRACT random variations in the phase and amplitude of a radio
signal as it passes through small-scale plasma density
Ionospheric scintillations cause RF signal amplitude irregularities in the ionosphere [Conker et al., 2003].
fading and phase variations as GPS satellite signals pass
through the ionosphere. This is a particular concern for The small-scale irregularities in electron density can
GPS operations in high latitude regions, such as Canada, diffract the signal, leading to rapid fluctuations in signal
where scintillations are associated with strong aurora – intensity, known as amplitude scintillation [Skone et al.,
effects which persist even during solar minimum. In 2000]. Amplitude scintillation can be severe enough that
general, scintillations can cause degraded receiver the received GPS signal intensity drops below a receiver’s
tracking performance and, in extreme cases, loss of lock threshold, forcing the receiver to reacquire the signal.
navigation capabilities entirely. Such effects are an issue Amplitude scintillation is measured by the S4 index,
for reliable GPS operations in the northern United States which is essentially a normalized standard deviation of
and Canada. The University of Calgary currently operates high-frequency (e.g. 50 Hz) detrended signal intensity
the Canadian GPS Network for Ionosphere Monitoring observations over 60 seconds. Phase scintillations are
(CANGIM), which makes scintillations observations at observed as high phase dynamics and may lead to cycle

Presented at the ION GNSS 2005 Conference, Long Beach, CA, September 13 – 16 1
slips and loss of lock for receivers as they track the signal equatorial region [Secan, 1996]. The phase scintillation is
[Datta-Barua et al, 2003]. Phase scintillation is quantified characterized as a Gaussian distribution with zero mean,
by σ δϕ which is the standard deviation of high-frequency where σ δϕ is the standard deviation of the scintillation
detrended phase observations over an interval of up to 60 phase variations. The value of σ δϕ can vary from 0.05
seconds. radians for very weak scintillation to 1.0 radians for a
strong scintillation case.
GPS receivers suffer from strong scintillation effects in
the equatorial region (±10° geomagnetic latitude). Many This paper investigates how scintillation affects
researchers have analysed effects in this region [Doherty performance of the phase tracking loop for GPS receivers.
et al., 2000, Skone et al., 2000 and Pullen et al., 1998] By using observations from ionospheric scintillation
and conducted simulation studies to determine the impact monitors in Canada, the magnitudes of high-latitude
on receiver tracking performance [Morrissey et al., 2004, scintillations are quantified. These physical characteristics
El-Arini et al., 2003 and Ganguly et al., 2003]. In the low are then used to simulate realistic scintillation effects on
latitudes, amplitude scintillations and long-term fades are the GPS L1 signal using an intermediate frequency (IF)
a dominant issue. Alternatively, large phase variations software signal generator. Tracking loop performance is
(phase scintillations) are present in the auroral region (65- then assessed for a variety of scintillation conditions,
75° geomagnetic latitude), and polar cap (>75° using a software receiver with variable discriminators and
geomagnetic latitude) regions [Pi et al., 2002]. tracking loop bandwidths.

Several studies have investigated GPS receiver tracking CANGIM OBSERVATIONS OF HIGH LATITUDE
performance during high-latitude scintillations. For IONOSPHERIC SCINTILLATION
example, observations from codeless (Norway) and semi-
codeless (Sweden) receivers indicate that there are higher The Canadian GPS Network for Ionosphere Monitoring
percentages of missing L2 phase or cycle slips in high- (CANGIM) was developed for the purpose of ionosphere
latitude regions [Skone et al., 2000]. Observations of monitoring for GPS applications. The CANGIM currently
auroral scintillations were also made within the FAA’s consists of three stations in Western Canada: Calgary
National Satellite Test Bed, using an ionospheric (51.08°N, 114.13°W), Athabasca (54.72°N, 113.31°W)
scintillation monitor at Fairbanks, Alaska during 1999- and Yellowknife (62.48°N, 114.48°W) – see Figure 1.
2000. Few amplitude scintillations were observed near These stations allow latitude profiling of both the auroral
Fairbanks, but significant phase scintillations were found. and sub-auroral regions. This network is augmented with
In the period of observation, the S4 for amplitude two additional stations in the northern United States: MIT
scintillation varied from 0.2 to 0.4 and the percentage of Haystack Observatory and Moorhead, Minnesota.
occurrence was less than 0.2%. Phase scintillation (low
and moderate level), however, occurred up to 20% of the
time [Doherty et al., 2000].

The GPS signal at the receiver can be expressed as


[Pullen et al., 1998]

E = Ae jφ = E 0 δE = ( A0 δA)e j (φ0 +δφ ) (1)

where E 0 = A0 e jφ0 is the nominal received signal, and


δE = δAe jδφ is the scintillation signal with amplitude δA Figure 1. Locations of CANGIM sites for high-latitude
and phase δφ . The scintillation parameter S4 is the ratio of scintillation monitoring.
the standard deviation of the intensity of the received
The CANGIM sites are equipped with NovAtel
signal to its mean.
Modulated Precision Clock (MPC) receivers and NovAtel
600 antennas. These units contain dual-frequency Euro4
σI E[ I 2 ] − (E[ I ])2 cards with an internal integrated PC and precise oscillator.
S4 = = (2) A user-interface command structure allows direct access
E(I ) E(I )
via modem or internet connection. The CANGIM
receivers have specialized firmware which provides
where I = A 2 . scintillation parameters extracted from 50 Hz L1 phase
observations, in addition to raw GPS code and phase
S4 can vary from 0.1 for very weak ionospheric observations, rates of change of total electron content
scintillation to 0.9 for strong amplitude scintillation in the (TEC), absolute TEC values, and WAAS messages. The

Presented at the ION GNSS 2005 Conference, Long Beach, CA, September 13 – 16 2
firmware version (scintw) was developed by A.J. Systems The relationship between high-latitude phase and
[Van Dierendonck et al., 1996]. amplitude scintillations is shown in Figure 3, where
scintillation parameters are plotted for all satellites in
The CANGIM network does not currently operate in real- view from the Calgary CANGIM site for two severe
time. Data are downloaded at periodic intervals and ionospheric storm events combined: October 30, 2003 and
archived at a central processing facility at University of May 15, 2005. Figure 3 demonstrates little correlation
Calgary. After several hardware upgrades, the three between high-latitude phase and amplitude scintillations.
installed stations have been operating reliably since May For high phase scintillations, S4 values vary from less than
2003. Figure 2 shows ionospheric scintillation 0.05 (low) to 0.3 (moderate). This relationship, and
measurements at Calgary on May 15, 2005. realistic values of S4 and σ δϕ , are used to develop
simulations in the following sections.

IF SIGNAL SIMULATOR AND SOFTWARE


RECEIVER

To simulate the GPS signal, an L1 GPS signal software


simulator is used (Figure 4). This signal simulator was
developed at University of Calgary and has been
described previously [Julien et al., 2004].

Figure 2. Ionospheric scintillation parameters for SV 7


during the period 5:40 to 11:40 UT on May 15, 2005.

Figure 4. Digitized IF GPS signal simulator structure


[Dong, 2003].

Required input parameters are IF frequency (15.42 MHz),


sampling frequency (4.75 MHz), C/N0 (49 dB-Hz), front
end bandwidth (2 MHz), noise density (-203 dBW/Hz)
and quantization bits (1 bit). The simulated GPS L1 IF
signal includes signal amplitude, IF carrier (15.42 MHz),
C/A code, navigation data and noise.
Figure 3. The relationship between intensity and phase
scintillations for October 30, 2003 and May 15, 2005. The required signal amplitude is determined from the
following relation [Dong, 2003]:
A scintillation event is observed in Figure 2 for the period
7:30 to 8:30 UT. Note the lack of data during the most A2
P= (3)
severe storm period. Intensity scintillations (as reflected 2
in S4 values) are less than 0.2. Such values are considered
to be low amplitude scintillation. In contrast, phase where P is the signal power. The carrier to noise ratio
scintillation parameters are in the range 0.5 to 1.0 radians, (C/N0) is frequently used in receiver design and
indicating severe phase variations. The scintillation values performance estimation. The signal power is related to the
are also dependent on the elevation angle and the region noise density and C/N0 as:
of sky through which the GPS signal travels.

Presented at the ION GNSS 2005 Conference, Long Beach, CA, September 13 – 16 3
C / N0 + N0
P = Pe = 10 10 (4)

where Pe is the effective carrier power and N0 is the noise


density. These equations are used to determine the
appropriate signal amplitude. The C/A code is modulated
on the IF signal, and the propagation time and appropriate
Doppler for the given satellite are included in the
simulation. Navigation data is obtained from real
ephemeris records and the noise is derived as the product
of noise density and front end bandwidth. Figure 5 shows
the IF sampled signal without noise or scintillation effects
added, prior to quantization. Figure 6. The architecture of the software receiver
[Ma et al., 2004].

SIMULATION OF IONOSPHERIC
SCINTILLATION

In previous studies of scintillation effects, the ionospheric


scintillation was modeled as a Nakagami-m distribution
for intensity and a Gaussian distribution with zero mean
and spectral shaping for phase [Hegarty et al., 2001]. The
Nakagami-m distribution is given by

m m δI m −1
f (δI ) = m
e − mδI / Ω (5)
Γ ( m) Ω
where δI = (δA) 2 and Ω = E (δI )

Figure 5. Simulated GPS IF signal without The intensity of the scintillation signal can be generated
scintillation. by a bivariate gamma random variable. The gamma
marginal density function used here is given as follows
The software receiver used to process the simulated GPS [Hegarty et al, 2001]:
signal was also developed by the University of Calgary
and has been described previously [Ma et al., 2004]. This ( x / β ) (α −1) e − x / β
f ( x) = (6)
receiver allows an unlimited number of channels for L1 βΓ(α )
C/A code observations, and includes the following
functions: acquisition, code and phase tracking loops,
It is assumed that x in Equation (6) represents the
sub-frame synchronization, navigation solution, inner
intensity of the scintillation signal with
clock generation and measurement derivation as shown in
Figure 6. The Intermediate Frequency is 1.17 MHz and
1
the sampling frequency is 4.75 MHz. A conventional 2nd- α= and β = S 42 (7)
order code delay lock loop (DLL) with E-L envelope S 42
discriminator and conventional 3rd-order phase lock loop
(PLL) with ATAN discriminator are used by default. All the probability density function of the Nakagami-m
properties and settings are easily reconfigured for the distribution is the same as that of the bivariate
scintillation studies. gamma(α , β ) distribution (note that E (δI ) = 1 ) [Hegarty
et al., 2001; Pullen et al., 1998]. The two-parameter
As shown in Figure 6, the software receiver can be used
gamma(α , β ) function can then be used to simulate the
with a NovAtel Euro-3M RF front-end to acquire IF
samples from real GPS signals. This front-end uses a 20 intensity of scintillation signal.
MHz sampling frequency and 3-bit ADC, and will be
used in future studies of measured scintillations at high- The standard deviation of the phase variations is
latitudes. The results presented here, however, focus on IF described by σ δϕ . Therefore, the intensity and phase
signal simulations with no hardware component required. variations are simulated as a function of the measured
scintillation parameters: S4 and σ δϕ . By choosing

Presented at the ION GNSS 2005 Conference, Long Beach, CA, September 13 – 16 4
appropriate values for these parameters, as observed at 30 s interval for the simulated signal. Note that realistic
CANGIM sites, realistic high-latitude scintillation effects high-latitude values of amplitude scintillation (S4) are in
can be simulated. the range 0 to 0.3, with phase scintillation values ( σ δϕ ) in
the range 0 to 1.5 radians (Figure 3).
Figure 7 shows the IF simulated L1 GPS signal with
scintillation effects added, prior to quantization (note that For the largest phase and amplitude scintillations, results
noise effects are not included for clarity purposes). The IF in Figure 8 indicate that a phase error of 20 degrees, and a
is 15.42 MHz and the sample rate is 4.75 MHz. The signal code error of 0.05 chips, may occur. For phase errors
includes L1 C/A code and navigation data. exceeding 15 degrees, it is considered that the PLL is
susceptible to loss of lock [Kaplan, 1996]. For moderate
scintillation effects ( σ δϕ = 0.5 radians), PLL tracking
errors reach this threshold. In the remaining sections of
this paper, various tracking loop properties are
investigated for two cases of high-latitude scintillations:

(1) S4 = 0.10 and σ δϕ = 0.5 radians (moderate), and


(2) S4 = 0.175 and σ δϕ = 1.15 radians (severe).

Figure 7. Simulated GPS IF signal with scintillation:


S4=0.1 and σ δϕ =1.0 radians.

As stated earlier, the scintillation parameter S4 is the ratio


of the standard deviation of the intensity of the received
signal to its mean. The phase scintillation σ δϕ is the
standard deviation of the scintillation phase variations. In
deriving the Gaussian random function (with standard
deviation σ δϕ ) to simulate the GPS signal phase, and by
Figure 8. The standard deviation of PLL error for a
multiplying the signal amplitude by the square root of the 30 second interval, as a function of phase and
two-parameter gamma function (with standard deviation amplitude scintillation.
S4), the simulated GPS signal with scintillation was
generated.

Initial results of tracking performance under scintillation


conditions were generated using default settings for the
software receiver: 2nd-order E-L DLL discriminator, 3rd-
order PLL with ATAN discriminator, PLL bandwidth of
15 Hz, and predetection integration time of 1 ms. As will
be demonstrated in the following sections, these settings
allow continuous L1 phase tracking without loss of lock.
Errors in phase and code tracking were then generated for
a number of simulated signals, where S4 was varied from
0.1 to 0.9 in increments of 0.1, and σ δϕ was varied from
0.1 to 1.1 radians in increments of 0.1 radians.

Figures 8 and 9 show DLL and PLL errors, respectively,


as generated and by the software receiver as a function of Figure 9. The standard deviation of DLL error for a
scintillation parameters. The standard deviations in these 30 second interval, as a function of phase and
figures, for each pair of S4 and σ δϕ , are computed over a amplitude scintillation.

Presented at the ION GNSS 2005 Conference, Long Beach, CA, September 13 – 16 5
INVESTIGATION OF CBPLL PERFORMANCE code samples. The local carrier has a central frequency at
f + fˆ and the local code C (t − τ~ ) has been properly
if d k e
In this section, analyses of phase tracking errors are
conducted for a constant bandwidth PLL. In particular, shifted to align with the phase of incoming code. The
case studies are presented for scintillation effects defined variable τ~e is the code phase estimation error, as
in previous sections. Prior to presenting results of these compared with the incoming code phase, and
analyses, a detailed description of the software receiver fˆ represents the estimated Doppler.
d
models is presented.

PLL Model In order to improve the SNR (signal to noise ratio) of raw
measurements used for carrier and code tracking loops,
The default PLL system model is a Costas loop, and is coherent integration accumulates base-band samples
highlighted with light blue boxes in Figure 10. The Costas during the period Tcoh. This is herein referred to as the
loop is designed to be insensitive to data bit transitions by predetection integration time. Prompt integration from the
employing the I (in-phase) and Q (quadra-phase) channel. I and Q channels are fed into the PLL for carrier tracking.
Phase of the integration values at I and Q channels switch
by 180 degrees simultaneously once data bit transition
IPk = A
(
Tcoh sin π∆f k Tcoh ) ( ) ( )
Rτ~e, k  D tk cos ∆ϕ k + ni , k
occurs. N0 (
π∆f k Tcoh )
 

The IF input signal (as supplied by the software signal


generator in these studies) is shown at point A in Figure (9)
10 and is modeled as: Tcoh sin (π∆f kTcoh ) ~
QPk = A R (τ e, k )D(tk )sin (∆ϕ k ) + nq ,k
N 0 (π∆f kTcoh )
( )
x k = A × D(t k )C (t k ) cos 2π ( f if + f d )t k + θ 0 + v k (8)
(10)
where xk is the digital IF sample at epoch tk, the constant A
is the amplitude of signal, and D(t) is the navigation bit where R (τ~e,k ) is the correlation value with a time offset of
stream, which is a sequence of +1 or -1 that randomly
switches with a period of 20 ms. The sequence C(t) is the τ~ , T is the predetection integration period, N is the
e, k coh 0
GPS C/A PRN (pseudo-random noise) code with a period noise density, and ∆fk and ∆ϕk are, respectively, the
of 1ms. The nominal intermediate frequency is fif after the average frequency error and phase error over the
L1 carrier signal has been down-converted, fd is the accumulation interval. Given the condition that Tcoh >> Ts
carrier Doppler shift, θ0 is the initial phase angle, and the (sampling interval), ni,k and nq,k are the noise samples at I
term vk is measurement noise at epoch k, normally being and Q channels after predetection integration, whose
regarded as zero-mean white Gaussian noise with a properties are expressed as [van Dierendonck, 1996]:
constant PDF of N0/2.
( ) ( )
E ni2,k = E nq2, k = 1 (11a)
E (ni , k ni ,m ) = 0 (k ≠ m ) (11b)
( )
E ni ,k nq , m = 0 ∀k , m (11c)
B

I&D IPk
A Ts Tcoh

NCO D(z) Phase


detector
A
B
Figure 10. Default tracking loop model in software π/2
receiver.
Ts Tcoh
I&D QPk
The software receiver processing begins with the carrier
removal and code despreading by multiplying the
incoming samples with locally reconstructed carrier and Figure 11. Digital Costas loop model.

Presented at the ION GNSS 2005 Conference, Long Beach, CA, September 13 – 16 6

∫ H ( j 2πf )
2
Figure 11 exhibits all the components of a digital Costas Bn = df (17b)
loop. D(z) is the transfer function of the loop filter. The 0
NCO is the numerical controlled oscillator used to
generate up-to-date phase for Doppler removal at the next Therefore, PLL properties will change with different
coherent integration period. It can be modeled by ENB, and performance for various ENB under
scintillation conditions is investigated here.
K 0 z −1
N (z ) = (12) Results and Analysis
(1 − z ) −1

Two scintillation cases were identified for detailed


The transfer function of the loop filter D(z) is described as investigations.
(1) S4 = 0.10 and σ δϕ = 0.5 radians (moderate), and
N −1

∑b z n
−n (2) S4 = 0.175 and σ δϕ = 1.15 radians (severe).
These parameters are used to define simulations of
D( z ) = n=0
(13)
(1 − z ) −1 N −1 scintillation effects on an IF test signal in order to study
PLL performance for different tracking loop models and
settings. The tests are conducted using the Costas loop
where N is the order of the carrier tracking loop.
described previously, with a constant bandwidth PLL
(CBPLL). The tests include two major analyses:
Therefore, the transfer function of the entire loop can be
comparisons of four types of discriminators, and
expressed in the z domain as:
comparisons of different ENB for the PLL.

H (z ) =
K0 D (z )z −1 (14) Discriminator comparison
(1 − z −1
) + K 0 D(z )z −1
Four types of Costas discriminators [cf. Kaplan, 1996] are
In the case of a third-order PLL (as used for the studies evaluated here in the Costas framework:
conducted here), the optimal coefficients of loop filters in
tan −1  k 
Equation 13 can be expressed as QP
IPk 
(1) Arctangent (ATAN):

Tcoh 2ω03 a3( 0)ω02Tcoh (2) Conventional Costas: (IPk x QPk)
b0 = + + b3( 0)ω0 (15a) (3) Decision-Directed: sgn(IPk) x QPk
4 2
(4) Q/I: QPk/IPk
T 2ω 3
b1 = coh 0 − 2b3( 0)ω0 (15b)
2 For these tests, Bn = 10 Hz and Tcoh = 1ms. As will be
T 2ω 3 a (0)ω 2T shown in the next section, this ENB allows tracking of the
b2 = coh 0 − 3 0 coh + b3( 0)ω0 (15c) scintillation high phase dynamics and the PLL maintains
4 2 phase lock. Figure 12 shows phase errors for the various
discriminators, for the moderate scintillation conditions.
where a3( 0) = 1.1 and b3( 0) = 2.4 [cf. Kaplan, 1996] and The phase jitter is derived as standard deviations over 20
ω0 is the natural frequency of a closed-loop system. For a ms intervals (using twenty samples of the 1 ms I and Q
third-order loop, ω0 is related to the equivalent noise values for a given satellite).
bandwidth (ENB) Bn by [Kaplan, 1996]:
Variations between satellites in Figure 12 are due to
Bn different elevation angles in the simulations. The lower
ω0 = 0.7845
(16) C/N0 value corresponds to a low-elevation satellite. All
C/N0 values are relatively high, however, due to the low
The relation between ENB and H(z) is expressed as magnitudes of amplitude scintillation simulated here for
the high latitudes. The ATAN, Conventional Costas and
Decision-Directed discriminators perform equally well for
Bn =
1
2πj ∫ z =1
( ) dzz
H ( z )H z −1 (17a) this moderate scintillation scenario. The Q/I discriminator
has larger phase errors and is not recommended.

and the natural frequency ω0 is implied in H(z). By


replacing z with j2πf,

Presented at the ION GNSS 2005 Conference, Long Beach, CA, September 13 – 16 7
further simulations. Note that these results are valid
primarily for high-latitude scintillations where
S4 = 0.1 degradations in signal strength are minimal. Figure 14
σ = 0.5rad shows the predicted discriminator performance (from
rad Monte Carlo simulations) for the ATAN, Conventional
T = rad
1ms Costas and Decision-Directed discriminators. For C/N0
Bn=10 Hz values above 35 dB-Hz, the three cases are similar. This is
the range of C/N0 valid for high-latitude scintillations. For
lower signal strengths, the ATAN discriminator has larger
errors, and this would be a factor to consider in choice of
discriminator for strong amplitude scintillation in the
equatorial region.

C/N0: 43.9 – 44.6 dB-Hz


Figure 12. Estimated phase jitter as a function of
discriminator for moderate scintillations. Carrier to
noise density varies from 43.9 dB-Hz to 44.6 dB-Hz for
the different satellites.

S4 = 0.175
σ = 1.15rad
T = 1ms
Bn=10 Hz
Figure 14. Discriminator comparison as a function of
C/N0.

PLL ENB comparison

As discussed previously, PLL stability is associated with


loop bandwidth. In order to accommodate the high phase
dynamics caused by scintillation, it may be necessary to
use a higher bandwidth in the PLL. It is important to note
that phase scintillations are associated with random phase
C/N0: 38.7 – 39.6 dB-Hz variations which have similar characteristics to noise. In
Figure 13. Estimated phase jitter as a function of the case of scintillations, however, these variations are
discriminator for severe scintillations. Carrier to noise true signal dynamics and the PLL must track such
density varies from 38.7 dB-Hz to 39.6 dB-Hz for the variations to obtain a strong signal and maintain phase
different satellites. lock.

Figure 13 shows additional discriminator comparisons, In the analyses conducted here, bandwidths of 5 Hz, 10
with analysis identical to that shown in Figure 12 but for Hz and 15 Hz are evaluated for the PLL. The ATAN
the severe scintillation conditions. For these simulations, discriminator and a predetection integration time of 1 ms
C/N0 values are approximately 5 dB-Hz lower than for the are used. The simulated IF signal is acquired by the
moderate scintillation case, and the phase jitter has software receiver and, once data bit synchronization is
increased marginally for the ATAN, Conventional Costas completed, the receiver switches from a frequency lock
and Decision-Directed discriminators compared to Figure loop (FLL) mode to the PLL tracking. For the tests
12. These three discriminators are again similar in conducted here, PLL tracking is evaluated over ten or
performance, however, with the Q/I discriminator not more seconds, for a given satellite, to assess capabilities
recommended. of the receiver to maintain phase lock.

From the results shown here, and for ease of Eight different satellites are considered, for varying
implementation, the ATAN discriminator is used in elevation angles and Doppler values. Table 1 shows the

Presented at the ION GNSS 2005 Conference, Long Beach, CA, September 13 – 16 8
number of satellites (out of the eight test cases) where loss
of phase lock was detected for the simulations of severe
scintillation. Three of the eight satellites lost lock for the
lower bandwidth of 5 Hz. This indicates that such a
bandwidth, while beneficial for reducing noise, cannot
accommodate the scintillation phase dynamics.
Commercial receivers typically employ a 5 Hz (or lower)
L1 bandwidth to eliminate noise.

Table 1: PLL Loss of Lock as a Function of ENB


(S4 = 0.175 and σ δϕ = 1.15 radians)
Bn 5 Hz 10 Hz 15 Hz
Number of Channels 3 0 0
with Loss of Lock

An example of the scintillation effects on PLL


performance is given in Figures 15 and 16 for one of the
simulated satellite signals (SV 27). Figure 15 shows the
Doppler estimate and C/N0 for severe scintillations. In the
first 1.5 seconds the receiver is in FLL mode, with only
approximate Doppler estimates being derived. After
subsequently switching to PLL tracking, the Doppler
estimates diverge and the receiver eventually loses lock.
The PLL is unable to provide accurate phase
measurements. This is also reflected in the C/N0 values, Figure 15. Doppler estimates and carrier to noise
where signal strengths drop as the PLL diverges and the density for simulated SV 27 signal under severe
receiver is unable to track the true carrier signal. scintillations (S4 = 0.175 and σ δϕ = 1.15 radians) for a
PLL bandwidth of 5 Hz.
Figure 16 shows the Doppler estimates for the same
signal as in Figure 15, again in the presence of severe
scintillations, but with a PLL bandwidth of 10 Hz. In this
case the PLL is able to track through the scintillations and
Doppler estimates converge to approximately the correct
values. Phase lock in this case is maintained.

For comparison, the case of moderate scintillation is


shown in Figure 17 for the same satellite signal (SV 27)
and a bandwidth of 5 Hz. In this case the Doppler
estimate diverges over several seconds and then
subsequently converges to the correct value. Under these
scintillation conditions the receiver loses lock and
reacquires the carrier. These effects are not as large as
those for severe high-latitude scintillations, but it is
observed that PLL performance is degraded even for
moderate scintillation events. It is recommended that, for
a constant bandwidth PLL, the ENB should be 10 Hz or
more for reliable carrier tracking during scintillations.
Figure 16. Doppler estimate for simulated SV 27 signal
under severe scintillations (S4 = 0.175 and σ δϕ = 1.15
radians) for a PLL bandwidth of 10 Hz.

Presented at the ION GNSS 2005 Conference, Long Beach, CA, September 13 – 16 9
conducted here using software components). According to
the rule-of-thumb, PLL measurement error is computed as
[Kaplan, 1996]:

θe
σ tPLL + ≤ threshold (18)
3

where σ tPLL is the thermal noise and θ e is the dynamic


error. The thermal noise error can be expressed as

360 Bn  1 
σ tPLL = 1 +  (19)
2π C / N0  2T / N 
 coh 0 

and the dynamic stress for a third-order tracking loop is

d 3R
3
θ e = 0.4828 dt3 (deg) (20)
Bn

d 3R
where represents line of sight (LOS) jerk.
dt 3

The PLL is able to track phase changes arising from line-


of-sight dynamics between satellite and receiver.
Therefore, measurements from the carrier tracking loop
can be used to estimate the line-of-sight motion. An
ATAN discriminator, in theory, generates misalignment
between the local replica of carrier phase and phase of
incoming signal, but the raw phase difference is too noisy
Figure 17. Doppler estimates and carrier to noise
to be directly adopted for dynamic estimation. To extract
density for simulated SV 27 signal under moderate
accurate jerk, noisy measurements are accumulated for an
scintillations (S4 = 0.10 and σ δϕ = 0.5 radians) for a
interval of 200 ms:
PLL bandwidth of 5 Hz.
2 k
FAST ADAPTIVE BANDWIDTH PLL Accel (k ) = λL1K 0 ∑ b B ∑θ
i =0
3
i n
n =1
e (∞ )n Tcoh (21)

Based on results of the previous section, it may be


advantageous to adopt an adaptive approach to estimation where k = 200ms/Tcoh and bi are coefficients of the
of the optimal bandwidth for PLL tracking [Legrand et transfer function for a 3rd–order PLL. Tcoh is the coherent
al., 2000]. An option is to use the fast adaptive bandwidth predetection integration period used to derive IPk and QPk.
(FAB) PLL, where the bandwidth is automatically The value θe(∞) indicates the measurement from the PLL
adjusted and modeled by Equations 17a and 17b. The discriminator, Bn is the equivalent noise bandwidth of the
adaptive nature of this approach arises from comparing PLL, λL1 is the wavelength of GPS L1 carrier, and bi
phase errors to a pre-defined threshold. Specifically, the (i=0,1,2) are given by equations 15a, 15b, and 15c.
thermal phase noise plus dynamic stress at optimal
bandwidth must be less than the pre-defined threshold.
d 3R Accel (k ) − Accel (k − i )
3
= Jerk (k ) = (22)
FAB Model dt iT
t =k

There are three dominant measurement errors for carrier


Based on the dynamic stress estimated by Equations 20,
tracking loop: thermal noise, dynamic stress, and clock
21 and 22, and the thermal noise estimate given by
phase noise. Clock phase noise includes vibration-induced
Equation 19, an optimal loop bandwidth can be iteratively
phase jitter and Allan phase noise, which is not
produced by Equation 18.
considered in this paper (as all simulations were

Presented at the ION GNSS 2005 Conference, Long Beach, CA, September 13 – 16 10
Results and Analysis CONCLUSIONS

The FAB approach is implemented in the software Phase and amplitude scintillations can cause degraded
receiver and preliminary results are derived for receiver tracking performance. Such effects are generally
scintillation simulations. In particular, the signal tracking observed at high latitudes (auroral and polar regions) and
for SV 27 is conducted here using the FAB as opposed to near-equatorial regions. Observations from the CANGIM
the constant bandwidth approach. In Figure 15, it was indicate that large phase scintillations may occur at high
observed that Doppler estimates diverged for a constant latitudes but amplitude scintillations are minimal in this
bandwidth of 5 Hz for SV 27, under severe scintillation region. The major concern for reliable receiver tracking at
conditions. Even during moderate scintillation (Figure auroral latitudes is therefore high phase dynamics.
17), there were large errors in Doppler estimates for SV
27. The FAB approach instead allows the bandwidth to Simulated L1 signals were generated using an IF signal
vary and accommodate the scintillation phase dynamics. simulator. Scintillation effects were then added to the
signals, assuming a Nakagami-m distribution for
Figure 18 shows the carrier to noise density, PLL amplitude scintillations and a Gaussian zero-mean
bandwidth, and SNR for SV 27 under severe distribution for phase scintillations. The scintillation
scintillations. An ATAN discriminator is used to generate models were expressed in terms of S4 and σ δϕ which
these results, with a predetection integration time of 1 ms reflect magnitudes of amplitude and phase scintillations,
(consistent with results in previous sections). The FAB respectively. These scintillation parameters are measured
method is compared in this plot to results for a constant directly at CANGIM sites, and realistic simulations of
bandwidth of 10 Hz. For the FAB method, the initial PLL high-latitude scintillation effects were based on these
bandwidth is greater than 10 Hz to achieve convergence observed values.
for the Doppler estimates. After several seconds, the
bandwidth decreases to 3-5 Hz, which allows filtering of A software receiver developed at University of Calgary
higher frequency noise and an SNR of approximately 35 was used to assess the impact of scintillations on tracking
dB. For the constant bandwidth of 10 Hz, the SNR is only loop performance. Because phase scintillations are
30 dB. dominant at high latitudes, the analyses focused on PLL
performance. It was observed that even moderate high-
While a wider bandwidth (10 Hz or more) is required for latitude scintillations can cause phase tracking errors
the PLL to converge, a narrower bandwidth can be used greater than 15 degrees. For a constant bandwidth PLL,
to subsequently track phase variations associated with bandwidths must be 10 Hz or greater to derive correct
even severe phase scintillations. For a constant bandwidth Doppler estimates. For lower bandwidths, the PLL
PLL, the bandwidth must be set at 10-15 Hz to initially diverges with eventual loss of phase lock.
track the carrier signal. If a FAB is used, however, the
PLL bandwidth can be decreased over time to allow An option to improve PLL tracking performance is the
improved noise reduction. fast adaptive bandwidth (FAB) method. This approach
allows the PLL to converge to correct Doppler estimates,
with the bandwidth decreasing over time to eliminate
high-frequency noise. During severe scintillations, the
FAB method allows the PLL bandwidth to vary from an
initial value of 15 Hz to less than 5 Hz. This results in a 5
dB increase in SNR over a constant bandwidth of 10 Hz
in the PLL.

Results presented here are preliminary and suggest further


analyses to be conducted. Future plans include using a
hardware front-end with the software receiver to collect
real data during strong auroral activity (and strong phase
scintillations) near Calgary. It is intended that I and Q
samples will be recorded for several hours during
increased ionospheric activity, and that these samples will
be post-processed for various receiver configurations to
investigate and develop robust receiver tracking loops for
Figure 18. C/N0, PLL bandwidth and SNR comparison
the GPS and other GNSS signals.
for the FAB method (blue) and the constant
bandwidth PLL (magenta), for simulated severe
scintillations (S4 = 0.175 and σ δϕ = 1.15 radians).

Presented at the ION GNSS 2005 Conference, Long Beach, CA, September 13 – 16 11
ACKNOWLEDGMENTS accuracy by using fast adaptive bandwidth lock loops”,
Proceedings of the ION GPS 2000, pp. 19-22, Salt Lake
We would like to acknowledge support from the City, UT, 2000.
Informatics Circle of Research Excellence.
Ma, C., Lachapelle, G. and M.E. Cannon, Implementation
REFERENCES of a software GPS receiver, Proceedings of the ION GNSS
2004, Long Beach, CA, Sept. 21-24, 2004.
Conker, R. S., M. B. El-Arini, C. J. Hegarty and T. Hsiao,
Modeling the effects of ionospheric scintillation on GPS Morrissey, T. N., Shallberg, K. W., Van Dierendonck, A.
satellite-based augmentation system availability, Radio J. and M. J. Nicholson, GPS receiver performance
Science, Vol. 38, No. 1, 1001, 2003. characterization under realistic ionospheric phase
scintillation environments, Radio Science, Vol. 39,
Datta-Barua, S., P. H. Doherty and S. H. Delay, Rs1S20, 1-18, 2004.
Ionospheric Scintillation Effects on single and dual
frequency GPS Positioning, Proceedings of the ION GPS Pi, X., Boulat, B.M., Mannucci, A.J. and D.A. Stowers,
2003, Portland, OR, September 2003. Latitudinal characteristics of L-band ionospheric
scintillation, Proceedings of the ION GPS 2002, Portland,
Doherty, P. H., Delay, S. H., Valladares, C. E. and J.A. OR, September 24-27, 2002.
Klobuchar, Ionospheric scintillation effects in the
equatorial and auroral regions, Proceedings of the ION Pullen, S., Opshaug, G., Hansen, A., Walter, T., Enge, P.
GPS 2000, Salt Lake City, UT, September, 2000. and B. Parkinson, B., A preliminary study of the effect of
ionospheric scintillation on WAAS user availability in
Dong, L., IF GPS Signal Simulator Development equatorial regions, Proceedings of the ION GPS 1998,
and Verification, M.Sc. Thesis, UCGE 20184, 2003 (URL: Alexandria, VA, September, 1998.
http://www.geomatics.ucalgary.ca/links/GradTheses.ht
ml) Secan, J. A., WBMOD: Ionospheric Radiowave
Scintillation Model, Version 13.04, Northwest Research
El-Arini, M. B., Conker, R. S., Ericson, S. D., Bean, K. Associates, Inc., Bellevue, Wash., 1996.
W., Niles, F., Matsunaga, K. and K. Hoshinoo, Analysis
of the effects of ionospheric scintillation on GPS L2 in Skone, S., and K. Knudsen, Impact of ionospheric
Japan, Proceedings of the ION GPS 2003, Portland, OR, scintillations on SBAS performance, Proceedings of the
September, 2003. ION GPS 2000, Salt Lake City, UT, September, 2000.

Ganguly, S., Jovancevic, A., Brown, A., Kirchner, M. and Skone, S., Wide area ionosphere modeling at low
S. Zigic, Ionospheric scintillation monitoring and latitudes - specifications and limitations, Proceedings of
mitigation using a software GPS receiver, Radio Science, the ION GPS 2000, Salt Lake City, UT, September, 2000.
Vol. 39, No. 1, Rs1S21, 1-9, 2003.
Van Dierendonck, A.J., GPS Receivers, in “Global
Hegarty, C., El-Arini, M.B., Kim, T. and S. Ericson, Positioning System: Theory and Applications” edited by
Scintillation modeling for GPS Wide Area Augmentation B.W. Parkinson and J.J. Spilker Jr., Vol. 1, pp. 364-365,
System receivers, Radio Science, Vol. 36, No. 5, Washington, 1996.
September/October 2001.
Van Dierendonck, A. J., Hua, Q., Fenton, P., and J.
Julien, O., Zheng, B., Dong, L. and G. Lachapelle, A Klobuchar, Commercial ionospheric scintillation
complete software-based IF GNSS signal generator for monitoring receiver development and test results,
software receiver development, Proceedings of the ION Proceedings of the ION 52th Annual Meeting, Cambridge,
GNSS 2004, Long Beach, CA, Sept. 21-24, 2004. MA, June, 1996.

Kaplan, E. D., “Understanding GPS: Principles and Van Dierendonck, A. J., Klobuchar, K. and Q. Hua,
Applications”, Artech House Publishers, 1996. Ionospheric scintillation monitoring using commercial
single frequency C/A code receivers, Proceedings of the
Knight, M. and A. Finn, The effects of ionospheric ION GPS-93, Salt Lake City, Utah, September, 1993.
scintillations on GPS, Proceedings of the ION GPS 1998,
Nashville, Tennessee, September, 1998.

Legrand, F., Macabiau, C., Issler, J.-L., Lestarquit, L. and


C. Mehlen, Improvement of pseudorange measurement

Presented at the ION GNSS 2005 Conference, Long Beach, CA, September 13 – 16 12

You might also like