Вы находитесь на странице: 1из 12

Open Loop Tracking of Radio Occultation Signals

from an Airborne Platform



Tyler D. Lulich, Emergent Space Technologies
James L. Garrison, Jennifer S. Haase, Yu-Ming Yang, Justin Voo, Purdue University
Feiqin Xie, Joint Institute for Regional Earth System Science and Engineering
Paytsar Muradyan, Purdue University


BIOGRAPHY

Tyler D. Lulich is an Aerospace Engineer at Emergent
Space Technologies. He is currently supporting
development of the Navigator GPS receiver for the MMS
mission at the Goddard Space Flight Center in Greenbelt,
MD. He holds a BSME from Milwaukee School of
Engineering (2006) and an MSE from the School of
Aeronautics and Astronautics at Purdue University
(2010).

James L. Garrison has been a member of the faculty at
Purdue University since 2000, where he is currently an
Associate Professor in the School of Aeronautics and
Astronautics, with a courtesy appointment in the School
of Electrical and Computer Engineering. From 1988 to
2000 he was employed by the National Aeronautics and
Space Administration (NASA), first at the Langley
Research Center in Hampton VA, and later at the
Goddard Space Flight Center in Greenbelt, MD. He
earned a PhD from the University of Colorado at Boulder
in 1997, while on graduate study leave from NASA. He
also holds a BS from the Rensselaer Polytechnic Institute
and an MS from Stanford University.

Jennifer S. Haase received the B.S. degree in geophysics
from the California Institute of Technology, Pasadena, in
1985, and the Ph.D. degree in Earth science from the
Scripps Institute of Oceanography, La Jolla, CA, in 1992.
She was a Research Fellow at the California Institute of
Technology from 1993 to 1995. She is currently an
Assistant Professor in the Department of Earth and
Atmospheric Sciences, Purdue University, West
Lafayette, IN, where her primary research interests center
on atmospheric remote sensing including the use of
spaceborne and airborne GNSS radio occultation
techniques.

Yu-Ming Yang is a Ph.D. student in the School of
Aeronautics and Astronautics at Purdue University. He
received his M.S. degree in Aeronautics and Astronautics
Engineering from National Chen-Kung University,
Taiwan and his B.S. degree in Aeronautical Engineering
from Feng-Chia University, Taiwan.

Justin Voo obtained his B.S. and M.S. degrees in
Aeronautical and Astronautical Engineering at Purdue
University where he also began his Ph.D. work. His
research involves sea surface remote sensing using
satellite navigation signals.

Feiqin Xie received the B.S. degree in Atmospheric
Sciences from Lanzhou University, Lanzhou, China, in
1998, the M.S. degree from Peking University, Beijing,
China, in 2001, and the Ph.D. degree from the University
of Arizona, Tucson, in 2006. He is currently with the
Joint Institute for Regional Earth System Science and
Engineering (JIFRESSE), University of California, Los
Angeles, CA. He was previously a Postdoctoral Research
Scholar with the Department of Earth and Atmospheric
Sciences, Purdue University, West Lafayette, IN. His
primary research objective includes atmospheric remote
sensing using spaceborne and airborne GNSS radio
occultation technique and planetary boundary layer
dynamics.

Paytsar Muradyan received a B.S. degree in physics at
Yerevan State University, Armenia, in 2002, and an M.S.
degree in Atmospheric Sciences at Purdue University,
West Lafayette, IN, in 2009. She is currently a Ph.D.
student at Purdue University, with a research interest in
atmospheric remote sensing with airborne GNSS radio
occultation technique.

ABSTRACT

Radio occultation (RO) is a remote sensing technique that
uses signals from the Global Positioning System (GPS) to
determine electron density in the ionosphere, and
temperature and water vapor content in the neutral
atmosphere. In the lowest part of the atmosphere where
there are strong refractivity gradients due to boundary
layer structure, the signal may experience rapid phase
accelerations, extending beyond the operational range of
traditional receiver tracking loops. Open loop (OL)
tracking is a method for tracking occulted GPS signals by
replacing a traditional feedback loop with an algorithm
1049
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010
for tracking the signal using a model-based estimate of
Doppler frequency and a record of the GPS data bits. The
OL tracking method described in this work operates using
data recorded by an airborne receiver. Occultation data
were recorded by the GISMOS system aboard the NSF
research aircraft HIAPER during flights over the Gulf of
Mexico and Southeastern United States in February 2008.
Rising and setting GPS occultations were recorded during
five flight missions. The OL tracking method was
demonstrated on GPS signals recorded from setting
occultations during this campaign and was shown to
successfully track signals to lower elevations than a high-
quality commercial receiver using a traditional phase-lock
loop method. OL tracking was able to extract continuous
phase information down to approximately -3.9, -4.2, and -
4.5 degrees elevation relative to the local horizon, on
three setting occultations observed on 22 February 2008.
Finally, the accuracy of OL measurements, averaged over
a Fresnel zone, is shown to be on the order of 3 mm/s,
which is well within the limit for extracting useful
scientific measurements.

INTRODUCTION

The method of radio occultation (RO) has been
successfully adapted from planetary science to sensing of
the Earths atmospheric limb with signals from the Global
Positioning System (GPS) [1, 2]. Chief among the
successful spaceborne missions have been GPS-MET,
CHAMP, SAC-C, GRACE, COSMIC, Metop-A, and
TerraSAR-X/TanDEM-X [3, 4, 5, 6, 7]. Most prolific
among these experiments is the joint US-Taiwan venture,
COSMIC, a constellation of six spacecraft in low Earth
orbit, producing thousands of RO measurements per day.

The spaceborne experiments provide numerous daily RO
profiles. Even so, the distribution of the profiles over the
surface of the Earth are random, and at times, sparse.
This can lead to large uncertainty in numerical weather
prediction for critical weather areas, such as those which
are known to create hurricanes, or typhoons, and those
near dense urban areas. Airborne RO is a method which
would serve to supplement spaceborne measurements
with on-demand atmospheric profiling in desired areas.

Zuffada et al., 1999 presented the concept of atmospheric
profiling with a mountain-based or airborne receiver [8].
Promising results from mountain-based observations have
been reported from China [9] and Japan [10, 11].
Yoshihara, et al., 2005 described an airborne RO
measurement system and presented data collected with
this system [12].

Tracking the signal from an occulting GPS satellite (SV)
is complicated by two effects caused by multipath as a
result of the complex structure of refractivity in the lower
moist troposphere: (1) rapid, random phase accelerations
and (2) fading of the signal amplitude. Multipath will
also result in multiple ray paths through the atmosphere,
and therefore the received signal will no longer consist of
a single frequency term [13]. The magnitude of these
effects is greatest as the signal passes through the lower,
moist troposphere.

It is widely known that a signal affected in such a way by
the atmosphere cannot be properly tracked by a standard
phase lock loop (PLL). Further, it has been shown that
manually tuning a PLL to compensate for atmospheric
effects through the use of a best-guess climatological
model introduces a bias to the results. Instead, the
method of OL tracking is used, and has proven to be
especially effective in recovering scientific information
from occulted GPS signals used for numerical weather
prediction. [13, 14, 15]

OPEN LOOP TRACKING METHOD

Open loop (OL) tracking is applied to raw samples of the
GPS signal following downconversion to an intermediate
frequency. As mentioned in the introduction, OL tracking
was developed as a tool of interplanetary science. It has
been adapted for sensing the Earth's atmosphere from
LEO, and this work shows the potential for airborne RO
as well. The term open loop is used loosely here to
indicate operation without the use of the feedback loops
used in traditional tracking methods. The information
necessary for tracking the GPS signal is provided from
external sources, giving knowledge of the position and
velocity of the (GPS) transmitter and (airborne)
occultation receiver. The input to the OL tracking method
described in this work is a pre-computed Doppler
frequency model,

. The Doppler model is based


upon geometry alone; however, the model can also
include a nominal atmospheric reference model, reducing
the size of the residual phase that must be estimated as
well as increasing the SNR of the OL tracking
measurements. An atmospheric reference model was not
employed for this study.

The goal of OL tracking is to extract phase information
from the received and down-converted GPS signal


(1)
which consists of the amplitude at the time of reception,
, ranging code, , navigation date message, ,
and carrier wave at the intermediate frequency,

, and
time-varying phase , and the initial phase


( = 0). The ranging code and navigation message
are delayed by , the time-varying path-length delay,
which includes the total time to propagate through the
multi-layered atmosphere. The signal is modeled as
1050
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010
having been received in the presence of additive, white
Gaussian (AWGN) noise

.

The total received phase in this case is the sum of the
phase due to the geometry between the transmitter and
receiver, and the excess phase induced by the atmosphere.

(2)
The products of OL tracking are complex correlator sums
as formed using a local signal generated with a model
phase time series,

. The phase of this complex


correlation is called the residual phase

and represents
the difference between the total phase and model phase.

(3)
In this research, the model phase generated incorporates
only geometry, thus residual phase is an estimate of the
excess phase due to the atmosphere.

(4)

The following sections will present details of our
approach to OL tracking, applied to an airborne
measurement system in which the full data can be
recovered and post-processed (i.e., without the telemetry
bandwidth limitations of a spaceborne receiver). Our
approach is based upon that described in Beyerle, et al
2006 [16]. The essential stages of this process are;
Doppler prediction from a geometric model to produce a
model signal; code and carrier wipeoff, which involves
correlation of the received signal with this model signal,
using an integration time synchronous with the data bit
interval; residual phase estimation from the resulting
complex correlation and the phase unwrapping.

DOPPLER PREDICTION

A Doppler prediction algorithm provides the reference,

, required for OL tracking. The Doppler


frequency is a function of the relative position and
velocity of the receiver and transmitter. Finalized GPS
orbit products from the International GNSS Service (IGS)
provided GPS satellite positions to within 2.5 mm RMS
accuracy. The orbits were interpolated with a Lagrangian
method, shown to be sufficiently accurate for OL tracking
[18]. Receiver position and velocity were recorded by the
Applanix Company Position and Orientation System for
Airborne Vehicles (POS AV), providing velocity
accuracy on the order of 5 mm/s [25]. The Doppler
frequency created by the relative motion between a
receiver and transmitter is


(5)
in which

is a unit vector from receiver to


transmitter,

and

are velocities of the transmitter and


receiver, respectively, in the same reference frame, and

is the wavelength of the L1 carrier wave.



The IGS orbit files include a clock correction parameter
for each SV, as well. This correction is applied as very
small shift in the carrier frequency. Clock errors are on
the order of 10

sec/sec, which are manifested as errors


in range of less than 10 cm, and in velocity, less than
1/1000 mm/s. The effects of relativity, which are also
centimeter-level, are ignored.

The frequency error represents the modeling error
between the pre-computed Doppler frequency which was
the input to OL tracking and the actual Doppler frequency
(plus noise) apparent on the signal at the time of
reception. It is clear from [14] the input to OL tracking
need only be with 10-15 Hz of the received Doppler
frequency. The frequency errors of the three cases
studied in this work are at most 4 Hz, well within the
required limit.

CODE WIPEOFF

Code wipeoff involves multiplying the received signal by
a copy of the ranging code at an estimate of the delay ,
which is pre-computed from the Doppler model. It is
assumed is accurate enough so that received and
estimated ranging codes align to within a small fraction of
a code chip.
1 (6)
Therefore the signal following code wipeoff is

+ . (7)
The Doppler model is assumed to be sufficiently close to
the actual Doppler frequency so that the estimated delay
can be updated indirectly by adjusting the code rate,

,
based on the estimate of the Doppler frequency.

1 +

(8)
where

is the nominal code rate of 1.023 MHz, and


is the frequency of the carrier on the L1 band (1575.42
MHz). This method of updating the code delay is
sufficient to keep the code aligned and enable wipeoff for
the duration of an occultation. Recall that the code chips
are 293 m in length, so alignment to within a few meters
should be good enough for code wipeoff.

CARRIER WIPEOFF

During carrier wipeoff, the despread signal is multiplied
with the complex exponential

=


(9)
generated at the model frequency. Note that frequency is
the derivative of phase.

=
1
2


(10)
The local signal and the model phase, at time step

, is
the accumulated effect of Doppler from the 1
previous integration times.

= 2

(11)
1051
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010
Here,

and
,

, a
convention which is repeated. The product is integrated
over the period

to form the complex correlator sum

=
1

(12)
The integration time,

, is set to one code period, which


varies with time due to the Doppler frequency as in
Equation (7). In practice, the number of samples per 1 ms
integration period (code cycle) is updated by

. The
number of samples per code cycle

is

= 1023

(13)
where

is the sampling rate of the system [20].



Assume that frequency and amplitude can be
approximated as piecewise constant functions (

and
,

), the data bit is constant over the


period of integration (since it is synchronous with a code
cycle), and data bit transitions are aligned with integration
times. Incorporating these assumptions into Equation
(11), a model for the result of code wipeoff,

, is
obtained.

(14)

sinc
,


(15)

where
,
=
,

,

and

.

RESIDUAL PHASE

The complex correlator

consists of real and imaginary


components,

and

, known as
inphase and quadrature components, respectively. The
inphase and quadrature components are next coherently
summed over 1 data bit, reducing the data rate from 1
KHz to 50 Hz. This effectively increases the SNR and
reduces the probability of a cycle slip.


(16)
The residual phase is extracted from the coherent
correlator sums by use of a four-quadrant arctangent
function, which requires knowledge of the data bit

.
GPS data bits are downloaded and from the COSMIC Bit
Grabber Network [17].

(17)

= tan

(18)
The variable

unwraps the residual phase, to


accumulate values larger than [, ].

+ 2,

<

2,

> +

(19)
The total phase may then be compared with other phase
measurements, if available, for integrity checking. In this
case the phase recorded by survey-quality Trimble NetRS
receivers was used. The NetRS receivers employ high-
fidelity, proprietary tracking algorithms that can
successfully generate carrier phase measurements during
the initial portion of occultation events, when the satellite
was at a higher elevation. Once the ray path extends
deeper into the atmosphere, the NetRS receivers will lose
lock and no longer produce measurements.

EQUIPMENT

The GNSS Instrument System for Multistatic and
Occultation Sensing (GISMOS) was developed at Purdue
University, under Subcontract S05-39696 for the
University Corporation for Atmospheric Research
(UCAR) in Boulder, CO, for the purpose of recording
occulted and reflected GNSS signals. A standard rack
assembly for the HIAPER aircraft houses the patch panel,
Symmetricom ExactTime 6000 timing receiver, Applanix
Company Position and Orientation System for Airborne
Vehicles (POS AV), Trimble NetRS receivers, keyboard-
video-monitor assembly, master computer, GNSS
Recording System (GRS), and a Just-a-Bunch-Of-Disks
(JBOD) enclosure. Additionally, high-gain limb-looking
antennas were custom designed and constructed by the
First RF Corporation. The antennas were designed to
have a narrow vertical and wide horizontal gain pattern to
maximize the field of view for RO signals. One antenna
was hard-mounted to the inside of forward-most window
on each side of the aircraft. Fig. 1 depicts interior (left)
and exterior (right) views of the First RF antennas
mounted in the research aircraft.

1052
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010

Fig. 1 Interior (top) and exterior (bottom) views of
the First RF occultation antennas.
The centerpiece of GISMOS is the GNSS Recording
System (GRS). The GRS was built by Johns Hopkins
Applied Physics Laboratory to record GPS signals for
post-processing. It samples the GPS signal on L1 and L2
at 10 MHz, writing complex-interleaved samples at 1-bit
quantization to the JBOD. The GRS is capable of
recording simultaneously from three input channels. The
GRS stripes the data across the drives housed in the
JBOD. The JBOD has two separate 6-disk bays; one for
writing from the GRS, and the other for extracting the
data for processing.

The Symmetricom ExactTime 6000 timing receiver
provided the pulse-per-second and 10 MHz reference
frequency critical for removing the clock bias between the
individual GISMOS instruments and the GPS
constellation. The reference signals are controlled by an
oven-controlled crystal oscillator, that is phase-locked to
GPS time.

The Applanix Company Position and Orientation System
for Airborne Vehicles (POS AV) is an integrated
GPS/INS system mating an optical gyroscope with dual
frequency GPS measurements to provide accurate
position and velocity of the aircraft. It was responsible
for providing the 5 mm/s velocity accuracy needed for
successful inversion of the excess phase on the received
signal.

The flight rack was also equipped with geodetic quality
NetRS receivers. Two NetRS receivers received
concurrent feeds of the limb-looking antennas during the
research flights. The performance of OL tracking is
compared to NetRS performance in the results section.

The Purdue Software Receiver (PSR) provided the means
by which the recorded signals were processed [19]. The
PSR code is written in a mixture of the C and C++
languages, however most of the correlating functions are
written in machine assembly language designed for use
with Intel processor chips with Streaming SIMD
Extensions (SSE). The use of SSE instruction sets greatly
increases the processing speed of the PSR. The PSR can
operate 12 channels in OL or CL mode, and writes the
tracking information (e.g. Doppler frequency, phase,
pseudorange, etc.) to binary files. The PSR is only
capable of operating in post-processing mode.

Ventre, et. al., 2006 describes the implementation of the
OL code into the PSR, and presents results of testing the
code on data from high-elevation satellites observed from
a stationary receiver [20].

HIAPER

The High-performance Instrumented Airborne Platform
for Environmental Research (HIAPER) is a modified
Gulfstream V aircraft purchased by the NSF. The aircraft
is operated by the Research Aviation Facility (RAF),
based at the Rocky Mountain Metropolitan Airport in
Broomfield, Colorado. HIAPER has a published service
ceiling of 15,000 m (51,000 ft) and a range of about
11,265 km (6,083 nautical miles) [21]. During research
flights the aircraft is laden with personnel and
instruments, and therefore the cruising altitude is several
thousand feet less than the service ceiling. The cruising
height during the research flights was between 13,716 and
14,620 m (45,000 and 48,000 ft).

RESULTS

The research campaign spanned 11 days, from 13-23
February 2008. The data analyzed in this work was
recorded on 22 February 2008, day 53 of year 2008
(2008.053).

Results for the setting occultations of PRNs 5, 12, and 15
during the 2008.053 flight are presented here. The first
two, PRN 5 and 12, were simultaneously setting starboard
of the aircraft as it headed East over northern Alabama
and Georgia. PRN 15 was recorded as the satellite was
1053
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010
setting on the starboard side of the aircraft as the aircraft
headed North above the Florida peninsula. In Fig. 2, the
location of the aircraft at these two time periods are
identified with red arrows overlaid on the blue flight path.


Fig. 2 Flight Path of HIAPER on 22 February 2008.
Two recording periods are identified by red lines with
the direction of flight indicated by an arrow; the entire
flight path is traced in blue. Warning areas in the
Gulf of Mexico are indicated by darker areas with
white boarders. The edge of US/Mexican airspace is
indicated by the red line across the middle of the Gulf
of Mexico. (The Google Earth image displayed in
Fig. 2 is used in this work of scholarship under the
principle of "Fair Use.")
The performance OL tracking executed by the PSR is
compared to that of a NetRS receiver (ID: PU12); both
PU12 and the GRS were receiving signals from the same
antenna.

Table 1 lists the initial epoch of each occultation. In the
results that follow, timestamps on the x-axis of Figs.
correspond to time forward from these initial epochs.

Table 1 Initial epochs of the occultation events.
PRN ID Initial Epoch
5 22 February 2008, 19:16:28 GPST
12 22 February 2008, 19:16:28 GPST
15 22 February 2008, 23:15:52 GPST

Fig. 5, Fig. 7, and Fig. 9 are dual axis plots which show
the residual phases from OL tracking, PSR CL tracking,
NetRS CL tracking, and the elevation of the satellite.
Elevation is defined as an angle relative to the local
horizon, which is defined as the plane normal to the
WGS-84 ellipsoid at the phase center of the antenna.
From these plots, it is clear that the OL tracking method
loses lock after the CL methods, and hence tracks the
signal to a lower ray-path height. The focus is now turned
to a qualitative comparison between PSR OL and NetRS
CL tracking.

In comparing the NetRS CL and PSR OL measurements,
single and double differences are examined. Total phase

from receiver viewing satellite is


+ + +


(20)
in which the terms on the right hand side of Equation (20)
represent (from left to right) path delay, satellite clock
bias, receiver clock bias, integer ambiguity, ionospheric
delay, tropospheric delay, and noise. Also in Equation
(20) are the speed of light, c, the carrier wavelength, ,
and the signal transit time,. The single difference is
created by differencing the received phase of two
receivers tracking the same SV.

,

]
+


(21)
The right hand side of Equation (21) consists of (from left
to right) differential path length to the receiver, receiver
clock difference, a constant integer ambiguity, and the
AWGN noise terms. Notice that the SV clock difference
is removed. The ionospheric and tropospheric delays
would also be the same, since both receivers use the same
antenna. The receiver clock difference can be
approximated as a constant offset plus a rate term.

(22)
where

is the rate of the clock in receiver . The


single difference phase can be further approximated as

,


,,

(23)
where
,,

is the sum term of the values assumed to be


constant. The remaining components are the rate of the
receiver clock,

, and the differential random


noise

.

After creating the single difference between NetRS and
OL phase, a non-zero rate is evident for both PRN 5 and
12. Fig. 3 shows the single difference measurements for
all three signals.

1054
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010

Fig. 3 Single difference between OL and NetRS
phase measurements. The initial epoch for PRN 5 and
12 was approximately 4 hours before the initial epoch
of PRN 15.

A receiver clock bias, represented by the term

in Equation (23), between the phase measurements


of PRN 5 and 12 is most likely the cause for the constant-
slope offset in the difference between the PSR OL and
NetRS CL measurements seen in Fig. 3. No significant
bias appears for measurements from PRN 15, which were
recorded approximately 4 hours later. One hypothesis is
that the internal clock in the NetRS receiver had not fully
adjusted itself to the 10 MHz reference at the time that the
PRN 5 and 12 data were collected. Clock-steering
algorithms in the NetRS may have slowly adjusted the
oscillator to align with the 10 MHz reference. Another
possible explanation is that the Symmetricom oscillator,
though having indicated locked on its front panel, had
settled more completely before the PRN 15 occultation
event. With four hours separating the two recording
periods, the Symmetricom had substantially more time for
its oscillator to settle before the occultation of PRN 15
was recorded.

To test whether a time bias was the cause of the bias seen
in the phase measurements of PRN 5 and 12, a double-
difference measurement was computed. By double-
differencing the measurements, the receiver clock bias as
well as the satellite clock bias are both eliminated. The
double difference measurement is

,
,
=
,


=
,,


,,


(24)
where the superscript indicates the SV and the subscript
indicates the receiver. Notice the rate term drops out and
the remainders are a constant

and the sum of the


random noise terms

.

The double difference should produce a zero-mean signal
plus AWGN. The variance of the double difference
measurement (

) is thus

= 2

+2

(25)
where

and

are the variances of the OL tracking

and NetRS

phases.
Assuming the NetRS measurements are perfect

=
0 creates an upper bound for the variance for OL
tracking. The upper bound of the variance of the OL
phase measurements would be equal to half that of the
double difference measurement.

2

(26)
Fig. 4 shows a plot of the double difference measurements
using PRN 5 and 12 from OL and NetRS tracking.

If OL tracking was working correctly, after removing all
apparent clock biases through single- or double-
differencing, the standard deviations of all measurements
should be approximately the same. The standard
deviations of the single and double difference
measurements are shown in Table 2, and are indeed on
the same order of magnitude. As expected, the standard
deviation of the double difference is slightly higher than
the single difference measurement. This demonstrates the
OL tracking measurements agree with those from the
NetRS on the order of 0.016-0.018 cycles. The standard
deviation of the measurements corresponds to an error of
approximately 3.4 mm. Under nominal signal conditions,
PLL carrier phase measurements are on the order of 1-2
mm [22]. Thus, the OL occultation measurement noise,
expected to be higher than those from a closed tracking
loop, is still at an acceptable level.


Fig. 4 Double difference between OL and NetRS
phase measurements of PRN 5 and 12.

1055
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010
Table 2 Standard deviations of the single difference
measurement of OL and NetRS phase for PRN 15 and
the double difference measurement of OL and NetRS
phase for PRNs 5 and 12.

Continuing from the conclusion that PSR OL
measurements agree with the NetRS CL measurements,
notice once again Figs. 5, 7, and 9. From these figures it
is clear that NetRS CL tracking loses lock much earlier
than PSR OL tracking. To determine when OL tracking
loses lock a minimum signal-to-noise ratio (SNR)
threshold is set. Fig. 6 shows the SNR for PRN 5. A
threshold on the SNR of a 60-second simple moving
average, for useful retrievals, was assumed to be 1,

< 1, no further useful scientific . When the SNR


decreases below that threshold, the phase residual is
assumed to reduce to a uniformly distributed random
variable. Fig. 8 and Fig. 10 show SNR plots for PRN 12
and 15, respectively.


Fig. 5 PRN 05 residual phase and elevation. In this
and subsequent figures, the zero point corresponds to
the beginning of the recording epoch for the particular
event being described. The epochs may be found in
Table 1.

Fig. 6 PRN 05 signal-to-noise ratio.

Fig. 7 PRN 12 residual phase and elevation.

Fig. 8 - PRN 12 signal-to-noise ratio.
Measurement Standard deviation
[full cycles]
Single difference (PRN 15) 0.015580
Double difference (PRN 5 and 12) 0.017897
1056
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010

Fig. 9 PRN 15 residual phase and elevation.

Fig. 10 PRN 15 signal-to-noise ratio.
The SNR criterion above was used to determine a time for
loss of data. Referencing the plots of elevation vs. time,
the final elevation for which data could be recovered were
calculated, and presented Table 1. OL tracking
consistently tracked the signal to a lower elevation than
the NetRS.
Table 3 Minimum elevation of the straight line path
to which the signal was tracked to and minimum
height of ray path (HORP) values for NetRS CL and
PSR OL tracking. (Negative elevation indicates an
angle below the local horizon.)
Minimum Elevation (degrees)
PRN ID NetRS CL PSR OL
5
12
15
-0.9
-3.6
+0.7
-3.9
-4.2
-4.5
Height of Ray Path (km)
PRN ID NetRS CL PSR OL
5
12
15
6.64
>8
>8
4.19
4.26
2.54

A simulation was then conducted to produce a numerical
relationship between the height of ray path (HORP) and
excess phase, using the orbital geometry and atmospheric
profile at the time of the PRN 5 occultation. This function
was then used to compute the HORP at the time of data
loss, using the excess phase from Figs. 5, 7, and 9. These
results show the improvement from OL tracking, in terms
of the minimum HORP reachable, versus that from NetRS
CL tracking. These results are presented in Table 3

During the simulation, the GPS orbits were approximated
as circular and the altitude of HIAPER was fixed at 11.4
km. The receiver and transmitter velocities were assumed
to be 0.25 km/s and 3.83 km/s, respectively, and relative
motion was restricted to the Y-Z plane. The state of the
atmosphere at the time and location of the occultation was
simulated using scientific information gathered by a
nearby radiosonde launch. The GPS signal was
propagated through the simulated 1-D atmospheric model
by the ray-tracing method [23].

At the time this paper was written, the occultation of PRN
5 had been the only event simulated. Since the all of the
events had occurred in a geographically close region and
were separated by a short time span, it was assumed the
simulation parameters were sufficiently close to those of
PRN 12 and 15 such that their HORP trajectory could be
estimated from the same data set. Thus, the simulation
results for PRN 5 were used to determine the HORP for
all the three occultations studied in this paper.

Finally, the frequency error for each case was calculated
and compared to the specified requirements for RO
science. The residual phase is produced every 20 ms.
Retrievals, however, are limited to a resolution set by the
Fresnel zone. Therefore, the residual phase differences
can be averaged over a resolution cell of one Fresnel zone
(60 m). For the simulation described here, the tangent
point descended 60 m every five seconds (averaged over
the duration of the occultation). The frequency error,
computed by differencing the residual phase
measurement, was averaged over five seconds and the
results are shown in Figs. 11, 12, and 13.

A 5 mm/s velocity error will produce retrieval errors on
the order of 1 K at 10 km, comparable to the accuracy
observed by radiosondes [24]. In order to supplement
current observations, OL tracking would need to meet or
exceed this accuracy. The noise of the averaged OL
tracking measurements was found to be on the order of
0.015 Hz, equal to approximately 3.0 mm/s velocity error.
This implies useful retrievals could be made with
measurements from OL tracking.
1057
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010

Fig. 11 PRN 5 frequency error, averaged over one
Fresnel zone (60m).


Fig. 12 PRN 12 frequency error, averaged over one
Fresnel zone (60m).


Fig. 13 PRN 15 frequency error, averaged over one
Fresnel zone (60m).

CONCLUSIONS

In this work, an OL tracking method for airborne RO
signals has been described. It was shown that OL phase
measurements are consistent with phase measurements
from a closed-loop tracking receiver, with a frequency
bias, in the first set of results, most likely caused by a
difference in clock bias between the two receivers.
Probable hypotheses are that this bias was the caused by
either an error in the way the NetRS receiver adjusted to
the external reference frequency, or the stability of the
reference source improving over time. It was also
demonstrated that the phase noise of OL tracking is
approximately 3.4 mm.

It has been demonstrated that OL tracking is better for
tracking RO signals than traditional CL methods. Signals
received from an occulting SV are known to be
characterized by phase accelerations of up to 1000 Hz/s
and rapid fading of signal amplitude due to multipath
caused by water vapor in the lower moist troposphere
[15]. The OL tracking method is able to extract a
continuous phase time-history of RO signals to lower
elevation than traditional CL methods. For the three cases
studied, PRNs 5, 12, and 15, signals were tracked in OL
mode to a minimum height of ray path of 4.19km,
4.26km, and 2.54km, respectively, whereas CL tracking
yielded a minimum height of ray path of 6.64km for PRN
5 and >8km for PRN 12 and 15. Finally, frequency error
of the measurements averaged over the Fresnel zone
indicate noise in the OL measurements is less than 3
mm/s, below the 5 mm/s requirement for extracting useful
scientific measurements. The results demonstrate OL
tracking is the preferred method for tracking GPS radio
occultation signals.

ACKNOWLEDGMENTS

This work was funded under Subcontract S05-39696 for
the University Corporation for Atmospheric Research
(UCAR) in Boulder, CO. The authors would like to
acknowledge all personnel at the UCAR Research
Aviation Facility who had a part in making our research
flights possible and enjoyable. Also, we wish to thank
Ben Wilmhoff from the First RF Corporation for his
technical support with the radio occultation antennas.

REFERENCES

[1] E. R. Kursinski, G. A. Hajj, J. T. Schofield, R. P.
Linfield, and K. R. Hardy. Observing Earths atmosphere
with radio occultation measurements using the Global
Positioning System. J. Geophys. Res., 102(D19):23429
23465, 1997.

[2] R. Ware, M. Exner, D. Fend, M. Gorbunov, K.
Hardy, B. Herman, Y. Kuo, T. Meehan, W. Melbourne,
1058
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010
C. Rocken, W. Schreiner, S. Sokolovskiy, F. Solheim, X.
Zou, R. Anthes, S. Businger, and K. Trenberth. GPS
Sounding of the Atmosphere from Low Earth Orbit:
Preliminary Results. Bull. Am. Meteo. Soc., 77(1):1940,
Jan 1996.

[3] R. A. Anthes, P. A. Bernhardt, Y. Chen, L. Cucurull,
K. F. Dymond, D. Ector, S. B. Healy, S.-P. Ho, D. C.
Hunt, Y.-H. Kuo, H. Liu, K. Manning, C. McCormick, T.
K. Meehan, W. J. Randel, C. Rocken, W. S. Schreiner, S.
V. Sokolovskiy, S. Syndergaard, D. C. Thompson, K. E.
Trenberth, T.-K. Wee, N. L. Yen, and Z. Zeng. The
COSMIC/FORMOSAT-3 Mission Early Results. Bull.
Am. Meteo. Soc., pages 313333, Mar 2008.

[4] J. Wickert, C. Arras, C. O. Ao, G. Beyerle, C. Falck,
L. Grunwaldt, S. B. Healy, S. Heise, A. Helm, C. Y.
Huang, N. Jakowski, R. Konig, T. Mannucci, C. Mayer,
G. Michalak, P. Poli, M. Rothacher, T. Schmidt, R.
Stosius, and B. Tapley. CHAMP, GRACE, SAC-C,
TerraSAR-X/TanDEM-X: Science results, status and
future prospects. GRAS SAF Workshop on Applications of
GPSRO Measurements, pages 4352. GRAS SAF, Jun
2008.

[5] J. Wickert, G. Michalak, T. Schmidt, G. Beyerle, C.
Z. Cheng, S. B. Healy, S. Heise, C. Y. Huang, N.
Jakowski, W. Kohler, C. Mayer, D. Offiler, E. Ozawa, A.
G. Pavelyev, M. Rothacher, B. Tapley, and C. Arras. GPS
Radio Occultation: Results from CHAMP, GRACE and
FORMOSAT-3/COSMIC. Terr. Atm. Ocean. Sci.,
20(1):3550, Feb 2009.

[6] C.-J. Fong, N. L. Yen, C.-H. Chu, S.-K. Yang, W.-T.
Shiau, C.-Y. Huang, S. Chi, S.-S. Chen, Y.-A. Liou, and
Y.-K. Kuo. FORMOSAT-3/COSMIC Spacecraft
Constellation System, Mission Results, and Prospect for
Follow-On Mission. Terr. Atm. Ocean. Sci., 20(1):119,
Feb 2009.

[7] A. von Engeln, Y. Andres, L. Butenko, D. Klaes, C.
Marquardt, E. OClerigh, F. Sancho. GRAS Radio
Occultation Measurements Onboard Metop. In Proc.
EUMETSAT Met. Sat. Conf., Sep 2008.

[8] C. Zuffada, G. A. Hajj, and E. R. Kursinski. A novel
approach to atmospheric profiling with a mountain-based
or airborne receiver. J. Geophys. Res., 104:24435 24447,
1999.

[9] X. Hu, X.C. Wu, X.Y. Gong, C.Y. Xiao, X.X. Zhang,
Y. Fu, X.Y. Du, H. Li, Z.Y. Fang, Q. Xia, G.L. Yang, and
J.T. Mao. An introduction of mountain based GPS radio
occultation experiments in China. Advances in Space
Research, 42(10):17231729, 2008.

[10] Y. Aoyama, Y. Shoji, A. Mousa, T Tsuda, and H.
Nakamura. Down looking GPS occultation measurement
on the top of Mt. Fuji. In Proc. of Intl. Workshop on GPS
Met., pages 20812084, Tsukuba, Japan, Jan 2003.
MEXT/JISTEC.

[11] Y. Aoyama, Y. Shoji, A. Mousa, T Tsuda, and H.
Nakamura. Temperature and water vapor profiles derived
from downward-looking GPS occultation data. J. Meteo.
Soc. of Japan, 82(1B):433440, 2004.

[12] T. Yoshihara, N. Fujii, K. Hoshinoo, K. Matsunaga,
S. Saitoh, T. Sakai, T. Tsuda, Y. Aoyama, S. Danno.
Measurement System and Experimental Results of
Airborne-based Downward-Looking GPS Occultation. In
Proc. ION GNSS. Inst. Navig., Sep 2005.

[13] S. Sokolovskiy, C. Rocken, W. Schreiner, D. Hunt,
and J. Johnson. Postprocessing of L1 GPS Radio
Occultation Signals Recorded in Open-loop Mode. Radio
Sci., 44, Mar 2009.

[14] S. V. Sokolovskiy. Tracking Tropospheric Radio
Occultation Signals from Low Earth Orbit. Radio Sci.,
36(3):483498, May/Jun 2001.

[15] S. V. Sokolovskiy. Modeling and Inverting Radio
Occultation Signals in the Moist Troposphere. Radio Sci.,
36(3):441458, May/Jun 2001.

[16] G. Beyerle, T. Schmidt, J. Wickert, S. Heise, M.
Rothacher, G. Konig-Langlo, and K. B. Lauritsen.
Observations and simulations of receiver-induced
refractivity biases in GPS radio occultation. J. Geophys.
Res., 111(D12101), June 2006.

[17] Cosmic Bit Grabber Network Archive. [online]
http://cosmic-io.cosmic.ucar.edu/cdaac/index.html
[18] M. Schenewerk. A brief review of basic GPS orbit
interpolation strategies. GPS Solutions, 6:265267, 2003.

[19] G. Heckler and J. L. Garrison. Architecture of a
reconfigurable software receiver. In Proc. ION GNSS.
Inst. Navig., Sep 2004.

[20] B. D. Ventre, J. L. Garrison, and J. S. Haase M. H.
Boehme. Implementation and testing of the open-loop
tracking for airborne GPS occultation measurements. In
Proc. ION GNSS, pages 620633. Inst. Navig., Sep 2006.

[21] University Corporation for Atmospheric Research.
GV (HIAPER), 2008. [online]
http://www.eol.ucar.edu/instrumentation/aircraft/G-V/g-v.

[22] P. Misra and P. Enge. Global Positioning System:
Signals, Measurements, and Performance. Ganga-
Jamuna Press, Lincoln, Mass., 2
nd
edition, 2006.
1059
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010

[23] P. Muradyan, May 2010. Personal communication

[24] S. Xie, J. S. Haase, and S. Syndergaard. Profiling the
atmosphere using the airborne GPS radio occultation
technique: A sensitivity study. IEEE Trans. on Geosci.
and Remote Sensing, 46(11), Nov 2008.

[25] O. Lesne, J. Haase, G. Kirchengast, J. Ramsauer, W.
Poetzi. Sensitivity analysis for airborne sounding of the
troposphere by GNSS radio occultation. Phys. And
Chem. of the Earth, 27, 2002.
1060
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010

Вам также может понравиться