Sei sulla pagina 1di 9

AIR-GROUND CHANNEL CHARACTERIZATION FOR UNMANNED

AIRCRAFT SYSTEMS: THE OVER-FRESHWATER SETTING


David W. Matolak, Ruoyu Sun, University of South Carolina, Columbia, South Carolina

Abstract The Radio Technical Commission for


Aeronautics (RTCA) special committee (SC) 228 [1]
Wireless communication systems must operate
is responsible for CNPC specifications. These
in the presence of potentially dispersive and rapidly
specifications will be for two spectral bands: the L-
time-varying channels, and the unmanned aircraft
band (960-977 MHz) and C-band (5030-5091 MHz)
system (UAS) control and non-payload
[2]. RTCA recommendations will also likely guide
communication (CNPC) system is no exception. To
policy for the International Civil Aviation
ensure high reliability, the UAS air-ground (AG)
Organization (ICAO) [3].
channel must be quantitatively characterized. In this
paper we report on measurements of the AG channel An important element of any wireless
in an over-freshwater setting, for a project sponsored communication system is the wireless channel, which
by NASA Glenn Research Center. The measurements in the case of UAS is the air-ground (AG) channel.
on which we report were conducted with a ground Channel effects such as fading and dispersion can
site near the coast of Lake Erie, with flights over the significantly degrade link performance. The vast
lake. The flight test measurements collected majority of past research efforts on the AG channel
simultaneous channel impulse responses in two UAS considered tall ground sites (GSs) in wide open areas,
bands, and employed two receivers for each band to and short duration, narrowband signals. In these
enable assessment of antenna diversity as well as cases, fairly simple channel models suffice. For
inter-band channel characteristics. We provide a short planned wider bandwidth CNPC signals, and UAS
introduction and description of the measurements, AG communication in more complex settings (e.g.,
then describe results for propagation path loss, delay low altitudes, near buildings, etc.), more detailed AG
spread, and correlations across antennas for the over- channel characteristics are needed to ensure that
freshwater AG channel, for two example flight paths. high-performance CNPC radio system designs can be
Path loss generally follows the curved-earth two-ray tested [4] in the most realistic and challenging
model, with a weak surface reflection; multipath conditions.
components from buildings near the coastline are also Thus to obtain quantitative AG channel
present. Maximum root-mean square delay spreads characteristics, NASAs Glenn Research Center
range from 70 to 160 ns, and inter-band correlations (GRC) has sponsored a measurement campaign. The
are well represented as zero-mean Gaussian with measurements were taken in a number of GS
standard deviation 0.3. We also comment on the
environments (over-sea, hilly terrain, etc.), with two
connection with recent results for the over-sea types of flight paths: straight (toward or away from
setting, and the AG channel models we are
the GS), and oval shaped. Models for the AG channel
developing. will be developed for all GS settings from this
measurement data, and these models will be used to
I. Introduction evaluate performance of candidate CNPC radios. In
Unmanned aircraft systems (UAS) are expected this paper we describe results of the measurements in
to proliferate in the next few years. To ensure safety the over-freshwater setting.
and reliability, researchers and developers in
The remainder of this paper is organized as
industry, government, and academia are working on
follows: in Section II we describe the measurement
all aspects of UAS deployment. One important aspect
system and over-freshwater environment. Section III
is the control and non-payload communications
describes measurement results, including path loss,
(CNPC) link. Since safety and reliability are
delay dispersion, and the correlation among the
paramount, UAS CNPC systems are being specified
signals received on the two different frequency
to meet rigorous performance requirements.
bands. We also comment on comparison with recent

978-1-4799-4891-8/14/$31.00 2014 IEEE


K1-1
over-sea AG channel results and future AG channel (c) GS view looking east toward downtown
models. In Section IV we conclude. Cleveland, Lake Erie to left).

II. Measurements
A. Measurement System
The measurement system is a dual-band, single-
input/multiple-output (SIMO) system that transmits
an L-band and a C-band signal simultaneously. Two
receivers in each band are spatially separated to allow
evaluation of antenna diversity. The complete system
(two transmitters plus four receivers) is termed a
channel sounder. The sounder was custom
manufactured by Berkeley Varitronics, Inc. [5],
(a) (b)
according to specifications developed by the author
and NASA GRC. Figure 1. Measurement System Components
The transmitters were located at the GS and the
four receivers were on the aircraft (see Figure 1(a)).
Figure 1(b) shows the transportable tower and
antenna mast (both antennas at height ~ 20 m). Both
transmitters send a direct-sequence spread spectrum
signal, with chip rate 5 Mcps at L-band, and 50 Mcps
at C-band, corresponding to delay resolutions of 200
ns and 20 ns, respectively. The receivers are stepped
correlators, which allow us to obtain estimates of the
AG channel impulse response (CIR).
The center frequency at C-band was 5060 MHz,
and that for L-band was 968 MHz. Transmit power
for both bands is 10 watts. A high-power amplifier of
gain 7 dB, and a 30 dB low noise amplifier (LNA)
are employed in C-band; a 15.5 dB gain LNA is also (a)
used at L-band. The GS antennas have gains of 6 dB
for C-band, 5 dB for L-band, and elevation/azimuth
beamwidths are approximately 40/70 for C-band,
85/60 for L-band. Aircraft monopole antennas are
nearly omnidirectional in azimuth, with gain 5 dB.

B. Measurement Site
The measurement flight was conducted over
Lake Erie near downtown Cleveland on 22 October
2013. The GS was located at 4129'33.8" N,
81 44'5.48" W with elevation (AMSL) of 177.4 m.
The GS antennas were elevated 20 m above the
ground by the transportable tower (Figure 1).
Figure 2 shows photos of the measurement
environment. ((a) aircraft view from south, looking
toward GS, Lake Erie and downtown Cleveland; (b) (b)
aircraft east view from aircraft along coastline; and

K1-2
(c)
Figure 2. Photo of Measurement Environment (b)
The GS antennas were oriented to 352 from Figure 3. GS and Aircraft Flight Routes FT2
geographic north in azimuth, with zero elevation (Straight, Orange) and FT4 (Oval, Green)
angles. The water level of Lake Erie is on average FT2 is a straight flight path toward the GS from
174 m above sea level. 29,350 m to 2,511 m (972 m in L-band, since we
The aircraft flew both straight paths toward and restrict analysis here to signals within the antenna
away from the GS, and also oval-shaped paths with main beams). The elevation angle ranges from 1.4 to
the oval major axis perpendicular to the straight line 13 degrees. The average altitude difference between
from GS to oval center. The flight altitude was kept GS and aircraft antennas is 566.3 m, with maximum
approximately constant. Each path for which AG 578.2 m and minimum 553.7 m.
channel measurement data was taken is denoted a FT4 is an oval shaped path with link distance
flight track (FT). In this paper we present results only ranging from 16.314 km to 21.764 km. The elevation
from one FT of each type. Figure 3 shows the two angle ranges from 1.7 to 2.2 degrees. The average
FTs we address in, (a) Earth-Centered Earth-Fixed altitude difference between GS and aircraft antennas
(ECEF) coordinates; (b) Google Maps. is 567.4 m with maximum 578.4 m and minimum
. 555.2 m.
ClevelandOH***10-22-2013***FT2&4

6
III. Measurement Results
x 10
In this section we present results for propagation
4.23
path loss, root-mean square delay spread, and
correlation between the signals on antennas in the
4.22
two bands. We also show the effect of nearby large
buildings from downtown Cleveland. To obtain
Z

4.21
results strictly pertinent to the freshwater coastal
Tx (Ground)
4.2
Rx (Aircraft) setting alone (without the nearby urban center), we
-4.71 Rx (Aircraft)
Large RMS-DS
remove in post-processing multipath components
7
6
-4.72 Rx start point
6.9 (MPCs) attributable to reflections from the Cleveland
x 10 Rx end point
-4.73
6.8 x 10
5
urban center. We do provide some example results
-4.74
Y 6.7
X with the urban MPCs, for illustration. As we will
show, even when the long-delay MPCs from the
(a) urban center are removed, MPCs due to smaller
structures near the lake coast are still present in many
CIRs.

K1-3
A. Path Loss 150
ClevelandOH***10-22-2013***FT2***C-band Rxs

C-band Rx1
Path loss is computed using known transmit 145 C-band Rx2
power, antenna gains, and cable losses. Measured Free Space PL
140
received power for each receiver is logged, and via a Two Ray

Path Loss (dB)


standard link budget equation, path loss is estimated 135
for each received power delay profile (PDPthe 130
power version of the CIR).
125
For our analysis, we use the curved earth two 120
ray (CE2R) path loss model, and we also incorporate
115
surface roughness [6] and reflection divergence due
to the spherical earth [7]. The relative dielectric 110
2.5 3 4 5 10 15 20 2529.35
constant (permittivity) r = 81, conductivity = 0.01 Link Distance (km)
S/m, and relative permeability r = 1 are used for
fresh water [8]. The average wind speed for Lake Figure 4. C-Band Path Loss vs. Distance for FT2
Erie on 22 October 2013 was 11.16 mph [9]. This is (Straight)
applied to compute our water surface reflection ClevelandOH***10-22-2013***FT2***L-band Rxs
150
coefficient according to the Miller-Brown surface L-band Rx1
roughness model [6]. These material parameters and 140 L-band Rx2
Free Space PL
the Miller-Brown surface roughness model determine Two Ray
130
the magnitude of the two-ray loss variation, which
agrees well with the variation seen in the empirical Path Loss (dB) 120
data. The phase (relative surface reflection delay) of
110
the analytical two ray model differs from the result of
the empirical data. This is likely due to the 100
fluctuating lake surface (this result was also observed 90
in our over-sea measurements [10]).
80
Figure 4 shows the path loss vs. distance for the 1 2 3 4 5 10 15 20 29.35
Link Distance (km)
C-band receivers for the straight FT2. Figure 5 shows
the analogous results for L-band. The mean path Figure 5. L-Band Path Loss vs. Distance for FT2
losses follow the free-space value, but losses are (Straight)
larger at the shorter distances due to aircraft antenna
Figures 6 and 7 show similar results for the oval-
pattern variation (gain reduction) at the larger
shaped FT4. The CE2R model fits the L-band result
elevation angles. The two-ray peaks are most
fairly well. From observation of PDPs, we have
noticeable for the L-band results in Figure 5 since the
concluded that additional weaker MPCs from the lake
lake surface is relatively smoother for the longer-
surface are present, resulting in the more complicated
wavelength L-band signal.
shape of the loss vs. distance curves.

K1-4
150
ClevelandOH***10-22-2013***FT4***C-band Rxs The RMS-DS essentially measures the temporal
C-band Rx1 spread of the AG channel impulse response, and here
C-band Rx2 we primarily report on the instantaneous values.
145
Average Path Loss (dB)

Free Space PL
Two Ray In Figure 8 we show (instantaneous) RMS-DS
140 vs. link distance for the straight FT2. This is before
any post-processing, hence MPCs from large
135 buildings in the nearby downtown Cleveland area
have not been removed (we have applied a moving-
130 average filter to smooth the plots). Since these
urban MPCs are not inherently a part of the over-
125
freshwater channel, two separate analyses should be
16.4 17 18 19 20 21 21.7 done: one for the data with urban MPCs removed,
Link Distance (m)
and one for the data with urban MPCs included.
Figure 6. C-Band Path Loss vs. Distance for FT4 Except for the following example, the results of the
(Oval) latter analysis are not reported here, but will be
ClevelandOH***10-22-2013***FT4***L-band Rxs addressed in a future publication.
140
L-band Rx1 Figure 9 shows a sequence of PDPs (every 20th
135
L-band Rx2
PDP) for the entire flight track with the urban MPCs
Average Path Loss (dB)

Free Space PL
Two Ray included. Numerous MPCs are present for long
130 distances, and their relative delays decrease slightly
as distance increases, in agreement with the
125
geometry. Figure 10 shows the sequence of PDPs for
120
a segment of FT2 for the link distance near 10 km.
Figure 11 illustrates a close-up view of the RMS-DS
115 vs. time. The excess delay of the distant MPCs shown
in Figure 10 is 13.28 microseconds (3984 m) at 10
110
16.4 17 18 19 20 21 21.7 km, it gradually decreases to 13.22 microseconds
Link Distance (m) (3966 m) at 10.32 km, in excellent agreement with
the distance computed by reflection from the large
Figure 7. L-Band Path Loss vs. Distance for FT4
(Oval) buildings in downtown Cleveland.

B. RMS Delay Spread


Channel temporal dispersion can be quantified
in several ways, but the most common by far is the
root-mean square delay spread (RMS-DS) :
L1 2 2
k=0 k k
= L1 2 2 , (1)
k=0 k
where k is the kth MPC amplitude out of a total of L
MPCs in a PDP, k is the delay of the kth MPC, and k
is the mean energy delay, calculated via
L1 2
k=0 k k
= L1 2 . (2)
k=0 k
The denominator sum L1 2
k=0 k in (1) and (2) is Figure 8. C-Band Rx 1 RMS-DS vs. Distance,
the total power in the PDP. This power can be either FT2; Urban MPCs Included
the power for the average PDP for a given set of data,
or for a single PDP; for the latter case the parameters
in (1) and (2) are termed the instantaneous RMS-DS
and instantaneous mean energy delay, respectively.

K1-5
Analyzing the geometry for this Lake Erie
setting, we can estimate that the relative delay of
MPCs from the large buildings in downtown
Cleveland ranges from approximately 9.4 to 18
microseconds over the entire FT2. Based upon this,
we apply a relative delay threshold of 900 ns, and
exclude all MPCs above this threshold for the over-
freshwater AG channel analysis.
ClevelandOH***10-22-2013***FT2***C-band Rx1
80
Measured
70 Moving Averaged 100
Moving Averaged 1000
60

RMS-DS (ns)
50

Figure 9. Sequence of PDPs for FT2 (Straight), C- 40

Band Rx 1; Urban MPCs Included 30

20

10

0
5 10 15 20 25
Link Distance (km)

Figure 12. RMS-DS vs. Link Distance for FT2


(Straight) After Thresholding, C-Band Rx 1

Figure 10. Segment of Sequence of PDPs for FT2


(Straight), C-Band Rx 1; Urban MPCs Included
ClevelandOH***10-22-2013***FT2***C-band Rx1
2500
Measured
Moving Averaged 100
2000 Moving Averaged 1000
RMS-DS (ns)

1500
Figure 13. Sequence of PDPs for FT2 (Straight)
After Thresholding, C-Band Rx 1
1000
Figure 12 shows the RMS-DS vs. link distance
500
for FT2 after this processing (cf. Fig. 8). Without the
urban MPCs, the RMS-DS is much smaller, reaching
0
a maximum of only approximately 70 ns. In Figure
10 10.05 10.1 10.15 10.2
Link Distance (km)
10.25 10.3 10.35
13 we show the sequence of PDPs for the entire FT2
after the 900 ns delay thresholding. Since the LOS
Figure 11. RMS-DS vs. Link Distance Segment for component (chip) in each PDP is aligned to have a
FT2 (Straight), C-Band Rx 1; Urban MPCs delay of 100 ns (5 chips), the delay range in Figure
Included 13 extends from zero to 1000 ns.

K1-6
The RMS-DS vs. time for the entire oval-shaped
FT4 is shown in Figure 14. Large values are present
only in the middle of the measurement; the aircraft
was making a U-turn during this segmentsee the
blue dotted line in Figure 3(a). The sequence of PDPs
for this large RMS-DS segment is shown in
Figure 15.

Figure 16. Google Maps View Indicating


Potential Reflectors for MPCs in Figure 15
Figures 17 and 18 show histograms of RMS-DS
for the two flight tracks, both the original data set (all
instantaneous PDPs), and the moving-average filtered
version of the instantaneous RMS-DS sequences. The
Figure 14. RMS-DS vs. Time for FT4 (Oval) After filtering naturally reduces the maximum values, but
Thresholding, C-Band Rx 1 does not change the means. Although the
distributions may not be strictly heavy-tailed
(hyperbolic), they clearly indicate that a small
percentage of PDPs have RMS-DS values well above
the mean. Table II quantifies the RMS-DS statistics.
The maxima73 ns for FT2 and 161 ns for FT4
are of interest for communication signaling design.
6
ClevelandOH***10-22-2013***FT2***C-band Rx1Original RMS-DS
10

4
10

2
10

0
10
0 10 20 30 40 50 60 70

6
Moving Averaged over 100 PDPs
10

Figure 15. Segment of Sequence of PDPs for FT4


(Oval) After Thresholding, C-Band Rx 1
4
10
Density

Figure 16 shows a plan view of the area near the


2
10

lake coast where several large buildings, three 10


0

harbors with watercraft, water barriers, light houses


0 10 20 30 40 50 60 70

and oil tanks are present. The geometry indicates that


these obstacles are likely responsible for the MPCs in Figure 17. Histogram of RMS-DS for FT2
Figure 15. (Straight) After Thresholding, C-Band Rx 1

K1-7
10
6
ClevelandOH***10-22-2013***FT4***C-band Rx1Original RMS-DS means of the A vectors. The s denote the standard
deviations of the respective amplitude vectors.
10
4
Element Ak,i (with k=1 or 2 the Rx index) is the
10
2
amplitude sample in the ith delay bin, and each delay
bin is 2 ms here. The vector length n we selected is
10
0

0 20 40 60 80 100 120 140 160


25. Hence the time duration over which we compute
Moving Averaged over 100 PDPs
the correlation is 0.05 s, the reciprocal of the 20 Hz
CNPC signal update rate [1]. The correlation
6
10

10
4 coefficients that result from different values of n (up
to 10000) do not significantly differ.
Density

2
10
An example histogram of the correlation
10
0

0 20 40 60 80 100 120 140 160 coefficients between C-band Rx1 and L-band Rx1 is
shown in Figure 19; it is very well modeled as a
Gaussian distribution. Table III lists inter-band
Figure 18. Histogram of RMS-DS for FT4 (Oval)
correlation coefficients. Mean values for each
After Thresholding, C-Band Rx 1
receiver combination are smaller than 0.01, with
Table II. Over Fresh Water RMS-DS Statistics standard deviations less than 0.35 for FT2 and
approximately 0.2 for FT4. The results indicate that
RMS-DS (ns) Mean Max the signals in the two bands are uncorrelated.
Original 73.3
Moving Averaged 2.5
Rx1 9.7 16.2 Correlation Coefficients
Over 100
Normal Mean=0.001 Std=0.20
Moving Averaged 2
14.0
Over 1000
FT2
Original 62.2
1.5
Density

Moving Averaged
Rx2 9.9 17.7
Over 100
Moving Averaged 1
15.3
Over 1000
Original 161.0
0.5
Moving Averaged
Rx1 9.9 119.3
Over 100
Moving Averaged 0
77.3 -0.5 0 0.5
Over 1000 Correlation between C-band Rx1 & L-band Rx1, FT4
FT4
Original 135.8
Moving Averaged
Figure 19. Histogram of Correlation Coefficients
Rx2 10.0 116.6 Between C-Band Rx1 and L-Band Rx1 for FT4
Over 100
Moving Averaged
96.0
(Oval)
Over 1000
Table III. Inter-Band Correlation Statistics
FT2 FT4
C. Inter-band Correlation C1 C1 C2 C2 C1 C1 C2 C2
L1 L2 L1 L2 L1 L2 L1 L2
As shown in Fig. 2, four Rx antennas are mounted
Mean 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.00
on the bottom of aircraft with a small separation
Standard
(~0.4 m). The LOS signal amplitude correlation deviation
0.33 0.34 0.33 0.34 0.20 0.20 0.20 0.20
coefficient between any two antennas in different
bands can be computed as
D. Comparison with Over-Sea Results
EA1 A1 A2 A2
A1 ,A2 = (3) In [10], we reported on results of measurements
A1 A2 taken in the Pacific Ocean near Oxnard, CA, using
where E denotes expectation, the As are vectors of the same measurement system. The over-sea path
LOS signal amplitude samples, and the s are the loss also agreed with the curved earth two-ray model.

K1-8
The over-sea RMS-DS was less than 75 ns, with [4] D. W. Matolak, Air-Ground Channels &
mean near 10 ns, and maximum excess MPC delay Models: Comprehensive Review and Considerations
approximately 460 ns. for Unmanned Aircraft Systems, Proc. IEEE
Aerospace Conference, Big Sky, MT, 3-10 March
Both these environments are very nearly ideal two-
2012.
ray settings since the surface is smooth (relative to
rugged terrain, for example). Our over-freshwater [5] Berkeley Varitronics Systems, website
results showed multiple reflections and larger delay www.bvsystems.com, 12 March 2014.
spreads due to the proximity to downtown Cleveland,
[6] H. V. Hitney, Refractive Effects from VHF to
and it is difficult to remove these entirely from the
EHF Part B: Propagation Models, Advisory Group
data, but further processing to do so is underway. The
for Aerospace Research and Development (AGARD)
inter-band signals are uncorrelated in both over-water
Lecture Series, 1994.
settings.
[7] International Telecommunications Union,
We are working on generating a tapped delay line
Reflection from the Surface of the Earth, Report
(TDL) model for both the over-freshwater and over-
1008-1 (Question 1/5), 1986-1990.
sea settings. Such a model will represent the LOS
component as the first tap, and the earth surface [8] J. D. Parsons, The Mobile Radio Propagation
reflection as the second tap. The third and potentially Channel, 2nd ed., John Wiley & Sons, New York,
fourth taps will be modeled statistically, according to NY, 2000.
measured results for these components relative [9] National Climatic Data Center, website,
power, lifetime, excess delay and probability of www.ncdc.noaa.gov, 20 March 2014.
occurrence.
[10] D. W. Matolak, R. Sun, Initial Results for Air-
IV. Conclusion Ground Channel Measurements & Modeling for
Unmanned Aircraft Systems: Over Sea, Proc. IEEE
In this paper we reported on channel Aerospace Conference, Big Sky, MT, 1-8 March
measurements of the air-ground channel for an over- 2014.
freshwater environment. Two separate frequency
bands at 5060 MHz and 968 MHz were measured.
The curved-earth two-ray model fits the propagation Acknowledgement
path loss very well for link distances from 2.5 to 30 The authors would like to thank the UAS project
km. The RMS-DS in C-band was on average 10 ns team at NASA Glenn Research Center for all their
with maxima 73 ns for the straight flight track and assistance in this work. In particular we extend
161 ns for the oval flight path. The signals in the two special thanks to J. Griner, K. Shalkhauser, R.
bands are uncorrelated. Future work involves Kerczewski, and J. Ishac. Thanks also to Steven H.
additional analysis and processing of the data, and Walker, P.E., for the flight track graphics in Google
development of models for the two over-water air- Maps.
ground channels.
Email Addresses
References matolak@sc.edu
[1] Radio Technical Commission for Aeronautics, sun55@email.sc.edu
website www.rtca.org, 15 March 2014.
[2] International Telecommunications Union,
website, www.itu.int, 14 March 2014. 2014 Integrated Communications Navigation
and Surveillance (ICNS) Conference
[3] International Civil Aviation Organization,
website, www.icao.int, 14 March 2014. April 8-10, 2014

K1-9

Potrebbero piacerti anche