Channel Measurements and Characterizations for Long Range Air-to-Ground Communication Systems in the UHF Band

In this paper, we consider long range air-to-ground (AG) communication systems which support aeronautical platforms including unmanned aerial vehicles (UAVs) in the ultra high frequency (UHF) band. For such a system, we present the measurement based path loss analysis and multipath characteristic results at the distance of hundreds of kilometers. To this end, the experimental AG channel measurement system is implemented at the aircraft with various ground station (GS) environments. Through realistic flight tests, we observe the path loss behaviors for long range UHF channels with sea and ground earth surface reflections. By comparing the measurement results with the empirical path loss model and the spherical earth two-ray model, we demonstrate that our measurement results match well with the model. Moreover, for the multipath channel characterization, we provide field test results on the occurrence probability, delay, and power of multipath components in hilly and mountainous environments with various altitudes of the aircraft.

sensor information at high rates of hundreds of Mbps from the 23 airborne platform to the ground station (GS), while the second 24 one represents a link which sends and receives reliable C2 and 25 The associate editor coordinating the review of this manuscript and approving it for publication was Julien Sarrazin . telemetry messages for the UAV and its payload at the rate of 26 lower than tens of Mbps via the AG wireless channel. 27 Due to the limitation of available spectrum, the wideband 28 AG data link typically adopts high frequency bands such as X 29 or Ku bands even though it suffers from a huge propagation 30 loss [5]. Thus, it practically needs high-gain directional track-31 ing antenna systems to cover hundreds of kilometers. How-32 ever, the lower frequency bands such as ultra high frequency 33 (UHF), L, and C have superiority over higher frequencies in 34 terms of the reliability of the long range AG links, resulting 35 from much less multipath fading, attenuation, phase distor-36 tion, and delay spread [6]. Normally, the UHF band enables 37 to extend a coverage of the system, and it has been shown 38 that atmospheric effects such as precipitation do not affect 39 transmissions in the UHF band. These inherent characteristics 40 provide competitive advantages in harsh applications. Thus, 41 several military AG communications employ the UHF band 42 for long range environments [7], [8]. 43 UHF band at the aircraft and various GSs. We conduct 81 flight measurement tests under realistic AG channel scenar-82 ios. Throughout the flight test, we focus on examining long 83 range path loss behaviors with the sea and ground earth 84 surface reflections. Additionally, for the multipath channel 85 characteristic analysis, we investigate the occurrence prob-86 ability of multipath component, delay, and power relative 87 to the line-of-sight (LOS) component values in hilly and 88 mountainous GSs. We demonstrate the measurement result 89 to illustrate the multipath effects for different AG channel 90 environments. 91 Main contributions of this paper include the followings: 92 1) long range AG path loss analysis for the sea and ground 93 surface reflections, 2) measurement based spherical earth 94 path loss model and log-distance path loss model, 3) multi-95 path component statistics for various aircraft and GS settings, 96 and 4) quantification of delay spread characteristics for long 97 range AG multipath channels. Note that all measurements are 98 conducted by actual flights in the UHF band.

99
The remainder of this paper is organized as follows: In 100 Section II, we address a flight measurement setup to inves-101 tigate the AG channel properties. Section III provides the 102 measured path loss results with the sea and ground reflec-103 tions including empirical losses predicted by models. Under 104 various GS conditions, the observations of multipath effects 105 are presented in Section IV, and conclusions are presented in 106 Section V.

108
In this section, we present the flight measurement setup 109 for test campaigns which observe path loss and multipath 110 channel characteristics. For measurements, we implement the 111 experimental AG channel measurement system as shown in 112 Fig. 1. It consists of a transceiver with UHF antennas, a high 113 power amplifier (HPA), and the channel sounding configu-114 ration including a signal generator and a spectrum analyzer. 115 For the path loss measurement, the transceiver continuously 116   For the channel sounding input, we create a Zadoff-Chu 151 seqeunce, which is well-known as a constant amplitude and 152 zero auto-correlation sequence. This sequence usually pro-153 vides good auto-correlation performance in many fields. The 154 sounding signal is operated at a sampling clock of 14 MHz, 155 which allows for a time resolution of 71.4 ns in the mul-156 tipath measurement 1 . Then the sounding signal is filtered 157 by a square root raised cosine (SRRC) response with the 158 roll-off factor of 1.0. Unlike [11] with the roll-off of 0.3, 159 we adopt a larger roll-off value in order to further reduce 160 oscillations in the time domain. After passing through the 161 target AG channel, the spectrum analyzer in the GS samples 162 the received signal digitally. Finally, the PDPs are produced 163 by the postprocessing that includes the SRRC filtering and 164 the auto-correlation. 165 In the measurement setup, the center frequency is assigned 166 between 400 and 500 MHz depending on spectrum avail-167 ability. The flight speed is set to 270 to 300 km/h during all 168 flights. The flight trajectories have been predefined such that 169 clear radio LOS can be maintained in term of the antenna's 170 field of view. An airframe shadowing may occur when the 171 aircraft body itself obstructs the radio LOS toward the GS. 172 We prevent such circumstances for the measurements by 173 allowing only straight-and-level flights. The detailed environ-174 ments about the GSs and the trajectories of aircraft will be 175 described in Sections III and IV.

177
In this section, we provide the path loss measurements for 178 long range AG communication channels with both sea and 179 ground surface reflections. Then we compare with results 180 predicted by the international telecommunications union 181 (ITU) recommended model and the spherical earth two-ray 182 model. Here, we focus on the path loss analysis according 183 to a distance between the aircraft and the GS, mainly from 184 100 to 200 km. The path loss model can be divided into a site-general model 188 and a site-specific model [6]. First, recommendation ITU-R 189 1 While the delay between two paths is close to the time resolution of the measurement system, two paths are often unresolvable [22]. However, as discussed in [11], it is sufficient to identify most multipath components because the long range AG channel with a high altitude has a relatively sparse multipath environment.
Hence, the remaining distance 233 d 2,k is derived as d 2,k = d k − d 1,k . The effective earth radius 234 R e is defined as R e = αR where α denotes the effective earth 235 radius factor and the earth radius R = 6371 km.

236
In the terrestrial channel with radio waves traveling near 237 the surface, the effective earth radius factor α is normally cho-238 sen as 4/3 to account for the ray bending effect due to changes 239 in the atmospheric refractivity. In our approach, the effective 240 earth radius factor α can be obtained form the average radius 241 of the curvature method proposed in [26]. However, since the 242 overall effect against the altitude changes is negligible, α is 243 set to 1.4 based on a constant surface refractivity of the test 244 area.

245
The aircraft altitude, the GS antenna height and the slant 246 range between the aircraft and the GS are denoted as h A,k , h G 247 and r k , respectively, which are known in advance. The value 248 r k denotes the difference between the direct slant range r k 249 and the length of the reflected ray r 1,k + r 2,k as [27] 250 r k r 1,k + r 2,k − r k Also, the length of the direct path r k is computed as which is determined by the link geometry.

260
Now, the total received field strength at time k is given as 262 where E d,k is the direct wave field strength, the path-gain 263 factor F k = 1 + ρ k D k P k e −j φ k means that how the field at where θ k denotes the phase and r , σ , λ and ψ k represent where P T , G T , G R and L C stand for a transmit power, a trans-303 mit antenna gain, a receive antenna gain, and total cable loss, 304 respectively. Finally, the path loss L k is given as 305 L k = P T + G T + G R − L C − P r,k (7) 306 = L fs,k + L a,k − 10 log 10 |F k | 2 (8) 307 = L fs,k + L a,k 308 − 10 log 10 1 + |ρ k | D k P k e −j( φ k −θ k ) 2 (9) 309 = L fs,k + L a,k − 10 log 10 (1 + |ρ k | D k P k ) 2 whereL k is the log-distance path loss in dB, A represents a 319 constant, and n indicates the path loss exponent [10]. Both n 320 and A can be extracted from the measured data using a least 321 square (LS) curve-fitting technique.

323
In this subsection, we discuss on the received signal strength 324 gathered at a distance between 100 and 200 km via several 325 flights. Specially, we conduct two measurement campaigns in 326 order to predict the path loss behavior on both sea and ground 327 surfaces. To this end, we choose two GS locations such that 328 the radio wave is mainly reflected over the sea or ground 329 earth surfaces, which will be referred to as GS 1 and GS 2, 330 respectively. GS 1 is located at about 1100 meters AMSL in 331 an island area where can cause a sea water reflection. On the 332 other hand, to establish the ground reflection environment, 333 GS 2 is placed on an inland spot with about 900 meters 334 AMSL. Note that both GSs are located in open fields sur-335 rounded by hilly and mountainous terrains as shown in Fig. 1. 336 However, we can expect that the surface reflection effects are 337 more dominant than terrains around the GS in terms of long 338 range LOS path environments.  to the site-general model, it can provide a trend of the path 372 loss based on general information rather than specific path 373 parameters. In this regard, the P.528 model cannot account 374 for the lobing effect because it omits the inclusion of detailed 375 information such as the path-gain factor F k according to the 376 varying geometry of the two-ray path.

377
The lobing structure is highly dependent on surface char-378 acteristics such as roughness as well as electrical constants. 379 Particularly, it is observed in Fig. 7 that the lobing pat-380 tern of the ground reflection case is not perfect compared 381 to the sea reflection case of Fig. 6. Since the ground is a 382 poorly reflecting surface relative to the signal wavelength, 383 the ground reflection may be irregular or even not be present 384 due to blocking from terrain obstructions. Also, since the sea 385 is effectively smoother as the wavelength becomes longer, 386 the two-ray lobing effect is more distinct compared to the 387 ground reflection condition. Moreover, we can recognize that 388 the lobing effect becomes wider as the link distance increases.

389
For the spherical earth two-ray model, the reflection sur-  where a k,i , τ k,i , and N k denote the time-varying ith multipath 430 component's amplitude, delay, and the number of multipath 431 components at the kth time instant, respectively. In the fol-432 lowing, we present field test results for the occurrence prob-433 ability, delay, and power of multipath components in various 434 AG communication environments.

435
By applying the channel sounding system described in 436 Section II, the channel impulse responses are collected in an 437 extra location GS 3 as well as GS 2 over more than one 438 hour flight trial. Figure 8 shows the detailed measurement 439 scenarios and trajectories in GS 3 under the LOS conditions. 440 As stated in the previous section, although the pass loss 441 behavior is similar to that of free space with a strong surface 442 reflection, other multipath components from the sea water 443 surface might be weaker [7], [11]. Hence, we discard GS 1 for 444 this test case. 445 Finally, three typical scenarios are carried out. The GS 446 3 has a lower ground level of about 210 meters and a hilly and 447 mountainous terrain with some small buildings compared to 448 delay spread since it receives stronger multipath signals in the 489 second and third taps as shown in Table 2.

491
In this paper, we have addressed the experimental characteri-492 zation and the modeling of the long range AG communication 493 channel over the sea and ground at the UHF band. The 494 measured path loss has been compared with the predicted 495 losses based on the spherical earth two-ray model and the 496 ITU recommended model. The path loss analysis results have 497 suggested that the measured losses follow a similar trend to 498 that predicted by our two-ray model. In addition, the mul-499 tipath characteristic results have shown that the multipath 500 component exist in the long range AG channel on a fairly 501 high probability regardless of their sparse and intermittent 502 property. The results are important to design the long range 503 AG communication system of the UAV for modern military 504 and civilian airborne network applications.