1 of 7

PROPAGATION CHARACTERISTICS OF GROUND BASED URBAN COMMUNICATIONS

IN THE MILITARY UHF BAND

Jerry R. Hampton, Naim M. Merheb, William L. Lain, Douglas E. Paunil, Robert M. Shuford, Jason A. Abrahamson, and William T. Kasch

Johns Hopkins University Applied Physics Laboratory Laurel, MD

ABSTRACT

A series of experiments were conducted to characterize the propagation characteristics in an urban canyon setting between ground-based communicators operating within the military ultrahigh frequency band (225 to 450 MHz). The experiments were conducted over a two-day period in downtown Philadelphia. Profiles of received power versus distance were generated along a variety of straight and L-shaped paths for different radio frequencies and antenna heights. The results were analyzed to determine the dependency of the received power on street width, RF frequency, antenna height, and corner distance (the distance between the transmitter and the corner on an L-shaped path). The results showed well-defined dependencies on all of the parameters except antenna height, which did not appear to affect the results over the range of values considered in this study. The results are compared with similar measurements by other investigators operating at higher frequencies in an attempt to identify the impact of RF frequency on urban propagation. This study addresses an important but largely overlooked propagation environment.

INTRODUCTION This paper describes the results from an Internal Research and Development effort by the Johns Hopkins University Applied Physics Laboratory to experimentally investigate propagation behavior in an urban canyon environment between ground-based communicators operating in the military ultrahigh frequency (UHF) band, which is defined to be the range of 225 to 450 MHz. The topic of urban propagation has been the focus of much experimental and theoretical research over the years, resulting in a large body of literature that addresses the characteristics of radio wave behavior in the urban environment. Because the

motivation for much of this research has been in support of commercial wireless applications, most of the existing studies are restricted to geometries and RF frequencies that match those likely to be encountered in such applications. In practice this means that most existing urban propagation measurements focus on propagation near 900 and 1800 MHz and assume geometries where one end of the communications path is elevated and the other end is located near ground level to simulate transmission between a cellular base station and a mobile unit. In recent years, however, there has been a growing interest in urban propagation by the U.S. military because of the emerging importance of urban warfare [1]. Unfortunately, because of the geometries and RF frequencies of interest to the military, much of the published literature is not directly applicable to the military urban communications environment. For example, it is critical in urban warfare for soldiers to be able to communicate directly with each other, whether on foot or in vehicles. In addition, communication between ground-based, unmanned sensors is a growing part of warfare in general and of urban warfare in particular. In both cases, the focus is on ground-based communications where users are near ground level (1-3 meters above ground) in contrast to cellular applications where one end of the communications path is an elevated base station. In addition, military tactical users often operate in the military UHF band in contrast to the cellular and PCS bands normally assumed in the literature. The motivation for conducting this study was to address this deficiency by collecting propagation measurements using geometries and frequencies matched to military rather than commercial applications.

2 of 7

DESCRIPTION OF EXPERIMENT Overview In this study, a series of experiments was conducted to measure the path loss between a transmitter and receiver in an urban environment. Testing was conducted during a 2-day period in June 2004 in downtown Philadelphia by fixing a transmitter at a variety of locations in the city and operating a mobile receiver that was driven over various paths within the general vicinity of the transmitter. The received signal, together with a record of the receiver’s travel distance along its collection path, was stored in the vehicle for later processing. The environment in Philadelphia where the data were collected is characterized by a rectangular grid of streets with typical widths equal to about 17 m, surrounded on both sides with tall buildings having heights that vary from about 30 to 60 m - conditions typical of an urban canyon. The transmit signal consisted of an unmodulated carrier at 225, 450, and 900 MHz with an effective isotropic radiated power (EIRP) of 1 W at 225 and 450 MHz and an EIRP of 0.1 W at 900 MHz1. The EIRP values used in this study were driven by a combination of factors, including test equipment constraints and the desire to minimize interference. The transmitter and receiver used identical wideband discone antennas with nominally omni directional gain patterns. The receive antenna was placed on top of a mobile van at a height of 2.3 m. The transmit antenna, which was adjustable to 1, 2, or 3 m, was varied to examine the effect of transmit antenna height on received power level. Antenna heights were chosen to span the range associated with soldier and vehicle-mounted antennas. During each data collection run, the transmitter was fixed and the receiver was driven over a predetermined path where received power was computed as a function of distance along the receiver’s path. Two types of collection paths were employed in this study: line-of-sight (LOS) paths where the transmitter and receiver were always on the same street as each other, and non-line-of-sight

1 Although the focus of this study was the 225- to 450-MHz band, preliminary measurements suggested that propagation behavior was only weakly dependent on frequency over this relatively narrow band. Additional measurements at 900 MHz were therefore added to the test plan to make it easier to identify frequency dependencies in the measured data.

(NLOS) paths where the receiver traveled down the same street as the transmitter at the outset and then turned down a perpendicular street. Because of the rectangular grid pattern of the streets, the NLOS paths were L-shaped and, consequently, are referred to as L-paths in this paper. The distance between the transmitter and the corner on an L-path is called the corner distance, which is denoted by dc. Results shown later indicate that the characteristics of path loss are strongly influenced by the value of dc. Approximately 70 independent data sets were gathered over the two days that this experiment was conducted. Data were collected over a variety of different routes involving different street widths and different corner distances. On selected routes, multiple collection runs were conducted in which the frequency (f) and transmit antenna height (Ht) were varied. The width of the street in the LOS portion of a run is denoted by w1, and the width of the street after the turn on an L-path route is denoted by w2. The baseline set of conditions was defined to be f = 225 MHz, Ht = 2 meters, and w1 = w2 = 17 m (the width of most of the downtown streets). Excursions about the baseline were conducted by varying f, Ht, and by collecting data on routes where the street widths were either narrower or wider than the baseline. Test Hardware A functional block diagram of the test equipment suite used in this study is shown in Figure 1. The transmit side on the left consisted of two sets of equipment, depending on which frequency was used. At 225 and 450 MHz, the signal generator was an Agilent 4421B set to a CW output and fixed at an output power level of -10 dBm. The resulting output signal was fed to a MiniCircuits ZHL-5W-1 power amplifier (PA), which provided nominally 40 dB of gain. At 900 MHz, an Anritsu Wiltron 68177B signal generator was used, set to deliver a CW output waveform, and fixed at its maximum output power level equal to +10 dBm. The output waveforms from both transmitters were fed directly to an Antenna Research Associates SAS-230/a discone wideband antenna with a nominally omni directional gain pattern. An identical antenna was used on the receive side. The suite of equipment for the receiver consists of an antenna, radio frequency (RF) filters used to reduce interference effects, a low noise amplifier (LNA), an

3 of 7

SignalGenerator

+ 40 dB

Mini Circuits

ZHL-5W-1Agilent

4421B

CW

+20 dBm

SignalGenerator

Wiltron

68177B

-10 dBm

CW

225

&

450 MHz

900 MHz

ARA

SAS-230/a

+30 dBm

SignalGenerator

+ 40 dB+ 40 dB

Mini Circuits

ZHL-5W-1Agilent

4421B

CW

+20 dBm

SignalGenerator

Wiltron

68177B

-10 dBm

CW

225

&

450 MHz

900 MHz

ARA

SAS-230/a

+30 dBm

+ 40 dBVoltage

File

Distance

File

Matlab

Post

Processing

Distance

Wheel

DDCADC

Host

Processor

Power

Versus

Distance

RFI

FilteringLNA

pulse count

+ 40 dB+ 40 dBVoltage

File

Distance

File

Matlab

Post

Processing

Distance

Wheel

DDCADC

Host

Processor

Power

Versus

Distance

RFI

FilteringLNA

pulse count

Figure 1. Functional Block Diagram of Test Equipment

analog down converter, an analog-to-digital converter (ADC), a digital down converter (DDC) that translates the received signal from an intermediate frequency down to baseband, and a wheel used to measure distance along the collection route. The output from the DDC consists of voltage samples that are stored in the host computer’s hard drive. A separate set of files is used to store the output from the distance wheel. In the final step shown in Fig. 1, post-processing is performed on the voltage and distance wheel samples using MATLAB scripts to generate sequences of received power versus distance. The received voltage signal was sampled at 250 kilo-samples per second. System Verification A series of initial measurements were made in a flat open field adjacent to JHU/APL’s main campus for the purpose of validating the test equipment and the data processing procedure. The tests consisted of placing a fixed transmitter at one end of the field and driving the receiver van away from the transmitter over an approximately 500 meter path. The transmit and receive antenna heights were fixed at 2 and 2.3 m, respectively; and data were collected at 350 and 450 MHz. Representative results at 350 MHz are shown in Figure 2, which presents the measured path loss plotted in blue as a function of distance from the transmitter. Shown superimposed on the plot is a red curve that corresponds to the two-ray theoretical path loss model. The results show that the two-path model closely matches the data. Similar comparisons were obtained at 450 MHz. The good agreement between theory and the measurements provided confidence in the proposed data collection procedure.

Figure 2 Comparison between theoretical and measured path loss in open environment (f=350 MHz)

EXPERIMENTAL RESULTS This section presents experimental results from the study. The results for LOS geometries are presented first, followed by the results for L-path geometries.

LOS Results A series of runs was made over a variety of different LOS paths. The purpose of these runs was to generate measurements that could be compared against the open space results (see Figure 2), and to use this comparison to assess the impact that buildings have on propagation. Experiments were conducted at 225, 450, and 900 MHz. In each case, the transmit and receive antenna heights were fixed at 2 and 2.3 m, respectively. The baseline street width was equal to 17 m; however, data was also collected on a narrow street with a width of only 6 m and on a wide street with a width equal to 34 m. A representative profile of received power versus path

4 of 7

distance is shown in Figure 3 where f = 450 MHz and w1 = 17 meters. Predictions based on the two-ray model are shown superimposed on the measured data in green. In this plot, the local mean of the data tends to follow the two-ray theory fairly closely. In particular, the measured received power tends to decrease in proportion to the square of the distance at relatively short distances and to decrease in proportion to the fourth power of the distance at longer ranges in agreement with two-ray theory. A comparison of these results with the open space results in Figure 3 shows that buildings primarily have the effect of introducing spatial fading in the received power profile, but the average trend is not significantly affected. Similar results were observed at 225 and 900 MHz and for other street widths. These results show that while the two-ray model does not predict detailed spatial fading effects, it provides a useful tool for estimating the mean received power level even in an urban canyon setting. This finding is consistent with what other investigators have reported for frequencies at 900 MHz and above [2], [5].

101

102

103

-100

-90

-80

-70

-60

-50

-40

-30

-20

-10

0

Distance (m)

Received Power (dBm)

Measurements

Two-Ray Model

Figure 3 Representative urban LOS profile of received power versus path distance

L-Path Results The characteristics of received power versus distance on an L-path are, as expected, significantly different than on a LOS-only path. Among other things, L-path geometry adds an additional degree of freedom to the space of independent parameters, the corner distance. Figures 4-6 are illustrative examples that show received power plotted versus distance over three L-paths with different corner distances. The data for these three plots were collected at 225 MHz with Ht = 2 m and w1 = w2 = 17 m. In each plot,

the receiver van collected data by first driving down a street within LOS of the transmitter and then turned a corner and continued down a street perpendicular to the original one. The beginning portion of each of these routes was identical. The beginning and ending of each corner turn are denoted by a red vertical line (i.e., left line indicates beginning of turn and right line indicates completion of turn), and linear regressions are shown in black for the data before the corner (LOS component) and also after the corner (NLOS component). The first observation is that the falloff rate of the received power (i.e., slope) is greater after the turn than before. This is consistent with observations made by other investigators at 1.8 GHz ( [3], [4] ) and has been explained by assuming that most of the energy arriving at the receiver is due to specular reflections; only those transmitted rays with specific takeoff angles at the transmitter get reflected down the side street to the NLOS receiver [3]. The second observation is that, in general, there is a discontinuity in the received power level between the beginning and end of the turn. This effect has also been observed by other researchers (e.g., [3], [4], and [6]), which is referred to as the corner loss. Thirdly, a careful comparison of these figures shows that the NLOS slope increases as the corner distance gets larger. This effect has also been noted by the same researchers at frequencies higher than those considered in this study. From the preceding observations, it is clear that corner loss, Lcor, and NLOS falloff slope, mNLOS, are important parameters for characterizing the received-power-versus-distance profile on an L-path. It is natural then to ask how these two parameters are affected by other system parameters such as corner distance, street width, antenna height, and frequency. To examine these dependencies, it is first necessary to define the term “corner loss” because its meaning is open to interpretation. For the purpose of this study, corner loss is defined by fitting models to the data in the LOS and NLOS segments of each run and computing the receive power levels where these models intersect the corner boundaries. The point where the LOS fit intersects the leading edge of the corner boundary is denoted by A, and the point where the NLOS fit intersects the trailing corner boundary is denoted by B. The corner loss, Lcor, is defined to be Lcor ! PA – PB; where, PA and PB denote the received power levels at points A and B, respectively. In the NLOS segment, a power law fit (which becomes a linear fit when plotted on a log scale) is used; in the LOS segment, the two-path propagation model is used and scaled in such a way that the mean square error

5 of 7

Startof turn End

of turn

Startof turn End

of turn

Figure 4 Representative L-Path power level

profile (dc = 132 m)

Startof turn

Endof turn

Startof turn

Endof turn

Figure 5 Representative L-Path power level

profile (dc = 265 m)

Startof turn

Endof turn

Startof turn

Endof turn

Figure 6 Representative L-Path power level

profile (dc = 400 m)

between the model and the data is minimized. To discern dependencies between the various system

parameters and the variables, Lcor and mNLOS, it is useful to display the measured values of these parameters graphically. Fig. 7 shows the NLOS slope plotted versus corner distance for all three frequencies. Although there is considerable scatter among the points, the trend is clear: NLOS slope increases as the corner distance gets larger. Furthermore, there appears to be a frequency dependence indicating that for a given corner distance, the slope increases with increasing frequency. Over the range 25 ≤ dc ≤ 400 m, the slope appears to increase relatively linearly with corner distance. Least mean square error linear regressions to the data are shown superimposed on the plot along with the respective equations of fit for each frequency. Fig. 8 shows the dependence of mNLOS on street width, w2. Fig. 8 is similar to Fig. 7 except that the data points are differentiated in terms of street width after the turn instead of frequency. The data in this figure indicate that as w2 decreases, the NLOS falloff rate increases. Unfortunately, because of limited data at w2 = 6 m (diamond points), it is not possible to fit a linear curve through the data at the narrowest width. Nevertheless, the three data points for w2 = 6 m are consistent with the general conclusion that NLOS falloff increases as the width of the street onto which the receiver is turning gets narrower. The dependence of mNLOS on w2 can be seen by examining Fig. 9, which is a plot of mNLOS versus w2 for dc = 106.7 m, the mean value for the diamond points. The mean value for mNLOS at w2 = 6 m is 7.33, and the corresponding values of mNLOS for w2 = 17 and 34 m are obtained by plugging 106.7 into dc in the linear regressions shown in Fig. 7. Fig. 9 suggests that, over the range 6≤ w2≤ 34 meters, mNLOS can be modeled as a linear function of w2. Frequency Dependence of NLOS Slope Versus Corner Distance

y = 0.0333x + 2.7

y = 0.045x + 2.7

y = 0.061x + 2.7

0.00

5.00

10.00

15.00

20.00

25.00

30.00

0.00 50.00 100.00 150.00 200.00 250.00 300.00 350.00 400.00 450.00

Corner Distance (meters)

NL

OS

Slo

pe

225 MHz

450 MHz

900 MHz

Linear (225MHz)Linear (450MHz)Linear (900MHz)

Figure 7 NLOS slope versus corner distance separated

out by frequency (w2 = 17 m, Ht = 2 m)

6 of 7

Street Width Sensitivity of NLOS Slope Verus Corner Distance

y = 0.02x + 2.7

y = 0.0333x + 2.7

0.00

2.00

4.00

6.00

8.00

10.00

12.00

14.00

16.00

18.00

0.00 50.00 100.00 150.00 200.00 250.00 300.00 350.00 400.00 450.00

Corner Distance (meters)

NL

OS

Slo

pe

Width=6

Width=17

Width=34

Linear(Width=34)Linear(Width=17)Linear(Width=6)

Figure 8 NLOS slope versus corner distance separated

out by street width, w2 (f = 225 MHz, Ht = 2 m)

Figure 9 NLOS Slope Versus w2 at dc = 106.7 m

(f = 225 MHz, Ht = 2 m) NLOS Slope Versus Tx Antenna Height

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

9.00

10.00

0.5 1 1.5 2 2.5 3 3.5

Transmit Antenna Height (meters)

NL

OS

Slo

pe

Individual

Average

Figure 10 NLOS Slope Versus Transmit Antenna

Height ( dc ≅ 130 m f = 225 MHz, w2 = 17)

The last figure that addresses the behavior of mNLOS is Fig. 10, which shows the NLOS slope plotted versus transmit antenna height at f=225 MHz, w2 = 17 meters, and dc nominally equal to 130 meters. Results for individual runs are denoted by the diamonds and the average slope at each

antenna height is denoted by the solid squares. Although there is considerable scatter among the data points, the mean values fail to show any discernable transmit antenna height dependency. These results suggest that mNLOS is independent of Ht over the range 1 ≤ Ht ≤ 3 meters. Plots similar to Figs. 7 – 9 were generated for corner loss. Unlike mNLOS, however, which was found to exhibit clear dependencies on the independent parameters, the data from this study failed to show clear relationships between Lcor and either dc, f, or w2. When averaged over all runs, the mean corner loss was found to equal 8 dB with a standard deviation equal to 3 dB. Other investigators ( [3], [4] ) studying propagation at higher frequencies (≥1.8 GHz) have found Lcor to increase with corner distance. Whether the failure of this study to observe a similar trend is attributable to operating at lower frequencies or to differences in measurement techniques, is unclear.

COMPARISON WITH OTHER INVESTIGATORS

Because of the lack of published propagation results for the conditions of interest in this study, it has proven difficult to perform a direct comparison with other investigators. Two papers that come closest are those by Erceg, et al. [3] and Masui, et al. [4], which present measured values for mNLOS and Lcor at frequencies considerably higher than those of interest in this study. Despite these differences, it is instructive to compare their results with the measurements in this study. Table 1 summarizes this comparison. The top set of results in the table is from Erceg’s paper, which was conducted at 1956 MHz, but on wider streets than in this study. Below those results are data from Masui’s research, which was collected at 3350, 8450, and 15750 MHz. Values for mNLOS and Lcor (dB) are listed for two different corner distances (dc = 64 and 429 meters). The bottom section of the table contains results from the present experiment. An examination of the table results in several observations of interest. First, we note that all three sets of results show a clear dependence of mNLOS on corner distance. The rightmost column lists the ratio of mNLOS at 429 meters to mNLOS at 64 meters. Despite the differences in conditions, all three sets of data show similar values for this ratio. A second observation regards the frequency dependence of mNLOS. Although we observed a clear tendency for mNLOS to increase with frequency (see Figure 7), Masui’s results fail to show a similar monotonic trend. The last observation relates to the corner distance. The table shows that Erceg and Masui both measured larger

7 of 7

corner losses than were observed in this investigation. Erceg found a strong dependence on corner distance, but Masui’s results fail to show a similar dependence. In general, there appears to be a tendency for corner loss to become smaller as the frequency is reduced. This trend is particularly clear for the shorter corner distance equal to 64 meters. This may be the result of stronger diffraction at lower frequencies, which would allow more of the transmit power to propagate down perpendicular streets.

Table 1 Comparison of test results with other investigations

dc = 64 m dc = 429 m Frequency

(MHz) mNLOS Lcor mNLOS Lcor

mNLOS Ratio

Erceg: w1 ≅ 20m, w2 = 30m 1956 4.0 11.0 10.6 24.5 2.7

Masui: w1= 11&27m, w2 = 35&44m 3350 4.6 17 12 16 2.6 8450 4.9 22 28 28 5.7 15750 4.1 23 15 22 3.6

APL: w1 = 17m, w2 = 17m 225 4.8 8 17 8 3.5 450 5.6 8 22 8 3.9 900 6.6 8 29 8 4.4

CONCLUSIONS

In this study, a series of measurements was conducted to characterize outdoor urban propagation among ground-based communicators in the military UHF band. Measurements were collected on a variety of routes in downtown Philadelphia over both straight and L-shaped paths. The results show that the mean power on LOS paths is insensitive to street width and matches two-ray theory fairly accurately. L-shaped routes exhibited the same three distinct regimes observed by other investigators in higher frequency bands: an LOS segment, a corner loss, and an NLOS segment where the falloff rate of the received power-versus-distance profile is greater than it is before the corner. The falloff rate after the corner was found to increase linearly with corner distance and frequency, and to decrease linearly with the width of the perpendicular street. In contrast, the corner loss was found not to exhibit clear dependencies on any of the independent parameters, contrary to what other investigators have reported at higher frequencies. It is hypothesized that the lower corner loss in this experiment may be due to there being enhanced diffraction at UHF compared with SHF frequencies where the other measurements were collected. The measurements reported in this paper address an important but largely ignored propagation environment.

REFERENCES [1] Military Operations In Urban Terrain Focus Area

Collaborative Team (MOUT FACT), https://www.moutfact.army.mil/research.asp.

[2] H. Xia, H. Bertoni, L. Maciel, A. Lindsay-Stewart, and R. Rowe, “Radio Propagation Characteristics for Line-of-Sight Microcellular and Personal Communications,” IEEE Transactions on Antennas and Propagation, Vol. 41, No. 10, October 1993.

[3] V. Erceg, S. Ghassemzadeh, M. Taylor, D. Li, and D. Schilling, “Urban/Suburban Out-of-Sight Propagation Modeling,” IEEE Communications Magazine, June 1992.

[4] H. Masui, M. Ishii, K. Sakawa, H. Shimizu, T. Kobayashi, and M. Akaike, “Microwave Path-Loss Characteristics in Urban LOS and NLOS Environments,” Proceedings of the Vehicular Technology Conference, 2001.

[5] A. J. Rustako, et al., “Radio propagation at microwave frequencies for line-of-sight microcellular mobile and personal communications,” IEEE Transactions on Vehicular Technology, Vol. 40, pp. 203–210, February 1991.

[6] A. Tang, J. Sun, and K. Gong, “Mobile Propagation Loss with a Low Base Station Antenna for NLOS Street Microcells in Urban Area,” Proceedings of the Vehicular Technology Conference, 2001.