You are on page 1of 9

IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. EMC-l 9, NO.

3, AUGUST 1977 301

An Empirical Propagation Model (EPM-73)


M. N. LUSTGARTEN, SENIOR MEMBER, IEEE, AND JAMES A. MADISON, MEMBER, IEEE

Abstract-A propagation model is described which has the advan- A. Tenninology


tage of simplicity of manual calculations of basic transmission loss
(Lb), and which provides a degree of accuracy which is similar to that Before proceeding, it should be noted that EPM-73 is a
obtained with other more sophisticated models which compute Lb. combination of two models which account for different types
The model uses a minimum number of parameters and is based on of physical phenomena.
both theoretical and empirical considerations. Also, inverse solutions However, the overall model is a "groundwave" model
may be obtained. That is, given a value of Lb and, for example, antenna
heights and frequency, the appropriate value of distance may be accompanied by troposcatter considerations for long distances.
calculated. High-altitude situations are not considered. In the line-of-sight
The model provides an estimate of mean basic transmission loss. region, ground-wave propagation can often be described by
in dB, with an associated standard deviation. It has been compared ray-theory descriptors, such as direct ray, reflected ray and
with measured values over a frequency range of approximately 20- "Norton Surface Wave" [2]. These terms lose significance
10 000 MHz. Approximately 7 000 paths have been considered in many in the diffraction region where ray theory breaks down and
different areas. Comparison with other more sophisticated models
indicates comparable results, including predictions for sea water propagation must be described by a rigorous diffraction theory.
paths and for frequencies down to 1 MHz (but not substantiated by At greater distances, propagation phenomena involve a com-
measurements between 1 and 20 MHz). bination of diffraction and troposcatter effects [2] .
A general model of the type described in this note, appli-
I. INTRODUCTION cable for relatively large distances beyond some minimum
value, cannot be described solely;as a ray-theory model, or
ELECTROMAGNETIC COMPATIBILITY (EMC) analysts
have a continuing requirement to predict basic transmission
a surface-wave model, or a diffraction model, or a troposcatter
model, for it is all of these combined. Accordingly, it was
loss [1] for thousands of paths. Basic transmission loss is decided not to attach any of the above terms to the composite
defined as the loss between two isotropic antennas for the model. The two models discussed in this paper are sensitive to
same separation and heights as the actual antennas. Antenna the dimensionless parameter h/X, where h is the antenna height
impedance effects are not considered. In many cases, precise and X is the wavelength; h and X are given in the same units.
environmental data (i.e., structural antenna heights and site When this parameter is large (e.g., 25) the high h/X model is
elevations) and detailed terrain data are not available. Further, the correct model to use. Similarly, when h/X is small (e.g.,
in many cases, mobile situations are involved. Complications 0.5) the low h/X model is appropriate. The limits of applica-
involving effects of foliage, man-made obstacles, reflections, bility of the two models and a transition procedure will be
and ducting phenomena should eventually be considered. discussed.
Detailed models of each of these effects can be (and have
been) used for specific situations, but there is also a need for II. ELEMENTS OF THE MODEL
a model which, with minimal input information, can provide
reasonably accurate estimates of expected losses, together Portions of the low-h/X model are similar, in some respects,
with associated uncertainties. EPM-73 is such a model. Antenna to Bullington's formulation [3]. Wave polarization, geophy-
structural heights, frequency, and path-length information sical constants and frequency are considered in calculating the
are required. Foliage, obstacles, etc., are not considered at "effective antenna height." Three distance regions are defmed
this time. by appropriate boundary formulas. Separate loss formulas
This paper will outline the basic elements of the model, are specified for each region.
the engineering formulas, a comparison of predictions obtained The high-h/X model is, essentially, a simplified version of
with the model (and other more complex models) relative -to smooth-earth theory [4]. Wave polarization and geophysical
a large set of measured data, and a brief discussion of apparent constants are considered to a limited extent. Three distance
model limitations. Appendices describe step-by-step proce- regions are specified, with appropriate loss formulas. A transi-
dures for directly calculating losses and inversely calculating tion region, where both models may be applicable, is con-
distances, given loss values. sidered by using both models. The two predicted losses are
compared and the model resulting in the larger loss is used to
provide the best estimate. This conclusion is based- on results
Manuscript received July 28, 1976; revised June 24, 1977. This reported in later sections of this paper.
work was performed under contract F19628-76-C-0017 with the Elec-
tronic Systems Division, Air Force Systems Command. Structural antenna heights are employed, although effec-
M. N. Lustgarten is with IIT Research Institute at the Department tive antenna height above "average" terrain can also be con-
of Defense Electromagnetic Compatibility Analysis Center, North sidered if this information is available. In general, the engi-
Severn, Annapolis, Md, 21402, (301) 267-2554.
J. A. Madison was formerly with IITRI. He is now with Westing- neering formulas were developed from existing theoretical
house, Baltimore, Md. approaches; some simplifications were made when it was
302 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. EMC-19, NO. 3, AUGUST 1977

The following empirical formulas are used for dl:


dFSURI'
I>_ Fre__ e Space Loss
1.lh1h2f
(3)
3.47 X 105
c:

d
d, = 1.lpdLos (hlh2f>A), (4)
;71
0
TROPONPIIERTC
\I I SCAlR.1 where
REFLECTION REGION REGION

DI FRACTI 2.08 X 108dLoS


C- REGION A= (5)
C--)
F_ 103'- 3.75dLos
p =0.6 + 1.08 X 10-8hlh2f. (6)

The maximum permissible value of p is 0.9; the maximum


permissible value of dl is 0.99 dLOS.
Empirical formulas for d2 are as follows:
d2
d2= dLoS-48.3 logf+ 163
DISTANCE (logrifthrtic scale)
Fig. 1. High h/X model.
(40 Sf < 160 MHz) (7)
d2 = dLoS-16.1 logf+ 91.8
determined that losses were relatively independent of certain (f > 160 MHz). (8)
parameters.- A few empirical estimates were formulated after
a limited amount of measured data was reviewed. Finally, For situations involving vertical polarization and sea-water
the model was tested to determine how well it would predict conditions, (9) is also applied and the larger distance value is
basic transmission loss by comparing its estimates of Lb with used
large amounts of measured Lb data.
d2 = -4dLoS - 1.29f + 406. (9)
III. MODEL EQUATIONS
The distance d2 is not calculated for f < 40 MHz. (Values
A. The High-h/X Model calculated using (9) at frequencies above 255 MHz will not be
significant.)
Fig. 1 provides a graphical portrayal of the high-h/X model. It should also be noted, in the case of (8), that the constant
As can be seen, the three regions of interest are the reflection 91.8 may not be appropriate for "very rough" terrain, e.g.,
region, the 'diffracti-on region, and the troposcatter region. the Colorado mountains. A value of about 48 will provide
The term d'FSRF is the distance at which diffraction effects better results, but sufficient, adequate data are not available
become noticeable. The model uses the parameter dj1, which is to resolve this difficult problem, which will be noted again
scaled to' be' 10 percent larger than dFSRF. The parameter below.
d2 is said 'to be the distance at which tropospheric scatter In regard to the loss expressions we define
effects become significant [5]. Note that distance'(d) values
are expressed in kilometers, the transmitter- and receiver Lb LR +LD +LTS- (10)
antenna heights (hl, h2 respectively) are in meters, frequency
(f) is in MHz, and wavelength (X) is in meters. The apparent
The bars indicate mean values. Distributions are assumed
lack of dimensional homogenity in many of the following to be log normal [6], [7] . All parameter units are in dB.
formulas is due to inclusion of (unstated) dimensional factors
in numerical constants contained in the formulas. The fre- Lb : basic transmission loss
quency range of applicability is 40-10 000 MHz. The sug-
LR : reflection region loss
gested maximum antenna height is 3 km. LD : diffraction region loss
The standard' expression for the -radio line-of-sight dis- LTS: tropospheric scatter region loss.
tance (dL OS) is used [4] namely ,

Then if
dLos ±h Ji7i2.
2 (1) I d<dl,LD =LTS =0

As indicated and
d, -
I-ldFSRF- (2) Lb =LR =L FS +5 (I11)
LUSTGARTEN AND MADISON: EMPIRICAL PROPAGATION MODEL 303

f --100 MHz
hi =h2 6m.i
z

R GION ~ ~CO-SITE
~ 40 dB/ DECADE 4 d/ ~~~~~~~HORIZONTAL
ECDEPOLARIZATION
-


0)
0 PLANE EARTH
J
z REGION
0
120
0
0)
z SPHERICAL EARTH
I-
140-
0

z
49
ia
I ISO -

TROPOSPHERIC
SCATTER REGION

ISO

0.01 0.1 l lo 100 I0oo


DISTANCE IN KM.
Fig. 2. Sample prediction of low hlk model.

where d
1 - <4, x=20
LFS: free space loss d2
then d Sd
4<- <8, X=-
d2 d2
LFS = 33 + 20logf+ 20 logd. (12)
d
If - >8, x=40. (17)
d2
d1 < d< d2,LTS =0
B. The Low-hIX Model
and
The low-h/X model is somewhat more complex than the
Lb=LFS+ 5 +LD (13) high-h/X model in that it includes additional parameters. The
applicable frequency range is 1-1000 MHz. The suggested
SO[d-dl] (14) maximum antenna height is 300 m. Fig. 2 provides a sample
[d2-d1] prediction. Note that four regions and three boundary-distance
values are supplied, dCF, dc, and d2. The cosite region will
And if not be discussed here. As indicated, in Fig. 2, the distance
d>d2, LD = 50 dc determines the boundary between the region over which
the earth is assumed to be a planar surface and where it is
and assumed to be spherical; d2 has been discussed previously.
The effective height h' is based on Bullington's formula-
Lb = LFS + 55 +LTS (15) tion [3], namely

LTS = 20 log [d1d2] . (16)


h= Vh2 +h2 (18)
Expressions (15) and (16) were found to be adequate for
distances less than, say, 400 km. Normally, for EMC purposes, where ho is the minimum effective-antenna-height parameter,
additional accuracy is not required for distances of this magni- and h is the structural antenna height, i.e., the height of the
tude. However, if greater accuracy is desired at larger distances, feedpoint above ground.
then the following model can be used: For horizontal polarization, ho is assumed to be zero for
all conditions.
_ d Precise formulas are available to calculate ho for all values
Lb =LFS + 55 xlogd
2 of f, given e (relative dielectric constant) and a (ground
304 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. EMC-19, NO. 3, AUGUST 1977

TABLE I f< 100 MHz, is


DEFINITION OF MINIMUM EFFECTIVE ANTENNA HEIGHT
PARAMETER (ho) FOR VERTICAL POLARIZATION M= 0.5 logf -0.35. (27)
log h0 (ho in meters)
Terrain Type 1 MHz < f < 20Mgz 200iz <f. f 1000 I'hz The formula for vertical polarization over marshy land, 1
1. Sea Water -1.6 log f + 4.69 -2.1 log f + D. 34 MHz f 10 MHz, is
2. Marshy land -2.0 log f + 3.98 -1.6 log f + 3.46
M= 0.25 logf+ 0.06. (28)
3. "average" land -2.0 log f + 3.61 -1.33 log S + 2.74
4. "plains" -2.0 log f + 3.55 -1. 33 log f + 2.68 The formula for all other polarization and land-condition
5. desert -2.0 log f + 3.48 -1.33 log f + 2.61 combinations is
Note: The "plains" designation was used in all comparisons of M=flJ3/7. (29)
predictions and measurements reported.
Mm ax for all conditions is 0.5.
conductivity) [31. However, for convenience, simplified ex- C. The Transition Region
pressions are provided in Table I for calculating ho under five The low-h/X mode, as indicated, is limited in applicability
different types of terrain conditions for vertical polarization. to certain combinations of frequency, distance and antenna
The formulas for the distance values are height. The high-h/X model also is restricted to certain ranges
log'dCF = logf+ 0.75 log hl 'h2' -3.92 (19) of the key parameters. For most situations, one or the other
model will be applicable. However, in some instances, a
dC = 129/P'2 (f IO00 MHz) (20) transition region will arise where one of the models will be
applicable over certain distances. As indicated, in this transi-
dc =59.9/f-1/3 (f> 100 MHz), (21)
tion range, the model which predicts the larger loss will
generally give a more-accurate prediction. Consequently, both
and d2 is defined above; i.e., (7), (8), and (9). The effective
models should be used for this situation.
heights hl' and h2' are for the transmitting and receiving an- Three examples are given, (f = 100 MHz) in Figs. 3-5,
tennas respectively. The model will be applicable only when
dC F < dC and when d > dC F If the calculated loss is less
indicating two situations where it is quite apparent that one
of the models always predicts greater loss than the second and
than LFS + 5, it is taken to be LFS + 5. The loss formulas are
is, therefore, applicable and a third situation where both are
applicable. The figures include plots of loss greater than free
dCF < d < dc space loss (Lb - LFS) versus distance for the "high" model
Lb= I1-l1S log hl'h2' + 40 log d (22) and the "low" model. Since heights may be unequal, the
parameter (hlh2)/(X2) is used.
dC < d <: d2 In Fig. 3, h1h2/X2 = 400 and the high-h/X model applies.
d In Fig. 4 h1h2/X2 = 4 and the low-h/X model applies. Fig. 5
Lb = L(dC) + 20 log - + 0.62M(d - dc) (23)
dc is an example of the transition region. The low model is said
to be applicable for distances less than 48 km; the high model
L(dC) is the mean loss calculated for d = dc using (22) is applicable for distances greater than 48 km.
or (25) and where M is delineated below In general, for a computer-automated version of EPM-73,1
the most straightforward method will be to calculate both
d>d2 high- and low-h/X model predictions and select the larger loss.
LEb =L(d2) + 40 logd2 - (24) This procedure is subject to the restriction that, above 1000
MHz, the low model should not be used. Also, below 40
where L(d2) is the mean loss calculated for d = d2 using (23). MHz the high-model loss prediction should revert to free
An alternate loss formula is supplied for the distance range space plus 5 dB for comparison with the low-model loss. The
dCF Sd.dc,whenhl >h2, namely larger value obtained should be used.
D. Standard Deviation of the Model
Lb = 1 -10 log hl' -20 log h2'+ 40 log d.
O1 (25)
The following formulas are used to estimate the standard
Limited data suggest that (25) may be preferable to (22) deviation of model predictions in dB. These have been derived
for h1 k>h2. from a consideration of the standard deviation values obtained
For the factor M, the following formulas are supplied. The from comparisons of the measured and predicted data (given
formula for vertical polarization over sea water, 1 MHz <f< in Table II and discussed in the next section). Additional
10 MHz, is effort is needed in this area. For example, larger values are
expected over mountainous terrain and in certain transitional
M=Pf/2/20. (26) situations.

The formula for vertical polarization over sea water 10 MHz < I
Programs have been written for several "desk-top" calculators.
LUSTGARTEN AND MADISON: EMPIRICAL PROPAGATION MODEL 305

.0d
,0 r
I
1, 0 III. COMPARISON OF MEASUREMENTS
o

102 AND PREDICTIONS


w
20 ._ _ HORIZONTAL
A large amount of propagation-loss data has been assembled
a
. _ ~~~~~~~~~~~POLARtIZATION at the U. S. Department of Defense Electromagnetic Compati-
:
30 bility Analysis Center (ECAC) and automated so that it may,
among other things, be used to validate various models. The
40 _
in

c
40- - - "HIGH" MODEL validation process involves checking the recorded measured
"LOW" MODEL
loss for each path and subtracting the measured value from
60
the predicted value. A positive difference implies that the
z
10 30 a0 70 90 110 13 model is predicting too much loss. An average of all of the
9 DISTANCE IN KM. differences is taken (m). Ideally, this value would be close to
Fig. 3. Loss predictions for "high" h/X values. zero. The standard deviation from the mean (U,,) also is
2
10^ calculated.
Several models were considered in addition to EPM-73.
20
h2

6m The NX model [8] is a complex theoretical smooth-earth


co model. IPS2 is an ECAC system model which provides results
30
'
as100
tOMHz that are generally similar to those predicted by AIX, involving
N HORIZONTAL
N0 - \
40
POLARIZATION
less computer time. Neither of these models uses terrain data.
The terrain-integrated rough-earth model (TIREM)2 is a
80
more complex ECAC model which includes terrain-data con-
60 siderations. As will be seen, the reported version (which has been
H_IGH" MODEL revised) is not applicable to frequencies below 100 MHz. The
LOW MODEL
70
Longley-Rice model [6], [7] 3 is statistical in nature and does
10 30 50 70 90 I0o 130 not consider terrain features in detail. It uses a factor zAh
DISTANCE IN KM. which is an estimate of terrain roughness, the interdecile
Fig. 4. Loss predictions for "low" h*X values. range of the terrain elevations for the area of interest.
10
All of the noted models are considerably more complex
than EPM-73. The following tables summarize the comparative
15\I results. The "plains" designation was used in all comparisons
10 of EPM-73 predictions and measurements. A terrain rough-
20 ___h2
DO
Z ness factor Ah of 90 m was used as an average value to obtain
co;
V)
25 all of the following results. All smooth-earth model (EPM-73,
0
-j IPS, and Longley-Rice) predictions were made using the
"structural antenna height" option. For the Longley-Rice
w
!I
model, the siting criteria (randomly, carefully, or very-care-
0)
w 35 _N

ui
sr
zI.- fully-located antennas) was generally chosen for each run that
40
-C
x
I.-
_ _\' _

yielded the most accurate predictions (smallest m and aUr).


w
1.-
4
la
45
HG"MODEL A. The High-h/X Region
w

a)
9)
{S0 - LOW" MODEL N Table III summarizes results obtained using the high-h/X
0
SS o sx~~~~ model for a large number of paths reported in [9] involving
-i
30 I
I_ the frequency range 910-9190 MHz. As can be seen, the
IV 3u :>0 TO 90 11 O 130I am values obtained with EPM-73 and the Longley-Rice
DISTANCE IN KM.
models were almost identical.4 The mean values obtained with
Fig. 5. Comparison of predictions in transition region: f = 100 MHz,
hl=h2= 30m. EPM-73 were somewhat smaller than those resulting from the
other models. The am values obtained with TIREM were
smaller than those calculated using the other models. Note
TABLE II
STANDARD DEVIATION (urm) OF COMPARISONS BETWEEN that precise antenna-location data were available for all paths.
EPM-73 PREDICTIONS AND MEASURED VALUES No attempt has been made to determine m and am values
Frequency Rarge (rfUizL) Standard Deviation (dO) which would result if location and site elevation data had not
1 < f < 20 am =6.5
been available.
20 < f c 14rn °m 5 log f B. TheLow-h/XRegion
100 < f < 200 am = l3og f - 16 Table IV presents results for the low-h/X region for vertical
200 <7
f S LO0 0 = 14 and horizontal polarization. Data given in [10] were used to
4lo <.f <600 ma = -20 log f + 66
600 < f < lo0 a =
2 Documentation on IPS and TIREM is being prepared.
10 'The later version was used for predictions reported in this paper.
iooo < f < 10,000 am 8=8log f - 14 rTheseam values are, of course, influenced by the scatter of the
measured data as well as the model error.
306 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. EMC-i9, NO. 3, AUGUST 1977

TABLE III
COMPARISON OF PREDICTIONS AND MEASURED DATA FOR
HIGH VALUES OF h/A
RECEIVER EPM-73 IPS TIRE'HI LONOGLEY-RICE
J
FREQ. -I.IGHT NO. 0? m o m U m Ur m a
(NHz) (m, PATHS (dB) m (d) (dO3) (dB)
910 6 105 1.9 iP.6 3.8 12.7 3.7 7.3 1.- 12.6
910 9 103 4.0 11.9 3 12.1 5.1 i.5 .° 11.6
910 12 105 3.7 11.8 1.3 12.3 4.4 7.3 3.7 11.7-
1666 6 1-15 -2.6 14.6 L,3 15.7 -4. 1 00.9 -3.0 14.
1846 9 129 -2.0 15.5 -3.6 17.0 -5.0 11.7 -2.7 15 . 0
1846b 12 129 -2.6 13.8 -5. 6 1, .6 -3.5 11.1 -3.6 13.5
4595 6 178 -1.4 18.1 -2.0 H. 1 -5.0 13-0 -1.9 17.3
4595 9 206 -3.1 18.3 -6.5 20.7 -6.8 13.3 -3.1 18.1
4595 12 179 -3.1 18.1 -7.8 20.6 -6.6I 12.1 -3.2 16.0
9190 6 213 0.7 20.5 -2.3 25.3 -3.8 16.6 1.7 20.2
9190 9 227 -0.3 20.5 -6.1 25.6 -4.9 13.9 0.3 20.2
9190 12 233 -0.9 18.8 -3.4 23.6 -4.8 13.5 -0.5 18.8

Note: These statistics in all cases are for Colorado Plains measure-
ments taken from [91 with a horizontal polarization; transmitter an-
tenna height of 6 m; careful siting assumed for Longley-Rice results.

TABLE IV
COMPARISON OF PREDICTIONS AND MEASURED DATA FOR
LOW VALUES OF h/X
A=NO7IDA DIST. EP?pH-713 IPS TTIRRI LO,DHLEY-RIREC
FREQ. HEIGH"IS PARESE NO. OF m or a Ur
a
(tRz) (m) POLARIZATION (1m) AREA PAIHSE) (dB(d (dB) _(dB__ am
20 3.3 3 V 1-50 Col. 393 1.7 5.9 -5.1 5. 0 - - -6 .0 6.1
50 4 3 V 1-50 Col. 393 -0.7 8.2 -1.7 8.3 -11.3 15.3 -4.3 S.6
100 3 6 H 1-8G Col. 447 0.8 9.7 .9 Q.' -7.2 12.5 -1.0 10.0
100 3 9 H 1-80 Col. 444 1.5 10.0 5.7 10.1 -6.2 11.8 0.0 10.3
100 4 6 H 10-52 Ohio 475 0.5 8.5 4.5 12.7 -3.4 11.9 -1.7 8.2
Note: These statistics are for measurements taken from [101;
random siting assumed for Longley-Rice results.

test the low h/X model in two geographic areas: Colorado TABLE V
COMPARISON OF PREDICTIONS AND MEASURED DATA FOR
and Ohio. The values of m resulting from EPM-73 predictions INTERMEDIATE VALUES OF h/X AND UNEQUAL
were less than 2 dB in all cases; AUr values were not greater ANTENNA HEIGHTS
than 10 dB. EPM-7 3 IES TIRER LONGLEY-RICI
m U U rnm o m U
(d.) m (dB) m (dB) m (dB) rn
C. The Transition Region
-3.7 8.6 -6.0 8.8 -4.0 21.5 -3.7 8.1
As indicated, the transition region will require the use of
both parts of the EPM-73 model. Several types of measured Note: These statistics in all cases are for Colorado Plains measure-
ments taken from [91 with 1100 samples; horizontal polarization; fre-
data were available in this range involving 1) unequal heights quency = 100 MHz; antenna heights: 69 m (transmitter) and 3, 6, 9 m
at 100 MHz, Colorado plains; 2) approximately equal heights (receiver); distance range: 6.5 to 121 km; random siting assumed for
(6-9 m) at 230 and 410 MHz, Colorado plains; and 3) very Longley-Rice results.
low heights (less than 1 m) in three areas, Wyoming, Idaho,
and Washington. in other portions of the spectrum. The data used were derived
from [I],; more data in this range would be of interest. As
D. Transition Region: Unequal Heights indicated in Table VI, all of the models predicted too much
1100 data samples were available for a situation (in Colo- loss. The TIREM model values of arn were relatively small,
rado), see [9], involving unequal heights in the transition region, indicating that terrain considerations are significant in this
f _ 100 MHz. As noted in Table V, the mean and an values band.
obtained from all of the models were comparable, except for
somewhat larger Un values using TIREM. Results (not shown) F. Transition Region: Very Low Antenna Heights (230 MHz)
using the modified low-h/X model (25) yielded a smaller Table VII summarizes several groups of data taken from
mean value. As indicated, additional data of this type would [11] involving antenna heights as low as 1 m. The approxi-
be desirable to determine whether the modified model con- mate frequency was 230 MHz. The three geographic regions
sistently provides better results. involved were Wyoming, Idaho, and Washington. For the
Wyoming and Idaho data, EPM-73 exhibited mean differences
E. Transition Region (200400 MHz) of less than 4 dB. The differences exhibited by the other
Prediction of losses in the 200-400 MHz band appears to models were larger. The Urn values for EPM-73 were, in general,
represent a more difficult problem than those encountered somewhat less than those produced by the other models.
LUSTGARTEN AND MADISON: EMPIRICAL PROPAGATION MODEL 307

TABLE VI
COMPARISON OF PREDICTIONS AND MEASURED DATA FOR
TRANSITION REGION
APPROX. EPM-73 IPS TIREI LONGLEY-RICE
FREQ. NO. OF m am am m a m ¢
(Mz) PATO (d) m (dB) (dB) m (dB) n
230C 117 9.1 12;6 11.9 13.8 7.9 7.2 11.7 12.3

410 121 4.8 15.6 7.2 16.3 5.4 11.5 9.1 15.6

Note: These statistics in all cases are for Colorado Plains measure-
ments taken from [11] with horizontal polarization; antenna heights;
6 m (transmitter) and 9 m (receiver); distance range: 0.5 to 119 kmi;
random siting assumed for Longley-Rice results.

TABLE VII
COMPARISON OF PREDICTIONS AND MEASURED DATA FOR
VERY LOW ANTENNA HEIGHTS
DIST. EPM-73 IFs TIREM LQIHGLEY-RICE
N4O.OF RANCGE m o m a m o m u
AREA PATHS (km) (dB) in (dB) m (dE) (dB)
Washington '488 1-6C 12.4 13.9 20.2 14.6 3.8 15.8 9.2, :14.1
Idaho 232 10-34 -3.6 9.8 4.6 10.0 -10.8 16.3 -7.1 10.3

1N1yorln; 348 3-44 -2.9 11.2 4.0 11.7 -7.5 15.4 -6.1 11.4

Note: These statistics are for measurements taken from [ 11] with
vertical polarization; frequency = 230 MHz; antenna height combina-
tions of 0.6, 1, 2, and 3- m; random siting assumed for Longley--Rice
results.

For the Washington data, the am values for all models were TABLE VIII
COMPARISON OF PREDICTIONS AND MEASURED DATA FOR
quite similar. The mean difference produced by EPM-73 PATHS OVER SEVERAL GEOGRAPHIC AREAS
was about 12 dB; the difference exhibited by TIREM was
considerably smaller. EPM-73
m(dB) am(d.B)
IPS
m(dB) dm(dB)
LONIGLEY-RICE
m(dB) sm(dB)
It is of interest to note that much of the Washington data 8.0 19.3 8.1 18.3
9.1 17.1
involved paths over very rugged terrain. Of more significance,
perhaps, is the fact that the sites were "very carefully selected," Freauency Range: approximately 40-10,000 MHz
Distance Range: approximately 10-965 km.
according to [11] . This implies relatively high effective antenna Assumed siting: random (for Longley-Rice predictions)
heights and a reasonably large number of line-of-sight paths,
mixed with a number of obstructed paths. The fact that TABLE IX
EPM-73 tended to predict too much loss is probably due to COMPARISON OF PREDICTIONS AND MEASURED DATA FOR
the larger preponderance of line-of-sight paths. If a value of VHFAND UHF PATHS
about' 4.5 m had been added to the structural heights, the EPM-73 IPS LONGLEY-RICE
mean difference would have been close to zero. Similarly, m(dB) om(dB) m(dB) dm(dB) m(dB) sm(dB)
use of the more appropriate siting criteria would have im- 5.3 10.5 3.2 9.6 4.0 9.2

proved the Longley-Rice predictions for the Washington Frequency Range: approxijrately 45-1050 MHz
data. Distance Raname: approximately 30-640 Lon
Assumed Siting: random (for Longley-Rice predictions)

G. World-Wide Tropospheric Data


Reference [12] provides a large amount of data relative to sequently, it was decided to use (17), although (15) and (16)
paths in many portions of the world in addition to the U.S. will essentially provide the same results for distances less than
e.g., Great Britain, Ireland, Germany, Japan, and Australia. (approximately) 400 km.
Table VIII summarizes the results using the 593 path-loss
measurements available. H. FCC Data
The line-of-sight paths included in the Table VIII results Reference [13] presents VHF and UHF data for which
were examined separately. Mean and am values of approxi- the transmitter antenna heights recorded are "effective"
mately 30 dB and 25 dB respectively were indicated by all antenna heights, with Pike's Peak among the site locations
three models. Use of the "effective antenna height" option included. Table IX summarizes the results using the 188
for all three models would improve these results, provided path-loss measurements supplied.
that terrain information were available.
In order to test the tropospheric-scatter-model (17), 346 IV. DISCUSSION
paths were selected from [12] which were clearly examples
of this phenomena. Using (15) and (16), m = 2.3 dB, arm = Evaluation of results indicate that EPM-73 may result in
10.7 dBm. Using (17), m = 0.03 dB, am = 8.6 dB. Con- gross errors in three types of situations: 1) where line-of-
308 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. EMC-19, NO. 3, AUGUST 1977

sight or near-line-of-sight conditions prevail at relatively large Calculate


distances, i.e., prediction of too much loss (these may, in 1) ho (Table I), hl' and h2' (18), dCF (19), dc (20, 21),
general, be accounted for by considering site elevation above M(26, 27, 28, 29),LFS (12).
surrounding terrain; 2) situations where antennas are "very Notes: 1) The model will not be applicable if d0F > dc,
close" to obstacles; i.e., predictions of too little loss (addi- or if d < dCF. 2) If the loss obtained is less than LFS + 5,
tional effort is required in this area-even terrain-dependent the loss will be LFS + 5.
models appear to have difficulties for this type of situation); 2) If d F d < dc: use (22) for loss.
and 3) very rough terrain, such as the Colorado mountains 3) If dc <d d2:
(all known models will result in very large uncertainties, a) Use (1) for dLos-
e.g., am > 20 dB). b) Use (7), (8), and (9) for d2. For vertical polarization
As indicated, certain features of EPM-73 can be extended over sea water:
and improved, given additional measured data. However, for 1) if 40 MHz S f < 160 MHz: use the maximum
the frequency range 20-100 MHz, present EPM-73 predictions value of (7) and (9); 2) if 160 MHz < f- 255 MHz:
are as good as those produced by any known model. At higher use the maximum value of (8) and (9); 3) if f> 255
frequencies, terrain-dependent models, such as TIREM, will MHz: use (8).
usually result in smaller values of amrn The frequency range For all other situations and f > 40 MHz, use (7) and (8)
of 200-400 MHz evidently represents a particularly difficult for d2 as indicated.
prediction problem for all models, as does the "very rough" c) If f 40 MHz: use (23) for loss.
terrain situation. 4) If d > dc, f < 40 MHz: use (23) for loss.
When precise antenna-location data are not available and, 5) If d>d2, f> 40 MHz: use (24) for loss.
similarly, if detailed terrain data are not available, there is
little justification for using a more complex model. Further, B. High-h/X Model
even if detailed data are available, use of a model such as Limitations: 40 MHz < f < 10 GHz; maximum antenna
EPM-73 (in conjunction with an appropriate safety factor, height 3000 m.
e.g., 1.3a subtracted from the calculated value, or approxi- Input parameters: f, hl, h2, d, polarization (V or H) and
mately 90 percent confidence) will result in considerable
type terrain (both only if vertically polarized over sea water
savings in computer time. If a few paths are in doubt, use of a and 40 MHz <f< 225 MHz).
more complex model is justifiable for them.
Calculate:
Also, the format of EPM-73 permits construction of inverse
solutions. That is, given a specified loss value and, for example, l)LFS (12), dLos (1),A (5)
2) If h1hJf AA: use (3) for d1 .
antenna heights and frequency, the analyst can calculate the
3) If h1hJf >A: use (6) p and (4) for d1
applicable distance value [see Appendix II]. Frequency/ 4) If d < d1, use (11) for loss.
distance-separation curves which specify acceptable system 5) Calculate d2. See instruction for low-h/X model: a)
EMC performance can also be calculated, given appropriate if d, < d 6 d2: use (13) and (14) for loss; b) if d > d2: use
constraints.5
Briefly, the use of complex propagation-loss models has (17) for loss.
its place, but a model such as EPM-73 can also be useful C. Transition Region
for many practical situations. Results obtained to date in-
dicate that, if appropriate precautions are taken, the model In numerous situations, it will be apparent that only one
has wide applicability. of the two models need be considered. For example, if f >
1000 MHz, only the high model is applicable. Similarly, if
ACKNOWLEDGMENT f < 40 MHz, only the low model is applicable, with the restric-
The work contained herein was performed under Contract tion that a loss less than LFS + 5 is not permitted. Antenna-
No. F19628-76-C-0017 with the Electronic Systems Division, height restrictions are noted. In general, if h/X is less than
Air Force Systems Command. 0.5 andf< 1 GHz, the low model will be applicable. Similarly,
if h/X > 25, the high model will generally be applicable in the
APPENDIX I. range 40 MHz-10 GHz.
In the transition region, both models should be used and
STEP-BY-STEP PROCEDURE FOR CALCULATING the larger loss taken as the best estimate.
BASIC TRANSMISSION LOSS (Lb)
A. Low-h/X Model APPENDIX II
Limitations: frequency range 1-1000 MHz, maximum
antenna height 300 m. Input parameters: f, hl, h2, d, polari- METHOD FOR CALCULATION OF
zation (V or H), type terrain (see Table I). INVERSE SOLUTIONS
This appendix will outline procedures for using EPM-73
IA report on this subject is being submitted by U.S. CCIR Study if a loss value (L) is specified and an applicable distance is
Group 1 in partial answer to CCIR Question 44/1: ."A method for cal-
culating adjacent band interference (frequency-dependent rejection to be calculated. (The bars above the L-parameter values are
and frequency-distance)." omitted for convenience.)
LUSTGARTEN AND MADISON: EMPIRICAL PROPAGATION MODEL 309

A. The Lov-h/X Model As in the case of (A2-3), which may be put in the form of
Limitations: h1, h2 <300 m; 1 MHz <f< 1 GHz (A2-4), d may be solved for graphically.
Input parameters: f, hl, h2, polarization (V or H), type and c) if L >L(d2)
terrain (see Table I), and L.
Calculate d = antilog [(L -88 + 20 log d2)/20]. (A2-8)
1) ho (Table I), h,1 and h2' (18), dCF (19), dc (20,
21), d2 (7)49) and M (26)-(29) A lengthy procedure for using the tropospheric-scatter-region
formulation (17), applicable to distances greater than 400
2) L(dCF), which is identical to LFS (12) when d = kmi, will not be given here.
dcF'
3) L(dc) from (22), setting d = dc; and L(d2) from C. Transition-Region Considerations
(23), setting d = d2.
4) a) If L < L(dCF) + 5, the model is not applicable; As indicated in Appendix I, both models should be con-
b)ifL =L(dCF)+ 5 sidered for situations where f is between 40 and 1000 MHz,
antenna heights less than 300 m and h/X values are between
d antilog [(L - 38 - 20 log f)/20] (A2-1) 0.5 and 25. For these situations, inverse solutions for both
c) if L(dCF) + 5 < L < L(dc): models should be calculated; the smaller distance will be the
appropriate solution.
d = antilog [(L - 11 1 + 15 log h1 '2 ')1401 (A2-2)
REFERENCES
and d) if L(dc) < L <L(d2), 1] CCIR, Recommendation 341, Int. Radio Consultative Com-
mittee, Int. Telecommun. Union, vol. III, pp. 30-31, Geneva,
20 log d + 0.62 Md =L -L(dc) + 20 log dc + 0.62 Mdc Switzerland, 1963.
[2] K. A. Norton, "The calculation of ground-wave field intensities
(A2-3) over a finitely-conducting spherical earth," Proc. IRE, Dec.
1941.
[3] K. Bullington, "Radio propagation above 30 megacycles,"PProc.
Note: The expression (A2-3) is of the form: IRE, Oct. 1947.
[41 H. R. Reed and C. M. Russell, Ultra-High Frequency Propaga-
A log d + d B (A2-4) tion. New York: Wiley, 1953.
[5] J. A. Madison, "Refinement of Yeh's method as used to predict
tropospheric scatter propagation loss values," IEEE Int. Electro-
where A and B are constants, M is a positive number, and the magn. Compat. Symp. Rec., July 1971.
expression may be used to solve for d graphically. [61 A. G. Longley and P. L. Rice, "Prediction of tropospheric trans-
mission loss over irregular terrain, a computer method-1968,"
5) If L >L(d2) ESSA Tech. Report ERL 79-ITS 67, U.S. Dept. of Commerce,
Boulder, CO, July 1968.
d = antilog [(L L(d2) + 40 log d2)/40]. (A2-5) [7] A. G. Longley, G. H. Hufford, R. Reasoner, and J. Montgomery,
"A statistical propagation model for the improved interference
-

prediction model," OT-TM 67, U.S. Dept. of Commerce, Office


B. The High-h/X Model of Telecommunications, Boulder, CO, Dec. 1971.
[81 M. Maiuzzo and W. Frazier, "A theoretical ground wave propa-
Limitations: h1, h2, < 3000 m; 40 MHz <f < 10 GHz. gation model-NX Model," ECAC Rep., ESD-TR-68-315, Electro-
Input parameters: f, hl, h2, polarization (V or H); and magn. Compat. Analysis Center, Annapolis, MD, Dec. 1968.
type terrain. (Note that polarization and type terrain are of [91 P. L. McQuate, J. M. Harman, and A. P. Barsis, "Tabulations of
propagation data over irregular terrain in the 230-9200 MHz
concern only if sea water and vertical polarization are involved frequency range, Part I: Gunbarrel Hill receiver site," ESSA
at a frequency between 40 and 255 MHz.) Tech. Rep. ERL 65-ITS 58, U.S. Dept. of Commerce, Boulder,
Calculate CO, Dec. 1968.
[10] M. E. Johnson, et al., "Tabulations of VHF propagation data
1) d1(3)-(6) and d2 (7)49). obtained over irregular terrain at 20, 50, and 100 MHz, Part
2) L(d1), using (1 2), setting d = dl. III: Ohio data," ESSA Tech. Rep. IER 38-ITSA-38-3, U.S.
3) L(d2), using LFS(d2) + 55. LFS is given by (12). Dept. of Commerce, Boulder, CO, Dec. 1967.
[11] A. G. Longley and R. K. Reasoner, "Comparison of propaga-
4) a) If L <L(dj) tion measurements with predicted values in the 20 to 10 000
MHz Range," ESSA Tech, Rep. ERL 148-ITS 97, Dept. of
d = antilog [(L -
38 -
20 log f)/20] Commerce, Boulder, CO, Jan. 1970.
[12] A. G. Longley, R. K. Reasoner, and V. L. Fuller, "Measured
and predicted long-term distributions of tropospheric trans-
b) if L(d1) < L < L(d2) mission loss,' OT/TRER 16, Dept. of Commerce, Boulder,
CO, July 1971.
[13] J. Damelin, et a!., "Development of VHF and UHF propagation
20 log d + =L-38-20logf+ Od-d curves for TV and FM broadcasting," FCC Rep. No. R. 6602,
d2- d2- Federal Communications Commission, Washington, D.C., Sept.
1966. (Available from NTIS, Springfield, VA, as PB 174 288.)

You might also like