IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 31, NO.

1, FEBRUARY 1988

Coverage Prediction for Mobile Radio Systems Operating in the 800/900 MHz Frequency Range
IEEE VEHICULAR TECHNOLOGY SOCIETY COMMITTEE ON RADIO PROPAGATION

Preface
Samuel R. McConoughey, Board member and former President of the IEEE Vehicular Technology Society, recognized the need for industry-accepted propagation models for predicting reliable service areas and co-channel interference in the 806-947 MHz segment of the radio spectrum. To address the growing need for these models in private and common carrier services with conventional, trunked, cellular and radio paging systems, in 1982 Mr. McConoughey presented a resolution to the Board of Directors of the Vehicular Technology Society. This resolution proposed the formation of an Ad Hoc Committee to study existing radio frequency propagation models and to make appropriate recommendations. Neal H. Shepherd served as Chairman of the committee since early 1983. The Secretary for the first eight meetings was Floyd Shipley, followed by Robert P. Eckert for the remaining 28 meetings. Special credit is due Henry Bertoni, who assumed the enormous task of editing this publication to provide continuity. Members of the committee submitted documents in accordance with an outline, which were then reviewed before acceptance by the Ad Hoc Committee. Following is a list of contributors.

Contributors
Nadia S. Adawi

Affiliation
Vision Systems, Inc. 850 N. Burlington Street Arlington, VA 22203 Center for Advanced Technology in Telecommunications Polytechnic University 333 Jay Street Brooklyn, NY 11201 Computer Sciences Corp. 3328 Glenmore Drive Falls Church, VA 22041 Federal Communications Commission 1919 M Street, N.W. Washington, DC 20554 Dettra Communications Inc. 2021 K Street, N.W., Suite 309 Washington, DC 20006 Federal Communications Commission 1919 M Street, N.W. Washington, DC 20554 Consultant 13634 Braemar Circle Dallas, TX 75234 Communications Technology Assoc., Inc. P.O. Box 4579 Lynchburg, VA 24502

Henry L. Bertoni

Joseph R. Child

William A. Daniel+

John E. Dettra

Robert P. Eckertt

Earl H. Flath, Jr.

Robert T. Forrest

' The views expressed here are those of the authors and do not necessarily reflect the views of the Federal Communications Commission

4 William C. Y. Lee

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 31, NO. 1, FEBRUARY 1988

Samuel R. McConoughey

John P. Murray

Herbert Sachs

George L. Schrenk

Neal H. Shepherd

Floyd D. Shipley

PacTel Mobile Companies 2355 Main Street, P.O. Box 19707 Irvine, CA 927 14 Federal Communications Commission 1919 M Street, N.W. Washington, DC 20554 John Murray Associates 1823 Folsom Street Boulder, CO 80302 SachdFreeman Associates, Inc. 14300 Gallant Fox Lane Bowie, MD 20715 Comp Comm, Inc. Station House, Suite 412, 900 Haddon Ave. Collingswood, NJ 08108 Consultant 1914 McGuffey Lane Lynchburg, VA 24503 Consultant 1201 Ridgewood Terrace Arlington, TX 76012

The Committees major objective was to recommend modeling approaches which would be suitable for predicting service areas for mobile systems operating in the 806-947 MHz band. Although each model studied had salient features, which met some of the needs, none were accurate enough to warrant a recommendation. The following list provides deficiencies noted in various models: none of the models studied provided an adjustment for reflections from buildings or hills; none of the models studied made an allowance for transmission loss due to foliage; the Longley-Rice model underestimated the transmission loss for most typical mobile situations; the Okumura urban model generally overestimated the transmission loss for most United States cities; both TIREM and Longley-Rice models provided serious discontinuities in the transmission loss values when changing from one mode of propagation to another; none of the models studied provided data on sector transmission loss distributions due to multipath reception, except to assume a loss deviation of 8.2 dB. Definitions have been included in Section I1 to provide a

basis for comparing propagation models and field measurements. Other members of the Committee attending one or more of the meetings were: Tom Aitchison, Virgil R. Arens, Robert Bultitude, Rick Burke, Darnyl DeLawder, George Dewire, William E. Frazier, Ernest Freeman, Simon Goldman, Vick Graziano, George Hagn, Ed Hanley, David Hodgin, Keith W. Kaczmarek, William H. Keller, Ken Kelly, William K. Kokorelis, David Land, John B. Lomax, R. Singh Lunayach, Don Mazak, James Mikulaski, John A. Moffet, Joseph Moffitt, Phil Rice, J. T. Roussos, Fred Schaefer, L. G. Schimpf, Victor Tawil, Mark Whitty , William Wickline and Don Yost. The Propagation Committee has been made a permanent technical committee of the Vehicular Technology Society, whose purpose will be to provide a propagation model for predicting reliable service areas and co-channel interference for all types of mobile radio systems; prepare standard methods of measurement of radio frequency propagation in land mobile bands.

Neal H. Shepherd

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

I. Introduction
URING THE PAST DECADE, broad applications of land mobile services have been authorized by regulatory agencies in the 800 and 900 MHz portions of the radio spectrum. These decisions opened the door for trunked conventional land mobile; for cellular radio telephone; for 800 MHz conventional and nationwide paging; and for other applications. It is generally recognized that there is no clear consensus of methods to evaluate coverage of mobile systems in these bands. There are many reasons for this, but perhaps the most significant factor is the limited experience with application of radio propagation to land mobile systems in the 800/900 MHz portion of the spectrum. This constraint applies to the entire land mobile community; there are no procedures within the mobile engineering community that are as yet generally acknowledged or accepted as providing accurate and reliable 800/900 MHz propagation information. This constraint on engineering analysis, particularly in the frequency range of 800/900 MHz, has been identified by the Institute of Electrical and Electronics Engineers as an area requiring immediate attention. As a result, early in 1983 the IEEE Vehicular Technology Society organized an Ad Hoc Committee on Radio Propagation Models for land mobile applications in that band. The committee was supported by a broad group of technical experts from government, industry, and universities whose common goal was the improvement of UHF propagation loss calculations. This is the Final Report of the committee, which became a standing committee in 1986. The Committees task was to recommend radio propagation models for land mobile radio services operating in the 800/900 MHz radio spectrum. These propagation models were to provide the industry and governments with statistical methods
Manuscript received February 4, 1988.

of predicting reliable service areas and interference. The radio services to be addressed included conventional and trunked private land mobile, common carrier cellular land mobile, and private and common carrier radio paging systems. Common carrier air-ground and satellite mobile systems were to be addressed at a later time. Based on the above task statement, the objectives of the Committee on Radio Propagation Models were as follows: 1) to identify current capabilities to estimate propagation in the 800/900 MHz band; 2) to define propagation modeling requirements for 806/ 947 MHz land mobile applications; 3) to recommend propagation modeling approaches to meet the defined requirements. These objectives will be explored in depth in this report, through a comprehensive evaluation of todays needs and capabilities in 800/900 MHz propagation loss determinations. More specifically, the report will:

IEEE Log Number 8821063.

1) identify the technical factors that affect 800/900 MHz propagation; 2) summarize current knowledge in making 800/900 MHz propagation estimates; 3) explore the relationship among the propagation path, the environment, and the equipment involved; 4) address such issues as prediction accuracy, automated versus manual calculations, the relationship of reliability criteria to system compliance testing; 5) indicate current deficiencies in prediction methods and the collection of propagation data, and outline a program that addresses these limitations; 6 ) recommend modeling approaches that are most appropriate to employ when evaluating particular land mobile systems.

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37. NO. 1, FEBRUARY 1988

11. Definitions
To assist the reader, the following list of definitions is included. The reader is encouraged to proceed to subsequent material, and return to these definitions as needed rather than mastering the definitions now. Analytical: Using, subject to, or capable of being subjected to a methodology involving algebra and calculus. Area propagation model: A propagation model in which median transmission loss calculations are based on generalized characteristics of the areas surrounding the transmitter and receiver, as well as the intervening area. Features such as local environment, terrain roughness, building density, etc. are used to modify a median transmission loss equation, in order to adapt it to the service area. Base station noise: A composite of individual noise sources assumed to be uniformly distributed around the base station antenna. Bit error rate (BER): Ratio of the number of bits of a message incorrectly received to the number of bits transmitted [I]. Calling probability: 1) The calling probability is the ratio of expected successful selective calls relative to the total number of trials [2]; 2) the standard calling probability is defined as an 80 percent probability of successful calling [3]. Cornu spiral: A plot in the complex plane of the Fresnel integral. The Fresnel zone clearance value U is the running parameter along the curve. Coverage area: A collection of sectors for which there is a stated probability of receiving a signal exceeding a given level. It may sometimes be convenient to describe the coverage area in terms of a bounding contour within which lower signal level subareas (holes) are identified. Note: the stated probability does not always guarantee a grade of service. See Service Area. Coverage probability: The probability, applying to a designated collection of sectors, that a specified signal level will be available for reception during (at least) a specific percent of the time. For example, F(50, 90) means that in 50 percent of the sectors designated, the specified signal level will be exceeded 90 percent of the time. For most systems, time variability is less significant than location variability and loss deviation, and may be neglected by assuming 100 percent for the time fraction. Ducting: Confinement of electromagnetic wave propagation to a restricted atmospheric layer by steep gradients in the index of refraction with altitude [I]. Effective antenna elevation: The height of the radiation center of the base station antenna above the average elevation of the ground between distances of 3 and 15 km from the base station in the direction of the mobile. This is equivalent to height above average terrain, as defined by the Federal Communications Commission. The mobile antenna height is defined as the height of its radiation center above ground level (adapted from [4]). Error burst: A group of bits in which two successive

Fig. 1.

Fresnel zone about the line connecting two antennas.

erroneous bits are always separated by less than a given number ( X )of correct bits. The last erroneous bit in the burst and the first erroneous bit in the following burst are accordingly separated by ( X ) correct bits or more. Number ( X ) should be specified when describing an error burst [I]. Fresnel ellipse: The locus of points formed by the intersection of the surface of the first Fresnel zone surrounding the line connecting transmitting and receiving antennas with: 1) any plane containing the two antennas; or 2 ) a plane such as the surface of the earth, that intersects the first Fresnel zone. Fresnel integral: An integral function used for numerical calculations of diffraction results. Fresnel radius: The radius r of the circle formed by the intersection of the first Fresnel zone with a plane perpendicular to the line between transmitting and receiving antennas. Referring to Fig. 1 , for transmitter and receiver at distances dl and d2 from the plane, and assuming that they are large compared to wavelength, the value of r in meters is given by

r=548

4 -

f( d ,+ d2)

dl d2

dl, d2 in km and f in MHz.

The radius of the nth Fresnel zone is found by multiplying r by &. If d2 is about 0.1 d,, or less, and iff = 900, then r 2: 18.3 Fresnel zones: 1) The region in space between successive ellipsoids of revolution whose foci are the transmitting and receiving antennas. The surfaces of the ellipsoids are defined by the condition that the combined distances from any point on the surface to the receiving antenna and to the transmitting antenna be greater than the direct distance between antennas by an integer multiple of one-half wavelength. The first Fresnel zone is the region within the innermost ellipsoid. 2) The area formed by the intersection of the regions of definition 1 with any plane perpendicular to the straight line between the transmitting and receiving antennas. The first such zone is a circular disk, and higher zones are circular rings. Fresnel zone clearance ratio: The effect of diffraction is evaluated in terms of the Fresnel zone clearance. The Fresnel zone clearance ratio is the ratio of the distance from the direct line between antennas to the path obstacle (distance h in Fig.

a.

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

7 Rayleigh density function: A probability density function, describing the behavior of some variable, given by [5]

1 ) and the Fresnel radius r. It is expressed in terms of the parameter U used with the Cornu spiral, where U = &!h/r. Height above average terrain (height AAT): See Effective antenna elevation. Location variability: The standard deviation of a sector transmission loss distribution. The sector transmission loss distribution is typically found to be log normal. Log normal distribution function: The integral of the log normal probability density function. Log normal probability density function: A probability density function describing some random variable, given by [11

f(x a)=-

X
U2

exp

[ (-) ] .
1 X 2
--

f(X)=--Xu&

exp [-(In X ) 2 / ( 2 a 2 ) ]

Loss deviation (LD): The decibel difference between the 50 and 90 percent values of the transmission loss distribution. When the transmission loss distribution is described by the Weibull variation, loss deviation can be found from any two points n = 1 , 2 on the distribution using the formula [2]:
LD = where

1.884(X, -X2) In [In ( l / P l ) ] -In [In ( l / P 2 ) ]

P,, percent1100 X,, transmission loss at P,,. Median transmission loss: The median value of the sector transmission loss distribution for a collection of sectors located in similar environments and at equal distances from the base station. Message acceptance rate (MAR): The number of correctly received messages of specified length, divided by the total number of messages sent through a fully loaded system. Message error rate: 1-(message acceptance rate). Open environment: See Rural environment. Point-to-point propagation model: A propagation model in which median transmission loss calculations are based on specific characteristics of the path along the great circle between the transmitter and the receiver. Point-to-random multipoint: A mode of operation in which one fixed station is in a defined location, and the other fixed stations are randomly located. This mode combines the effects of mobile variability, where the dominant variations are primarily affected by location, and point-to-point variability, where the dominant variations are primarily affected by time. Propagation loss: The total reduction in radiant power density. The propagation loss for any path traversed by a point on a wave front is the sum of the spreading loss and the attenuation loss for that path [ l ] . Propagation model: An empirical or mathematical expression used to compute propagation path loss [ 13. Radio propagation path: For a radio wave propagating from one point to another, the great-circle route between the transmitter and receiver antenna sites [ 11.

Rayleigh fading: Radio fading due to multipath which follows the Rayleigh probability curve stating the natural distribution of random variables [6]. Rural environment: A radio environment comprised of an area where there are few obstacles like tall trees or buildings in the propagation path, and with cleared areas approaching 300 to 400 m across (as for instance, farm land, open fields, etc.) [71, PI. Sector: A region over which the variations of transmission loss can be used to construct a transmission loss distribution. A sector can have any shape; typically the minimum and maximum dimensions are 4 and 20 wavelengths, respectively 121. Sector transmission loss (STL): The transmission loss corresponding to the 50 percent location or median value of the distribution of transmission loss within a sector. When the transmission loss distribution for a sector is described by the Weibull variation, sector transmission loss can be found from any two points n = 1 , 2 on the distribution using the formula 121:
STL =XI + ( X i - X2) where

0.367 +In [In ( l / P , ) ] In [In ( l / P , ) ] -In [In (l/PZ)]

P,, percent1100 X,
transmission loss at P,,.

Sector transmission loss distribution: The probability distribution of the sector transmission loss for sectors located in similar environments and at equal distance from the base station. The 50 percent value of the sector transmission loss distribution is the value used for propagation models for the specified environmental parameters [2], and is called the median transmission loss. Service area: 1 ) The area within which a radio system provides either generally satisfactory service or a specific quality of service [ l ] ; 2) the area within which radio service is required. Service probability: The probability that a designated collection of sectors will receive a specified grade of service for a specific percent of the time. For example, S(50, 90) means that in 50 percent of the sectors designated, the specified grade of service will be exceeded 90 percent of the time. Signal-to-interference ratio (S/I): The ratio of the magnitude of the signal to that of the interference or noise. Note: The ratio may be in terms of peak values or root mean square (rms) values, and is often expressed in decibels. The ratio may be a function of the bandwidth of the system [I]. Signal-to-noise ratio (S/N): The ratio of the value of the

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 31, NO. I . FEBRUARY 1988

signal to that of the noise. Notes: 1) This ratio is usually in terms of peak values in the case of impulse noise and in terms of the rms values in the case of random noise. 2) Where there is a possibility of ambiguity, suitable definitions of the signal and noise should be associated with the term, as for example, peak signal to peak noise ratio, root mean square signal to root mean square noise ratio, peak-to-peak signal to peak-to-peak noise ratio, etc. 3) The ratio may be often expressed in decibels. 4) This ratio may be a function of the bandwidth of the transmission system. 5) In mobile systems, the S/N may be defined as the ratio of a specified speech energy spectrum to the energy of the noise in the same spectrum [l]. Signal strength: The level of radio signal presented to the antenna terminals of a receiver, expressed in dBW. SZNAD ratio: A measure expressed in decibels of the ratio of 1) the signal-plus-noise-plus-distortion to 2) noise-plusdistortion produced at the output of a receiver that is the result of a modulated-signal input [ 11. SZNAD sensitivity: The minimum standard modulated carrier-signal input required to produce a specified SINAD ratio at the receiver output [l]. Spectrum amplitude: The vector sum of the voltages produced by an impulse in a given bandwidth divided by that bandwidth. Note: Spectrum amplitude is usually expressed in volts/Hz or in dBpV/MHz (across a resistance of 50 a) see Section 1 of [9]. Strip noise: A composite impulsive noise source formed by a long narrow boundary. The most common source results from automobiles on a multilane highway. Suburban environment: A radio environment comprised of a village or highway with scattered houses, small buildings, and trees, often near the mobile unit [7], [8]. Transmission loss: In a system consisting of a transmitting antenna, receiving antenna, and the intervening propagation medium, the ratio of the power radiated from the transmitting antenna to the resultant power that would be available from an equivalent loss-free receiving antenna, For this report, both antennas are assumed to be vertical half-wave dipoles and the ratio is expressed in decibels [l]. Transmission loss distribution: The distribution of the percentage of locations within a sector where the transmission loss is less than the indicated value. It is best described by a Weibull distribution [2].

Transmission quality: For mobile communications, the measure of the minimum usable speech-to-noise ratio, with reference to the number of correctly received words in a specified speech sequence [I]. Urban environment: A radio environment comprised of an area which is heavily built up, crowded with large buildings and multistory residences, or a large village closely interspersed with multistory houses and thickly grown trees [7], PI. Vegetation effects: The effects of biomass on transmission loss. These effects may be due to reflection, attenuation, or diffraction by local vegetation, or by vegetation along the path. Weibull distribution: A class of distributions, the cumulative distribution functions of which have the form:

f(X; U, X)=1-exp [-(X/a)h]

where U and X are the scale and shape factors, respectively. The Rayleigh distribution function is a special case of the [51. Weibull function, with X = 2 and U = 40 Weibull fading: The variation of transmission loss at different locations within a sector due to multipath reception. REFERENCES
F. Jay, Ed., IEEE Standard Dictionary of Electrical and Electronic Terms, 3rd ed. IEEE Std. 100-1984, New York, 1984. N. H. Shepherd, Radio wave loss deviation and shadow loss at 900 MHz, IEEE Trans. Veh. Technol., vol. VT-26, pp. 309-313, 1977. Pub. 489-6, Int. Electrotech. Commission (IEC), Geneva, 1974. Propagation, CCIR Xth Plenary Assembly, Geneva, Rep. 239, vol. 11, 1963. A. M. Wood, F. A. Graybill, and D. C. Boes, Introduction to the Theory of Statistics, 3rd Ed. New York: McGraw-Hill, 1974. E. E. Smith, Glossary of Communications. Chicago, IL: Telephony, 1971. W. C. Jakes, Jr., MicrowaveMobile Communications. New York: Wiley, 1974. Y. Okumura et al., Field strength and its variability in VHF and UHF land-mobile radio service, Rev. Elec. Commun. Lab., vol. 16, pp. 825-873, 1968. Draft Pub. 489, Subcommittee 12F, pt. 9, Int. Electrotech. Commission (IEC), Geneva, 1983. Some of these references are reprinted in Land-Mobile Communications Engineering, D. Bodson, G. F. McClure, and S. R. McConoughey, Eds. New York: IEEE Press, 1984. The table of contents of this book is listed in Appendix I.

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

111. Background
A . Types of Systems
A variety of communication systems employ radio propagation in the 800/900 MHz frequency band. A listing and brief description of a number of these systems is given below. I ) Trunked and Conventional: A trunk is a one- or twoway channel provided as a common artery between switching equipment. A trunked radio system is a method of operation in which a number of radio frequency channel pairs are assigned to mobile and base stations in the system for use as a trunk group. Conventional land mobile systems operate on one or more radio frequency channels, but are not employed as a trunked group. In general, conventional systems are manually switched between channels, while trunked systems are automatically switched. Propagation prediction requirements for trunked and conventional land mobile systems are similar, since both normally operate with a single base station and require the greatest practical communication range. 2) Cellular Mobile Radio: An advanced land mobile system characterized by the ability to accommodate large numbers of subscribers through efficient frequency reuse [ 11[3]. The area to be served by a cellular system is subdivided into smaller service areas called cells. A cell is the area reliably served by one transmitter location. The cells are configured in such a way as to permit handoff, that is, the smooth transfer of a call from a channel in one cell to a different channel in an adjacent cell as the subscriber moves through the service area. Each cell is assigned a number of discrete frequencies which may be reused in other cell sites that are sufficiently far away to avoid interference. Because many channels are available for use by subscribers in each cell, a significantly higher number of subscribers per channel can be achieved than in conventional land mobile systems. This increased trunking efficiency coupled with frequency reuse makes it possible to serve a great many subscribers and still maintain a grade of service comparable to that of a land line network. A cellular system can grow internally by cell splitting or sectoring. Cell splitting often involves the introduction of smaller corner-excited (directional) cells into prior clusters of circular cells. This involves additions of cell sites. Cell splitting can be accomplished by adding many smaller cells to the existing larger cells serving the same area. This enables a system to accommodate a higher traffic density because a cell sites entire complement of channels, which previously served a larger cell area, can now be devoted to a smaller area. Cell sectoring is the use of multiple directional antennas at a single cell site to decrease the distance required between cells for frequency reuse without interference. The original cell is partitioned into a number of wedge-shaped sectors, each with its own set of channels. These directional cells and the sectorized cells require less separation for frequency reuse than omnidirectional cells would because the radiation in each sector is suppressed in all directions except for the narrow arc defining that sector. 3) Air/ground: Aidground land mobile systems are similar to cellular systems, except the mobile station is aboard an aircraft. The cells are much larger in size than ground mobile cellular systems. 4) Paging: A landmobile system that provides one-way signaling between one or more base stations and personal receivers. It may provide tone-only, tone-with-optical-reador tone-plus-voice. For tone-only service, oral, visual or tactile signaling devices are received. For tone-plus-voice service, the audio circuit in the addressed receiver is activated after the tone signal is received. Transmission loss to paging receivers is greatly affected by the environment near the receiver. 5) Mobile-to-Mobile: Communication between two mobile stations without the use of a base station repeater in a mobile-to-mobile system. Due to a relatively large transmission loss even for short distances between mobile stations, such a system has limited use. 6) Point-to-point (e.g., Control Links): Point-to-point transmissions between base stations are used to control the operation of remote base stations. The prediction of transmission loss between such stations can usually be done with few or no errors. 7) Satellite: A satellite mobile system provides communication between earth stations and active satellites for communication between mobile and base stations [4], [ 5 ] . The earth station may be fixed or mobile, and is intended to be used while either in motion or at unspecified fixed locations. 8) Radiating Cable: Radiating cable, also referred to as leaky cable, can be used in systems requiring communication inside buildings, mines, and tunnels where shielding by the environment restricts mobile radio coverage [ 6 ] . Such radiating cables are usually connected to a remotely operated base station, but can also be used as part of a passive or active repeater using outside antennas. In the 900 MHz frequency range, it is important to compare the transmission loss in the environment with the attenuation of a radiating cable, in order that the mode with the least attenuation is selected. 9) Simulcast (Paging and Mobile): Simulcast base stations can be used to extend the coverage area of a paging or mobile system without the requirement of additional channel frequencies being assigned. The operating frequency of each base station must be adjusted and maintained to within 3 parts in lo9 while the audio group delay is held within 100 p s .

B. Modes of Propagation
Propagation in the 800/900 MHz band, for the most part, takes place via space waves. At this frequency groundwaves are attenuated very rapidly with distance, and skywaves pass readily through the ionosphere with little energy being

10

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37, NO. 1, FEBRUARY 1988

reflected back to earth. Space waves are subject to absorption, reflection, refraction, and scattering by the troposhere and by the surface of the earth and obstacles in their paths. For the relatively short service ranges of land mobile systems using this band, propagation is usually by direct or reflected wave with some diffraction over obstacles. Interference, however, may occur at distances beyond the normal horizon via diffraction over the earth or by refraction or scattering by the troposphere. Because of the multiplicity of factors involved, most coverage and interference predictions depend on a statistical approach. However, deterministic analyses are still required to account for isolated and gross terrain features. Following is a brief discussion of the various propagation mechanisms and factors pertinent to the 800/900 MHz band. More detailed equations and qualifications are discussed in connection with specific propagation models described in a later section. I) Free Space Propagation: In free space, electromagnetic energy spreads out uniformly in all directions from a source. The amount of energy available to an antenna of a given effective area is inversely proportional to the square of the distance from the source [7]. Generally, free space conditions can be deemed to prevail if there is clearance for the first Fresnel zone for the path between transmitter and receiver and no reflections are present from the surface of the earth or from hills, trees, buildings, etc. In land mobile operations, free space conditions are rarely encountered at the mobile unit, unless it is located on an elevated roadway or near the crest hill. 2) Reflections from the Surface o the Earth: Signals f which arrive at the receiving antenna after reflection from the surface of the earth may interfere constructively or destructively with the direct wave depending on the relative phase of the two waves [7]. The relative phase and amplitude of the reflected wave is determined by the difference in path length and reflection coefficient of the surface. When the surface is smooth, specular reflections occur and the angle of incidence is equal to the angle of reflection. For most land mobile systems, reflections occur at small grazing angles so that the magnitude of the reflection coefficient is approximately unity, indicating no loss or attenuation, and its phase is 180. For path lengths where the curvature of the earth has to be considered, the divergence of the reflected rays causes a reduction of the specularly reflected signal. When surface roughness is comparable to a wavelength, the reflected signal has a specular component which is coherent with the incident wave and a diffuse component whose amplitude is Rayleigh distributed. The reflected signal can also be blocked or partially blocked by terrain or structures along the path. For paths a few miles in length over water or smooth terrain and with a first Fresnel zone clearance of 0.577 or more, the resultant signal decreases monotonically with distance and the theoretical value can be found from nomograms developed by Bullington [ 7 ] . At 900 MHz for normal mobile antenna heights, measured values of transmission loss are usually 10 to 20 dB greater than the theoretical value. In air-to-ground systems, multiple reflections may take

place from surface features such as mountains, flat terrain, and buildings. Generally, the signal will exhibit a series of predictable nulls as the separation distance between the ground station and aircraft increases (see Appendix IV). 3) Diffraction Over a Smooth Earth: Diffraction of radio waves around the curvature of the earth is affected by the frequency and polarization of the waves, the properties of the ground, and the geometry of the path. The path geometry is affected by the properties of the atmosphere through which the waves pass. At 900 MHz the path loss increases rapidly with distance beyond the horizon. A series of papers beginning with one by Sommerfeld in 1909 and one by Watson in 1918 developed analytical expressions for computing the field in the diffraction zone. The history of these developments was reviewed in a paper by Ekhert in 1986 [8]. The mathematics are complex and require a computer for any degree of precision. 4) Refraction: The speed of electromagnetic waves through the earths atmosphere is less than the speed through a vacuum. The refractive index of the atmosphere is the ratio of the speed in a vacuum to that in the atmosphere. This ratio at the surface of the earth in temperate climates is approximately 1.000340. It varies with the temperature, pressure, and moisture content of the air and can be calculated from measured values of these quantities or can be measured directly with a refractometer [9]. For computational convenience, radio refractivity N is defined in terms of the refractive index of the atmosphere n through the following [9]:
N = (n- 1) x 106.

In terms of the physical properties of the atmosphere, the radio refractivity is given by

N = 7 7 . 6 ( P / T ) + 3.73(e/T)
where

P atmospheric pressure (in millibars)

e water vapor pressure (in millibars) T absolute temperature
(OK).

Radio refractivity decreases with height above the earth being an average of about N = 301 at the surface and about 260 at 1 km above the surface in temperate climates. This gradient causes radio waves passing through the atmosphere to be bent back toward the earth, so that signals radiated at angles above the horizon may be bent around the curvature of the earth and received at distances beyond the normal horizon. In designing communications systems, this bending effect is taken into account by drawing path profiles assuming the radius of the earth to be 4/3 of its actual value [ 101. With this assumption, ray paths can be represented as straight lines, greatly simplifying ray tracing procedures. If more precise calculations are needed, an effective earths radius is calculated based on the refractive gradient for a specific location and season of the year. Contour maps showing surface refractivity and the refractivity gradient for all parts of the world can be found in a number of references [ 111.

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

11

Relatively low levels of energy are propagated great distances beyond the horizon via scattering from local irregularities in the refractive index [ 1 2 ] . For this mode of propagation, called troposcatter, energy is scattered from parts of the atmosphere visible to both the transmitter and receiving antennas with an attenuation proportional to the inverse distance between antennas. 5) Superrefraction and Ducting: In coastal areas, the movement of large masses of air due to advection and subsidence result in layers of air with refractive indexes significantly different from those of layers above and below. These steep gradients in refractivity can cause radio waves to bend sharply back toward earth and can produce greatly enhanced fields and cause interference at distances well beyond the normal radio horizon. Superrefraction occurs where the decrease in N per kilometer exceeds 100. Ducting, wherein a radio wave is trapped between elevated layers of air or between an elevated layer and the surface of the earth, can occur when the rate of decrease of N exceeds 157 per kilometer. This phenomenon can produce signals at or above the free space value at distances of hundreds of miles beyond the radio horizon. Superrefraction and ducting conditions are present for significant percentages of the time in the Southern California and Gulf Coast areas of the U.S. C . Effects of Buildings and Trees In urban areas propagation is generally dominated by shadow loss and reflections caused by buildings and trees. In the case of land mobile systems these are usually in the environment surrounding the mobile. Signals arrive from all directions with random amplitude and phase, and with a spread in arrival times of several microseconds. It is impossible to describe the received signal using deterministic models. The signal at a vehicle moving through the resulting standing wave pattern exhibits a Weibull distribution. The statistical treatment is discussed in detail in Appendix 111. I ) Reflections from Buildings and Trees: In a given situation a reflection from an individual building may be dominant. This may happen because of the building's height, size, orientation, or specific path configurations. Reflections may make a high signal available in areas deeply shadowed by buildings or terrain. The gain produced by a large reflecting surface can even result in signal levels in excess of free space values, as observed by Shepherd 1131. The strength of the reflected signal is determined by the height and width of the portion of a building visible to both terminals and is affected by obstacles within the first Fresnel zone of the incident or reflected ray and by the nature of the reflecting surface-type of material, size, shape, and orientation of sheathing elements. The vertical and horizontal width of the reflecting surface will determine the effective beamwidth of the reflected signal in the vertical and horizontal directions, respectively. In the case of hills, the factors that influence the amplitude and beamwidth of the reflected fields include the size, shape, and orientation of the reflecting surface and its coefficient of reflection which may be affected by trees or other vegetation. Composite signals reflected from wind-disturbed foliage or moving automobiles may exhibit a short-term variability of

several decibels when measured between fixed terminals or a base station and a stationary vehicle. 2) Attenuation Due to Trees: At 900 MHz shadowing, scattering, and absorption by vegetation can introduce substantial path losses. Over the past 45 years a number of studies have been carried out to characterize and model attenuation from vegetation. Fifty published works covering reports of measurement data and empirical and theoretical models were reviewed by Weissberger in 1982 1141. Results of recent measurements can be found in [ 1 5 ] . Weissberger concluded that a traditional exponential decay model was appropriate for those situations where propagation is likely to occur through a grove of trees rather than by diffraction over the top. After reviewing several exponential decay models, which are based on specific attenuation in terms of dB per meter of path length, and comparing them with several sets of available data at frequencies from 230 MHz to 95 GHz, he developed a modified exponential decay model (MED) which is applicable where a ray path is blocked by dense, dry, in-leaf trees found in temperate climates.

L = 1.33Fo.284d:588, for 1 4 s d f 5 4 0 0
= 0.45P.284df,

for O l d f < 14

where L is loss in dB, F is frequency in GHz, and df is depth of trees in meters. The difference in path loss for trees with and without leaves has been found to be about 3 to 5 dB. For a frequency of 900 MHz, the foregoing expressions reduce to L = 1.291df, for 1 4 s d f s 4 0 0 for O s d f < 14.

=0.437df,

An empirical-theoretical model by Kinase 1161 was found to be applicable to situations in which one antenna is located well above the foliage and the area in the vicinity of the second antenna is covered by a given percentage of clutter, which may be either buildings or trees. A recent extension of this model can be found in CCIR vol. 5 [17]. When both antennas are clear of trees in their immediate vicinity, the principal mode of propagation occurs by diffraction over the trees. A criterion for determining sufficient clearance is based on the takeoff angle to the tops of the trees. Estimates of this angle range from 8" to 26". There is considerable uncertainty among investigators as to when a knife-edge versus a rounded obstacle model is appropriate. 3) Building Penetration: Attenuation of signals penetrating a building is measured by taking the difference between the median signal level of several sectors located on the streets in front and beside the building in question, and the median signal level of sectors located inside the building. Building attenuation is dependent on the type of construction and the materials used, as well as the size of the building [17]-[25]. Table I lists construction elements and corresponding attenuations that are typical [ 2 5 ] . Given these element attenuations, the attenuation of a multifloor office building can be calculated by applying the method illustrated in the following example. A corridor is located on the first floor and 15 m from the nearest exterior

12

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37, NO. 1, FEBRUARY 1988 TABLE I ATTENUATION OF CONSTRUCTION MATERIALS
Construction Element
8 in. concrete block wall Wood and brick siding Aluminum siding Metal walls Attenuation past office furnishings (dB/m)

Attenuation (dB1
7 3

Standard Deviation (de) 1

0.5 0.5
4

2 12 1

0.3

40'
m h

'

' '

. 30 5

z m

0I
I I

-10.

'

'

'

B
Fig. 1 .

8 10 FLOORLEVEL

1 2 1 4

Variation with height of the signal reduction inside tall buildings (average over seven buildings in downtown Washington, DC).

is frequently assumed to be Rayleigh distributed. Such an assumption is not valid under all physical conditions [26]. The Nakagami distribution (also called the m distribution), however, assumes that the received signal is a sum of vectors with both random amplitude and random phase, which leads to a better fit with experimental data. The Weibull distribution affords the same flexibility as the Nakagami distribution and closely matches experimental data for those transmission loss measurements within a sector that are greater than the median value. It further provides the advantage of giving straight line plots on Weibull graph paper, which can be identified as a single slope parameter called loss deviation. In the 800/900 MHz bands, loss deviation varies from about 3 to 30 dB. This variation is considerable compared to the Rayleigh distribution value of 8.2 dB. In general, loss deviation can be related to the following propagation conditions [ 131. Loss de,viation: increases as the size of either antenna decreases; increases as the median loss for a given sector increases above the median value for all sectors at a given distance from the base station; increases in locations where the absolute area of a reflecting surface is greater than one-half the distance from the reflector to either the transmitter or receiver antenna; is greatest when several (2-4) signals of nearly equal amplitude are present; decreases in locations where a strong single source signal is present; decreases in locations where large numbers of signal sources are present. A further discussion of the statistics of path loss is given in Appendix 111. 2) Between Sectors: Although a sector is defined as being within a relatively small area, i.e., up to 20 wavelengths, the signal variation in adjacent sectors and sometimes in a large number of those nearby, may have about the same median value of signal amplitude and loss deviation. In other cases, however, both the median value and the loss deviation from one sector to the next can change rapidly due to such environmental factors as reflectors, obstructions, and foliage. Reflectors, such as tall buildings and hills, reradiate signals by passive reflection which is similar to radiation or reception by very high-gain antennas. This radiation, due to high gain (narrow beamwidth), frequently causes large variations in signal amplitude from one sector to the next.

wall of a 50 m square concrete block building. A single wall board partition separates the corridor from the exterior wall.
Attenuation (dB)
~I

E1em ent
~~~~~~~ ~

Exterior wall Office furnishings Interior oartition Total Attenuation

7 15 2

24

On higher floors the signal exhibits a height gain relative to the signal at street level. This height gain can be estimated from Fig. 1, which is based on attenuation measurements made on different floors of seven buildings in downtown Washington, DC.

D . Statistical Variability of the Desired Signal

Measurements of transmission loss between base station and mobile antennas, made as the mobile position varies, form a statistical distribution. The median of this distribution for a given sector is called the sector transmission loss, and can have long-term variations, such as season-to-season and yearto-year. However, within a given sector the statistical variation about the median is relatively constant. Values of transmission loss greater than the median have more significant effect on signal quality than do values less than the median. Therefore, the statistical model which best describes those values of transmission loss greater than the median value will be used. I ) Within a Sector: The variation of the signal in a sector

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

13

Further variations in a sector's median signal value occur behind a reflector since it also acts as an obstruction. The sector transmission loss for a large number of sectors located at a specified distance from the base station antenna can be combined to obtain a sector transmission loss distribution. In general, such a distribution is log normal, and its standard deviation gives the location variability for the specified environmental parameters. Both foliage and trees without foliage will cause the sector transmission loss to increase for each sector located in or behind the trees.

g3 0 5
2 20
U
W

r------y
RECEIVER NOISE--

E. Interference Various sources of interference are found in the 800/900 MHz band. These include noise, both impulsive and nonimpulsive, other communications systems, and intermodulation within the transmitter and receiver. These sources are discussed below. I ) Impulsive Noise: The rms amplitude of impulsive noise decreases at the rate of 28 dB per decade of frequency in the frequency bands above 200 MHz, which results in a noise level of about 8 dB lower at 800 MHz, as compared to 450 MHz. The lower impulsive noise level at 800 MHz permits the effective use of lower noise figure receivers than is possible at lower frequencies. a) Typical receiver noise figure of mobile and base receivers is about 10 dB. The effective noise figure of a receiver and its antenna is degraded by loss in the transmission line. It is the effective noise figure at a receiver antenna which must be used to determine degradation of performance due to impulsive noise or consideration for attenuation of the transmission line must be used for determining the spectrum amplitude of impulsive noise at the receiver input. Transmission lines for base receivers normally have greater attenuation than for the case of a mobile receiver. This greater attenuation can usually be offset by RF amplifiers with a noise figure as low as 3 dB. The best available receivers, transmission lines and RF amplifiers will provide effective noise figures of 1 1 dB. A noise figure of 11 dB should provide a receiver sensitivity of 0.25 pV for narrow-band FM receivers. For wider band width cellular receivers the effective sensitivity will be slightly greater. b) Characterizationof impulsive noise, for the purpose of determining degradation of the performance of receiving systems, utilizes two basic parameters: spectrum amplitude; and impulse rate. These two parameters are effectively displayed by a graph of the noise amplitude distribution (NAD). e) Impulsive noise sources of sufficient amplitude to be considered a potential for creating degradation of performance of receiver systems are spherics, electrical power lines, and ignition-typeengines. Impulsive noise radiated by spherics can exceed received spectrum amplitudes of 60 dBpV/MHz. Its impulse rate is very low, and as a result does not create significant interference except for digital transmission where very low error rates are required. Impulsive noise radiated by electrical power lines can produce received spectrum amplitudes near a line exceeding 40 dBpV/MHz at impulse rates of 200 i/s. However, the

!I
10 100
1,000 10,000 IMPULSE RATE (]/SI
Fig. 2. Noise amplitude distribution (NAD) for strip and receiver noise. Reference sensitivity is 0.25 pV, and signal and noise frequencies are 850 MHz.

400
Fig. 3 .

600

800 FREQUENCY (MHz)

1,000

Degradation of reference sensitivity due to ignition noise. Reference sensitivity is 0.25 pV, and signal frequency is 850 MHz.

occurrence of such noise is not often enough to be further considered. Impulsive noise radiated by ignition-type engines is the only source that will be considered of significant spectrum amplitude and impulse rate to cause degradation of receiver performance. d) Ignition noise must be considered both for vehicles near the base station and for vehicles in the vicinity of the mobile (strip noise). Two examples of the NAD of strip noise for 10 000 and 1000 vehicles per hour are illustrated by Fig. 2. Also shown in Fig. 2. are the spectrum amplitude and impulse rates for receiver noise, which have an effective sensitivity of 0.25 pV. Comparison of the receiver noise with the strip noise will indicate degradation for any spectrum amplitude of the strip noise exceeding the receiver noise. The NAD of base station noise for high vehicle densities of lo00 vehicles/km* is about equivalent in interference to the NAD for strip noise of C = loo0 vehicledh. e) Degradation of receiver performance by ignition noise is shown in Fig. 3. This figure gives the probable degradation of receiver performance for both base station and two types of strip noise as a function of operating frequency.

14

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37, NO. 1, FEBRUARY 1988

TABLE I1
SOURCES OF INTERFERENCE

Source of interference Television transmitter . sideband Transmitter noise Transmitter intermodulation Receiver intermodulation

- -

- - - - - - -

Degree of Interference to

- -

- -

- - - - - - - -

- - -

- -

Base station receiver Strength depends frequency separation

Mobile station receiver Strength depends on related location of station

- _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ - - - - - - - - - - - - - - - - - Strong on frequencies near band edges Strength depends on relative location of base station Strength depends on design of transmitter combiner Strong i f the design of the transmitter combiner has reduced transmitter intermodulation Not important unless transmitter noise has been filtered out Strength depends on design of transmitter combiner

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Not a serious problem except when a common antenna is used with two or more transmitters Not important except on frequencies near band edges

- - - - - - - _ _ -. - - - - - - - - - - - - - _ - - - - - - - - - - - - Receiver desensitization

The frequency range from 400 to 1000 MHz was included to more serious, since the frequency spacing can be as close as 25 illustrate the decreased degradation as a function of frequency. kHz. At close frequency spacing, less than 1 MHz, the 2) Nonimpulsive Interferenee: Interference to both the interference range can be as great as ten miles. base and mobile station receivers can come from a number of e) Transmitter nonlinearities in the output stage can different sources. In the case of a base station receiver the generate third-order products when an external carrier is interference can be co-site, from a distance, or some combina- received through the transmitting antenna. The carrier fretion of both. Interference to mobile receivers is subject to quency of the IM product is given by 2fB - fA where fA is the wider variation, which is dependent on receiver location and frequency of the transmitter of the external carrier and fB is the time of day. The more important types of interference are frequency of the transmitter that generates the IM product. The amplitude of the product varies from about 0 to 6 dB listed in Table I1 and discussed below. In general the potential for interference has been reduced by below the amplitude of the external carrier. A similar product two approaches. The first of these was accomplished by the can also be generated by transmitter A, in which case its design of equipment and systems with state-of-the-art per- frequency will be 2fA -fB. The amplitude of either product is formance. Transmitter characteristics, such as sideband noise a direct function of the attenuation between the output of one and intermodulation-produced radiation, along with receiver transmitter and that of the other transmitter. d) Receiver nonlinearitiescan cause interference on the characteristics, such as selectivity, intermodulation rejection, and desensitization immunity, have been controlled. The same channel frequencies as those mentioned above for second approach was to assign transmitter and receivers with transmitter generated intermodulation. Although difficult, it is usually possible to locate the source of the intermodulation 45 MHz frequency separation. a) Television transmitters operating on UHF channels products through the use of attenuators, isolators, or filters. 67-69, adjacent to the land mobile base receiver frequencies After a source is located, the effect of interference can be starting at 806 MHz, radiate sideband noise covering a wide reduced by adding isolators, filters, or relocating antennas. frequency range. The resulting interference can be significant Intermodulation products generated in receivers are more for a base station receiver located within six miles of a channel numerous than those generated in transmitters. Higher order 68 transmitter and within 100 miles from a channel 69 products, as well as low-order products, can cause serious transmitter. Mobile receivers are less susceptible to TV interference. The potential for interference increases rapidly transmitter noise due to a wider frequency separation. as the number of radio systems at a given site increases. b) Transmitter noise in typical systems decreases by The intermodulation products generated in base station about 65 dBc/lO kHz close to the operating frequency. It receivers, due to the 45 MHz transmitterheceiver spacing, is decreases in amplitude beyond 1 MHz and is no longer a not usually a serious problem. Base station receivers, when problem at 20 MHz. The use of 45 MHz frequency spacing operated on a common antenna with two or more transmitters, between the base station transmitter and receivers of a duplex can receive high-order generated product interference from channel should prevent noise from being a problem at a base any of a number of possible generators in the antenna system. station for the channel. However, the frequency spacing 3) Interference due to Multipath Fading Within a Sector: between the transmitter of one channel and the receiver for Rapid fading of a received signal due to motion of a mobile another channel can be much less than 45 MHz, leading to the station or any reflective objects in the transmission path causes possibility of interference due to transmitter noise. a degradation of the receiver sensitivity at both the base and Transmitter noise interference to mobile receivers is much mobile station. The degree of sensitivity degradation is a

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

15 actual radiation patterns, as relative power gain (often just gain) patterns are more often employed in systems design. These show the ratio of the power gain of an antenna as a function of spatial coordinates to the power gain of a reference antenna in its reference direction. Here, power gain refers to 47r times the ratio of radiation intensity [27] in a given direction to the net power accepted by the antenna from a connected transmitter. Although an isotropic pattern is sometimes used as a reference, it is more common in land mobile work to reference power gain patterns to that of the omnidirectional H-plane pattern of a free space half-wave dipole with equivalent excitation. Relative power gain is then expressed in dBd, where the subscript indicates the dipole reference. Manufacturers specifications often quote antenna power gain to the nearest 0.01 dB. It is well known, however, that the best accuracy that can be expected from simulated free space antenna measurements [28] is 0.5 dB. Specificationof antenna gain to greater accuracies in system design documents is therefore not a recommended practice. The number of different shapes of relative gain patterns that can be radiated by different antennas is limitless. Often, however, the patterns of directive antennas exhibit one prominent radiation lobe. This lobe is visualized as the main beam of the antenna, leading to a subsequent classification of directive antennas by the width of this main beam (beamwidth) in a specified plane at its half-power points. One immediately obvious use for a knowledge or specification of the relative gain pattern of an antenna is to protect a particular radio system against unwanted interference. More involved requirements for specific pattern shapes are discussed in later paragraphs. Antenna polarization is perhaps the next important characteristic of an antenna. An antenna is said to be polarized identically to the electromagnetic waves radiated by it. Wave polarization can be linear or elliptical and is defined by the locus formed by the tip of the electric field vector as the wave propagates away from the antenna. Linearly polarized waves are classified by the direction of their electric field vector with reference to some reference plane. The land mobile industry takes this reference plane to be the earths surface. A wave having an electric field directed horizontally is therefore classified as horizontally polarized. This classification is extended to the antenna from which it is radiated and similarly for vertically polarized antennas. The sense of polarization most often employed at mobile system base stations is vertical. While polarization is unchanged by reflection or transmission at plane homogeneous dielectric or conducting interfaces [29], measurements indicate that both vertical and horizontal components of electric fields are present at the receiver. The rotated component of field arises through scattering from nonplaner and rough surfaces [30]. It has been found that the statistics of the orthogonal components of field are uncorrelated [3 I]. Polarization diversity has therefore been suggested as a means for the reduction of fading. An associated phenomenon, though not a polarization change, is a reversal of the direction of rotation for elliptical polarization. This can occur on reflection from a plane conducting interface, because the tangential

1
Fig. 4.

5
Loss

10

15

DEVIATION(DB)

Degradation of SINAD sensitivity due to multipath fading.

function of the loss deviation of the received signal. In FM receiver, the improvement ratio also decreases from a typical 3: 1 to 1: 1 ratio, when the loss deviation increases from 0 to about 10 dB. Fig. 4 shows typical values of degradation of SINAD sensitivity for loss deviations between 1 and 14 dB. Each curve gives the required increase in the median value of a desired signal level to maintain the indicated SINAD sensitivity. F. Antenna Characteristics The design of an 800/900 MHz mobile radio communication system must include considerations for the characteristics of antennas used in fixed base stations, as well as those used on the mobiles. Specific characteristics of the antennas chosen for use in a system have different effects upon system operation. These must be well understood if optimum system design is to be achieved. The following paragraphs highlight considerations related to the choice of antennas for use in 800/900 MHz mobile radio systems. For completeness, references are given where more detailed information can be obtained from the literature. I ) Antennas in Mobile Radio: One of the primary characteristics of an antenna is its radiation pattern. This is a graphical representation [27] of the radiation properties of the antenna as a function of spatial coordinates. Normally, radiation patterns are referred to as isotropic (equal radiation in all directions), omnidirectional (equal radiation in all directions in a given plane), or directional. Although an isotropic radiator is not physically realizable, its theoretical radiation pattern is often used as a reference for comparison of different antennas. Unless received power measurements are being made in the field, it is unlikely that a systems engineer will encounter

16

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37. NO. I , FEBRUARY 1988

component of an electric field undergoes a phase reversal, whereas the normal component remains unchanged. Another very important antenna characteristic is bandwidth. This is the range of frequencies over which the performance of the antenna, with respect to some characteristic, conforms to a specified standard [27]. The specifications for each characteristic vary, and are set in order to meet the requirements of each particular application. Pattern bandwidth and impedance bandwidth are often distinguished from each other. Associated with pattern bandwidth are gain, sidelobe level, beamwidth, polarization, and beam direction. Impedance bandwidth influences feed point impedance and radiation efficiency, and is of particular importance in the new 800/900 MHz band where there is a 45 MHz separation between allocated transmit and receive frequencies. This requirement for large antenna bandwidths complicates antenna designs and has been the subject of recent investigations [32]. Detailed information regarding this subject can be found in the reference. It should be noted, however, that manufacturer specified voltage standing-wave ratio (VSWR)/bandwidth curves are often smoothed and the detailed shape of such characteristics, although extremely important in some digital applications, is often obscured. The impedance characteristic of an antenna as seen from the transmitter or receiver is also greatly influenced by the connecting transmission line, and the operating environment of the antenna. Possible mismatches in feed lines and connectors are important considerations. Bandwidths specified by manufacturers are typically the frequency range over which the maximum VSWR measured at discrete frequencies at the antenna feed point is less than 2: 1. 2) Base Station Antenna Considerations: Factors which must be considered in the specification and siting of a base station antenna include: relative power gain as a function of azimuth, polarization, transmission line losses, height, impedance matching, proximity to support towers, buildings and other antennas, environmental noise, and coverage ability. While some of these and the methods by which they can be dealt with are obvious, others require further considerations. These will be discussed in the following paragraphs. Unless polarization diversity is planned [3 11, nearly all base station antennas utilize vertical polarization. This is reflected by typical commercially available designs which include vertical dipole fed corner reflectors and vertically spaced arrays of vertically polarized folded dipole elements. Base station antenna gain in any particular direction is dependent upon the proximity of the antenna to other potential radiators, the multipath environment, and the surrounding topography (including man-made obstacles). The antenna gain specified by manufacturers is measured on an antenna range in a simulated free space environment. The best accuracy that can be expected from such measurements is 0.5 dB. However, 3 dB variations in the measured azimuthal pattern of land mobile system omnidirectional antennas are not uncommon. If the antenna is side mounted on a conducting support tower [33] its azimuthal gain pattern may be altered significantly in a manner which is difficult to predict. Patterns may also be altered by the presence of transmission lines running up the tower behind the antenna. It is obvious that top

mounting the antenna is a means by which these problems can be eliminated, but the top position on any tower is not always available. It has been suggested that the proximity problems can be solved by installing four co-fed directive antennas so as to produce a quasi-omnidirectional pattern. This method has been employed successfully in Europe. One disadvantage, however, is that the element spacings required for such a system are greater than that which can be achieved using tower mounted antennas. For this reason, antenna mounting at the top of orthogonal faces on tall buildings has been found to be necessary. Another proximity problem that can influence the pattern of base station antennas is that of nearby antennas which form parts of other co-sited systems. Normally, at a horizontal spacing of four wavelengths, antenna isolations of approximately 25 dB can be attained. More isolation can be achieved through vertical spacing of vertically polarized antennas, but precautions must be exercised to keep vertically spaced antennas within each others cone of silence. The influence of multipath propagation is pronounced and can reduce the gain of phased array antennas to the extent that any design advantage in effective radiated power is eliminated. This is due to the lack of phase coherence in the multipath field and subsequent destruction of required phase relationships at the antenna elements. To avoid this problem base station antennas should be mounted high above surrounding obstacles, which reduces multipath, provides higher effective gain, and also allows advantage to be taken of a reported [34] 6 dB gain advantage for every doubling of the antenna height. Although this gain can probably not be realized in an urban environment, coverage to any parts of the system that are in a suburban or rural area separated from the base station by flat or smooth rolling terrain should be improved. 3) Mobile Antenna Considerations: Some of the constraints on antennas for mobile operation such as size, ruggedness, appearance, and cost are obvious and require little discussion. There are, however, a number of more subtle constraints and characteristics, particularly in the 800/900 MHz band, some of which can be used to gain unexpected advantages in mobile system operation. In a previous section, it was explained that if vertical polarization is used at the base station, most of the waves impinging on the mobile antenna can be considered to be essentially vertically polarized. For this reason most commercially available mobile antennas are vertically polarized, and this sense of polarization is assumed for the antenna characteristics discussed in the following paragraphs. The quarterwave rooftop antenna [34], which is a good mobile antenna for 450 and 150 MHz, exhibits a significantly poorer performance in the 800/900 MHz band. This is because the effective area of such an antenna in the band of interest is less by a factor of about thirty. To achieve higher theoretical gain, commercially available antennas for this band are usually of the collinear array type, consisting of 5/8 or 1/2 wavelength top sections, a phasing coil, and a 1/4 wavelength lower section. When measured in a multipath environment, however, the same antennas exhibit substantially less gain [32]. As for the base station antenna, this is a result of the lack

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 31,NO. 1, FEBRUARY 1988

17

TABLE 1 1 1 ATTENUATION OF SEMIFLEXIBLE CORRUGATED COPPER LINES

Foam Filled

Air Filled
1- 5 / 8
0.0
7/8

LOSS/100 ft LOSS/IOO m

1/2 2.2
6.0

7/0

1.2 4.1

1.2
3.8

1-5/0 0.6

2.5

2.1

of phase coherence in the multipath fields. It is therefore recommended that systems designers anticipate vehicular antenna performance that falls somewhat short of published manufacturers range measurement results. In addition, it can be expected that the in situ azimuthal gain pattern of a vehicular mounted antenna will be dependent upon mounting position and will invariably be different from that measured on a range. Although changes in the azimuthal pattern of a vehicular antenna as a result of the proximity of the vehicle body are reported [34] to have little effect on average received field strength, azimuthal pattern shape has been found [35] to influence the observed received signal fading rate. The elevation angles from which signals arrive at a mobile unit in an urban environment have been measured [34] to be limited to about 30. Advantage can therefore be taken of extra antenna gain offered by antennas with vertical directivity without a significant reduction in received signal power due to the exclusion of waves impinging from high elevation angles. One final point that might be considered with regard to mobile antennas is the possibility of obtaining some power gain by increasing height. For a large range of frequencies [36] and for several base station antenna heights, it has been observed that 3 dB gain advantage can be obtained for a 3 m high antenna as compared with the gain of a 1.5 m high antenna. Due to obvious constraints on mobile antenna height, however, it is considered impractical to attempt to achieve more than about 1 dB height gain advantage.

receiving locations may be used if the received signals are processed by a voting receiver that electronically selects the best signal. In cellular radio, the frequency pair is automatically switched to the frequency of the next cell to expand the area of coverage.

G . System Characteristics for Multisource Signals

Most systems are designed to operate at virtually the maximum effective radiated power permitted in that service, and from an antenna that is located as high above average terrain as possible for that geographical area. Therefore, in order to expand the area of coverage of the base station, additional antenna locations may be used. Paging systems and signaling transmissions (on two-way systems) may be simulcast, all transmitters broadcasting the same signal at the same time. For operation of such simulcast systems, consideration should be given to: the frequency stability of the transmitter 3 parts in lo9 or better); use of a radio control link from a central control point to avoid appreciable audio delay; identical modulation characteristics in all base transmitters; audio group delay within 100 ps at each base transmitter; and the frequency of each transmitter adjusted to within one or two Hertz of the reference transmitter. Mobile (two-way) systems have additional considerations because of the mobiles. If the power output of the mobile unit is considerably less than that of the base station, remote

H. Fixed System Losses Far more attention must be given to the passive components between the transmitter RF power output stage and the input to the antenna in the 800/900 MHz bands than is required in the VHF band. Transmission line losses are greater in the 800/900 MHz band, and the quality of mechanical installation, as well as aging of the components (especially connectors), can easily lead to a high VSWR, thus reducing the RF power radiated from the antenna. The number of connections should be kept to a minimum. Air filled transmission lines have less loss than solid or foam dielectric filled lines, per unit of size and length; however, air lines will require more maintenance after installation to keep the lines pressurized with dry air or gas to prevent condensation. I ) Cable/Connector Characteristics: All transmission lines, even so called jumper-cables used in combiner cabinets, should be of solid copper outer conductor rather than braided to reduce leakage. Typical losses (at 900 MHz) of lines are listed in Table 111. Although not having any effect on the electrical characteristics of the transmission line, all exterior transmission lines should be jacketed with a polyethylene sheath. Not only will the jacket protect the copper from the elements, but it will also provide a bond to the tape wrapped around the connectors to make the connections waterproof. All connectors in the RF portion of the system should be of the N type. Not only is it good engineering, but it may be required at many antenna sites to insert one bandpass cavity with each transmitter. The cavity filter reduces the RF pollution due to out-of-channel emissions and prevents channel signals from mixing in the output stage of the transmitter. The typical insertion loss of a cavity is 0.5 dB; however, many filters have settings of 1, 2, or 3 dB which will give a greater selectivity to the response curve of the cavity. If the transmitter is to be used in an environment which has other transmitters operating in the same band, an isolator should be inserted in the system. An isolator has an insertion loss of approximately 0.5 dB at the desired frequency but offers an attenuation of more than 25 dB to an unwanted signal external to the system. An isolator is a circulator with an appropriate load termination attached. A harmonic filter should be used with all isolators to reject the second harmonic. Normal loss for a harmonic filter is 0.2 dB. Duplexers, used to

18

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 31, NO. I , FEBRUARY 1988

isolate the transmitter and receiver when using a common transmission line and broad-band antenna, are either of a bandpass or band-reject type. For a single transmit frequency and a separate receive frequency system, a duplexer with a combination of bandpass and band-reject sections is preferred. The bandpass portion gives the advantages of a cavity and the reject portion offers a higher rejection of the unwanted signal than a cavity. In multichannel systems, as in trunked and cellular systems which require a very large bandwidth, only the cavity type is used. 2) Duplexers and Diplexers: The isolation between the transmitter and receiver is a function of the number of cavities used in either branch of the duplexer. A diplexer is essentially a duplexer used as a transmitter combiner when the frequency separation between the two transmitters is approximately 2 MHz or more. 3) Combiners (Transmitters): There are two basic types of transmitter-combiners depending on the frequency separation between the transmitting frequencies. If the transmitting frequencies are closely spaced, less than 0.25 MHz-0.30 MHz, a hybrid ferrite type combiner must be used. However, the hybrid device has an insertion loss of 3.2 dB. If the transmitting frequencies are widely spaced, greater than 0.30 MHz, a junction combiner (multiport connector) can be used. The junction itself has no real significant loss; however, the isolating components necessary can make the loss for a two-transmitter combiner comparable to a hybrid type combiner. a) Hybrid-ferrite type transmitter combiners can be used with frequency separations as close as 0.025 MHz and are always used with at least one isolator and harmonic filter. Minimum isolation is 70 dB. If additional isolation is needed, another isolator may be inserted in series with the first. If three or four transmitters are to be combined, the outputs of two hybrid couplers can be fed into a third hybrid coupler. The insertion loss of the hybrid, 3.2 dB, is additive, thus making the loss of a four-transmitter hybrid combiner 7.1 dB. The hybrid combiner can be further expanded for eight transmitters and 16 transmitters by adding hybrids to the pyramid. b) Junction type transmitter combiners can be used if the frequency separation is 0.30 MHz or greater. The junction is a multiport connector with one port for the antenna and a port for each transmitter. Isolation is accomplished by the number of isolators and cavities in each leg of the combiner. In a multifrequency system, rather than using a separate duplexer in each paired channel, all receivers should be connected to one multicoupler which is fed by one antenna. Since receive systems are plagued by intermodulation and receiver desensitization, it is usually more manageable to deal with the location of one antenna and a band-pass filter for all the receiving frequencies. Most multicouplers have built-in preamplifiers which give an overall gain to the system. REFERENCES
[I]

Special Issue on Advanced Mobile Phone Service, Bell Syst. Tech. J., vol. 58, no. 1, 1979.

W. C. Y. Lee, Mobile Communications Engineering. New York: McGraw-Hill, 1982. J. F. Whitehead, Cellular system design: An emerging engineering discipline, IEEE Commun. Mag., vol. 24, pp. 8-15, 1986. R. E. Anderson, R. L. Frey, I. R. Lewis, and R. T . Milton, Satelliteaided mobile communications: Experiments, applications and prospects, IEEE Trans. Veh. Technol., vol. VT-30, pp. 54-61, 1981. G. H. Knouse and P. A. Castruccio, The concept of an integrated terrestrial/land mobile satellite system, IEEE Trans. Veh. Technol., vol. VT-30, pp. 97-101, 1981. Q . V. Davis, Guest Ed., Special Issue on Leaky Feeder Radio Communication Systems, The Radio and Electronic Engineer, vol. 45, no. 5, 1975. K . Bullington, Radio propagation variations at VHF and UHF, Proc. IRE, vol. 38, pp. 27-32, 1950. R. P. Eckert, Modern methods for calculating ground-wave field strength over a smooth spherical earth, FCC/OET R86-1, Feb. 1986. The formula for the radio refractive index, CCIR XV Plenary Assembly, Geneva, Recommendation 453, vol. 5, 1982. Effects of large-scale tropospheric refraction on radiowave propagation, CCIR XV Plenary Assembly, Geneva, Rep. 718-1, vol. 5, 1982. Radiometeorological data, CCIR XV Plenary Assembly, Geneva, Rep. 563-2, vol. 5, 1982. Effects of Small-Scale Spatial or Temporal Variations of Refraction on Radiowave Propagation, CCIR XV Plenary Assembly, Geneva, Rep. 881, vol. 5, 1982. N. H. Shepherd, UHF radio wave propagation in Dallas, Texas base to mobile stations for vertical polarization, Gen. Elec. Tech. Inform. Series, R75-MRD-1, Mar. 1975. M. A. Weissberger, An initial critical summary of models for predicting the attenuation of radio waves by trees, ESD-TR-81-101, Electromagn. Compat. Analysis Center, Annapolis, MD, July 1982. W. J. Vogel and J. Goldhirsh, Tree attenuation at 869 MHz derived from remotely piloted aircraft measurements, IEEE Trans. Antennas Propagat., vol. AP-34, pp. 1460-1464, 1986. A. Kinase, Influences of terrain irregularities and environment clutter surroundings on the propagation of broadcasting waves in the UHF and VHF bands, Japan Broadcasting Corp., Tokyo, Japan, NHK Tech. Monograph 14, Mar. 1969. Methods and statistics for estimating field strength values in the land mobile services using the frequency range 30 MHz to 1 GHz, CCIR XV Plenary Assembly, Geneva, Rep. 567, vol. 5, 1982. P. 1. Wells and P. V. Tryon, The attenuation of UHF radio signals by houses, IEEE Trans. Veh. Technol., vol. VT-26, pp. 358-362, 1977. H. H . Hoffman and D. C. Cox, Attenuation of 900 MHz radio waves propagating into a metal building, IEEE Trans. Antennas Propagat., vol. AP-30, pp. 808-811, 1982. D. C . Cox, R. R. Murray, and A. W. Norris, Measurements of 800MHz radio transmission into buildings with metallic walls, Bell Syst. Tech. J . , vol. 62, pp. 2695-2718, 1983. E. H. Walker, Penetration of radio signals into buildings in the cellular radio environment, Bell Syst. Tech. J . , vol. 62, no. 9, pp. 2719-2734, 1984. D. C. Cox, R. R. Murray, and A. W . Norris, 800 MHz attenuation measured in and around suburban houses, Bell Labs. Tech. J . , vol. 63, pp. 921-954, 1986. D. C. Cox, R. R. Murray, and A. W. Norris, Antenna height dependence of 800 MHz attenuation measured in houses, IEEE Trans. Veh. Technol., vol. VT-34, pp. 108-115, 1985. J. Horikoshi, K. Tanaka, and T. Morinaga, 1.2 GHz band wave propagation measurements in concrete buildings for indoor radio communications, IEEE Trans. Veh. Technol., vol. VT-35, pp. 146152, 1986. General Electric Systems Application Manual, sec. 80-A 1, Table IV-3, General Electric Corp., Lynchburg, VA, Dec. 1972. N. H. Shepherd, Radio wave loss deviation and shadow loss at 900 MHz, IEEE Trans. Veh. Technol., vol. VT-26, pp. 309-313, 1977. C. A. Balanis, Antenna Theory. New York: Harper & Row, 1982, ch. 2. Unpublished report to the EIA committee on the calibration of the gain reference standard antenna for EIA standard RS-329 by Nat. Bur. Stand., Boulder, CO. S . R. Seshadri, Fundamentals of Transmission Lines and Electromagnetic Fields. Reading, MA, Addison Wesley, 1971, ch. 6. F. G. Bass et al., Very high frequency radio wave scattering by a

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

19
using a directive antenna, IEEE Trans. Veh. Commun., vol. VC15, pp. 16-24, 1966. [36] Y. Okumura et al., Field strength and its variability in VHF and UHF land mobile service, Rev. Elec. Comm. Lab., vol. 16, pp. 825-873, 1968. Some of these references are reprinted in Land-Mobile Communications Engineering, D. Bodson, G . F. McClure, and S. R. McConoughey, Eds. New York: IEEE Press, 1984. The table of contents for this book is listed in Appendix I.

[31] [32] [33] [34] [35]

disturbed sea surface, IEEE Trans. Antennas Propagat., vol. AP16, pp. 554-559, 1968. W. C. Y. Lee, Mobile Communications Engineering. New York: McGraw-Hill, 1982, ch. 5. J . S. Belrose, Vehicular antennas for 800 MHz mobile radio, in Conf. Rec., IEEE Veh. Technol. Conf., Toronto, May 1983. W. A. Wickline, Cellular demands superior antenna radiation control, Mobile Radio Technol., vol. 2, Feb. 1984. W. C. Jakes, Microwave Mobile Communications. New York: Wiley, 1974, ch. 3. J. R . Stidham, Experimental study of UHF mobile radio transmission

20

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37. NO. 1 , FEBRUARY 1988

IV. Propagation Models

A variety of experimentally or theoretically based models the respective horizons of the transmitting and receiving have been developed to predict radio propagation in land antennas. Computer versions of these nomograms are in common use. mobile systems. Many of the models that have been reported in the literature are summarized and compared in this section. The model is an oversimplification since vegetation, buildWe start by listing and summarizing the major features of the ings, and other terrain factor are not accounted for. 4) Egli: This model consists of empirical formulas that models. The models are then compared in terms of the environmental and propagation factors they account for, and in provide a terrain factor to be applied to the theoretical terms of the output information they provide. Finally, the plane-earth field strength [4]. The terrain factor for 900 MHz median transmission loss predictions of various models are has median value of 27.5 dB. The variation of field strength compared for a given base antenna site. A detailed evaluation with antenna heights and distance is that of the underlying plane earth model. Median transmission loss for a 1.5 m high of some models is given in Section V. mobile antenna is equal to A . Summary of Current Models L = 139.1 -20 log h+4O log D (dB). In listing the various models, a common notation and set of units have been adopted for all. The distance D between base and mobile antennas is in kilometers, while the height of h of the base antenna is in meters. L is used for median transmission loss in dB between half-wave dipoles for 900 MHz operating frequency. I ) Free Space: This fundamental theoretical model predicts the losses due to radial spreading for propagation in an ideal region without boundaries [ 11, [2]. Transmission loss predicted by this model is given by L = 87 + 20 log D (dB). The Egli model is a systematic interpretation of measurements covering a wide span of frequencies (90 to 1000 MHz) with some sacrifice of the accuracy that might be obtained by examining narrower bands individually. 5) Carey: Curves give the F(50, 50) and F(50, 10) field strength versus distance for propagation under average terrain conditions with: 1) mobile antenna height of 1.8 m; 2) base station antenna heights ranging from 30 to 1500 m above average terrain; and 3) distances up to 130 km for the F(50, 50) curves, and up to 240 km for the F(50, 10) curves [5]. The frequency bands explicitly covered are those used for common carrier mobile service at the time. Although 900 MHz frequencies are not mentioned, the Carey curves for 450-460 MHz were based on a CCIR recommendation (Geneva, 1963) covering the entire band from 450 to 1000 MHz. Median transmission loss for this model can be found for engineering purposes from the following expressions that have been fitted to the published curve: L=110.7-19.1logh+55logD L=91.8-18 log h + 6 6 log D (dB), (dB), for8 for 48

This model is useful as a reference for comparing transmission loss given by other models. When the Fresnel clearance is 0.6 or greater, this model can be used to estimate the loss associated with the fields that propagate directly (without reflection off the surface of the earth) from transmitting to receiving antenna. 2) Plane Earth: The plane earth model has been derived theoretically taking into account the presence of an idealize plane earth with finite conductivity [3]. For 900 MHz, median transmission loss to a 1.5 m high mobile antenna is given by L=111.6-20 log h+4O log D (dB).
=

< D<48 < D<96.

This expression is equal to the free space loss when D

h/ 17

(W.
This model is also useful as a reference for comparing other models. Calculated results for land mobile service at 900 MHz are not sensitive to the values chosen for the ground constants. The model is an oversimplification in that it does not include important factors such as terrain profile, vegetation and buildings. 3) Bullington Nomograms: In addition to free space and plane earth effects, the Bullington nomograms give field strength and median transmission loss as a function of antenna heights for certain idealized types of intervening terrain, specifically smooth curved earth with: 1) no obstructions; 2) a single sharp ridge; and 3) hills [ 1J , [2]. The smooth earth and single ridge results come directly from physical theory. The treatment of hills is based on an approximate solution to the theoretical problem of diffraction by two knife edges located at

The Carey curves themselves are included in FCC Rules and Regulations and are being used for cellular radio licensing applications (the analytic fit to the Carey curves given above should not be used for application purposes). The curves are derived from CCIR curves for television broadcasting. The latter were adjusted downward by 9 dB to account for the 1.8 m height of mobile station antennas. Note that F ( L , T ) denotes field strength exceed at L percent of locations during T percent of the time. At distances involved in land mobile service, only the location variability is significant, and the time variability can be considered as T = 100. 6) Tech. Note IOI: This model consists of curves, theoretical equations, and empirical formulas for predicting cumulative distributions of sector transmission loss for a wide range of frequencies over almost any type of terrain and in several cliniatic regions [6]. Application of this model usually requires a succession of fairly detailed calculations, and at each step one must take care to find the curves or formulas that are exactly appropriate.

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

21
the form of a computer program, this model was developed by the Electromagnetic Compatibility Analysis Center (ECAC) [ 101. It predicts propagation loss between two points taking into account the frequency, atmospheric and ground constants, and the characteristics of the terrain profile between the two points. A digitized data base of terrain elevations gives necessary information on the profile, and the model then selects the appropriate propagation algorithm and calculates loss. TIREM is one of seven point-to-point propagation models included in [lo]. The other six, however, are not appropriate for 900 MHz land mobile propagation. The inputs to the program consist of the frequency (40 MHz to 20 GHz), polarization, ground permittivity and conductivity, atmospheric refractivity modulus, absolute humidity, and transmitter and receiver antenna structural heights, site elevations, latitudes and longitudes. The program gives as outputs the path loss and fading statistics. It may be difficult to adapt the computer program to operate at a particular facility because of the requirement to read a digitized data base. This model is discussed in greater detail in Section V. 12) General Electric Slide Rule: This model is in the form of a slide rule for calculating the maximum range for mobile, portable, and point-to-point communications in the land mobile bands; alternatively, it may be used to calculate required transmitter power for a given range [ l l ] . Actual mobile antenna height is used unless effective height (read from a table) is higher. Percentage of locations to be covered may be specified from 50 to 99.9. Received signals (microvolts across a 50 Q input impedance) are related by the slide rule to transmitter output power after the device has been set for the desired range, percentage of locations, antenna heights, and antenna system gain/loss adjustment. 13) EPM-73, High Antenna/Low Antenna: Step-by-step procedures are given for estimating median transmission loss with an associated standard deviation [ 121. Input parameters are antenna heights, frequency, distance, polarization and type of soil. The procedure differs according to whether the ratio of antenna height to wavelength is large or small, and for land mobile applications it may be necessary to make both calculations and use the larger of the resulting loss estimates. EPM-73 was developed with the intention of providing a model that would give, with minimum input information, reasonably accurate estimates of expected loss together with associated uncertainties. 14) Lee: The Lee model predicts point-to-point transmission by using two components [ 131. The first is an area-to-area path-loss prediction, while the second component gives pointto-point prediction for each distance. The constants provided are for 30 m base antenna height and 3 m mobile antenna height. Transmission loss expressions are different for different cities. A typical example is for Philadelphia where median transmission loss is given by

Many specific ideas presented in Tech. Note 101 are utilized in other models, most notably Longley-Rice and TIREM. See Section V for further discussion. 7) R-6602-LM: Calculation procedures for F(50, 50) and F(50, 10) field strength as a function of distance and base station antenna height are given in this report [ 7 ] . Mobile antenna height is 1.8 m, but an equation is given for adjusting heights within the range 1.8 to 9 m. This method was developed to assist rulemaking activities in Docket 18261 having to do with land mobile sharing with TV broadcast curves appearing in the FCC Rules. At distances involved in land mobile service, only the location variability is significant and the time variability can be considered as T = 100. 8) Longley-Rice Point-to-point Prediction: This model is in the form of a computer code for predicting long-term median transmission loss over irregular terrain [8]. The method is applicable for radio frequencies above 20 MHz. The point-to-point prediction procedure requires detailed terrain profiles. From the profiles, one must determine the distances to the respective radio horizon, the horizon elevation angles, and effective antenna heights. These distances, angles, and heights are then supplied as input to the computer program. The report [8] covers two distinct situations. In one of these, both ends of the radio path are known and fixed so that a terrain profile can be prepared. For such cases of point-topoint prediction, the computer program is basically an automated version of the procedures of Tech. Note 101 (see above). The computer programs have been revised several times. This model is discussed further in Section V. 9) Longley-Rice Area Prediction: For random paths in an area where the variations in terrain elevation are characterized statistically, this model predicts long-term median transmission loss [8]. Estimates of variability are provided as program output. These estimates are of the variability with respect to location and time, and an estimate is also given of the standard error of prediction. This report [8] introduces methods foi estimating effective antenna heights, horizon distances and angles. The estimates are made in terms of a terrain roughness factor, the heights of the antennas above ground, and the type of antenna siting (random or preferred). IO) Okumura: This model gives a method for predicting field strength and service area for a given terrain for frequencies in the range 150-2000 MHz, distance of 1 to 100 km, base station effective antenna heights of 30 to 1000 m, and receiver antenna heights typical of land mobile applications [9]. The basic median field strength curve for 900 MHz applies to an urban area. Correction factors are given for suburban, open and isolated mountain areas, rolling hills, sloping terrain and mixed land-sea paths. The model is based on measurements made in Tokyo, Japan, and surrounding suburbs at 200,453,922, 1310, 1430, and 1920 MHz. Statistical analysis of measurements was used to determine distance and frequency dependence of median field strength, location variability and antenna height gain factors. The urban curves with suburban correction factors seem to be most suited for cities in the U.S. 11) Terrain Integrated Rough Earth Model (TIREM):In

L = 142.3 - 20 log h + 36.8 log D

(dB).

Corrections for sloping terrain and a path obstruction are also given.

22

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 31, NO. I , FEBRUARY 1988

TABLE I SUMMARY OF MODEL FEATURES

1
L

I n p u t P a r a m e t e r s TTeated Antenna H e i g h t

Propagation Factors Included

output

I 1 1 1 I

i=Yi=

Egli

R-6602-LM

Bertoni & Walfisch

E N

N E
=

*R a t i n g

Scale: N = not treated; L = limited treatment; E

extensive treatment

It is difficult to know how to apply a given expression to an actual case, other than the three cities discussed. A mobile

where

antenna height of 3 m is in errors up to 6 dB. given to this parameter and may resulttwice the value usuallyThis model is A = 5 log [ ( f ) 2 + ( h b - h m ) z l described in detail in Appendix VI. - 9 log d + 2 0 log {tan- [ 2 ( h b - 2 h , ) / d ] } . IS) CCIR: Curves give field strength at 900 MHz frequencies for 50 percent of locations and 50 percent of the time in Here h, is the height of the mobile antenna, hb is average urban areas (to be used with caution in other areas) for mobile building height, and d is the center-to-center spacing of the antenna heights of 1.5 m and base antenna heights between 30 rows of buildings (typically one-half the narrow dimension of and lo00 m [ 141. To adjust for mobile antenna heights of 3 m the blocks). For example if d = 50 m, hb = 12 m and h, = instead of 1.5 m, a height-gain factor of 3 dB is suggested. 1.5 m, thenA = -9 dB. Standard deviations are given as a function of distance and In an alternate form to that given above, the model accounts terrain irregularity, and this information is to be used in for local terrain slope in the vicinity of the mobile. It does not, conjunction with the assumption that propagation variations however, incorporate terrain roughness factors or treat obwith location and time are characterized in decibel quantities structing terrain features, such as hills. The model is discussed by the Gaussian distribution. in detail in Appendix V. Treatment of the median transmission loss for frequencies near 900 MHz in this report is based on the work of Okumura B. Comparison of Factors Treated et al. 1 1 (also see Okumura model above). 9 16) Bertoni and Walfish: This theoretical model was The models listed above were developed from various developed specifically to predict the effect of buildings on the perspectives and with different original intentions. As a result, median transmission loss [15], [16]. It applies to those urban the models have for their inputs different environmental and and suburban environments where the buildings are of fairly path parameters, and account for different propagation factors. uniform height and are built in rows with small separation While all give median transmission loss as an output, some between neighboring buildings. For level terrain and well also give location variability and time fading information, but within the radio horizon, the median transmission loss at 900 none give loss deviation. A comparison of the extent to which MHz between half-wave dipoles is predicted to be the models make use of various input parameters and propagation factors, and the output information they give, is L = 147.2+A - 18 log ( h - hb) + 38 log D (dB) shown in Table I. The letter N is used to indicate that a
L 1

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

23
I 1

01

--I

MEASURED

ULL

I I I I I

i I I 1 1 ~ 3 0

1
Fig. 1.

10

20

50

100

DISTANCE (KM)
Comparison of measured median transmission loss for level terrain and open areas with predictions of the Okumura and Longley-Rice models.

parameter or factor is not treated, while limited or extensive treatments are indicated by L or E , respectively. Certain input parameters have significant influence on the accuracy of predicting median transmission loss by a given model. In particular, terrain data, including hill shape, building data and foilage data are individually and jointly indispensible for making accurate predictions. It should be noted that all of the models evaluated have failed to give extensive treatment to all three parameters. Even though two models give limited treatment of foliage data, such data is seldom available, nor in the proper format to use as input to a computer prediction program. While foliage always increases transmission loss, large isolated buildings and hills may either increase or decrease the transmission loss, depending on whether they block or provide an additional scattering path between the base and mobile antennas. When any one of the three types of input data cited above is neglected, the resulting error in the prediction of median transmission loss is usually unacceptable. Both the accuracy of predicting the output parameters, and the type of parameters given, determine the usefulness of the model for predicting service probability in a given area. The most important output parameters for determining service probability are median transmission loss, location variability and loss deviation. While all models give median transmission loss, and some give the location variability, none give an extensive treatment of the loss deviation. As a result, errors can be expected in determining the service probability, no matter which model is used.

C. Comparison of Median Transmission Loss Predict ions

The values of median transmission loss predicted by several of the models described above have been compared with each other, and with measurements, for various radials away from the Sears Tower in Chicago. This base station site offers

different terrain conditions ranging from the smooth lake surface to the highly irregular terrain South and West of the Sears Tower. Figs. 1-3 compare Longley-Rice, Okumura, slide rule, Carey, and TIREM models with measured median transmission loss data. Each value of median transmission loss shown is relative to free space transmission loss. The total median transmission loss for each distance can be found by adding the free space loss values to those given for each model. Fig. 1 shows five median transmission loss curves measured for different antenna heights. These curves are typical of open areas with smooth terrain. Starting from a point close to the base station, the median transmission loss initially decreases at a rate of about 6 dB per doubling of distance out to a distance of about 25 times the base antenna height h. This decrease is due to ground reflections and to the patterns of the vertically oriented dipole antennas. Beyond 25 times the base antenna height, there is an initial sharp increase of median transmission loss as the mode of propagation switches from that of free space to propagation over a plane earth. In the plane earth mode, median transmission loss increases at a constant 6 dB per doubling of distance out to about 60 km, where diffraction loss due to the earths curvature increases the slope to 25 dB per doubling of distance. These five median transmission loss curves, which have been smoothed to show average conditions, are useful as standards of comparison with predicted losses for the models shown on the three figures. The predictions of the Okumura open model for h = 380 m are also shown in Fig. 1. It is seen to give lower values of median transmission loss than measured for distances out to 60 km. Predictions obtained from the Longley-Rice model using terrain data for propagation over Lake Michigan are also shown in Fig. 1 for h = 380 m and 600 m. The resulting curves illustrate three serious problems in the Longley-Rice model. First, for distances less than about 55 km, the predicted median transmission loss is considerably less than would be

24

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 31, NO. I , FEBRUARY 1988

01

40

---LONGLEY-I C E R
OKUMURA SUBURBAN
I
I

50

.... ... F(95,95)

I I
I

F (50,50)
U E

60 .

SLIDE
I I I l l

RULE-"1

I
( KM)

I I I

10

20
DISTANCE

50

130 100

. 2. Comparison of the median transmission loss predicted by several models for propagation to the south of the Sears
Tower for base station antenna height of 380 m.

measured. Secondly, discontinuities occur at distances beyond 55 km, which are probably caused by changes in the mode of propagation assumed by the model. Finally, for large distances the slope of the Longley-Rice predictions increases to 55 dB per doubling of distance, which is 25 dB greater than that measured. It should be noted that the curves have been plotted for the F(95, 95) locatiodtime probability levels since the predictions for median probability F(50, 50) give even smaller transmission loss than the F(95, 95) predictions. Fig. 2 shows a comparison of median transmission loss relative to free space as predicted by five models for h = 380 m. The two Longley-Rice curves were obtained using terrain data for the Chicago area and averaging the results computed along seven radials located in a southerly direction from the

Sears Tower. The Okumura suburban model has been used to obtain the predictions shown in Fig. 2, rather than the urban model, which gives values of median transmission loss that are as much as 20 dB higher. The suburban model gives values closer to those measured in most U.S. cities. Predictions made by the G.E. slide rule closely follow the Okumura suburban curve. For distances between 5 and 65 km, the F(95, 95) Longley-Rice predictions are also close to the Okumura suburban model. On the other hand, the plane earth and F(50, 50) Longley-Rice models give predictions that differ by more than 25 dB from those of the Okumura suburban model. A comparison between TIREM, Carey, and measured data is shown in Fig. 3 . The measurements were taken for propagation between the Sears Tower and a mobile traveling south-

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

25
transmission over irregular terrain. A. Computer method- 1968, ESSA Tech. Rep. ERL 79-ITS 67, U.S. Government Printing Office, Washington, DC, July 1968. Y. Okumura et al., Field strength and its variability in VHF and UHF land-mobile radio service, Rev. Elec. Commun. Lab., vol. 16, pp. 825-873, 1968. Master Propagation System (MPS 11) Users Manual, U.S. Dep. Commerce, Nat. Telecommun. Inform. Serv., NTIS Accession no. PB 83-178624. (Computer program tape available from NTIS as PB173971.) Range and transmitter power calculator, General Electric, Mobile Radio Dep., Lynchburg, VA, 1977. M. Lustarten and J. Madison, An empirical propagation model (EPM-73), ZEEE Trans. Electromagn. Commun., vol. EMC-19, pp. 301-309, 1977. W. C. Y. Lee, Mobile Communications Engineering. New York: McGraw-Hill, 1982, ch. 3, 4. Methods and statistics for estimating field strength values in the land mobile services using the frequency range 30 MHz, CCIR Plenary Assembly, Geneva, Rep. 567-2, 1982. r151 H. L. Bertoni and J. Walfisch, A diffraction based theoretical model for prediction of UHF path loss in cities, in Proc. AGARD ConJ. Terrestrial Propagat. Characteristics in Modern Syst., Ottawa, 1986, pp. 8-1-8-9. J. Walfisch and H. L. Bertoni, A theoretical model of UHF propagation in urban environments, ZEEE Trans. Antennas Propagat., to be published.

ward in a sector centered on a bearing angle of 200. The TIREM results were obtained by averaging computations made using terrain data along seven radicals lying between the bearing angles 168.75 and 236.25. The TIREM model predictions of median path loss are seen to be in close agreement with the measured data, while the Carey median transmission loss values are on the average about 15 dB greater. These differences are caused in part by the gradual rise of the terrain south of Chicago. REFERENCES
K. Bullington, Radio propagation above 30 megacycles, Proc. IRE, vol. 35, pp. 1122-1136, 1947. , Radio propagation for vehicular communications, ZEEE Trans. Veh. Technol., VT-26, pp. 295-308, 1977. C. Burrows, Radio propagation over plane earth - field strength curves, Bell. Syst. Tech. J., 16, Jan. 1937. J. Egli, Radio propagation above 40 Mc over irregular terrain, Proc. IRE, vol. 45, pp. 1383-1391, 1957. R. Carey, Technical factors affecting the assignment of facilities in the domestic public land mobile radio service, FCC, Washington, DC, Rep. R-6406, 1964. P. L. Rice, A. G. Longley, K. A. Norton, and A. P. Barsis, Transmission loss predictions for tropospheric communication circuits, U.S. Government Printing Office, Washington, DC, NBS Tech. Note 101, issued May 1965; revised May 1966 and Jan. 1967. [71 J. Damelin, W. A. Daniel, H. Fine and G. Waldo, Development of UHF propagation curves for TV and FM broadcasting, FCC, Washington, DC, Rep. R-6602, 1966. A. G. Longley and P. L. Rice, Prediction of tropospheric radio

Some of these references are reprinted in Land-Mobile Communications Engineering, D. Bodson, G. F. McClure, and S. R. McCononghey, Eds. New York: IEEE Press, 1984. The table of contents for this book is listed in Appendix I.

26

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37, NO. 1, FEBRUARY 1988

V. Evaluation of Current Models

In this section we evaluate in detail several of the propagation models as they apply to 8001900 MHz land mobile communications. In order to evaluate how various propagation models perform, we will discuss predictions obtained in the following basic situations:
1) smooth terrain between base station and mobile;

2) smooth terrain with a knife-edge obstacle located close

to the mobile; and 3) smooth terrain with a knife-edge obstacle at the midpoint of the path. The case of smooth terrain between base station and mobile gives insight into how models handle the various propagation modes, such as line-of-sight, diffraction, and scatter, and how the models make transitions between the modes. Of particular interest are those models that employ detailed terrain path profiles and are sufficiently defined that they have been implemented as computer programs so that all engineering judgements surrounding the interpretation of terrain path profiles have been reduced to computer algorithms. Accordingly, these models can be interfaced to digital terrain data bases for predicting coverage on a point-to-point basis at various receive locations surrounding the base station. In studying the predictions of different models it is convenient to utilize the median transmission loss relative to the free-space transmission loss. The separation out of the free-space transmission loss removes the basic minimum transmission loss that is common to all models and leaves the specific loss that is predicted by each model. Thus, in order to obtain the total median transmission loss, the appropriate freespace attenuation must always be added to the results presented here. Two basic antenna heights will be considered, namely 500 ft (152 m) and 200 ft (61 m). In all cases the mobile antenna is assumed to be 6 ft (1.8 m) above street level. Both configurations will be evaluated over smooth-earth paths from 1 to 100 km. In addition, as appropriate, all results in this section are for: frequency of 900 MHz; vertical polarization; surface refractivity of 301 (N-units); a 413 effective earth radius; continental temperature climate; relative ground dielectric constant of 15; ground conductivity of 0.005 slm. As appropriate, the prediction of signal variabilities will also be discussed, in which case the predictions deal with corrections to be made to the basic median transmission loss predictions. In order to act as a reference point for the various models, results for the two basic antenna heights are presented in Figs. 1 and 2 for the Bullington smooth-earth model, and for the

1G

I00

DISTANCE (km)

Fig. 1. Median transmission loss relative to the free-space transmission loss for Bullington smooth earth and Okumura urban, suburban, quasi-open, and open at 900 MHz; HT = 500 ft (152 m); H R = 6 ft (1.8 m).

v)

4
z

40
53

I
U)

60

a a
70

IO
DISTANCE (km)

103

Fig. 2. Median transmission loss relative to the free-space transmission loss for Bullington smooth earth and Okumura urban, suburban, quasi-open, and open at 900 MHz; HT = 200 ft (61 m); H R = 6 ft (1.8 m).

Okumura model with all four environments (urban, suburban, quasi-open, and open).

A . National Bureau of Standards (NBS) Technical Note

IO1
NBS Tech. Note 101 [l] was first published in 1965, then revised in 1966 and again in 1967. The two-volume report describes detailed prediction methods for radio systems using propagation through the troposphere. It is intended for pointto-point transmission paths with frequencies from 40 MHz to 100 GHz over all kinds of terrain and long time intervals. Distances range from a few kilometers to hundreds of kilometers, with an emphasis on long radio links.

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

27 namely: effective antenna heights; horizon distances of the antennas; horizon elevation angles; the angular distance for a transhorizon path; terrain irregularity of the path. If a detailed terrain path profile is available, these pathspecific parameters can be determined from the specific terrain path profile. The resulting prediction is referred to as a point-to-point mode prediction. If a detailed terrain path profile is not available, this report gives techniques for estimating these path-specific parameters. The resulting prediction using these estimated parameters is referred to as an area mode prediction. The 1968 Longley-Rice Report [2] recognizes the importance of procedures for determining the effective heights of the transmitting and receiving antennas. On [2, p. 121 they state:
Further study of definitions of effective antenna height appears to be the most urgent requirement for improving these predictions for low antenna heights over irregular terrain. Different definitions of he1,2 may be found appropriate for lineof-sight and diffraction formulas.

The detailed prediction methods described in Tech. Note 101 utilize the path geometry of the terrain profile and the atmospheric refractivity of the troposphere to predict median transmission loss. For transmission paths within the radio horizon, the transmission loss is computed using geometric optics (ray) theory. Fresnel-Kirchhoff knife-edge diffraction is used for common horizon paths and is extended to predict the diffraction loss over isolated round obstacles. A modification of the Van der Pol-Bremmer method is utilized to compute the loss in the far diffraction region of double horizon paths whose distance is only slightly longer than the radio horizon. For longer distances, forward scatter theory is used to make troposcatter predictions of the median transmission loss. Tech. Note 101 describes these techniques in detail with many equations and figures. The report, however, does not contain any listing of a computer implementation of these techniques. Because there is no publicly available computer program that specifically implements Tech. Note 101, no results are presented here for comparison. Tech. Note 101 is included here because of the extent to which it is referenced in radio propagation literature. The significant contribution of Tech. Note 101 for 900 MHz propagation is the in-depth discussion of propagation fundamentals, particularly as they relate to radio transmission through the troposphere. Many of the specific ideas presented in Tech. Note 101 are utilized in other prediction methods. Most notable, the Longley-Rice computer method is a direct application of the Tech. Note 101 information. Also, various parts of TIREM are based on Tech. Note 101 concepts.

B. Longley-Rice Methods
In 1968, Longley and Rice published an ESSA Technical Report [2] presenting both the theory and an associated computer program for predicting long-term median transmission loss over irregular terrain for frequencies between 20 MHz and 10 GHz. Many of the equations used were derived from Tech. Note 101. This report presents methods for calculating the median transmission loss relative to the freespace transmission loss (often referred to as the reference attenuation, A,(dB)) from characteristics of the propagation path. The Longley-Rice model is sometimes referred to as the ITS irregular terrain model where ITS stands for the Institute for Telecommunication Sciences. For a particular transmission path, a series of parameters must be defined, namely: frequency; polarization; path length; antenna heights above ground; surface refractivity; effectuve earths radius; climate; ground conductivity; ground dielectric constant. In addition, a series of path-specific parameters are required,

On pages 3-1 and 3-2 of the report they define the effective antenna heights to be the elevations of the antennas above the dominant reflecting plane, and go on to discuss how the dominant reflecting plane might be determined. It is important to note, however, that this 1968 report does not give a computer program that implements the concepts they discuss for determining either the effective antenna heights or the terrain irregularity. The computer program they give in this report assumes that the effective antenna heights have already been determined as they are to be specified as input to the program. It is also important to note that this report does not describe how to compute the various types of signal variabilities, nor does it provide a computer program for calculating signal variabilities. No attempt will be made here to describe the basic LongleyRice procedure. The reader is referred to the original 1968 report [2] and to a more recent 1982 NTIA Report [3] that lists computer programs that determine effective antenna heights and the terrain irregularity parameter from terrain profiles. There are a series of computer program versions of the Longley-Rice model, which are described in the following quote taken from the 1982 report [3]:
The original Longley-Rice model was published in 1968. Shortly afterward a new version was developed which improved the formulation for the forward scatter prediction, and later the computer implementation was changed to improve its efficiency and increase the speed of operation. Since then, minor but important modificationshave been made in the line-of-sight calculations. To keep track of the various versions, most of which are presently being used at some facility, we have recently begun numbering them in serial fashion. Following the original (which might be called version 0 ) , here is a list of the more important versions, together with approximate dates when they

28
were first distributed:
1.0 January 1969 1.1 August 1971 1.2 March 1977 1.2.1 April 1979 2.0 May 1970 February 1972 2.1 September 1972 2.2

IEEE TRAhlSACTlONS ON VEHICULAR TECHNOLOGY, VOL. 31, NO. 1, FEBRUARY 1988

and the least square elevation values hxLl and hXL2 are determined. With these values, the effective antenna heights are then computed from h,,,,+DMIN (ground level at ant1,2, ~ x L I , L ~ ) (3) .

Version 1.2.1 corrects an error in version 1.2; it is the currently recommended version and is the one whose implementation is listed in Appendix A. The second series, beginning with version 2.0, used considerably modified diffraction calculations and tried to incorporate a groundwave at low frequencies. It is not now recommended and is no longer maintained by its developers. Subsequent to the 1982 report, a 1985 Memorandum [4] containing several modifications was issued. This memo refers to the newly altered model as version 1.2.2 and gives it the date of September 1984. All results presented here for the Longley-Rice model are based upon version 1.2.2. I ) Definition of Path Parameters: Because of the importance of effective antenna heights and terrain irregularity, we shall now discuss in detail exactly how they are determined in version 1.2.2 of the Longley-Rice model. Program QLRPFL first defines a range of interest for determining the effective antenna heights and terrain irregularity. The distance from is each antenna XLl,2 computed as follows:

To determine terrain irregularity, program DLTHX selects up to 25 points using linear interpolation on the path profile between XLl and XLzas a basis for fitting a straight line. The routine QTILE is then used to determine the interdecile range Ah(d) of terrain heights above and below this straight line. The terrain irregularity parameter is then computed by applying the following correction:

A h = A h ( d ) / [ l - 0 . 8 exp (-min (20, d / 5 x lo4))] (4)

where

d = (path distance - X L Z- XL1. )

(5)

The 1982 report [2] also contains a computer program called AVAR for computing the various types of signal variabilities. 2) Median Transmission Loss Predictions: The LongleyRice model and associated computer programs have two distinct parts that will be considered separately. The first part is the prediction of median transmission loss. The second part is the prediction of signal variabilities. The basic median transmission loss prediction techniques are described in the original 1968 Longley-Rice Report [ 2 ] .Between the original X~1,2=min (15hc1.2, O - ~ ~ L I J ) (1) 1968 report and the 1982 report there have also been several where hcl,zis the elevation of each antenna above ground level publications proposing modifications to the basic LongleyRice procedure to resolve difficulties that have been observed and dL1,2 the horizon distance of each antenna. is The range of interest starts at XLl from the transmitter and when comparing median transmission loss predictions to goes to XLz from the receiver. For a 6 ft (1.8 m) mobile actual measurements. A 1970 ESSA Tech. Report [5] proposes an equation that antenna, 15 hG2= 27.45 m, so that XLz = 27.45 m unless dLz is less than 274.5 m, which will not often be the case. For a is obtained empirically from a series of measurements for 500 ft (152 m) transmitting antenna, 15 hGl = 2,286 m, and calculatingthe reference attenuation for line-of-sightpaths and XLl = 2286 m, unless dLl is less than 22.86 km, which, if a revised procedure utilizing Fresnel-Kirchhoff knife-edge there is an obstacle, will often be the case. Once the range of diffraction along with an allowance for ground reflections for interest is determined, the determination of effective antenna calculating the reference attenuation for single-horizon paths. In an appendix to her 1978 OT Report [ 6 ] ,Longley states heights falls into one of the following situations. Line-Of-Sight: A least squares fit to the terrain profile is that The Longley-Rice model was originally developed for made over the range of interest XLl to XL2 and the least use with rather low antennas in irregular terrain, and was squares elevation values hxLl and hxL2are determined. Using tested against data from measurements made with low antenthese values, the effective antenna heights are then computed nas at frequencies from 20 to 100 MHz. Very few of these original test paths had terminals within radio line-of-sight. from Later tests against data obtained with higher antennas and at He1,2= hc1,2+ DMIN (ground level at antl,2, hxLI,2) (2) higher frequencies showed rather poor agreement with the line-of-sight area predictions. This section describes in where DMIN is the Fortran positive difference function that detail a series of changes to the original Longley-Rice model returns the first argument minus the minimum of the two to overcome these deficiencies. arguments. Using these effective elevations, new line-of-sight The 1978 OT Report [6] deals with radio propagation in distances, DL are computed. If the sum of the new DLis less urban areas and contains a correction factor to account for the than the total path distance, then He and DL have a urban environment. As described in this report [6, p. 311, correction applied to them. most of the data used in the formulation of the original Transhorizon: A least squares fit of the terrain profile is Longley-Rice model came from open areas, towns, and small made over the range X L l to 0.9 dL1for the transmitter, and cities. This report developed an allowance for the additional another least squares fit of the terrain is made over the range attenuation due to urban clutter near the receiving antenna. This allowance was developed by comparing results from the (path distance - o.9dL2)to XL2

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

29

w
W

2@

E
(D

40

5 t f f
a c
1

50

60

so

IO

GO

DlSTANCE(km1

DISTANCE(km)

Fig. 3. Median transmission loss relative to the free-space transmission loss for Longley-Rice over smooth earth at 900 MHz; HT = 500 ft (152 m); HR = 4, 6, 10, and 30 ft.

Fig. 4. Median transmission loss relative to the free-space transmission loss for Longley-Rice over smooth earth at 900 MHz; H r = 200 ft (61 m); HR = 4, 6, 10, and 30 ft.

1977 modified version of the Longley-Rice computer method with Okumuras curves. All comparisons utilized smooth earth. The Okumura predictions consistently showed additional power loss in urban areas compared to the LongleyRice predictions. Accordingly, an empirical equation was developed that gives the difference between the two models; this difference is referred to as the urban factor (UF). Longley states [6, pp. 32-33] that the Longley-Rice computer model, with the urban factor added, should adequately predict the median attenuation for moderately large cities in rather smooth terrain. In order to understand better how the Longley-Rice model functions in predicting median signal levels for 900 MHz mobile communication systems, a series of smooth-earth test cases have been run using version 1.2.2 of the Longley-Rice program in the point-to-point prediction mode. All test cases predict the median transmission loss relative to the free-space transmission loss. It should be noted that version 1.2.2 does not utilize either the corrections proposed in [5] or the urban factor proposed in 161. The first set of test cases we will consider are predictions of the median transmission loss relative to the free-space transmission loss over smooth-earth as a function of path length. Results obtained for a transmitting elevation of 500 ft (152 m) and receiver elevations of 4, 6, 10, and 30 ft are shown in Fig. 3. Similar results obtained for a transmitting elevation of 200 ft (61 m) and receiver elevations of 4, 6, 10, and 30 ft are given in Fig. 4. Examination of these results shows that the Longley-Rice point-to-pointprediction method gives a smooth monotonically increasing median transmission loss relative to the free-space transmission loss as the path distance over smooth earth is increased from 1 to 100 km. The transition to the scattering region for large path distances is evident in both figures-particularly in Fig. 4 for the 200 ft transmitting elevation. Comparison of Figs. 3 and 4 with the Okumura and Bullington results presented in Figs. 1 and 2 shows that Longley-Rice 6 ft predictions essentially agree with Bul

lington smooth-earth predictions for distances where the median transmission loss relative to the free-space transmission loss is small. The Longley-Rice predictions start to increase over the Bullington smooth-earth predictions as distance increases in order to provide a continuous transition into the diffraction region. We have also compared LongleyRice diffraction predictions with Bullington diffraction predictions and find agreement within the accuracy to which the Bullington nomograms can be read. Because of the difficulties in accurately reading the Bullington diffraction nomograms, results obtained from them are not shown in Figs. 3 and 4. Thus, Longley-Rice predictions and Bullington predictions essentially agree when the path loss mechanism is either totally smooth earth or totally diffraction. The Longley-Rice procedures provide a continuous transition between these two modes whereas Bullington procedures do not. Longley-Rice 6 ft predictions disagree with all four environmental modes of the Okumura smooth-earth predictions. The shape of the curves, as well as the values, is significantly different. The second set of test cases we will consider are point-topoint mode predictions of median transmission loss relative to the free-space transmission loss over smooth earth as single knife-edge obstacles are introduced into specific paths. These results give insight into how the Longley-Rice point-to-point prediction mode handles the introduction of obstacles into the Fresnel zone. Results obtained for a transmitting elevation of 500 ft (152 m), a receiver elevation of 6 ft (1.8 m), and a path distance of 12 km are shown in Fig. 5. A knife-edge obstacle is introduced at a distance of 1 km from the receiver and the height of the knife-edge obstacle is increased from 0 to 20 m. This figure shows that there is a 3.3 dB discontinuity in the median transmission loss at the precise point where the Longley-Rice model changes from the line-of-sight mode to the transhorizon mode. This occurs at the point where the knife-edge obstacle intersects the central ray connecting the transmit and receive antennas. Fig. 6 shows results obtained for the same case as above, except that the knife-edge obstacle is introduced at a distance

30
0

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37, NO. 1, FEBRUARY 1988

W E OF SYIHT

m
U
W

I w
2
0

&

TRANS HORRON

- 0

a 2 U
U)

U)

U w 1
laL a w
U) U)

w Y
4

a
w
IO

a
U) U)

0 U) 2 s a U) z
U

15

P 5
U) U)

t--+--5
6

20

OBSTACLE

2
4

a
IO
IO I /

a
12
13
14
I C

c
16

25
7

17

18

I O

KNIFE EDGE OBSTACLE HT (m)

KNIFE EDGE OBSTACLE HT (m)

Fig. 5 . Median transmission loss relative to the free-space transmission loss for Longley-Rice over a 12 km smooth earth path as a function of the height of a knife-edge obstacle located 1 km from a 6 ft (1.8 m) receiver at 900 MHz; Hr = 500 ft (152 m).

Fig. 7. Median transmission loss relative to the free-space transmission loss for Longley-Rice over a 58 km smooth earth path as a function of the height of knife-edge obstacle located at 1 km from a 6 ft (1.8 m) receiver at 900 MHz; Hr = 2000 ft (610 m).

c .

( 1 5

a
W w

U)

Id

a a w

lL

IS

50

55

60

65

70

75

80

10

KNIFE EDGE OBSTACLE HT (m) Fig. 6. Median transmission loss relative to the free-space transmission loss for Longley-Rice over a 12 km smooth earth path as a function of the height of a knife-edge obstacle at 6 km (the midpoint of the path) from a 6 ft (1.8 m) receiver at 900 MHz; Hr = 500 ft (152 m).

DISTANCE(km1

Fig. 8. Median transmission loss relative to the free-space transmission loss for Longley-Rice over a 74 km smooth earth path as a function of the height of a knife-edge obstacle located at I km from a 6 ft (t .8m) receiver at 900 MHz; HT = 2000 fi (610 m).

of 6 km from the receiver (the midpoint of the path) and the height of the knife-edge obstacle is increased from 0 to 80 m. This figure shows that there is a 3.8 dB discontinuity in the median transmission loss at the precise point where the Longley-Rice model changes from the line-of-sight mode to the transhorizon mode. As before, this occurs at the point where the knife-edge obstacle intersects the central ray connecting the transmit and receive antennas. In Fig. 7 we show results obtained for a transmitting elevation of 2000 ft (610 m), a receiver elevation of 6 ft (1.8 m), and a path distance of 58 km. A knife-edge obstacle is introduced at a distance of 1 km from the receiver and the height of the knife-edge obstacle is increased from 0 to 10 m. This figure shows that there is a 22.1 dB discontinuity at the

point where the knife-edge obstacle intersects the central ray connecting the transmit and receive antennas. Similar results are shown in Fig. 8 when the path distance is changed to 74 km. In this case there is a 17.3 dB discontinuity at the point where the knife-edge obstacle intersects the central ray connecting the transmit and receive antennas. The 2000 ft transmitter, 6 ft receiver, 58 km path case of Fig. 7 was also studied with midpath knife-edge obstacles. At the point where the knife-edge intersected the central ray connecting transmitting and receive antennas, the LongleyRice program returned an error flag indicating that internal calculations show parameters out of range. The program, however, went on to produce results and returned a discontinuity significantly larger than that shown in Fig. 7. Investiga-

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

31

tion of the range criteria in the Longley-Rice LPROP subroutine show that the distances between transmitted receiver and the nearest obstacle that defines the actual horizon of each antenna must be: 1) not less than 0.1 times the smoothearth line-of-sight transmitter/receiver distances; and 2) not greater than three times the smooth-earth line-of-sight transmitterheceiver distances. Some consequences of the foregoing conditions are seen for a 6 ft mobile antenna and a 500 ft base station antenna. The 6 ft antenna has a 5.6 km horizon over a smooth 413 radius earth. Thus, a horizon-defining obstacle closer than 0.56 km to the receiver or further than 16.8 km will violate the restrictions in the Longley-Rice program and return an error code. For the 500 ft antenna, the horizon is 50.9 km over a smooth 413 radius earth so that a horizon-defining obstacle closer than 5.1 km or further than 152.7 km will violate the restrictions in the Longley-Rice program and return an error code. Thus for the geometries of Figs. 5 and 6, any single knife-edge obstacle that defines the horizon of both the transmitter and receiver must occur in the region that is between 5.1 km from the transmitter and 0.56 km from the receiver. The knife-edge obstacles used in Figs. 5 and 6 satisfy these conditions. If the transmitting antenna is raised to 2000 ft, its horizon is 101.8 km for a smooth 4/3 radius earth. Thus, a horizondefining obstacle closer than 10.2 km or further than 305.4 km will violate the restrictions in the Longley-Rice program and return an error code. For a 6 ft receiver, any single knife-edge obstacle that defines the horizon of both the transmitter and receiver must occur in the region that is both between 10.2 and 305.4 km from the transmitter and also between 0.56 and 16.8 km from the receiver. While the knife-edge obstacle used in Fig. 7 satisfies these conditions, a knife-edge obstacle at the midpoint of a 58 km path does not. Similarly for the 74 km path of Fig. 8, the knife-edge obstacle satisfies these conditions. A knife-edge obstacle at the midpoint of a 74 km path does not satisfy both of these conditions. A third test case we will consider are point-to-point mode predictions of median transmission loss relative to the freespace transmission loss over smooth-earth as a function of distance behind a knife-edge obstacle. These results give further insight into how the Longley-Rice point-to-point prediction model handles single knife-edge obstacles. Fig. 9 shows results obtained for a transmitting elevation of 500 ft (1 52 m) and a receiver elevation of 6 ft (1.8 m). Here a knifeedge obstacle of height 80 m is introduced at a distance of 6 km from the transmitter. Median transmission loss relative to the free-space transmission loss is computed for the receiver located at distances from the transmitter ranging from 8 to 100 km. The median transmission loss relative to the free-space transmission loss without the single knife-edge obstacle is also plotted. As readily apparent from Fig. 9, there is a 4.1 dB discontinuity at the point where the central ray between the transmit and receive antennas is no longer blocked by the knife-edge obstruction. The differences between the predictions at distances greater than 20 km are due to the way the computer program computes the effective antenna heights and horizon angles. As is readily apparent from the preceding results, the

-0

Fig. 9. Median transmission loss relative to the free-space transmission loss for Longley-Rice over a smooth earth path as a function of the path distance for a knife-edge obstacle located at 6 km from a 500 ft (152 m) transmitter at 900 MHz; HR = 6 ft (1.8 m).

Longley-Rice point-to-point prediction model gives a discontinuous transition from the line-of-sight mode to the transhorizon mode when the transition is caused by single knife-edge obstacles. The Longley-Rice point-to-point prediction model does not provide for the increase in attenuation that would be expected as an obstacle penetrates the first Fresnel zone prior to the obstacle penetrating the central ray between the transmit and receive antennas. This failure to utilize first Fresnel zone penetration is significant as 900 MHz mobile propagation almost always has first Fresnel zone penetration near the mobile receiver. 3) Signal Variability Predictions: The original 1968 Longley-Rice Report [2] and associated computer programs do not describe how to compute signal variabilities. Several subsequent publications have treated location variability. A 1976 OT Report [7] develops equations for predicting location variability as a function of wavelength and the terrain irregularity parameter. A 1978 report [6] compared location variability results based on the 1976 report with results reported by Okumura and Egli and concluded that the 1976 equations predict more variability than that observed by Okumura in Japan but agree with the relationship shown by Egli in 1957. Finally, the 1982 report [3] contains a detailed discussion on Statistics and Variability and also includes computer code (as part of the Longley-Rice model, version 1.2.1) that implements the procedures discussed. The report (on page 28) specifically excludes the short-term or small displacement variability that is usually attributed to multipath propagation. Three basic types of variability are defined:

1) time variability-variations of local hourly medians on a specific path with time; 2) location variability-variations in long-term statistics that occur from path to path; 3) situation variability-variations in location variability that occur from situation to situation. The report and associated computer program define four

32

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37, NO. I , FEBRUARY 1988

Qc=0.84

different variability modes for combining these three basic types of variability, namely:

1) single message mode-time, location, and situation variability are combined together to give a confidence level; 2) individual mode-reliability is given by time availability, while confidence is a combination of location and situation variability; 3) mobile mode-reliability is a combination of time and location variability, and confidence is given by the situation variability; 4) broadcast mode-reliability is given by the two-fold statement of at least qrof the time in qLof the locations, with confidence given by the situation variability.
In addition, they provide an option whereby location variability is eliminated, as it should be when a well-engineeredpath is being treated in the point-to-point mode. A second option is also provided for eliminating direct situation variability, as it should be when considering interference problems. With this last option, it must be noted that there may still be a small residual situation variability. The procedures for predicting variability in the program AVAR depend significantly upon the type of climate specified: equatorial; continental subtropical; maritime subtropical; desert; continental temperate; maritime temperate over land; maritime temperate over sea. These prediction procedures also depend significantly upon the terrain irregularity parameter, whose determination is discussed in detail in Section 5.2 of [3]. On page 15, they give an equation that is used by their computer program to convert from the terrain irregularity parameter determined over a specified path distance to the asymptotic terrain irregularity parameter. It should be noted that version 1.2.2 of the Longley-Rice model does not utilize the location variability results published by Longley in 1976 [7]. In order to understand better how the Longley-Rice model functions in predicting signal variabilities for 900 MHz mobile communication systems, a series of test cases have been run using version 1.2.2 in the point-to-point prediction mode. All test cases predict the signal variability relative to the median transmission loss for a 500 ft (152 m) base antenna and a 6 ft (1.8 m) mobile antenna for path distances from 1 to 100 km. Local terrain irregularity parameter values of 0, 1, 5, 10, 50, and 100 m were utilized; unless otherwise noted, these pathspecific terrain irregularity values were corrected by the program to their asymptotic values. The first set of signal variability test cases consider predictions of the single message mode variability (also referred to as the confidence level variability) as a function of terrain irregularity and path length. Fig. 10 shows the standard deviation in decibels of the single message mode variability (as determined by setting Qc = 0.84). As readily apparent from
1

IO

1820

DISTANCE(km)

Fig. 10. Longley-Rice single message mode variability as a function of path lengthat900MHz;HT= 500ft(152m);HR = 6 f t ( 1 . 8 m ) ; A h = 0 , 1 , 5 , 10, and 50 m with asymptotic correction; Ah = 100 m curve virtually indistinguishable from 50 m curve.

Ah
om

10

GD

DISTANCE(km)

Fig. 11. Longley-Rice time variability as a function of path length at 900 MHz; HT = 500 f (152 m); H R = 6 f (1.8 m); 50 percent confidence t t level; Ah = 0, 1, 5 , 10, 50, and 100 m with asymptotic correction.

this figure, the standard deviation of the single message mode variability (confidence level) varies significantly (approximately 4 dB) as the terrain irregularity goes from 0 to 5 m, whereas the difference from 50 to 100 m is so small that the 100 m curve is not plotted as it is virtually indistinguishable from the 50 m curve. The second set of signal variability test cases consider predictions of the time variability as a function of terrain irregularity and path length. Fig. 11 shows the standard deviation in decibels of the time variability for a confidence level of 50 percent (as determined by setting Qt = 0.84; Q1= 0.5; Qc = 0.5 in the broadcast mode). As readily apparent from this figure, the standard deviation of the time variability does not vary as a function of terrain irregularity. Third, location variability predictions are considered as a function of terrain irregularity and path length. The standard

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

-2

33

(D

7
Om

yc

>
Y A ( 1

s
$.
Ah

1m

z 0
c
U 8 0 0
-1

om 1m 6m

em 1O m 6om

Om

12
1

-50m

IO
DISTANCE(km1

1oc

1 on

DISTANCE(km)

Fig. 12. Longley-Rice location variability as a function of path length at 900 MHz; H? = 500 ft (152 m); HR = 6 ft (1.8 m); 50 percent confidence level; Ah = 0, 1, 5 , 10, and 50 m, both with and without asymptotic correction; Ah = 100 m curve virtually indistinguishable from 50 m curve. Right margin gives the values of location variability determined from equations published by Longley in 1976.

Fig. 13.

Longley-Rice location variability as a function of path length at 900

MHz; Hr = 500 ft (152 m); HR = 6 ft (1.8 m); 9 0 percent confidence level; Ah = 0, 1, 5, 10, and 50 m with asymptotic correction; Ah = 100 m
curve virtually indistinguishable from 50 m curve.

deviation in decibels of the location variability for a confidence level of 50 percent (as determined by setting Q1 = 0.84; QI = 0.5; Qc = 0.5 in the broadcast mode) are shown in Fig. 12. In order to understand the significance of the difference between asymptotic terrain irregularity values and pathspecific terrain irregularity values, results were obtained for both types of terrain-irregularity values. Fig. 12 shows how location variability is impacted by terrain irregularity. The margin in Fig. 12 also shows the values of location variability determined from the equations published by Longley in 1976. As readily apparent from this figure, the standard deviation of the location variability depends significantlyupon whether the terrain irregularity parameter is the path-specific value or the 14 asymptotic value. Also, the location variability for a confidence level of 50 percent varies significantly (almost 9 dB) as DISTANCE (km) the terrain irregularity goes from 0 to 5 m, whereas the Fig. 14. Longley-Rice effective combined timehcation variability as a difference from 50 to 100 m is too small to plot. Fig. 13 shows function of path length at 900 MHz; HT = 500 ft (152 m); HR = 6 ft (1.8 that the standard deviation of the location variability for a m); 50 percent confidence level; Ah = 0, 1, 5, 10, and 50 m with asymptotic correction; Ah = 100 m curve virtually indistinguishable from confidence level of 90 percent (as determined by setting QI = 50 m curve. 0.84, QI = 0.5; Qc = 0.90 in the broadcast mode) also varies significantly (approximately 9 dB) as the terrain irregularity goes from 0 to 5 m, whereas the difference from 50 to 100 m is significantly (approximately 9 dB) as the terrain irregularity goes from 0 to 5 m, whereas the difference from 50 to 100 m is too small to plot. The fourth set of signal variability test cases consider the too small to plot. The fifth and final set of signal variability test cases we effective combined time/location variability as a function of terrain irregularity and path length. Fig. 14 shows that the consider are predictions of the mobile mode variability as a standard deviation in dB of the effective combined time/ function of terrain irregularity and path length. Fig. 16 shows location variability for a confidence level of 50 percent (as that the standard deviation in decibels of the mobile mode determined by setting Q, = 0.84; QI = 0.84; Qc = 0.5 in the variability for a confidence level of 50 percent (as determined broadcast mode) varies significantly (almost 9 dB) as the by setting Qr = 0.84; Qc = 0.5 in the mobile mode) varies terrain irregularity goes from 0 to 5 m, whereas the difference significantly (almost 9 dB) as the terrain irregularity goes from from 50 to 100 m is too small to plot. Fig. 15 shows that the 0 to 5 m, whereas the difference from 50 to 100 m is too small effective combined time/location variability for a confi- to plot. Fig. 17 shows that the mobile mode variability for a dence level of 90 percent (as determined by setting Q1 = 0.84; confidence level of 90 percent (as determined by setting Qr = Qt = 0.84; Qc = 0.90 in the broadcast mode) also varies 0.84; Qc = 0.90 in the mobile mode) also varies significantly

34

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37, NO. 1 , FEBRUARY 1988

(lh

om

5m 1Om SOm
I

IO

100
I

10

,30

DISTANCE L m )

DISTANCE(km)

Fig. 15. Longley-Rice "effective" combined time/location variability as a function of path length at 900 MHz; HT = 500 ft (152 m); HR = 6 ft (1.8 m); 90 percent confidence level; Ah = 0, 1, 5, 10, and 50 m with asymptotic correction; Ah = 100 m curve virtually indistinguishable from 50 m curve.

Fig. 17. Longley-Rice mobile mode variability as a function of path length at 900 MHz; HT = 500 ft (152 m); HR = 6 ft (1.8 m); 90 percent confidence level; Ah = 0, 1, 5, 10, and 50 m with asymptotic correction; Ah = 100 m curve virtually indistinguishable from 50 m curve.

tion of the base station antenna relative to ground level will usually lead to erroneous results. Digital terrain data bases do not provide accurate elevations at all points; rather they provide accurate elevations only at discrete data points and use interpolation procedures to provide elevations at other points. Thus ground elevations at other than the actual discrete data points are subject to error. If an antenna elevation is specified relative to ground level with the ground level being determined by interpolation from the digital terrain data base, the resulting antenna elevation will usually be in error. To ensure that the antenna have the proper elevation, either the transmitter height must always be specified in AMSL, or the actual ground elevation of the transmitter antenna must be input and utilized. The Longley-Rice point-to-point prediction model utilizes effective antenna heights as part of its computation of 1 > I00 transmission loss. These predictions are quite sensitive to the DISfANCHkm) actual value determinedhsed for the effective height of the Fig. 16. Longley-Rice mobile mode variability as a function of path length receiver. In mobile system, the receive antenna is low so that at 900 MHz; HT = 500 ft (152 m); HR = 6 ft (1.8 m); 50 percent any small change in the surrounding terrain can result in a confidence level; Ah = 0, 1, 5, 10, and 50 m with asymptotic correction; Ah = 100 m curve virtually indistinguishable from 50 m curve. rather large change in the effective height of the receive antenna and a corresponding sizeable change in the predicted median transmission loss. Depending upon the accuracy of the (approximately9 dB) as the terrain irregularity goes from 0 to digital terrain information utilized, sizeable variations in the 5 m, whereas the difference from 50 to 100 m is too small to effective height of the receiver can often be due to digital data plot. base factors, such as whether the elevation information is The preceding results show that the terrain irregularity rounded to the nearest 10 ft or nearest 10 m rather than due to value, if less than 10 m, can significantly impact signal actual physical situations. It should be noted that there are two variability predictions. It is important to note that signal versions of the National Geophysical and Solar-Terrestrial variabilities also depend upon the effective transmitting and Data Center (NGSDC) 30-s data base in use; the original receiving elevations. version had its elevations rounded to the nearest 10 m whereas 4) Digital Terrain Data: Several important matters must be the revised version has its elevations rounded to the nearest 10 noted concerning the use of the Longley-Rice point-to-point ft. It is our experience that significantly different results can prediction model with digital terrain data bases. The Longley- often be obtained for both median transmission loss and signal Rice point-to-pointprediction model expects both the transmit- variability predictions as a result of the difference between ter and receiver antenna elevations to be input relative to these two digital terrain data bases. ground level. While this is appropriate for a mobile, specificaIn addition to effective antenna height, small changes in
8 "

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

35
the form of time-sharing computer programs to interested agencies and organizations on a cost-reimbursable basis. As described in NTIAs Telecommunications Analysis Services Reference Guide [8], there are three different propagation analysis programs: CSPM-communication system performance model RAPIT-radio propagation over irregular terrain COVERAGE. The communication system performance model program is described in Section 10 of this reference as follows: The CSPM program creates detailed contour plots of basic transmission loss, field strength, power density, etc. from one or more transmitters in a given geographical area. It uses the ITS irregular terrain model (ITM) in the point-to-point mode for determining path loss along radials of about one degree azimuthal intervals around each transmitter. The ITM propagation model is applicable to analyze mobile, broadcast, or radar coverage and interference problems in the 20 MHz to 20 GHz band. ITM is the same propagation model that is used in both COVERAGE (section 6) and RAPIT (section 5) where it is discussed in more detail. The RAPIT program is described in Section 5 of this reference. There are two models in program RAPIT, the area model and the point-to-point model. The area predictions model is described in Section 5.3 as follows: Program RAPIT contains two different models (or algorithms) which are to be used under different circumstances. The first of these is the Area Prediction model designed by Longley and Rice (1968), and meant to be used when only moderate information about the propagation paths is known .... A recent ITS report (Hufford, et al., 1982) gives additional guidance on the use of the Area Prediction model. The point-to-point predictions model is described in Section 5.4 as follows:
The second propagation model is designed to be used on a fixed path where certain gross features of the intervening terrain profile are known. The model is known as the point-topoint mode of the irregular terrain model and is based upon the ESSA 70 model (Longley, et a/., 1971).

terrain elevations near the mobile receiver can intercept the central ray between the base transmitter and the mobile receiver and thus introduce a sizeable change in median transmission loss relative to the free-space transmission loss. Moreover, if the terrain obstacle is too close to the receiver, an error flag will indicate that program internal parameters are out of range. Depending upon the accuracy of the digital terrain information utilized, the presence or absence of an obstacle can often be due to digital data base factors rather than due to any actual physical situation. When a single horizon-defining obstacle is present, the elevation of the horizon-defining obstacle is excluded from the determination of effective antenna heights. In order for this to function properly in version 1.2.2, the path profile must utilize an increment that is somewhat finer than 0.1 times the shortest actual horizon distance. Otherwise, the elevation of the horizon-defining obstacle may impact the determination of the effective antenna heights and thus the resultant computation of the median transmission loss relative to the free-space transmission loss. The Longley-Rice model is also relatively sensitive to where horizon-defining obstacles are located. Accordingly, it is imperative that the error-flags built into the program be utilized. The program will return an error flag that indicates that internal parameters are out of range and still go on to return the median transmission loss. When computing mobile system coverage, i.e., coverage over a number of receive points, a sizeable number of situations will occur that are beyond the stated range of applicability of the model and thus give rise to error flags. Dependence upon median transmission loss values without monitoring the value of the built-in error-flag can easily be fraught with peril. The signal variability predictions are extremely sensitive to the terrain irregularity parameter when terrain irregularity is less than approximately 10 m. Depending upon the accuracy of the digital terrain information utilized, if the terrain is relatively smooth, sizeable variations in signal variability predictions can be expected-variations that are due to digital data base factors, such as whether the elevation information is rounded to the nearest 10 ft or nearest 10 m, rather than due to actual physical situations. Proper use of the model requires that median signal level predictions be compared against field strength measurements of median signal levels. Signal variability corrections should not be used as means for modifying model predictions before comparison with measured median signal level data. To use such a method for modifying the models predictions of median signal level totally defeats the usefulness/applicability of the model. 5) Significant Factors Not Considered The Longley-Rice prediction model does not contain any provisions for determining corrections due to local environmental factors near the mobile receiver or for predicting the effects of either buildings or foliage. Moreover, it does not contain any provisions for predicting the short-term variability (loss deviation) that is due to multipath propagation.

The COVERAGE program is described in Section 6 of this reference as a Telecommunications Analysis Services program which predicts the field strength coverage contours and populations within these contours for broadcast or base stations. Section 6.4 states: Three propagation models are available in COVERAGE. The first model is based on the FCC propagation curves for TV and FM broadcast stations. . . . The second and third models are based on the ITS irregular terrain (ITM) model, successor to the Longley-Rice model (Longley and Rice, 1968; Hufford, Longley and Kissick, 1982). The two models are applicable to ground-based systems from 20 MHz to 20 GHz. The second model is commonly referred to as the ITM in the area prediction mode and the third as the ITM in the point-to-point mode. The difference between the later two models is that the point-to-point model

C. NTIA Propagation Models

The National Telecommunications and Information Administration (NTIA) provides propagation analysis techniques in

36

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 31, NO. I , FEBRUARY 1988

requires detailed terrain information about the particular propagation path involved while the area prediction model requires less information. The foregoing quotes indicate that NTIAs model is a version of the Longley-Rice model. Regrettably, however, NTIAs documentation does not relate their version of the Longley-Rice model to the version numbers defined in [3]. Absent this information, and absent copies of the actual Fortran programs used by NTIA, it is not possible to comment in depth on NTIAs specific implementation of the Longley-Rice model. Longley-Rice methods have been covered extensively in this Section, and no further discussion will be presented. All concerns and findings previously presented in this section about Longley-Rice methods should also apply to the NTIA implementation of this model.

of the first Fresnel zone at that point r as depicted in Section 11, fig. 1. The three line-of-sight modes are:

h / r >, 1.5 h/r

< 0.5

0.5 < h / r < 1.5

Line-of-sight, free-space mode (mode 8). The ray is well above the terrain and free-space loss is used. Line-of-sight, rough-earth mode (mode 7). The ray is very near the earths surface, and an empirical, rough-earth formulation is used to compute the loss. Line-of-sight, transition mode (mod? 5). A weighted combination of the free-space and rough-earth losses is used.

D. The TIREMModel
TIREM is an acronym for terrain integrated rough earth model. It is one of a series of point-to-point propagation models in the Master Propagation System (MPSll) developed by the office of United States Department of Commerce, National Telecommunications and Information Administration, Annapolis, MD [9]. As part of this system, the model is referred to as TIREMll; the computer program is available from NTIS [9]. For propagation problems in the 40 MHz to 20 GHz band where the latitude and longitude of the transmitter and receiver are known, TIREMll is the preferred model in the master propagation system. The TIREM11 model is described in chapter 2 of the MPS 11 documentation [9]. Unfortunately, this documentation is rather sketchy and does not always agree with the computer program. According to comments in the computer program, the TIREM11 program is based on the propagation model TIREM used at the Electromagnetic Compatibility Analysis Center (ECAC). The computer program is still evolving, but until early in 1987, there were no version numbers output on runs; accordingly, it is very difficult to track the evolution of this program. In January 1987 version number 1.3 appears on TIREM output from a revised version wherein significant changes were made to beyond-the-horizon prediction techniques. All discussions here that refer to the computer code utilize a version obtained from NTIA during the fall of 1986. I ) TIREMll Propagation Modes: As described in chapter 2 of the MPS11 documentation, TIREM11 computes the median basic transmission loss in a series of steps. First, the terrain profile is examined and a series of basic parameters are determined, e.g., radio horizon distances, effective antenna heights, path angular distances, etc. Using the radio horizon distances, a determination is made whether the given path is within the horizon or beyond it. The ultimate selection of a propagation mode is based upon a number of additional parameters such as Fresnel zone clearance, etc. Chapter 2 of the MPSll documentation discusses 12 different TIREMll modes. Line-of-sight propagation modes: For line-of-sight paths, there are three TIREMll propagation modes. The selection of which mode to use is based on the minimum ratio along the entire path of the ray clearance above the terrain h to the width

Beyond line-of-sight propagation modes: For beyond line-of-sight paths, chapter 2 of the MPS11 documentation discusses nine different TIREMll modes, namely:
knife-edge diffraction, beyond line-of-sight (mode 2); rough-earth diffraction (mode 6); effective knife-edge diffraction (mode 1); effective knife-edgehough-earth diffraction (mode 4). This is a weighted combination of effective knife-edge diffraction and rough-earth diffraction; tropospheric scatter (mode 9); effective double knife-edge (mode 12); diffraction-scatter 1 (mode lo). This mode combines diffraction and troposcatter losses; diffraction-scatter 2 (mode 11). This mode combines diffraction and troposcatter losses with effective double knife-edge losses; diffraction-scatter 3 (mode 13). This mode combines troposcatter and effective double knife-edge losses.

2) TIREMI 1 Median Transmission Loss Predictions: Since line-of-sight propagation modes are of considerable importance, we will examine in detail exactly how TIREMll computes the median transmission loss for line-of-sight paths. As previously described, there are three different modes used for line-of-sight paths. For each of these modes, there are two parts to the median transmission loss-the free-space transmission loss and the median transmission loss relative to the freespace transmission loss. Let us first consider the line-of-sight, rough-earth case (mode 7). Study of the TIREMll computer code, in conjunction with the TIREMll documentation and references, shows that in this case TIREMll computes the median transmission loss relative to the free-space transmission loss utilizing either a version of the Longley-Rice method or an empirical equation described in [ 5 ] . Longley and Reasoner developed this expression to fit experimental communication link data and report that it gives excellent agreement with measured values for the 84 line-of-sight paths they studied. This empirical expression for the attenuation in decibels below free space, A,, is
A,=9[1 +exp (-O.OlAh)]
- 3.5 loglo(min h e l , Z / X+0.07d )

(6)

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

37

where d is the path length in kilometers, Ah is the terrain roughness in meters, and h is the wavelength in meters. The terms are the effective antenna heights in meters of each antenna along the great circle path between the antennas. As detailed in this report, page 5 1, all of these parameters are to be computed using methods described by Longley and Rice [2] and Rice et al. [l]. Study of the computer program shows that TIREM11 does not compute as defined by these reports, but rather computes each effective antenna height as the maximum of the structural height above terrain and the corrected height above the local average terrain, where the local average is computed over the first ten miles from each antenna (if d < 10 mi, the entire distance between the antennas is used). In contrast, Longley and Rice [2] define the effective antenna heights to be elevations above the dominant reflecting plane (or the structural heights, should they be greater), as determined by a least squares fit to terrain elevations that are visible to both antennas over the entire path. Study of the computer program shows that TIREM11 computes the terrain roughness parameter by determining the interdecile range of terrain heights, above and below a straight line fitted by least squares to elevations above sea level, over the entire path between the transmitter and receiver. It also uses the asymptotic correction for terrain roughness that was developed by Longley and Rice. Thus, TIREM11 uses the same techniques for determining terrain roughness as described by Longley and Rice [2]. It should be noted, however, that in version 1.2.2 of the Longley-Rice program, the determination of effective antenna heights and the terrain roughness parameter are somewhat different; they define a concept they call range of interest and use this concept to determine the range over which effective antenna heights and terrain roughness are determined. The logical branches of the TIREM11 computer program show that the TIREM11 program computes the median transmission loss relative to the free-space transmission loss as follows. If the path distance is greater than 1 km, the effective heights of both the transmitter and the receiver are less than 3000 m, and the frequency greater than 200 MHz, then the median transmission loss relative to the free-space transmission loss is computed from the Longley-Reasoner empirical equation. If the frequency is less than 150 MHz, then the median transmission loss relative to the free-space transmission loss is computed using a version of the Longley-Rice method. Finally, if the frequency is between 150 MHz and 200 MHz, then the median transmission loss relative to the freespace transmission loss is computed as a weighted average of the value determined from the Longley-Reasoner empirical equation and the value determined from the Longley-Rice method. 0.5 (mode 7), the For line-of-sight paths where h / r TIREM11 median transmission loss is the sum of the freespace transmission loss and the value determined above for the transmission loss relative to the free-space transmission loss. For 900 MHz land mobile communication systems, the TIREM11 median transmission loss is always determined by the Longley-Reasoner empirical equation. For line-of-sight

<

paths where 0.5 < h / r < 1.5 (mode 5), a weighted combination of the free-space transmission loss and the roughearth transmission loss determined above is used. For line-ofsight paths where h / r 2 1.5 (mode 8), only the free-space transmission loss is used. 3) Physical Consequences f o r 900 MHz Land Mobile Communication Systems: For 900 MHz land mobile communication systems, it is important to understand the consequences of TIREM 11s use of the Longley-Reasoner empirical equation and the above separation into three modes based on the parameter h / r . Since 900 MHz land mobile communication systems almost always have hel > he2,for line-of-sight conditions the base station antenna height is only used to determine h / r and, from this value, the regions where the 0.5, the various modes apply. Within mode 7, where h / r median transmission loss is determined solely by the effective height of the mobile receiver, the terrain roughness, and the path length. Changing the base station height, so long as the h / r remains less than 0.5, has no effect on the median transmission loss. As a result TIREMll has a sizeable line-ofsight region wherein changes to the transmitter height have absolutely no effect on the transmission loss, a result that is contrary to basic propagation theory. Mobile propagation measurements, e.g., Okumura [lo], show a very definite dependence on the height of the base station antenna. The Longley-Reasoner equation (6) has very little dependence on path length. Study of this equation shows that the path dependence is contained in only the last term, 0.07 d , where d is expressed in kilometers. Thus a change in path length from 10 to 20 km, assuming transmitter and receiver heights such that both paths have line-of-sight and h / r 0.5, only changes the median transmission loss relative to the freespace transmission loss by an additional 0.7 dB. For the same path change from 10 to 20 km, the free-space transmission loss increases by 6 dB; thus the total median transmission loss increase is 6.7 dB for this path change. This is considerably less than the change predicted by Bullington [ 111 smooth earth (12 dB) or that measured by Okumura [ 101. In order to understand better the Longley-Reasoner empirical equation, it is important to look at the data for the 84 paths they used as the basis for their empirical equation. As shown in Table 7 of their report [5], the minimum effective receive height of these 84 line-of-sight paths ranged from 1.5 to 286 m. Of particular relevance to land mobile communications, however, is the fact that only 10 percent of these 84 line-ofsight paths had effective receive heights less than 5.8 m, whereas 50 percent had effective receive heights greater than 38 m. Moreover, these paths covered a frequency range from 40 MHz to nearly 10 000 MHz. From the cumulative distribution information presented only 20 percent of the paths were in the frequency range from 516 to 1310 MHz. As readily apparent from this information, the primary focus of their work was for point-to-point communication paths. In such cases, the effective transmitter height may at times be less than the effective receive height; thus Longley-Reasoner utilized the minimum of hl or hZ.For point-to-point communication paths utilizing transmit and receive antennas wherein either may be closer to the ground, the effective elevation of

<

<

38

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37, NO. 1 , FEBRUARY 1988

H = 6

the antenna closest to the ground can reasonably be expected to be a more important parameter than the effective elevation of the taller antenna. Thus the Longley-Reasoner empirical equation did not utilize any statistically significant amount of data relevant to 900 MHz land mobile communication systems. Accordingly, use of this Longley-Reasoner equation for 900 MHz land mobile communication systems is a use far beyond the context for which the equation was developed. One consequence of this out-of-context use is the lack of dependence on the transmitter height noted above. The fact that the median transmission loss relative to the free-space transmission loss is dependent primarily on the effective receiver height for all cases where h / r < 0.5 is most significant. This means that a significant part of the entire impact of the path geometry is embodied in the determination of the numerical value of the effective height of the receiver. The balance of the impact is through the terrain roughness parameter. Thus, any change in the method of determining the f effective height o the receiver can significantly change the results. For example, the change in computation methods from those specified in the original Longley-Reasoner report to those actually used in the TIREMl 1 program does change the actual results. The impact of these changes on correlations between predictions and actual experimental data is unknown. Also, any modification that assumes a minimum effective receive height other than the actual structural height can have a significant impact on the results, as can the use of approximate conversion factors to convert the receiver height from feet to meters. Furthermore, depending upon the accuracy of the digital terrain information utilized, sizeable variations in the effective height of the receiver can often be due to digital data base factors, such as whether the elevation information is rounded to the nearest 10 ft or the nearest 10 m, rather than due to actual physical situations. 4) Illustrative TIREMI I Results: In order to understand better how the TIREM11 model functions in predicting median transmission loss for 900 MHz mobile communication systems, a series of smooth earth test cases have been run using program code made available by NTIA during the fall of 1986. The program code was adapted so that all test cases predict the median transmission loss relative to the free-space transmission loss. No corrections have been applied for signal variabilities; the prediction of signal variabilities will be discussed separately. The first set of test cases we will consider are predictions of median transmission loss relative to the free-space transmission loss over smooth earth as a function of path length. Fig. 18 shows results obtained for a transmitting elevation of 500 ft (152 m) and a receiver elevation of 6 ft (1.8 m). Results obtained for a transmitting elevation of 200 ft (61 m) and a receiver elevation of 6 ft (1.8 m) are shown in Fig. 19. Examination of these two curves shows that each has two sizeable discontinuities. As path length increases, the first discontinuity occurs when the TIREMI 1 program switches from the mode 7 computation using the Reasoner rough-earth equation to the mode 10 computation of diffraction-scatter. The next discontinuity occurs within the mode 10 diffractionscatter computations.

.8 ,i

; 53

DISTANCE(km)

Fig. 18. Median transmission loss relative to the free-space transmission loss for TIREM over smooth earth at 900 MHz; HT = 500 ft (152 m); H, = 6 f (1.8 m) t
H = 6

C I

IO

0 U U) Y 20 U 1

g
U)
U)

30
40

s z
I

9 U) Y
U)

50

z
U

60
70

10

1 GO

DISTANCE(km)

Fig. 19. Median transmission loss relative to the free-space transmission loss for TIREM over smooth earth at 900 MHz; HT = 200 f (61 m); HR = t 6 ft (1.8 m).

The next set of test cases we will consider are predictions of median transmission loss relative to the free-space transmission loss over smooth earth as a function of path length for a series of receive heights. Fig. 20 shows results obtained for a transmitting elevation of 500 ft (152 m) and receiver elevations of 4, 6, 10, and 30 ft. Similar results obtained for a transmitting elevation of 200 ft (61 m) are shown in Fig. 21. The curves in these two figures also contain the two sizeable discontinuities found in the previous two figures. Examination of the computer program shows that there is no attempt to blend the line-of-sight results into the beyond line-of-sight results. Accordingly, a discontinuity can be expected when the program switches from mode 7 to mode 10. In order to rule out program problems, the same smooth earth test cases were run by the NTIA program authors and the same two discontinuities were found. Test runs made by the NTIA

39
n o m
U

>o
Y-RICE
2G

U n
v)

a
Y

IREM

w 3u

v)

I 40 !
2

P
v)

50

U,
v)

I 60 z U a
70
I
1CJ

10

', 00

DISTANCE(km)

DISTANCEfkm)

Fig. 20. Median transmission loss relative to the free-space transmission loss for TIREM over smooth earth at 900 MHz; HT = 500 ft (152 m); H R = 4, 6, 10, and 30 A.
H

Fig. 22. Median transmission loss relative to the free-space transmission loss for TIREM, Longley-Rice, Bullington smooth earth, Okumura suburban, and Okumura quasi-open at 900 MHz; HT = 500 ft (152 m), HR = 6 ft (1.8 m).

'IREM

:o
1

i C l

10

100

DISTANCE(km1

DISTANCE(km1

Fig. 2 1. Median transmission loss relative to the free-space transmission loss for TIREM over smooth earth at 900 MHz; H r = 200 ft (61 m); HR = 4, 6 , 10, and 30 ft.

Fig. 23. Median transmission loss relative to the free-space transmission loss for TIREM, Longley-Rice, Bullington smooth earth, Okumura suburban, and Okumura quasi-open at 900 MHz; HT = 200 ft (61 m), H R = 6 ft (1.8 m).

program authors on TIREMll version number 1.3 showed the same discontinuity at the point of transition to the beyond lineof-sight region. Results in the beyond line-of-sight region, however, were somewhat different. Rather than the second discontinuity that existed previously, a series of oscillationlike discontinuities were found that started at the point of transition to the beyond line-of-sight region. The results of Figs. 18 and 19 can be directly compared with the same test results obtained for Okumura [lo], Longley-Rice, and Bullington [ 113. For easy comparison, the results for a 6 ft mobile receive antenna are shown in Figs. 22 and 23. Examination of these figures shows that TIREM11 predicts a sizeable region wherein the median transmission loss relative to the free-space transmission loss is nearly constant, followed by the two discontinuities discussed above. The second set of test cases we will consider are TIREM 11

predictions of median transmission loss relative to the freespace transmission loss over smooth earth as single knife-edge obstacles are introduced into specific paths. The results give insight into how TIREM 11 handles the introduction of single knife-edge obstacles into the Fresnel zone. Fig. 24 shows results obtained for a transmitting elevation of 500 ft (152 m), a receiver elevation of 6 ft (1.8 m), and a path distance of 12 km. A knife-edge obstacle is introduced at a distance of 6 km from the receiver (the midpoint of the path) and the height of the knife-edge obstacle is increased from 0 to 80 m. This figure shows that there is a 6.2 dB discontinuity in the median transmission loss at the precise point where the TIREMll model changes from the line-to-sight mode to the beyond lineof-sight diffractiodscatter mode. This occurs at the point where the knife-edge obstacle intersects the central ray connecting the transmit and receive antennas.

40
LINE OF SlQHT & T R A N S

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 31, NO. 1, FEBRUARY 1988

HORIZON

- 0

m
0

0
U) I O W W

a
W

a
U) 2 o

U)

s z
0

25

30

Y
U

I U) z

35
40
- r
i i

bn

65

70

75

80

KNIFE EDGE OBSTACLE HT(m)

Fig. 24. Median transmission loss relative to the free-space transmission loss for TIREM over a 12 km smooth earth path as a function of the height of a knife-edge obstacle located at 6 km (the midpoint of the path) from a 6 ft (1.8 m) receiver at 900 MHz; Hr = 500 f (152 m). t

m
U

0
5 13

0 W

a
U) W

Study of these results shows that the median transmission loss for a sizeable distance prior to the loss of line-of-sight is independent of the base station height and depends only upon the effective height of the mobile receiver and upon the path length. In this region, the increase in median transmission loss relative to the free-space transmission loss as a function of distance is very small. At the point where line-of-sight is lost due to the curvature of the earth, there is a sizeable and discontinuous increase in transmission loss. When obstacles are present, the TIREM11 model does not provide for the increase in transmission loss that would be expected as an obstacle penetrates the first Fresnel zone prior to the obstacle penetrating the central ray between the transmit and receive antennas. This failure to utilize first Fresnel zone penetration is significant as 900 MHz propagation almost always has first Fresnel zone penetration near the mobile receiver. 5) Signal Variability Predictions: The MPSl 1 documentation does not discuss the computation of signal variabilities. Also, the computer code of the MPSll Master Propagation System obtained from NTIA in the fall of 1986 does not contain routines to compute signal variabilities. 6) Signifcant Factors not Considered: The MPSll documentation and associated computer programs do not contain any provisions for determining corrections due to local environmental factors near the mobile receiver or for predicting the effects of either buildings or foliage. Moreover, there are no provisions for predicting the short-term variability (loss deviation) that is due to multipath propagation.

a a

U W
U)

:5

E. Okumura 3 Model
Full details of the Okumura technique for predicting the median transmission loss of land mobile radio systems are presented in [lo]. Okumuras paper presents the results of extensive propagation tests for land mobile radio service carried out in both the VHF and UHF bands over various situations of irregular terrain and environmental clutter. The results are then analyzed statistically to determine the distance and frequency dependences of the median field strength, location variabilities, and antenna height gain factors for the base and vehicular station in urban, suburban, quasi-open, and open areas over quasi-smooth terrain. Path-specific correction factors are also developed to consider irregular terrain, such as rolling hills, isolated mountains, general sloped terrain, and mixed land-sea paths. Okumura techniques are applicable for distances of 1 to 100 km and for base station effective antenna heights of 30 to lo00 m. Because of the way the Okumura propagation results were derived and analyzed, a number of significant points must be remembered when utilizing the Okumura procedures to predict median transmission losses. First, it is important to recognize that Okumura techniques are based on detailed propagation measurements for landmobile radio systems; they do not utilize theoretical models and/or theoretical predictions concerning how radio waves should behave. Second, Okumura techniques provide the basis for making not only area-to-area predictions, but also path-specific predictions. Many applications of Okumura overlook the pathspecific corrections. In order for comparisons of Okumura

20

U)

0 U) Y
I
U)

s z
z a
a
I-

25
30

35 401

!
55
00

51 3

65

70

75

80

KNIFE EDGE OBSTACLE HT (m)

Fig. 25. Median transmission loss relative to the free-space transmission loss for TIREM over a 30 km smooth earth path as a function of the height of a knife-edge obstacle located at 15 km (the midpoint of the path) from a 6 ft (1.8 m) receiver at 900 MHz; HT = 500 ft (152 m).

Fig. 25 shows results obtained for the same antenna heights, but for a path distance of 30 km. A knife-edge obstacle is introduced at a distance of 15 km from the receiver (the midpoint of the path) and the height of the knife-edge obstacle is increased from 0 to 80 m. Now there is a 12.3 dB discontinuity at the point where the knife-edge obstacle intersects the central ray connecting the transmit and receive antennas. As is readily apparent from the preceding results, the TIREM11 model provides a discontinuous transition from the line-of-sight region to the beyond line-of-sight region both when the transition i s caused by the curvature of the earth and when the transition is caused by single knife-edge obstacles.

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

41 In order to use these path-specific corrections on digital computers, a number of assumptions must necessarily be made and the resultant predictions would differ accordingly. The European Broadcasting Union has developed a very useful empirical way of applying path-specific corrections to median field strength prediction. This technique is described in the CCIR Study Group Recommendation 370-3, dated February 20, 1980. This method was adopted by CCIR and appears in the 1982 version of CCIR Report 239-5 (sec. 4.2, pp. 237-238). It defines and uses a terrain clearance angle together with an empirically developed correction factor to correct the median field strength prediction for the specific path under consideration. The terrain clearance angle is defined as the angle measured between the horizontal at the receiving antenna and the line which clears all obstacles within 16 km in the direction of the transmitter. This technique considers both negative angles (when the line to the obstacles is above the horizontal) and positive angles (when the line to the obstacles is below the horizontal). The empirical correction curve utilized by this method was developed from measurements of over 200 paths in the Federal Republic of Germany, Finland, France, and the United Kingdom. 2) Okumura Results re Theoretical Models: Walfisch and Bertoni [14] and Ikegami et al. [16] have developed models for predicting how buildings influence the signal at vehicular receivers. They both view 900 MHz land mobile propagation as taking place over the tops of the buildings, with diffraction at the rooftops being the mechanism by which the fields reach street level. The diffraction loss for such a process depends primarily on building height and the separation between the rows of buildings and secondarily on the shape and construction materials of the roof and upper building walls. The model of Walfisch and Bertoni [14] is discussed in detail in Appendix V. It considers the path loss to consist of three parts: the path loss between antennas in free space; reduction of the fields above the roof tops due to forward diffraction past many rows of buildings; diffraction of the roof top fields down to ground level. This theoretical model proposes a wave propagation mechanism to explain how the presence of buildings influences the distance dependence of the median transmission loss. The model applies only within the radio horizon of an elevated base station antenna. Within the radio horizon, the measurements of Okumura [ 101 show a distance dependence that varies with base station antenna height within the range of about 31 to 35 dB per decade [ 101, [ 121. The theoretical model also exhibits a variation of range dependence with antenna height, but the variation over typical ranges will be smaller and centered about 38 dB per decade. Measurements made in U.S. and English cities on flat terrain [13], [15] more typically suggest range dependence between 35 and 40 dB per decade. In the Okumura model, the buildings in the vicinity of the mobile are accounted for via a correction factor that corrects the basic urban model to suburban, quasi-open, and open environments. This factor can be over 20 dB, and there is currently no objective way of selecting its value based on the

with other point-to-point models to be valid, all applicable path-specific corrections must be applied before the comparisons are made. This is particularly true for those who use Hatas empirical formulas [ 121. Hatas empirical formulas were developed from Okumuras in order to make the use of Okumura procedures simpler (it is much easier to compute median transmission losses from equations rather than by reading curves). With computation simplicity, however, also come some restrictions. Most notable, Hatas equations are severely limited in the range of applicability-they consider only the frequency range 100-1500 MHz, distance 1-20 km, base station antenna height 30-200 m, and vehicular antenna height 1-10 m. In addition, Hatas empirical formulas do not consider any of the path-specific corrections. Third, Okumura techniques adopt as their standard propagation curves the propagation curves for the urban area environment. This was done not because the most typical propagation situation is urban, but because of computation considerations and the fact that the greatest degree of accuracy could be obtained if all other results were based on the use of the urban propagation curves as the standard. The most typical situation in the United States is far from the urban situation. Fourth, caution must be used when applying the Okumura definitions of environmental types to United States situations. Okumuras measurements are valid only for the building types found in Tokyo. Experience with comparable measurements in the United States has shown that the typical United States suburban situation is often somewhere between Okumuras suburban and open areas. Okumuras suburban definition is more representative of residential metropolitan areas with large groups of row houses. The CCIR Report 567-3 (1986) has adopted the Okumura urban curve as its basic curve for land mobile 900 MHz propagation. Accordingly, all of the above cautions apply when using these curves to predict median transmission loss. 1) Path-Specific Corrections: Okumuras path-specific corrections are not easily implemented for use on digital computers. A number of assumptions need to be made. One such implementation has been done by Allsebrook and Parsons [13]. They considered and introduced a set of definitions applicable to computer codes for the following quantities: effective base station antenna height; rolling hilly terrain correction; vehicular ground slope correction. It should be noted that the assumptions and definitions in their computer implementation are not necessarily unique; other definitions could be used and the predictions would differ accordingly. Okumura also includes path-specific corrections for the following additional effects: isolated mountain ridge (shadow loss); mixed land-sea paths; orientation of urban streets relative to transmitter location; effects of vehicular antenna height gain.

42

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37, NO. I . FEBRUARY 1988

measurable properties of the buildings near the mobile. Moreover, it is not clear how the correction factors obtained from measurements in Tokyo might apply in other cities with different building styles and construction materials. Okumuras classification of environmental areas is discrete. Moreover, it applies only to the location of the mobile unit. The path-loss slope is independent of the type of environment, except in those few close-in situations wherein the environmental correction would result in a path-loss less than the freespace loss. Several authors [17], [18] have expanded this concept to include a continuum of values through the introduction of a ground occupancy density parameter. Because the theoretical model of Walfisch and Bertoni and the work of Ikagami et al. [ 161 give a mechanism by which the buildings influence the signal at the mobile, it may be possible to use them to predict the environmental correction for the Okumura model from the physical properties of the buildings. Their mechanism makes use of building height and the separation between rows of buildings as the key parameters for determining the environmental correction factor. These parameters are significantly different from those used for a ground occupancy density factor [ 171, [ 181. 3) Comparison of Path-Specific Prediction Methods: Several recent articles have been published that compare land mobile path-specific prediction techniques. Aurand and Post [ 191 published a study that gives a practical comparison of 12 prediction techniques with an emphasis on three essential criteria to consider in the selection of a model: 1) type of terrain or geography covered by each method; 2) the form of prediction provided (transmission loss estimates, field strength contour maps, etc.); and 3 ) the implementation difficulty and degree of sophistication. This study compares the features of Okumura techniques to the 11 other models. They conclude that the typical accuracy that can be expected from various methods is determined primarily by the degree to which a method accounts for terrain characteristics. This study, however, does not compare predictions with measured data on specific paths. Delisle et al. [20] published a study that compared Okumura/Hata predictions with those of four other models. In addition, they made comparisons with experimental data they took in Ottawa at 910 MHz from a base transmitting antenna at 33.5 m AGL and a mobile antenna at 3.8 m AGL. They found general agreement between the various predictions provided that they were compared in the conditions where they are applicable and provided they are corrected for specific conditions-such as losses over hilly terrain or due to buildings or vegetation. Paunovic et al. [21] measured actual field strengths over a series of well defined path profiles over irregular terrain and compared them with the predictions of various models for VHF frequencies. This study also considered the clearance angle method developed by the European Broadcasting Union. Based on a comparison of predicted and experimentally determined field strength levels along different path profiles over irregular terrain, they found that the clearance angle method yielded the most reliable prediction values. They also concluded that the clearance angle method was very

operative, not complex, deprived of subjective factors, and thus was exceptionally well suited to computer applications. Accordingly, they chose the clearance angle method as the most suitable method for engineering land mobile communications systems. 4) Applicability: Okumura methods are uniquely different than most other propagation prediction techniques for land mobile communication systems. They are based on measurement data obtained from actual land mobile communication systems rather than on abstract theoretical models. Their use, however, requires that considerable engineering judgement be used, particularly in the selection of the appropriate environmental factor. More research is needed in order to be able to predict the environmental factor from the physical properties of the buildings surrounding a mobile receiver. In addition to the appropriate environmental factor, path-specific corrections must also be applied to convert Okumura median transmission loss predictions to predictions that apply to the specific path under study. Okumuras techniques for correcting for irregular terrain and other path-specific features require engineering interpretations and are thus not readily adaptable for computer use. In contrast, the European Broadcasting Unions Clearance Angle approach is a relatively simple approach for correcting for irregular terrain features; it is easily adaptable for computer use. Accordingly, clearance angle techniques provide a most useful way of determining the path-specific corrections required to make path-specific Okumura median transmission loss predictions.

F. Lees Model Lees model is described in detail in [22, ch. 21, and is also summarized in Appendix VI of this report. Review of these documents shows that this model is applicable only for mobile receivers when they are located within the radio horizon of the base station antenna. This model does not consider the additional losses that are encountered as the mobile unit approaches the radio horizon of the base station antenna. Accordingly, the various curves and examples presented in Lees book consider only distances out to 10 mi. Lees model has two components. The first component is an area-to-area path-loss prediction model. The second component uses the area-to-area path-loss prediction as a base and develops a point-to-point prediction therefrom. Three parameters are needed to make an area-to-area path-loss prediction:
the power at the 1-mi point; the path-loss slope; an adjustment factor. The power at the 1-mi point is the way Lees model takes into account the environment of the transmitting antenna. Lees model presumes that this parameter is determined empirically and no procedures are given for calculating it. He does give some representative values for different cities. The path-loss slope is the way Lees model takes into account the characteristics of the environment over which the signal is propagating. Typical ranges are from 30 dB/dec to 40 dB/dec, and he gives representative values to aid in the selection of this value. The path-loss slope is usually presumed

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

43

constant over the entire propagation path, but procedures are given for treating changes in the path characteristics through changes in the path-loss slope at specific points along the path. Lee states that Okumura found 30 dB/dec. Close study of Okumuras paper [lo], however, shows that this is an approximation. Okumura found 30 dB/dec up to distances of approximately 15 km; at greater distances, this value increased significantly. Okumura (fig. 12 of Okumuras paper) also found a dependence upon the effective height of the base station. For a 30 m antenna height he found approximately 34 dB/dec, while for a 250 m antenna height he found approximately 30 dB/dec, and for a 700 m antenna height he found approximately 26 dB/dec. It is important to note that the Okumura model has the same path-loss slope for all environments, urban, suburban, etc., provided the path-loss exceeds free-space, whereas Lees model takes this parameter to be a characteristic of the environment. The adjustment factor is the way Lees model accounts for a number of different factors-specifically : base-station antenna height (standard given was 30 m or 100 ft); mobile-unit antenna height (standard given was 3 m or 10

ft) ;
transmitter power (standard given was 10 W); base station antenna gain (standard given was 6 dB above dipole gain); mobile-unit antenna gain (standard given was 0 dB above dipole gain). The area-to-area prediction gives a median path loss, or signal level, over flat terrain in general. In order to apply this to specific paths, corrections for the nature of the actual path must be made. Lees model considers two specific types of situations: 1) nonobstructed paths; and 2 ) obstructed paths. Nonobstructed paths are defined to be those paths within the radio horizon that have no obstacles between the base station and the mobile unit. Lees model presents a technique for correcting the area-to-area prediction by an effective base station antenna height that is determined by using the slope of the terrain in the vicinity of the mobile unit. This technique is relatively easy to apply to specific paths when they are being analyzed by engineers. However, it is not readily amenable for computer analysis. Any realistic implementation of this concept as a general computer algorithm requires a number of assumptions to be made so that the resultant computer algorithm can realistically analyze all possible types of terrain situations that can be encountered. Obstructed paths are defined to be those paths that have a single terrain feature that obstructs the line-of-sight from the base station antenna to the mobile unit. Lees model utilizes knife-edge diffraction to compute the shadow loss from the obstacle, which is computed as if the basic propagation were free space. No corrections are made to take into account the reflections that occur in the foreground of both the base antenna and the mobile unit. Bullington has presented techniques for computing shadow losses that consider the reflections that occur when the antennas are relatively close to the ground; Lees model does not utilize these techniques.

Comparison with Okumura: Both Lees model and Okumura use empirical data to predict signal strengths. Lees model is restricted to distances well within the radio horizon of the base station antenna whereas Okumura applies out to distances of 100 km. Lees model explicitly considers the effects of the slope of the terrain in the vicinity of the mobile unit and also the effects of terrain obstacles. Okumura also has provisions for the treatment of a series of path-specific corrections, e.g., slope of the terrain in the vicinity of the mobile, mixed land-sea paths, isolated terrain obstacles. In order for a proper comparison to be made, it is necessary for all appropriate path-specific corrections to be made to both techniques. Okumura without path-specific corrections must never be compared to Lees model with its path-specific corrections. Both Lees model and Okumura consider the effects of the environment, but there is a very important difference in the way they do so. Lees model considers the effect of the environment surrounding the base station antenna in the determination of the power at the 1-mi point. The environment along the receive path is treated by the path-loss slope. Changes in the environment along the receive path are treated by introducing changes in the path-loss slope. It must be recognized, however, that in Lees model the overall impact of environmental changes along the path are relatively small. The model applies only over distances up to approximately 10 mil decade from the l-mi point. The changes in path-loss slope as a function of environment are relatively small (usually several decibels). Okumuras results, on the other hand, show no change in path-loss slope as a function of environment. Lees model does not consider the effect of the environment in which the mobile unit is actually located except through changes in the path-loss slope. Thus, Lees model does not predict significant changes in signal strength as the environment surrounding the mobile unit changes; the environment of the base station is the primary controlling factor. In contrast, Okumura utilizes the environment of the mobile unit to predict signal levels. Okumura has no provision for treating the environment surrounding the base station antenna. Also, there is no change in the path-loss slope as a function of environment. All environmental effects are determined solely by the environment surrounding the mobile unit. Thus, Okumura predicts significant changes in signal strength as the environment surrounding the mobile unit changes. This difference in treatment of the environment between Lees model and Okumura is significant only when the mobile unit is in an environment significantly different than the base station antenna. When the base station antenna and the mobile unit are in the same type of environment, this difference essentially disappears.
REFERENCES
P. L. Rice, A. G . Longley, K. A. Norton, and A. P. Barsis, Transmission loss predictions for tropospheric communication circuits, NBS Tech. Note 101; two volumes; issued May 7 , 1965; revised May 1 , 1966; revised Jan, 1967. A. G . Longley and P. L. Rice, Prediction of tropospheric radio transmission loss over irregular terrain; A computer method- 1968, ESSA Tech. Rep. ERL 79-ITS 67. G. A. Hufford, A. G . Longley, and W. A . Kissick, A guide to the use

44

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 31, NO. 1, FEBRUARY 1988

[4] [5] [6] [7] [8] [9] [IO]

[ll]

[I21 [13] [I41

[15]

of the ITS irregular terrain model in the area prediction mode, NTIA Rep. 82-100, Apr. 1982. G. A. Hufford, Memorandum to users of the ITS irregular terrain model, Jan. 30, 1985. A. G. Longley and P. K. Reasoner, Comparison of propagation measurements with predicted values in the 20 to 10,OOO MHz range, ESSA Tech. Rep. ERL 148-ITS 97, Jan. 1970. A. G. Longley, Radio propagation in urban areas, OT Rep. 78-144, Apr. 1978. , Local variability of transmission loss-land mobile and broadcast systems, OT Rep., May 1976. Telecommunications Analysis Services; Reference Guide, NTIA. Master Propagation System (MPSII); Users Manual. (Users Manual available from NTIS under No. NTIS-PB83-178624. Computer program tape available under No. NTIS-PB83- 173971.) Y.Okumura, E. Ohmori, T. Kawano, and K. Fukuda, Field strength and its variability in VHF and UHF land-mobile radio service, Rev. Elec. Commun. Lab., vol. 16, pp. 825-873, 1968. K. Bullington, Radio propagation variations at VHF and UHF, Proc. IRE, vol. 38, pp. 27-32, 1950. M. Hata, Empirical formula for propagation loss in mobile radio services, IEEE Trans. Veh. Technol., vol. VT-29, pp. 317-325, 1980. K. Allsebrook and J. D. Parsons, Mobile radio propagation in British cities at frequencies in the VHF and UHF bands, ZEEE Trans. Veh. Technol., vol. VT-26, pp. 313-322, 1977. J. Walfisch and H. L. Bertoni, A theoretical model of UHF propagation in urban environments, IEEE Trans. Antennas Propagat., to be published. G. D. Ott and A. Plitkins, Urban path-loss characteristics at 820 MHz, ZEEE Trans. Veh. Technol., vol. VT-27, pp. 189-197, 1978.

[16]

1171 [I81

~ 9 [20]

1211

[22]

F. Ikegami, S. Yoshida, T. Takeuchi, and M. Umehira, Propagation factors controlling mean field strength on urban streets, ZEEE Trans. Antennas Propagat., vol. AP-32, pp. 822-829, 1984. A. Akeyama, T. Nagatsu, Y. Ebine, Mobile radio propagation characteristics and radio zone design method in local cities, Rev. Elec. Commun.Lab., vol. 30, pp. 308-317, 1982. S. Kozono and K. Watanabe, Influence of environmental buildings on UHF land mobile radio propagation, ZEEE Trans. Commun.,vol. COM-25, pp. 1133-1 143, 1977. (Correction: IEEE Trans. Commun., vol. com-26, pp. 199-200, 1978.) J. 1 F. Aurand and R. E. Post, A comparison of prediction methods for 800 MHz mobile radio propagation, IEEE Trans. Veh. Technol., vol. VT-34, pp. 149-153, 1985. G. Delisle, J. P. Lefevre, M. Lecours, J. Y. Chouinard, Propagation loss prediction: A comparative study with application to the mobile radio channel, ZEEE Trans. Veh. Technol., vol. VT-34, pp. 86-96, 1985. D. S. Paunovic, Z. D. Stojanovic, and I. S . Stojanovic, Choice of a suitable method for the prediction of the field strength in planning land mobile radio systems, IEEE Trans. Veh. Technol., vol. VT-33, pp. 259-265, 1984. W. C. Y. Lee, Mobile Communications Design Fundamentals. Indianapolis, IN: Howard W. Sams and Co., 1986.

Some of these references are reprinted in Land-Mobile Communications Engineering, D. Bodson, G . F. McClure, and S. R. McConoughey, Eds. New York: IEEE Press, 1984. The table of contents of this book is listed in Appendix I.

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37, NO. 1, FEBRUARY 1988

45
U

VI. Measurement and Modeling: Recommendations

The following recommendations on measurement procedures in the 800/900 MHz band are offered here by the Committee on Radio Propagation. Factors affecting the choice of a propagation model for particular applications are discussed below.
-97
h

m
v

ij -100
v,

0
W

A . Measurement Recommendations
Mobile radio signals in the 800/900 MHz band are subject to rapid and large amplitude fluctuations [1]-[4]. These fluctuations have a greater variation than those experienced in point-to-point communication systems due to the fact that the mobile station antenna is located close to the ground and is usually in motion. By being located close to the ground, most of the received signals are reflected from surrounding objects. These multiple signals are caused by multipath propagation, and arrive at the antenna with different phases and amplitudes and thus interfere with one another. A small change in the location of the mobile antenna (such as a movement of a few inches) and/or any change in the surrounding environment can cause a large change in the received signal resulting from the combination of all reflected signals. Obviously, the character of the environment surrounding the mobile receive antenna can be expected to determine the character and depth of the resultant signal fluctuations [3], [5], [6]. Similarly, if the base antenna is surrounded by small reflectors nearby, or large reflectors at a distance, multipath signals can be generated in the vicinity of this antenna. In such situations, the depth of the resultant signal fluctuations can be significantly greater, depending on the size and distance to each reflector, than signal fluctuations that result from multipath only near the mobile antenna. Superimposed on these multipath signal fluctuations are variations that result from shadowing by terrain features or man-made propagation obstructions. I ) 900 MHz Mobile Signal Fluctuations: Fig. 1 shows a typical signal that a mobile receiver experiences at 900 MHz when the receiver moves a distance of 35 ft. If the vehicle is moving at a speed of 24 mi/h, the mobile receiver experiences the variations depicted in Fig. 1 within a time period of 1 s. Fig. 2 shows a typical signal that a mobile receiver experiences at 900 MHz when the receiver is stationary: the time period in this figure is 20 s. Mobile propagation measurements and theoretical predictions must take into account the rapid and large amplitude fluctuations depicted in these figures. It is not sufficient to characterize such rapidly varying signals by the median transmission loss alone; the variabilities of the loss must also be characterized. Accordingly, transmission loss measurements, if they are to produce meaningful results, must utilize measurement procedures that are designed to determine the characteristics of the signal level variabilities. As readily apparent from Fig. 1, a single average signal measurement made every 5 or 10 ft is not sufficient to characterize such signals. 2) Basic Signal Characteristics:In order to analyze mobile

B 2
m W

-103

-106

-109
0

1 4

21

25
nlSTANCE (FEET)

35

Fig. 1 .

A typical signal received by a mobile at 900 MHz when the receiver moves over a distance of 35 ft within a time period of 1 s.

10

15
T I M E (SEC)

20

Fig. 2.

A typical signal received by a mobile at 900 MHz when the receiver

is stationary within a time period of 20 s.

signals it is convenient to separate the observed signal variations into the following three main components: median transmission loss; location variability; loss deviation. As a mobile receiver moves through an area, the received signal fluctuates rapidly as a result of multipath propagation and wave interference. These rapid signal variations are often referred to as fast fading, multipath fading, or Rayleigh

46

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 31, NO. 1, FEBRUARY 1988

fading. The theoretical analysis of fast fading, however, is complicated by the fact that there is much evidence in the literature that shows that fast fading seldom follows true Rayleigh behavior. The most significant properties of fast fading can be characterized by determining the 50-90 percent cumulative fade depth (i.e., the loss deviation) of the instantaneous signal strength over small areas (sectors). For fast fading to be Rayleigh distributed, the cumulative small area distribution of the instantaneous signal, i.e., the transmission loss distribution, must be characterized by a Weibull distribution with a 50-90 percent fade depth of 8.2 dB. Assuming a Weibull distribution, fast fading is characterized by specifying the value of the loss deviation, i.e., the 5090 percent cumulative fade depth. Coupled with fast fading, there is also a second and much slower variation from sector to sector of the median signal within each sector as a result of shadowing andlor reflections by terrain features or man made propagation obstructions [4][ 6 ] .This slower variation is frequently referred to as location variability. In contrast to the fast fading situation wherein the distribution is not always Rayleigh, there is general agreement that this slower component of fading can be accurately modeled by a log-normal probability distribution [3]. This means that the logarithm of the local sector median signal, or sector transmission loss, varies in accordance with a normal probability distribution. Location variability is characterized by specifying the standard deviation of the log-normal distribution. The third component of propagation analysis between a transmitter and a mobile receiver is the determination of the dependence on distance from the base station of the median transmission loss, which is obtained by averaging the sector transmission loss over many sectors located at the same distance from the base station. 3) Signal Measurements: In order to obtain values for the basic propagation parameters described above, it is essential that signal measurements be taken in a manner that permits these basic parameters to be extracted with an appropriate level of accuracy. This means that sufficient measurements must be taken so that the variabilities of the signal can be determined. The basic measurement procedure must be capable of measuring, recording, and analyzing the actual instantaneous received signal and how it changes with time. In order to develop a procedure to measure the actual instantaneous received signal, it is necessary to account for the effects of modulation. For the FM modulation techniques utilized in 900 MHz cellular system, energy is taken from the primary carrier and dispersed into sidebands. In order to make accurate measurements, it is necessary that:

the total power radiated by modulated signals from such transmitters varies with the characteristics of the signal used to modulate the carrier. Thus, modulated signals do not meet the criteria that the power radiated by the transmitting antenna be known and fixed during the interval when signal strength measurements are being made. This criteria can be met only by making measurements on an unmodulated carrier. The second requirement is that the measurement equipment must be designed to capture all of the energy impinging on the receiver antenna from the desired signal; this means that when the desired signal is modulated the measuring equipment must have sufficient bandwidth to capture the signal and all of its sidebands without introducing any sideband attenuation from the bandwidth of the measuring equipment. Typical cellular receivers do not have sufficient bandwidth to meet this requirement; their bandwidth, in order to reject adjacent channels, is sufficiently narrow so that they introduce distortions from their attenuation of the various sidebands. Accordingly, typical cellular receivers should not be used when measuring modulated signals. Wide-band receivers do exist, but the use of these special wide-band receivers introduces a series of complexities. First, they can be used only in environments where there are no other signals from adjacent transmitters, etc., within the bandwidth of the wide-band receiver. Second, wide-band receivers have noise levels higher than narrow-band receivers; thus the minimum signal that can be accurately measured with a wide-band receiver is significantlyhigher than possible with a narrow-band receiver. The problems inherent in making signal measurements on modulated mobile signals can be overcome by using unmodulated signals. When an unmodulated carrier is utilized, all power radiated is in the carrier. The problem of transmitter attenuation of sidebands disappears; the radiated signal power can be determined and held constant during the interval when measurements are being made. The bandwidth of the receiving equipment can be selected through the use of narrow-band filters to be considerably narrower than conventional receivers in order to reject stray signals and reduce the amount of noise present. In fact, modern filter technology is such that receiver and transmitter frequency stability (i.e., the drift of the basic carrier frequency) often controls the selection of the narrowest band filter that can be used. The only complexity with using unmodulated signals is the operational complexity of maintaining a continuous unmodulated test signal on an operating cellular system for the duration measurements. Fig. 3 shows a block diagram of an experimental arrangement based on the assumption that measurements are to be made on an unmodulated signal. Accordingly, a calibrated receiver and narrow-band filters are used to reduce both ambient noise and reject and adjacent channel signals. 1) the power radiated by the transmitting antenna be known The experimental arrangement depicted in Fig. 3 has and fixed during the interval when measurements are been implemented using a sophisticated state-of-the-art combeing made; puter controlled spectrum analyzer as the calibrated receiver. 2) the measurement procedure must be designed to capture The unit utilized, in addition to being a calibrated receiver, has all of the energy impinging on the received antenna from integral AID conversion, has programmable filters down to a the desired signal. 30 Hz bandwidth, and has a transient response capable of Transmitters utilized in mobile communications, under FCC capturing all variations of the received signal. The narrower Rules and Regulations, must be designed to provide sufficient bandwidth filters cannot be used at speeds of the mobile attenuation of the sideband signals they radiate. Accordingly, receiver greater than about 5 mi/h. The individual blocks of

SPECIAL ISSUE ON MOBILE RADIO PROPAGATIOK

47

BASE TPANSMITTER

WATTMETER LNMODULATCD CARRIER

NOBILE TEST EQUIPMEST'

Fig. 3 .

Block diagrani of signal strength mobile measurement equipment.

the mobile measurement equipment are under control of the data acquisition computer while segments of the instantaneous signal are captured, digitized, and magnetically recorded. A printer and video terminal are attached to the data acquisition computer that controls the analyzer; these devices give the operator real-time information about the signal being measured. The actual digitized waveform segments are recorded magnetically for subsequent analysis on a separate large computer. The computer controlling the equipment is programmed so that for each sector being measured, the entire instantaneous waveform is recorded with a resolution of approximately ten points per fine-structure/fast-fade variation. For moving measurements, each such sector waveform segment usually correspond to a distance of approximately 35 ft. Each such sector waveform is analyzed by computer to determine its median signal level and loss deviation. A series of sector measurements is continually taken by the measurement equipment under the control of the data acquisition computer as the vehicle moves throughout the area under study. Depending upon the path taken, different results can be obtained. There are two basic types of routes. Routes can be designed to obtain: 1) measurements along defined roadways; or 2 ) the location variability of the signal. Both cases gives information about the median, 90 percent level and lossdeviation characteristics of the signal along the route driven. In order to obtain location variability information, the route utilized must be designed so that multiple series (typically 100 or more) of sector measurements are taken in various small areas wherein the distance and direction from the transmitter are such that the median signal would be expected to be the same. When such routes are utilized, the series of consecutive sector measurements are computer analyzed to determine, in addition to the median signal and associated loss deviation characteristics, the location variability. Figs. 4 and 5 are photographs of field measurement vehicle

Fig 4 Photograph of d field measurement vehicle using the signal strength mobile measurement equipment shown in the block diagram of Fig 3 Photograph courtesy of Comp Comm, Inc . Collingqwood, NJ

Fig. 5 . Photograph o f the te\t equipment used iii the vehicle shown in Fig. 4. Photograph courtesy of Coiiip Comm. Inc.. Collingswood. NJ.

48

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37, NO. I , FEBRUARY 1988

Fig. 6. Photograph of the test equipment when used to make in-building signal strength measurements. Photograph courtesy of Comp Comm, Inc., Collingswood, NJ

99.9

95 90 80

50

20

10
5
- --

-t
. .
.
! -

.. .-

. .

I .5

4 I

6
I

-97,4

-101

-105 0
I

-108.8

-108,8

-105 0
I

-101 I 2
SIGNAL

-97,4

SIGNAL
Fig. 7.

(DBM)

(DBM)

(a) (b) Curnulativc log-normal plot (a) and ~ u i i i ~ l a t i v e Weibull plot (b) of the 500 point waveform segment shown in Fig. I . Supcriniposed on each plot is a linear least squares fit of all data points between 5 and 95 percent.

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

49

99,991
I

1
I

1
I

1
I

1
I

I
I

I
I

I
I

I
I I

99 95

90
80

50

20

10

5
1

I /

I / I

i
I

._
I

I
I

0IOl-C

1.5 I
I

2 1 I
1

a
I

k
I

-94 68
I

-99 53
I

-104 39 SIGNAL
I

-109 24

-109 24

-104.39

(DBM)

-99 53 SIGNAL
I

-94 68
I

(DBM)

(a) (b) Fig. 8. Cumulative log-normal plot (a) and cumulative Weibull plot (b) of the 500 point waveform segment shown in Fig. 2 Superimposed on each plot is a linear least squares fit of all data points between 5 and 95 percent.

and test equipment. As apparent from these photographs, all measurements are made from antennas mounted on a standard automobile. This use of a standard sedan is significant as it permits the receive antennas to be mounted in a manner identical to those in regular use (e.g., roof mount, trunk mount, etc.) rather than being mounted 3 ft or more higher on the top of a van. Technical literature shows that measurements made from antennas on such vans can easily produce signals with medians approximately 3 dB greater than measurements made from antennas on standard sedans. Most important, the loss deviation characteristics can also be different for different height receive antennas; sufficient information does not exist to permit any correlation of signal variabilities determined from van-height antennas to these determined from standard sedan-height antennas. Accordingly, it is most important that receive antennas for measurement purposes be located at the same elevation relative to the ground at which they will ultimately be used. The equipment configuration used is highly portable and independent of the measurement vehicle; thus as shown in Fig. 6 it can also easily be utilized to make in-building measurements. The series of figures discussed subsequently show some of

the types of analysis performed on the waveform segments recorded by this field measurement equipment. Fig. 7 shows cumulative log-normal and cumulative Weibull plots of the waveform segments shown in Fig. 1. This waveform segment is comprised of 500 digitized signal values. Superimposed on each plot is a linear least squares fit of all data points between 5 and 95 percent. As apparent from Fig. 7, the fast-fade variabilities of this waveform segment are better represented by a Weibull distribution (with a 50-90 percent loss deviation of 4.14 dB) than by a log-normal distribution. Fig. 8 shows cumulative log-normal and cumulative Weibull plots of the waveform segments shown in Fig. 2, which is composed of 500 digitized signal values. Superimposed on each plot is a linear least squares fit of all data points between 5 and 95 percent. As apparent from Fig. 8, the fastfade variabilities of this waveform segment are better represented by a log-normal distribution than by a Weibull distribution. Fig. 9 shows a plot of the mean power, 50 percent value, 75 percent value, and 90 percent value of the signals from a consecutive series of waveform segments (herein referred to as sweeps). Each waveform segment consisted of 500 digitized values plotted on this curve for the specified segment (sweep).

50

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 31, NO. 1, FEBRUARY 1988

-96 -98

; -100

f -102
-104
106

99
98
nc

-108
-110

-3J

90
I
10

-11211

1
30

I
40

I
50

I
SHEEP

1
60 NUMBER

1
70

20

1 1
80

80

Fig 9. A plot of the mean power, 50 value, 75 value, and 90 percent value of signal strength from a consecutive series of 500 point waveform segments.

50

20

10
5

1
SWEEP

NUMBER

Fig. 10. A plot of the actual 50-90 percent loss deviation values versus sweep number for the measurements presented in Fig. 8.

7-54

5,34

1 0 3,14 0,94 50% - 90% Loss D E V I A T I O N

This is a convenient way of representing how waveform characteristics vary with position along the route driven. Each sweep number corresponds to a specific sector measurement over a distance of approximately 35 ft (the location of each sector measurement is known by a separate recording that correlates position on a map to sweep number). Fig. 10 shows a plot of the actual 50-90 percent loss deviation values versus sweep number for the measurements presented in Fig. 8. Even a cursory examination of this information shows that the fast-fade characteristics of this signal are nowhere close to obeying a Rayleigh distribution. The observed values of the 50-90 percent loss deviation vary from a low of approximately 1.1 dB to a high of approximately 4.3 dB. These values are significantly less than the 8.2 dB value required by the Rayleigh distribution. Fig. 11 shows a cumulative log-normal plot of the fastfade (50-90 percent loss deviation) for each of 72 consecutive waveform segments, each of which is comprised of 500 digitized signal values. Superimposed on this plot is a linear least squares fit of all data points between 5 and 95 percent. As apparent from Fig. 1 1, the linear least squares fit to the fastfade (50-90 percent loss deviation) variability of these 72 consecutive waveform segments has a mean value of 4.32 dB and a standard deviation of 1.39 dB. Fig. 12 shows a cumulative log-normal plot of the actual median signal determined from a series of 72 consecutive sector measurements. Each waveform segment in each sector measurement is comprised of 500 digitized signal values. Superimposed on this plot is a linear least squares fit to all data

Fig. 11. A cumulative log-normal plot of the 50-90 percent loss deviation for each of 72 consecutive 500 point waveform segments. Superimposed on this plot is a linear least squares fit of all data points between 5 and 95 percent. This cumulative log-normal plot shows the variability of the lossdeviation.

points between 5 and 95 percent. As apparent from Fig. 12, the location variability of the medians of these 72 consecutive waveform segments can be represented by a log-normal distribution with a mean of - 100.07 dBm and a standard deviation, i.e., location variability, of 4.46 dB.

B. Factors Affecting Modeling Choices

Various factors influence the choice of a propagation model. A simple and obvious objective is to choose the most reliable model that can be afforded. However, this objective involves complex considerations such as accuracy, ease of use, generality, and availability and cost of input data, which taken together may not lead to sharply defined selection criteria. I ) Applicability: Mobile radio planning and design activities, which rely on a knowledge of signal propagation behavior, are carried out for a variety of purposes, each of which places different requirements on the propagation model. Model selection should consider the requirements of these diverse applications, including: policy development, broad application; regulation, national application, including frequency allocation;

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

51
I I

I
99,9 99 98 95

20

10

5
1
0.1 0
-86.58

O
-93 51
I

j
-100 I44 MEDIAN SIGNAL

I -107 38
I

(DBM)

Fig. 12. A cumulative log-normal plot of the actual median signal levels determined from each of 72 consecutive 500-point small-sector measurements. Superimposed on this plot is a linear least squares fit of all data points between 5 and 95 percent. This cumulative log-normal plot shows the location variability of the small-sector measurements.

regulation for specific application; frequency coordination and frequency allotment; licensing and application review; system design of general applicability; system design for specific applications. Broadly applicable propagation models are needed for policy development, which includes issues such as the best use for a new frequency band, the best modulation for a new service, or the total capacity for a proposed system or class of systems. Typical regulations f o r national application are those prescribing a standard distance beyond which frequencies may be reused. Such regulations may prescribe the method for computing propagation loss. Regulations f o r specific application may, for example, involve a narrow range of frequencies, or may consider limited geographic areas, such as the protection of national radio quiet zones or international borders. In such cases, special purpose propagation models may be required with characteristics unique to special circumstances. Frequency coordination and allotment require interference considerations, which involve phenomena that have a low probability of occurrence, as is typically encountered for

longer ranges. More reliable methods for predicting interference will be needed as increased demand produces mobile/ broadcast sharing and the introduction of new spectrum efficient technologies. Simplicity, or standardization combined with automation, is important for propagation models used in licensing and application review functions of regulatory agencies, so as to permit consistent predictions by both the applicants and the reviewing agency. Propagation models that apply to a wide variety of locations, but in a limited frequency band and for limited distances, are needed for generalsystem design, such as when systems are being developed that will operate in many locations. When a given performance objective is to be met in a known location, the specific system design requires a propagation model that accounts for relevant environmental and topographical information. 2) Model Basis and Form: The basis for a model may be either theoretical or empirical, or a combination of these two. Theoretical propagation models allow a recognition of the fundamental relationships that apply over a broad range of circumstances. They also allow definition of relationships that exist among any combination of input parameters. Empirical models are derived from measurements and observation, and offer a major advantage in that all environmental influences are implicit in the result regardless of whether or not they can be separately recognized and theoretically studied. Empirical models offer the opportunity to provide probabilistic descriptions of the propagation phenomena. The validity of empirical models is limited not only by the accuracy with which individual measurements are made, but also by the extent to which the environment of the measurements adequately represents the physical environment in which the model is to be applied. Whatever the basis for the model, it may take one of several forms, as discussed below. a) Analytical models, consisting of mathematical expressions and explicit procedures, such as analytic fit by Hata [7] to the measurements in Tokyo, usually allow greater accuracy, precision, breadth of application, and can account for the interaction of many variables. Analytic models give consistent results when applied by different users, and are readily implemented as computer programs. b) Graphical models, such as the Carey curves [8] and Bullington nomograms [9], allow problem solution without the need for computational machinery (or computational skills). Thus, they can be applied by less skilled personnel and can provide ad hoc estimates. However, graphical methods are limited in practice as to the number, range, and interrelationships of input variables than can be represented, and are of limited accuracy. c) Slide rules, such as those developed at General Electric, are accurate, portable and useful for many problem solutions. They can be used for rapid prediction of many different system parameters such as required base antenna height, percentage of area covered, and required transmitter power. Slide rules can be based on either average or specific topographic conditions. d) Computer programs lend themselves to automated applications, including the production of charts, graphs, or

52

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37, NO. 1, FEBRUARY 1988

nomographs for manual applications, and can provide an excellent design tool for many applications [ 101, [111. However, the computer model will be no better than the underlying phy sical/mathematical model and may, if poorly designed, be substantially poorer. 3) Accuracy and Validity: Accuracy is clearly a critical factor affecting the choice of models for our purposes. As distinct from precision, accuracy is a function of the fundamental design of the model, and not the precision with which computations are made. While maximum accuracy is important, it is both difficult to quantify and is the model characteristic most directly affected by any compromise elsewhere. A major consideration in the choice of a model is that accuracy be known, at least relative to the other models under consideration. Clearly, a knowledge of absolute accuracy is preferable. However, the highly variable character of radio propagation, the broad range of applications of concern here, and the limited amount of statistically reliable propagation measurement data makes absolute accuracy difficult to define. It is also important to know what accuracy compromises are incurred when the model is applied in circumstances for which it was not primarily intended. a) Technical accuracy refers to the ability of the model to predict received signal characteristics in a given environment. One should choose a model that provides the accuracy warranted by the application. The accuracy and resolution of input data should be consistent with the accuracy and reliability of the prediction methods. For example, using high resolution topographic data with a model that considers only approximate obstruction characteristics is usually not justified. b) Administrative accuracy refers to the case when design results must conform to regulations or specifications. Often in such cases, the required procedure is specified in the controlling regulations. Alternatively, the general form or character of the model may be specified in the rules. Models used in response to such requirements need to be verified as conforming with the standard. 4) Comprehensiveness:A useful model often needs to be of practical use in a number of the applications. Where possible, the model should allow extension to other, unforseen applications. (The current example of the Carey method application to 900 MHz planning and design is a prominent reminder of such a requirement.) In many cases, it will be important to be able to consistently analyze the performance of very different radio systems. Analysis of interference between different services sharing the same bands is one increasingly frequent example. A common approach to evaluating radio performance in different frequency bands is often desirable, as is the ability to evaluate new technology in radio systems design. The growing diversity of services and the increasing requirement to share spectrum between two or more services is an important consideration in propagation model choice. 5) Standardization: Even in applications where regulation or specificationsdo not dictate the use of a particular model, it may be desirable to choose a model that is in widespread usea de facto standard. In addition to greater acceptance of the

model, such a choice provides an increased likelihood that unusual behavior has been identified. In many cases a broader group of users will contribute a body of operational experience, which will enhance the confidence with which the model is applied. 6) Inverse Solution: There are many cases where the transmission loss between two known points is less important than is the distance required between a transmitter and receiver to achieve a specified loss value. Such a result requires an inverse solution to the model: given loss, find distance [12]. Unique solutions to this problem are only possible in those cases where the assumption that transmission loss is a monotonic function of distance is valid. However, many studies allow such an assumption and a propagation model capable of inverse solution is very helpful. 7) Input Parameters Treated by Model: Models need to be chosen with the assurance that required input data will be available in suitable form and at reasonable cost. Each input parameter should be considered as to the availability of input data and the extent of its influence on the results. The following input parameters should be considered: base antenna height above ground; effective base antenna height; mobile antenna height (1 to 3 m); terrain data; building dimensions and locations; foilage density and location; hill shape and location; distance.

8) Propagation Factors Treated by Model: Any model which is useful for most applications must consider all the important modes of propagation. Reflections from buildings and hills have greatest influence on loss deviation and location variability, while diffraction around buildings, hills, and trees will cause increases in transmission loss. Other modes of propagation which must be treated are the following:
free space; diffraction over a smooth earth; reflections from the surface of the earth; reflections, superrefraction, and ducting; building attenuation.

9) Computational Considerations:When computer methods are under evaluation, selection criteria should include some consideration of the size of computer required to reliably implement the model. Program complexity should also be considered, with simplicity (or at least clarity) an important criterion for reliable use. The cost of developing or installing the computer program (including any required data bases and related software) and the ongoing cost of program operation and maintenance are also important. Special requirements for computational precision or for special features (such as complex variables) should be explicitly recognized and, where possible, avoided. 10) Extent of Supporting Data Available: The choice of a propagation model must be governed by the data available for its development, evaluation, and application.

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

53
the ability to produce complex models in easy-to-use form, and improves the compromise with other aspects of model quality. Even with the ubiquity of computers, many situations dictate that results will be needed without access to machines. An important factor affecting the choice of a model (or of the way in which it will be implemented) is the extent to which manual methods, which may be graphical or algebraic, can be developed.

a) Environmental data to which radio propagation is sensitive should be included in the model. The most prominent environmental factors are topography, buildings, trees, manmade structures, and climatic factors. Topographic and climatic factors can be treated in most models, however, the influences of buildings, trees, and man-made structures are frequently ignored. Topographic data are available for the United States in a variety of forms, the most familiar being topographic maps. For a broad range of applications, however, the availability of topographic data in digital form is essential to computer analysis. The sensitivity of the propagation model to the accuracy and resolution of topographic data is an important consideration in model design and selection. Clearly, there is no need for a model that produces results of a quality superior to that which can be supported by affordable tipographic data. Similarly, topographic data quality need not be better than that to which the model is reasonably sensitive. A discussion of this trade-off is presented in [13]. b) Propagation measurements made in a particular location with a specific system are often used as the basis for predictions. Measurements may be made of radio system performance, or in the form of experiments conducted specifically to define propagation characteristics. Propagation measurements are desirable for the development and evaluation of propagation models, and are much more difficult to ,perform and to analyze than are radio performance measurements. As a result, available data are in short supply and generally do not support a comprehensive evaluation of available models. Nevertheless, it is imperative that such data as available be used to test available methods as the basis for their selection. c) Radio system performance measurement is a rapidly growing area, although not necessarily in a consistent or publicly available form. These data can serve a useful purpose in evaluating alternative models, although they are less useful in model development since they usually lack the detail needed for quantitative analysis. 11) Convenience of Use: Model selection must consider the extent to which the method is convenient to use by individuals and organizations expected to apply it. Of course, ease of use and simplicity can be achieved by sacrificing other characteristics of the method. Computer automation provides

REFERENCES
Y . Okumura, E. Ohmori, T. Kawano, and K. Fukuda, Field strength and its variability in VHF and UHF land mobile radio service, Rev. Elec. Commun. Lab., vol. 16, pp. 825-873, 1968. D. 0. Reudink, Properties of mobile radio propagation above 400 MHz, IEEE Trans. Veh. Technol., vol. VT-23, pp. 143-159, 1974. S. Kozono and K. Watanabe, Influence of environmental buildings on UHF land mobile radio propagation, IEEE Trans. Commun., vol.
COM-25, pp. 1133-1145, 1977. K. Allsebrook and J. D. Parsons, Mobile radio propagation in British cities at frequencies in the VHF and UHF bands, IEEE Trans. Veh. Technol., vol. VT-26, pp. 313-322, 1977. F. Ikegami and S. Yoshida, Analysis of multipath propagation structure in urban mobile radio environments, IEEE Trans. Antennas Propagat., vol. AP-28, pp. 531-537, 1980. F. Ikegami, S. Yoshida, T. Takenchi, and M. Umehira, Propagation factors controlling mean field strength on urban streets, IEEE Trans. Antennas Propagat., vol. AP-32, 1984. M.Hata, Empirical formula for propagation loss in land mobile radio service, IEEE Trans. Veh. Technol., vol. VT-29, pp. 317-325, 1980. R. B. Carey, Technical factors affecting the assignment of facilities in the domestic public land mobile radio service, FCC Rep. 4 6406, June 1964. K. Bullington, Radio propagation variations at VHF and UHF, Proc. IRE, vol. 38, pp. 27-32, 1950. E. A. Neham, An approach to estimating land mobile radio coverage, IEEE Trans. Veh. Technol., vol. VT-23, pp. 135-138. 1974. J. Durkin, Computer prediction of service areas for VHF and UHF land mobile radio services, IEEE Trans. Veh. Technol., vol. VT26, pp. 323-327, 1977. A. P. Barsis, Determination of service area for VHF/UHF land mobile and broadcast operations over irregular terrain, IEEE Trans. Veh. Technol., vol. VT-22, pp. 21-29, 1973. R. W. Loreny, Frequency planning of cellular radio by the use of topographical data base, presented at 35th IEEE Vehicular Technol. Conf., Boulder, CO, May 1985. Some of these references are reprinted in Land-Mobile Communications Engineering, D. Bodson, G . F. McClure, and S . R. McConoughey, Eds. New York: IEEE Press, 1984. The table of contents of this book is listed in Appendix I.

54

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37, NO. 1, FEBRUARY 1988

Appendix I Land-Mobile Communication Engineering

Listed here is the table of contents of Land-Mobile Communication Engineering, D. Bodson, G . F. McClure, and S. R. McConoughey, Eds. New York: IEEE Press, 1984.

Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Part I: Mobile Propagation.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ........... 3 Radio Propagation for Vehicular Communications, K. Bullington (IEEE Transactions Technology, November 1 9 7 7 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Comments o n Radio Propagation for Vehicular Communications, A . H. Badger (IEEE Transactions on .... 18 Vehicular Technology, August 1 9 7 8 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Field Strength and Its Variability in VHF and UHF Land-Mobile Radio Service, Y. Okumura, E. O h m T. Kawano. and K. Fukuda (Review of the Electrical Communication Laboratory, September-October 1968) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .................................. 19 Radio Propagation Above 40 M C Over Irregular Terrain, J. J. Egli (Proceedings of the IRE, October 1957) . . 68 Characteristics of Small-Area Signal Fading o n Mobile Circuits in the 1 5 0 MHz Band, H. W. Nylund (IEEE Transactions on Vehicular Technology, October 1968) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 Properties of Mobile Radio Propagation Above 400 MHz, D. 0. Reudink (IEEE Transactions on Vehicular Technology, November 1974) . . . . . . . . . . . . . ................................................. a4 Some Characteristics of Mobile Radio Propagation at 836 MHz in the Philadelphia Area, D. M. Black and D. 0. Reudink (IEEE Transactions on Vehicular Technology, M a y 1972) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 Radio Wave Loss Deviation and Shadow Loss at 900 MHz, N. H. Shepherd (IEEE Transactions o n Vehicular 108 Technology, November 1 9 7 7 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Correction to Radio Wave Loss Deviation and Shadow Loss at 900 MHz N. H. Shepherd (IEEE Transactions on Vehicular Technology, August 1 9 7 8 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1 2 Correlation Bandwidth and Delay Spread Multipath Propagation Statistics for 91 O-MHz Urban Mobile Radio Channels, D. C. Cox and R. P. Leek (IEEE Transactions on Communications, November 1 9 7 5 ) . . . . . . . . . 1 1 3 9 1 0 M H z Urban Mobile Radio Propagation: Multipath Characteristics in N e w York City, D. C. Cox (IEEE Transactions on Communications, November 1 9 7 3 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2 3 Multipath Delay Spread and Path Loss Correlation for 910-MHz Urban Mobile Radio Propagation, D. C. Cox (IEEE Transactions on Vehicular Technology, November 1977) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 Comparison of Mobile Radio Transmission at 150, 450, 900, and 3 7 0 0 Mc, W. R. Young, Jr. (The Bell System Technical Journal, November 1952) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 A Statistical Model of Urban Multipath Propagation, G. L. Turin, F. D. Clapp, T. L. Johnston, S. B. Fine, and D. Lavry (IEEE Transactions o n Vehicular Technology, February 1972) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144 Determination of Service Area for VHF/UHF Land Mobile and Broadcast Operations Over Irregular Terrain, A. P. Barsis (IEEE Transactions on Vehicular Technology, M a y 1973) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 Computer Prediction of Service Areas for VHF and UHF Land Mobile Radio Services, J. Durkin (IEEE Transactions o n Vehicular Technology, November 1 9 7 7 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 Mobile Radio Signal Correlation Versus Antenna Height and Spacing, W. C. Y. Lee (IEEE Transactions o n Vehicular Technology, August 1977) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 Studies of Base-Station Antenna Height Effects o n Mobile Radio, W. C. Y. Lee (/E Transactions o n Vehicular Technology, M a y 1980) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170 Microwave Propagation Measurements for Mobile Digital Radio Application, D. L. Nielson (IEEE Transactions on Vehicular Technology, August 1 9 7 8 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

1 . .

55

. . * . . . . . . . .. . . . . . . . . . . . . 195 Part 11: Mobile Data Communjcatjons.. . . . . . . . . . . . . . , . . . . . . . . . . . . . . . . A UHF Channel Simulator for Digital Mobile Radio, E. L. Caples, K. E. Massad, and T. R. Minor (IEEE . . . . . . . . . . . . . . . 196 Transactions on Vehicular Technology, May 1980) . . . . . . . . . . . . . . . . . . . . . . . . Characteristics of a Digital Mobile Radio Channel, T. Aulin (IEEE Transactions on Vehicular Technology, May 1981) .. . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 0 5 Error Rate Predictions and Measurements in the Mobile Radio Data Channel, R. C. French (IEEE Transactions on Vehicular Technology, August 1978) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 14 Burst Error Performance Encountered in Digital Land Mobile Radio Channel, K. Otanl, K. Daikoku, and H. Omori W E E Transactions on Vehicular Technology, November 19 8 1 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2 1 Calculation of Error Probability for MSK and OQPSK Systems Operating i n a Fading Multipath Environment, D. R. Hummels and F. W. Ratcliffe (IEEE Transactions on Vehicular Technology, August 1981)................................................................................... 226 Transmission of Digital Data Over a Rayleigh Fading Channel, M. R. Karim (/E Transactions o n Vehicular Technology, February 1982). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 3 5
+

Part I l l : Mobile Communications Above 800 M H z . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 4 1 800-MHz Band Land Mobile Telephone System-Overall View, S. /to and Y. Matsuzaka (IEEE Transactions on Vehicular Technology, November 1978) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 4 2 Radio Link Design of Cellular Land Mobile Communication Systems, M. Hata, K. Kinoshita, and K. Hirade (IEEE Transactions on Vehicular Technology, February 19 8 2 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 4 8 Vehicle Location i n Angular Sectors Based on Signal Strength, S.-B. Rhee (IEEE Transactions on Vehicular Technology, November 1978) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 5 5 A Periodic Switching Diversity Technique for a Digital F M Land Mobile Radio, F. Adachi, T. Hattori, K. Hirade, and T. Kamata (IEEE Transactions on Vehicular Technology, November 1 9 7 8 ) . . . . . . . . . . . . . 2 7 0 Periodic Switching Diversity Effect on Co-Channel Interference Performance of a Digital F M Land Mobile Radio, F. Adachi (IEEE Transactions on Vehicular Technology, November 19 7 8 ) . . . . . . . . . . . . . . . . . . . . . . 2 7 9 Mobile Radio Performance for a Two-Branch Equal-Gain Combining Receiver with Correlated Signals at the Land Site, W. C. Y. Lee (IEEE Transactions on Vehicular Technology, November 1 9 7 8 ) . . . . . . . . . . . 2 8 3 Multitransmitter Digital Signal Transmission by Using Offset Frequency Strategy i n a Land-Mobile Telephone System, T. Hattori and K. Hirade (IEEE Transactions on Vehicular Technology, November 1978) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 8 8 Advanced Mobile Phone Service, F. H. Blecher (IEEE Transactions on Vehicular Technology, M a y 1 9 8 0 ) . . . 2 9 6 Voice and Data Transmission, G. A. Arredondo, J. C. Feggeler, and J. I. Smith (The Bell System Technical Journal, January 1979) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 0 3 Cell-Site Hardware, N. Ehrlich, R. E. Fisher, and T. K. Wingard (The Bell System Technical Journal, January 1979). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 16 The Cellular Test Bed, G. C. Di Piazza, A. Plitkins, and G. I. Zysman (The Bell System Technical Journal, January 1979). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 4 0 The Developmental System, D. L. Huff (The Bell System Technical Journal, January 1 9 7 9 ) . . . . . . . . . . . . . . . 3 5 7 Energy Deposition i n Simulated Human Operators of 800-MHz Portable Transmitters, Q. Balzano, 0. Garay, and F. R. Steel (IEEE Transactions on Vehicular Technology, November 1 9 7 8 ) . . . . . . . . . . . . . 3 6 8
I

Author Index.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 7 7 Subject Index

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 7 9 Editors Biographies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 8 2

56

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 31, NO.

1, FEBRUARY 1988

Appendix I1 Conversions and Useful Expressions

The following conversions apply for frequencies in the 806947 MHz band, based on a midfrequency of 874 MHz. for 1) Field strength to received power (P) a half-wave dipole: radiated power (erp):

L = 193.9-E

(dB).

3) Effective area (A) of a half-wave dipole:

P = - 163.9-E

(dBw)

where E is in dBpV/m. 2) Field strength to transmission loss ( L ) for 1 kw effective

A =0.015 (m2). 4) Distance (0) the radio horizon for an effective earth to radius that is K = 4/3 the actual radius: D = 4 . 1 2 h (km) where h is the antenna height in meters.

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37, NO. I , FEBRUARY 1988

57

Appendix I11 Received Signal Fading Distribution

I. INTRODUCTION If a signal is transmitted between a base section and a mobile station that is moving through a multipath environment, wave interference among the multipath components results in severe fading of the received signal. This fading is rapid and its rate is a function of the frequency of operation and the velocity of the receiving antenna. In addition to the fast fading, there is a much slower variation from sector to sector of the median signal level in each sector as a result of shadowing by terrain features or man-made obstructions. There is a general agreement that the slower component of fading can be accurately modeled as having a log normal amplitude probability distribution. A number of different empirical and theoretical models have been proposed, however, for describing the statistics of the more rapid multipath fading. These include the widely known Rayleigh distribution [ 11-[5], the Rice distribution [6], the Weibull distribution [7], the Nakagami distribution [8], [9], and several other fading laws [lo], [ l l ] that can be applied in particular circumstances. This Appendix presents a literature review of multipath fading analyses and measurement. Models from the literature are discussed, and a review is made of previously reported measured fading distributions for mobile operation in the 8001 900 MHz frequency band. Conclusions are derived from the literature survey and recommendations are made with regard to a preferred model.
OF PROPOSED MULTIPATH FADING 11. ASSESSMENTPREVIOUSLY MODELS

The following paragraphs discuss the origins of the probability distributions that have been previously proposed for use in modeling multipath fading. This information is presented as a basis for recommendations concerning the preferred model for multipath fading in the 800/900 MHz mobile radio communications frequency band. a) The Rayleigh distribution was derived by Lord Rayleigh [I] in 1880 in response to a published work by Verdet in which it was concluded that the resultant of n vibrations of unit amplitude and arbitrary phase approaches a definite value n , when n is a very large number. Lord Rayleigh showed that this result is inaccurate, and that no matter how large n becomes, the value of the resultant is random. Further, he showed that for large n (no specific bounds were given), if n unit vectors of the same frequency and arbitrary phase are compounded, the probability density function is given by 2 p ( r )= - e-r2/nr. n (1)

It follows that the corresponding cumulative distribution function (CDF) is

Statistical models for multipath fading are generally developed using one of two different approaches. Often, a simplified physical model of the propagation environment is described and equations for the statistics are developed, based upon the assumptions of the physical model. For verification, results are then compared with results derived through the analysis of measured data. The second approach involves empirically fitting known probability distributions (often derived for other applications) to those computed from measured data. If the first type of model is considered for use, care must be taken to ensure that the physical assumptions upon which the model is based are not seriously violated in the environment of proposed operation. If the empirical type of model is to be employed, the use of extrapolations to predict signal behavior outside the region of operation for which the original data were measured must be closely monitored. Often, measured variances apply well to data collected in the median signal range but produce erroneous extrapolations when applied at the tails of the received signal amplitude distribution. In fact, it is possible [12] for several of the well-known distributions to provide similar fits to the same measured data near the rms value of the received signal.

Lord Rayleighs paper also showed that the expectation of power (time average ac power for an ergodic process) has a value of n. Further, it was shown that the basic exponential law is not altered even if the components have different amplitudes, provided that there are a large number of waves of each amplitude. In modern notation, the parameter n in (1) and (2) is often replaced by 2a2. This arises from a derivation of the Rayleigh distribution by computing the statistics of the sum of two orthogonal Gaussian random processes. In this case 2a2 is the time averaged ac power. It is common in current literature [5], [12] to derive the Rayleigh probability density function (pdf) based on the assumption that, from central limit theory considerations for large n, the resultant of n component waves can be decomposed into two orthogonal Gaussian random processes, which have zero mean, and the same standard deviation (U). However, Lin [12] has shown that the constraints of the complex Gaussian model are unnecessary (see Appendix 111-e below). His work showed that the relationship between the Rayleigh pdf and its underlying physical assumption is not unique. Physically, this means that if a received signal amplitude CDF derived from measurements follows a Rayleigh curve, it does not necessarily mean that there are a large number of interfering waves, or that the complex Gaussian decomposition is accurate. The relationship is, however,

58
unique under the assumptions:

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 31, NO. I , FEBRUARY 1988

1) the amplitude and phase of each component wave are statistically independent; 2) the phase of each component wave is a random variable which is uniformly distribution on [0, 2a].

b) The Rice distribution applies to the case when there is a direct line-of-sight or specularly reflected component which contributes to the received signal. This alters the amplitude probability distribution from that in the absence of such a steady signal component. If there is a single steady component and a large number of multipath signals, the situation is analogous to the case of a single sine-wave being received through additive Gaussian noise. The probability density function is

( r 2+ B 2 )
Here, Io is the Bessel function of zero order and imaginary argument, B is the amplitude of the steady signal, and = Ra/2 where Ro is the rms value of r. c) The Nakagami (or m) distribution was developed in the early 1940's by Nakagami et al. [8] who performed an experiment in which the fading of HF signals received over long propagation paths was monitored by connecting the same fading signal input to both the horizontal and vertical deflection plates of a cathode ray tube (CRT). The received signal amplitude probability distribution was then determined by measuring the emulsion density of CRT photographs taken after carefully controlled (for statistical stationarity) monitoring periods. The distributions were then plotted on log-log paper and compared with the results of a mathematical expression, the form of which was arrived at by inspection. Agreement between the measured and theoretical distributions was found to be reasonable, and after normalization and a change of variable, the corresponding probability density function was written as

Form = 1/2, (4) describes a one-sided Gaussian distribution. Finally, it can be shown [8], [9] that the Nakagami expression can approximate both the Rice and log normal distributions over certain domains given appropriate bounds on the parameter values. d) The Weibull distribution arises when results from radio propagation measurements are plotted on graph paper that is scaled such that a Rayleigh distribution appears as a straight line with a slope of - 1 . Measured 800/900 MHz mobile radio fading distributions [7] have also been found to be along straight lines on Rayleigh paper, but their slopes varied as a function of the radio propagation environment. To provide a fit to the straight line distributions derived from measured data, the use of the Weibull probability density function has been proposed [7]. This is an empirically derived pdf which has been used extensively in maintenance theory [14] and in radar sea clutter models [15]. The Weibull pdf can be written as

where a is a shape parameter which is chosen so as to yield a best fit to measurement results, ro is the rms value of r, and b = [ ( 2 h ) r ( 2 h ) ] is a normalization factor. For the special case in which a = 1/2, ( 5 ) describes a Rayleigh pdf, and when a = 1, it reduces to the exponential density function. The Weibull distribution affords the shape flexibility offered by the Nakagami distribution, but it lacks a theoretical basis. e) Other fading laws have been derived for small numbers of interfering rays. Although Lin [12] showed that the Rayleigh pdf, and its corresponding cumulative distribution function, are not uniquely related to a physical model (see Appendix 111-a above) he did not derive the CDF under less restrictive assumptions than those of the complex Gaussian model. Thus, when the number of interfering multipath components is small, so that the Gaussian model is inappropriate, analytical expressions in support of measurement (4) results must be drawn from the works of Slack [lo] and Norton et al. [ 1 I]. Here, = ( r 2 )is the time average power of the received Slack derived relationships for the probability density and signal and m = ( r 2 ) * / ( ( r- ( r ) 2 ) 2is the inverse of the cumulative distribution functions for the amplitude of the 2 ) normalized variance of r2. The parameter m is referred to as resultant of n cosine waves of equal amplitude and random the shape factor, and has a lower bound of 1/2, which has been phase, for n I12. Here work shows that the shape of the pdf derived mathematically, as well as ascertained from experi- varies considerably for n between 2 and 8, and thereafter ments [8]. follows the Rayleigh distribution. Nakagami was subsequently able to express p ( r ) of (4) in Norton et al. also presented probability distribution curves the form of a Hankel integral. Later, in 1953, he mathemati- for the resultant of the vector sum of a small number of cally derived the same integral form based on the assumption interfering waves whose amplitudes combined to give a that the received signal is composed of n( > 1 ) component constant sum squared. These distributions show a very rapid waves, the amplitudes and phases of which are both random convergence to Rayleigh behavior for n > 4. For the larger variables. It was also shown that the Nakagami distribution values of n , departures from Rayleigh behavior were found to can be applied to the case of a random superposition of random be greater for low probability values (high values of signal vectorial elements. level). The Nakagami pdf of (4) may be shown to be a more Norton et al. [ l l ] also applied the Rice distribution to general expression of other well known density functions. For analyze the case in which the received signal is the vector sum m = 1 , the Rayleigh probability density function is obtained. of a constant vector and a Rayleigh distributed vector. Results

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

59 Computed amplitude cumulative distributive distribution functions for single channel fading (the experiment also involved diversity measurements) show that the Rayleigh law is followed for strong signals, down to 15 dB below the median. Thereafter, the departure from Rayleigh fading is substantial, leading to about 6 dB signal level difference from the Rayleigh curve at 0.1 percent probability. d) Okamura et al. [16] made two sets of extensive measurements in the Kanto district of Japan, including the downtown area of Tokyo. Measurements were simultaneously made at frequencies of 463, 922, 1330, and 1920 MHz in the first set of measurements and 453, 922, 1317, and 1430 MHz in the second set of measurements. The 922 MHz transmitter was located at an elevated base station and was used to feed a corner reflector antenna. The mobile unit consisted of a school bus type vehicle with a roof mounted whip antenna at 922 MHz connected to a field strength meter. Received signal levels were recorded on a chart recorder, set for an average chart speed of 5 mm/s. The average vehicle speed was 30 km/h. To record fast fading over small sectors (50 m), the vehicle and chart recorder speeds were changed to 15 km/h and 250 mm/s, respectively. Small sector fading was measured in a number of urban and forested areas, and the corresponding amplitude CDFs were plotted on log-probability paper so as to make their median values coincide. At 922 MHz results from the urban measurements show agreement with the Rayleigh law for field strengths between + 13 and - 18 dB with respect to the median. Outside this range there is considerable departure from Rayleigh behavior. For the data derived from measurements in the forested area agreement with the Rayleigh curve is poor throughout the fading range. e) Shepherd [7] made small sector CW measurements at 900 MHz in Dallas, TX. The signal at the IF output of a mobile receiver was fed to an eight channel signal amplitude sampler. When triggered (at a rate of 50 times per meter of vehicle motion), each channel detector incremented a counter if the IF output exceeded its preset detection threshold. Since the thresholds on the eight channels were set at 6 dB intervals, the fraction of locations in which the signal level exceeded each of the eight preset thresholds could be determined. For the analysis of the experimental results, received signal amplitudes were converted to equivalent transmission loss (TL), .and transmission loss distributions [P(TL < ordinate)] were plotted on paper designed such that Weibull distributions appear as straight lines. The results showed that for different small sector measurements, the slopes of the distributions varied widely. In particular, it was shown that measurement results rarely coincided with results that would be expected if the multipath fading had followed a Rayleigh distribution. The measurement environment was not described in the referenced paper.
AND IV. SUMMARY CONCLUSION

were parametrized by the ratio of random to steady received signal power. This work shows that the slope of the amplitude CDF plotted on Rayleigh paper varies over a wide range as a function of the stated power ratio. As the ratio becomes large, (greater than about 4 dB) it is shown that the distribution follows a Rayleigh law.
1 1 FADING 1. STATISTICS COMPUTED MEASURED FROM DATA

This section details experiments that have been conducted by others to investigate multipath (fast) fading on 800/900 MHz mobile radio channels. Results are discussed and comparisons are made with results obtained through the application of the model distributions of Section 11. a) Young [2] made the first reported measurements at 900 MHz. The measurement system included a fixed base station transmitter located in downtown New York City and a receiver (10 kHz BW) mounted in a station wagon which was driven through various areas of the city, which were described as urban or suburban. The vehicle speed during these measurements was not reported. Instead of recording actual signal levels, a level distribution recorder was employed, which took 20 samples of signal strength per second. These samples were then automatically sorted by amplitude, and resulting data were used to plot received signal amplitude cumulative distribution functions. The referenced paper indicated that the plotted distribution functions show that whenever the sample measurements were confined to a relatively small distance (500-1000 ft), the results tended strongly to follow a Rayleigh curve. From a distribution curved presented in the paper, although only seven points are plotted, it appears as if the fit to the Rayleigh curve is good for signal amplitudes between 6 dB above and 15 dB below the median. The goodness of fit deteriorates for weak signal levels, where, from a systems design perspective, accuracy is most important. No difference was reported between fading statistics derived from data measured in urban and suburban environments. 6) Jakes and Reudink [3] made mobile CW fading measurements at 836 MHz in a suburban area. The transmitter was in the mobile unit, which was driven at moderate speeds. The measurement area was one where, in general, there was adequate terrain clearance to the fixed receive base station, but line-of-sight was obstructed by trees and houses to varying degrees. The analog receiver AGC output was recorded on magnetic tape for later digitizing and signal processing. Vehicle speed was maintained near 15 mi/h, and runs were made at a range of 1-2 mi from the base station. Cumulative probability distribution functions for the amplitude of the received signal computed from the measured data show that the Rayleigh fit is reasonable near the median signal level, and deviations increase at the extremities of the received signal range. Plotted level crossing rate and average fade deviation functions, however, agree well with curves computed from theory based on the Rayleigh fading model. c) Rustako [4] made CW fading measurements at 836 MHz in New Providence, NJ. The measurements were made in a residential area at a constant vehicular speed of 15 mi/h. Analog AGC levels were recorded on magnetic tape.

Five different CW multipath fading experiments from the literature have been reviewed. Due to the popularity of the Rayleigh fading model, signal amplitude probability distributions computed from the data collected during each experiment

60

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37, NO. I , FEBRUARY 1988

were compared with a Rayleigh distribution. In each case the Rayleigh curve was plotted so as to have the same median value as the measurement results. Generally, the comparisons show that the Rayleigh curve provides a good fit to the measurement results over only a limited signal amplitude range centered on the median value. In some cases the fit is tolerable for high signal levels, but it was never shown that the Rayleigh fading law was accurate for weak signals. Some of the experimental results in the literature were [7] reported to agree well with plotted Weibull distributions for different values of the shape parameter. From a study of the physical model which forms the basis for the Rayleigh distribution, and the results of other analyses [lo], [ll], it is clear that if multipath fading is to follow true Rayleigh behavior, the propagation environment must satisfy three assumptions, as follows:

another reason for the poor fit of measured distributions to the Rayleigh curve. Either the Weibull or the Nakagami model can be used in theoretical studies to introduce slope changes in model distributions to compensate for the obvious shortcomings of the Rayleigh model. The shape parameter of the Nakagami model is related to the average power of the received signal, whereas that of the Weibull model follows from the percentage of locations around the median in a sector. Both distributions have been recommended in the literature [7i1 ~ 7 1 . REFERENCES
Lord Rayleigh, On the resultant of a large number of vibrations of the same pitch and of arbitrary phase, Philosph. Magazine and J. Sci., S . 5 , vol. 10, no. 60, p. 73, Aug. 1880. W. R. Young, Jr., Comparison of mobile radio transmission at 150, 450,900, and 3700 Mc., Bell Syst. Tech. J., vol. 31, p. 1068, Nov. 1952. W. C. Jakes, and D. 0. Reudink, Comparison of mobile radio transmission at UHF and X band, IEEE Trans. Veh. Technol., vol. VT-16, no. 1, Oct. 1967. A. J. Rostako, Jr., Evaluation of a mobile radio multiple channel diversity receiver using prediction combining, IEEE Trans. Veh. Technol., vol. VT-16, no. 1, Oct. 1967. D. 0. Reudink, Properties of mobile radio above 400 MHz, IEEE Trans. Veh. Technol., vol. VT-23, no. 4, Nov. 1974. S. 0. Rice, Statistical properties of a sine wave plus random noise, Bell Syst. Tech. J., vol. 27, no. 1, Jan. 1948. N. H. Shepherd, Radio wave loss deviation and shadow loss at 900 Mhz, IEEE Trans. Veh. Technol., vol. VT-26, no. 4, Nov. 1977. M. Nakagami, Statistical Methods in Radio Wave Propagation, W. C. Hoffman, Ed. Oxford, Pergamon, 1960, p. 3. U. Charash, Reception through Nakagami fading multipath channels with random delays, IEEE Trans. Commun., vol. COM-27, no. 4, Apr. 1979. M. Slack, The probability distributions of sinusoidal oscillations combined in random phase, J . Inst. Elec. Eng., vol. 93, Pt. 111, p. 76, 1946. K . A. Norton et al., The probability distribution of the amplitude of a constant vector plus a Rayleigh distributed vector, Proc. IRE, vol. 43, Oct. 1955. S . H. Lin, Statistical behavior of a fading signal, Bell Syst. Tech. J., vol. SO, p. 3211, Dec. 1971. W. C. Y. Lee, Mobile Communications Engineering. New York: McGraw-Hill, 1982. M. Furuya ef al., Characteristics of the optimal maintenance theory based on Weibull model and their applications, Trans. IECE Japan, vol. E65, no. 4 , Apr. 1982. D. C. Scheler, Radar detection in Weibull clutter, IEEE Trans. Aerospace Electron. Syst., vol. AES-12, no. 6, Nov. 1976. Y. Okumura et al., Field strength and its variability in VHF and UHF land-mobile radio service, Rev. Elec. Commun. Lab., vol. 16, no. 9-10, p. 825, Sept.-Oct. 1968. H. Suzuki, A statistical model for urban radio propagation, IEEE Trans. Commun., vol. COM-25, no. 7, July 1977.

1) the interfering waves must be randomly varying in phase and be of near equal power; 2) the phases of the component waves must be uniformly distributed on [0, 2a]; 3) there must be a minimum of five interfering waves;
Assumptions 2) and 3) probably apply to most suburban and urban mobile radio environments in the 800/900 MHz frequency band. There are two conditions of the first assumption, however, which are not so easily satisfied. First, it is hard to envisage a situation in which all multipath components would have equal, or even nearly equal powers. Secondly, the randomness criterion can easily be violated by the presence of a steady direct signal component, or a strong specular reflection. These are believed to be the reasons that reported measurements results rarely compare well with a Rayleigh distribution throughout the entire fading signal range. The Rician model is often looked upon to account for the presence of a steady signal component. It is, however, difficult to find reported measurement results for 800/900 MHz urban mobile radio that support this model. The Gaussian noise substitution in place of the multipath signal contribution is the suspected inaccuracy in this case. It is interesting to note in the works of both Slack and Norton et al. how the shape of fading distributions can change significantly as a function of the number of interfering waves and the ratio of random to steady signal power. This could be

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

61

Appendix IV Air-To-Ground Radio Propagation

I. INTRODUCTION Air-to-ground propagation at 900 MHz exhibits some fading characteristics similar to those observed in land-based mobile radio systems, and other characteristics that are similar to terrestrial line-of-sight microwave systems. Air-to-ground propagation at 900 MHz, over long distances and at low vertical look angles, is governed primarily by obstacles blocking the path. At shorter distances the signal sometimes experiences multipath fading. The level varies slowly as the aircraft passes in and out of Fresnel zones. When the aircraft is over a ground station, the signal experiences rapid fading similar to land mobile systems. At 900 MHz, with a limited fade margin, the radio horizon is approximately equal to the optical horizon.
11. SYSTEM ARCHITECTURE The air-to-ground radio link uses frequencies in the 944 to 946 MHz band for the uplink and 899 to 901 MHz for the downlink. Single-sideband (SSB) technology is employed to achieve a high degree of spectral efficiency. SSB operation with narrow channel spacing must have enough guard band to protect against Doppler shift at jet speeds plus long-term equipment frequency drift. To achieve these objectives, high stability oscillators are required, on the order of three parts in 10O per day. Because the Doppler shift is proportional to the transmitter frequency, above 1 GHz greater frequency stability is required for a given channel spacing, resulting in higher costs. Operating in the SSB mode with several watts and omnidirectional antennas at both ends of the link, the usable line-of-sight distance is approximately 320 km. This is close to the radio horizon for aircraft flying at an altitude of approximately 30 OOO feet (9100 m). One of the most severe limitations of current VHF (100 to 400 MHz) air-to-ground systems used in aviation is multipath due to surface reflections. Aside from terrain blockage and flight path restrictions, an aircraft may assume any positive look angle from the ground antenna at any azimuth. This means all reflective surfaces on the ground have the potential of affecting the air-to-ground path. At the greater distances, where the signal is weaker, destructive multipath is an important issue. In this respect the shorter wavelengths have an advantage since they are more likely to be dispersed by ground objects of given size. This dispersion is referred to as apparent surface roughness. At low look angles, surface roughness is desirable. In as much as aircraft will statistically spend most of their time some distance from ground stations at low look angles, this roughness will be an important factor. This suggests higher frequencies are more suitable. 111. PROPAGATION FACTORS Of importance is the radiation pattern of the ground station. Real antennas may have unwanted pattern depressions, if only at the very top. One of the desirable antenna characteristics is

0,l

10
PERCENT
OF

100
TIME

Fig. 1 . Signal distribution nomogram: for specified ranges out to a given distance (verticle scale) and for a prescribed level in dBm, the curves give the percentage of time that the measured signal is less than the prescribed level over the specified range.

large gain in the horizontal direction for communications with aircraft at long distances. Unfortunately, antennas with high gains in the horizontal direction normally exhibit significant holes in the radiation pattern in upward directions. A reasonable compromise appears to be an antenna with a 5 dBi gain in the horizontal direction. This leaves enough illumination in the upward direction to handle aircraft passing overhead at 40 OOO ft (12 200 m). Generally, weather conditions do not affect propagation at 900 MHz. One exception is the condition in which the refractive index gradient is sufficiently positive to cause excessive upward beam bending. This is widely experienced in fixed microwave systems, whereby the signal passes over the intended receive antenna resulting in a loss of signal. Terrestrial microwave system designs pay close attention to this phenomenon. The methods of calculation must be modified for use on air-to-ground paths. Most of the air-to-ground radio path is higher than 300 m above ground level. Gradients are usually less severe at these higher altitudes. The coverage distance for aircraft at a given altitude is critically dependent on the look angle to the radio horizon from the ground station. A decrease of only 2 in look angle will approximately double the coverage distance. It is, therefore, a good practice to examine the look angle in all directions of interest from a site. Typically, the fading characteristics of the air-to-ground path are similar to a terrestrial microwave system in which the standard deviation of signal variations is small compared to land mobile radio systems. Data plotted to show the percentages of time at various signal levels are shown in Fig. 1. When flying directly over a site, airborne and ground antenna pattern nulls result in signal fading. Such fades can cause momentary outages depending on the aircrafts altitude and the available fade margin. The fades can also cause data transmission errors, but experience shows that outages that are
O

62

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37, NO. 1 , FEBRUARY 1988

and others similar to mobile radio systems. Fading on the airto-ground link is not normally as pronounced as it is in land 1 I mobile systems. RF fade margin requirements are, therefore, 4 rules Overhead 19 rules not as severe on an air-to-ground path. The radio horizon at Fig. 2. Signal recording exhibiting rapid fading. 900 MHz is coincidental with the optical horizon for purposes short in duration do not terminate normal voice conversations. of predicting the coverage area of a ground station, if the A record of signal level showing rapid fading that can exist system fade margin is small. during direct over flight is shown in Fig. 2. REFERENCES IV . CONCLUSION Fading on a 900 MHz air-to-ground link has some character- [l] J . R. Child, Air-to-ground propagation at 900 MHz, in Proc. 35th Ann. IEEE Veh. Technol. Conf., May 1985, pp. 73-80. istics that are similar to fixed, terrestrial microwave systems
1

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

63

Appendix V A Diffraction Based Theoretical Model for Predicting UHF Path Loss
I. INTRODUCTION The requirement for cellular systems to have coverage in one area, and isolation from other areas, makes prediction of path loss in cities an important aspect of system design. The usual basis for predicting path loss is measurements made in various cities [1]-161. One purely theoretical model for predicting path loss in cities has been developed by Bertoni and coworkers [7], [8]. Their model applies outside the highrise urban core, where the buildings are of nearly uniform height and lie in more or less parallel rows. Each row has the appearance of a bar of rectangular cross section lying on the ground, so that propagation is a process of diffraction past many bars. The model shows how the buildings result in the observed range dependence l/DM, with M between 3 and 4, and how building height, street width, etc. influence the received signal. Besides clarifying the range of validity of measurement based models, the theoretical model extends median path loss prediction to new situations, such as base antennas at rooftop level.

I
I

Fig. 1. Various ray paths for UHF propagation in the presence of buildings.

II. MODELING ASSUMPTIONS In many cities the high-rise buildings are clustered in a core region. Surrounding this core is a much larger area where the buildings have relatively uniform height, except for occasional tall buildings. In this surrounding area the buildings are organized by the street system into nearly parallel rows. Even when neighboring buildings are not physicallyjoined together, the passageways between them are typically smaller than the building width. Except for the special cases when the passageways are aligned .with the transmitter, propagation must take place through the buildings or over them. Because of reflection and attenuation by exterior and interior walls, propagation through the buildings can, at most, be significant only for the row immediately in front of the mobile. The foregoing considerations lead to modeling a row of buildings by an opaque rectangular bar lying on the ground. Propagation then takes place over the bars, as suggested by path 1 in Fig. 1 for the case when the elevated fixed antenna is the source. Here a few rows of buildings are shown in cross section. For D in the range 1-10 km the actual number of rows goes from about 20 to over 200. At each bar a portion of the field is diffracted down to the ground. The diffracted and subsequently reflected rays in the vicinity of the mobile, which are indicated by 2 in Fig. 1, account for the well-known multipath, or wavelength scale, variations of the field 191,
1101.

The field reaching the ground can rejoin that above the buildings after multiple reflections and refractions, as suggested by path 3 in Fig. 1. In addition to a loss in amplitude, these fields lose phase coherence with the original field above the buildings, and are scattered in many directions. For these reasons the fields reflected from the ground are neglected, so

that the height of the buildings does not influence propagation over their roofs, but only influences the final diffraction down to street level. For small grazing angles a,the forward diffracted field will maintain coherence with the incident field even when the tops of the buildings are irregular. However, fields that are multiply back-diffracted at the tops of the buildings will lose coherence with the phase and direction of the incident field due to irregularities of the buildings. Because of this loss of coherence, the back-diffracted fields are neglected. Since only forward diffraction is included in the model, and since for small angles a this diffraction is not strongly dependent on the cross section or electrical properties of the diffracting obstacle, each row of buildings is replaced by an absorbing screen of vanishing thickness. A further simplification is achieved by using the local plane wave approximation to find the influence of the buildings on the spherical wave radiated by the elevated antenna. The amplitude Q(a)of the field at !he rooftops due to a plane wave of unit amplitude incident at the grazing angle a on an array of building rows is first determined. The rooftop fields due to the spherical wave are then the product of Q(a),with a as shown in Fig. 1, and the spherical wave amplitude, which is inversely proportional to the range D in Fig. 1. This plane wave approximation is consistent with the concept of effective antenna height introduced by Lee [9], and is similar to that used when finding the field above a homogeneous earth. For a homogeneous earth, the spherical wave amplitude is multiwhere r(a)is the appropriate plied by the factor [l + r(a)], plane wave reflection coefficient. When computing the range dependence of the median signal, all buildings are assumed to have the same height hb equal to the median height of the actual buildings. The use of uniform height buildings to find the median signal strength is justified by the agreement obtained with measurements. Computations made assuming the height of the rows to be randomly distributed also confirm the validity of this approach. 111. DIFFRACTION A SERIES HALF-SCREENS PAST OF To find the plane wave amplitude factor Q ( a ) ,consider the diffraction of a plane wave by a series of half-screens labeled n

64

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 31, NO. 1 , FEBRUARY 1988

n=O

i . . .- . . . - . . . . .

INo = X/(da2)-J ,
Fig. 2. Ray reaching the edge n = N a n d the Fresnel zone about it.

0.05
0,03

0,Ol 0,02

0,05

0,l

0,2

as shown in Fig. 2. Diffraction past all preceding screens will alter the field incident on plane of the nth screen. Let H , ( y ) be the magnetic field incident on the plane of the n th screen. The field incident on the edge of the nth screen H,(O) will vary as n increases, but can be expected to settle to the constant value Q(a) independent of n for n large enough, since the plane wave is incident from above the edges. An estimate of the number of half-screens that must be considered in order to achieve settling can be obtained from the Fresnel zone for the plane wave field reaching the edge n = N in Fig. 2. For N large enough, the edges of No halfscreens will lie within the Fresnel zone about the ray reaching the edge n = N . For small a , NOis the integer part of d/ (ha2),so that
=
* *
e ,

0, 1, 2,

w m

0.5

1,O

Fig. 3.

Dependence of the settled field (solid curve) on the parameter

a m , where a is in radians. The dashed line has slope 0.9.

900 MHz, and a is in the range 0.1"-2", as is typical for cellular mobile radio, then a m ranges from about 0.02 to 0.4. As a result, the straight line may be used as a fit to the settled field, so that

Q=o*l

[TI
a m

0.9

No 2: d / ( ha2).

(1)

Because the edges for screens with n < N - NO, not lie in do the Fresnel zone, they should not significantly effect the field incident on the n = N edge. Thus one expects settling to be achieved after about No half-screens. These observations confirmed by the numerical results described below. Evaluation of the field H,,(O) incident on the edge of the nth screen has been discussed in detail in [8] for n as large as 200. It is found that the fields indeed settle to a nearly constant value for n > No. In fact, it is found that for small a the settling condition is achieved to within 8 percent of the final value after only O.lNo screens. The significance of this abbreviated settlings number is that the number of rows of building that must be passed over to achieve the settled field is only O.lNo, rather than No itself. For undulating terrain, the abbreviated settling distance gives the foreground that should be used to define the local slope at the mobile [9]. The computed value of the settled field is plotted as the solid curve in Fig. 3 versus the parameter a m , where a is in radians. Logarithmic scales have been used to make clear the > 1, No is unity and the power law dependence. For a@A settled field is found from diffraction past a single half-screen. For a m small, the curve has unit slope. The dashed straight line in Fig. 3 is drawn through the point (0.03,O. 1 ) with slope 0.9. It is seen to lie within 10 percent or 0.8 dB of the curve for a m up to 0.4. If d = 50 m, f =

In Fig. 1, a 2: h f / D which implies that Q of (4) varies as 1 / for fixed hf. This range variation is in addition to the 1 /D dependence of the spherical wave, so that the overall range dependence of the field is l/D'.9. Thus received power will vary as l/D3.*,which is in good agreement with many 0, Q is measurements [3], [4], [6]. In the limit as h f / D proportional to 1/D giving the field variation of 1/D2.
+

IV. PROPAGATION Loss PATH The propagation model developed in the foregoing sections can be used to predict the median path loss between the elevated antenna and the mobile. Median path loss consists of three factors: 1 ) the path loss between antennas in free space; 2) the reduction Q(a)of the roof top fields due to settling; and 3) diffraction of the roof top fields down to ground level. a) Thepath loss model consists of three factors, the first of which is the free space loss. Not accounting for antenna gain, i.e., for isotropic antennas, the ratio of received power to transmitted power for antennas in free space is given by (A/ ~ T D ) Expressing this ratio in decibels gives the free space ~. path loss LO.Iff, is the frequency in MHz and D is the range in km, then

Lo= 32,4 + 20 log f, + 20 log D.

(3)

Equation (2) is used to find Q(a),and through the definition of a , it is possible to account for terrain slope at the mobile and for the earth's curvature, provided that D is not close to

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION

65

For calculations we have assumed an average building height of 25 ft in determining the transmitter height hf above the roof (4) tops. The height of the mobile antenna was 5 fi, or about 1.5 m, and the separation between rows was taken to be d = 35 m. Finally, the earths curvature was ignored in expression where Re = 8.5 x lo3 km is the effective earth radius. Path loss associated with diffraction down to street level (9). Fig. 4 reproduces the measurements of [4] for the six depends on the shape and construction of the buildings in the vicinity of the mobile. A simple approximation to this process locations. The dots represent sector averages of the received for receiving antennas near street level is obtained by signal level in dBm plotted as a function of range in miles from assuming diffraction to occur at the center of the row of the transmitter. Transmitter height above the assumed roof top buildings before the mobile. In this case the path loss factor is level is indicated for each site. Superimposed on the data is a straight line representing the received signal predicted by the given by [IO] theory. di 1 The first two sites are located near downtown Philadelphia, (5) whose tall buildings may be responsible for the reduction in 27r [ ( ; ) 2 + ( h b - h m ) 2 ] 14 y-a 2 7 r + y - a the signal below the predicted values. The remaining four sites are away from the downtown region. For these sites the where hb is the height of the building before the mobile and h, average has range dependence that is in good agreement with is height of the mobile antenna, as indicated in Fig. 1. The theory. Combining the data for sites 2-6, Ott and Plitkins angles y and a are in radians with obtain a range dependence of 36.8 dB per decade, as compared y=tan- [2(hb-h,)/d]. (6) to the 38 dB per decade given by the model. An alternative way of comparing theory with measurements Expression ( 5 ) may be further simplified by neglecting 1/(27r is to plot excess attenuation as a function of the angle of + y - a ) as compared to l/(y - a ) ,and assuming that a is incidence given by (4). Okumura et al. [l] have measured small. Finally, to account for reflections from buildings, as median path loss for various frequencies and antenna heights. shown in Fig. 1, and other sources of multipath, a factor of d The free space path loss Lo of (3) was subtracted from their is included to arrive at the average field. measurements to obtain excess path loss, and (4) used to Combining (2) and (4)-(6), and including the factor d , convert range into angle. The resulting plots of excess gives the expression for the reduction of the field over that attenuation versus a are shown as the broken curves in Figs. experienced by the same antennas separated by a distance D in 5(a)-5(c) for antenna heights between 30 and 140 m. In Fig. free space. When expressed in decibels, this expression is the 5(a), data for fc = 1317 and 1430 MHz have been superimexcess path loss Lex,and is given by posed, while Fig. 5(b) is for fc = 922 MHz. Data at 453 MHz for heights 30,45, and 60 m deviate by less than 1 dB from the Lex=57.1+logfc+A - 18 log hf single curve used to plot them in Fig. 5(c). When excess loss is plotted versus a , antennas 140 and 30+18 log D - 18 log [ l - z ] . (7) 60 m high are seen to have the same excess loss, to within 3 17hf dB. This agreement supports the assumption that excess path The last term in (7) accounts for the curvature of the earth, loss is a function of a , rather than of D and hf separately. The while the term observed variation of excess path loss with h for a given a may be due to the fact that since D = ah, the measurements were made in different portions of Tokyo. A = 5 log [ ( ; ) 2 + ( h b - h , ) ] The solid curves in Fig. 5 represent excess path loss computed from the theoretical model. Bending of these curves - 9 log d + 2 0 log {tan- [2(h*-h,)/d]} (8) for larger values of CY results from the fact that the simple results from diffraction down from the rooftops. approximation (E-2) is not valid for Q ( a ) ,and it is necessary The overall median path loss L is found by adding Lexto the to resort to values of Q taken from Fig. 3 in the region where free space path loss Lo for isotropic antennas. Received signal curvature is significant. The computations have been made strength can be found by subtracting L from the radiated assuming h, = 3 m, which is the value used in making the measurements. We have also assumed that d = 35 m and power plus antenna gains, expressed in decibels. b) Comparison with .measurements has been made for building height hb = 12 m. two sets of data. Measurements of sector average received The predicted excess loss for fc = 1400 MHz is seen from signal were made in Philadelphia by Ott and Plitkins [4]. Six Fig. 5(a) to be within about 5 dB of the measured value over fixed antenna locations were used with antenna heights the range of shown. At fc = 922 MHz the agreement is about ranging between 45 and 255 ft above the curb. The measure- 4 dB, as seen from Fig. 5(b). However, at 453 MHz the ments were carried out at 820 MHz with radiated power and deviation is seen from Fig. 5(c) to be greater than 5 dB for a antenna gains totaling 23.3 dB. The maximum building height high antenna and low a . The primary disagreement is the slope in the vicinity of the mobile locations was reported to be 30 ft. between the measurements curve and the model.

the radio horizon. For level terrain, a in radians is given by

.[

66

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 31, NO. 1, FEBRUARY 1988

CO
rl

II

m
rl

II

4-1 c

e .

E
rl
0
P

II k G

67
20 -

30-

/ 14on

B
In
U )

330
T

! ai
In In Y

40

0.2

0.5 01.0

5
~DEGREES)

1 10

(b)

0.2

0.5

10

a ( m E E s )

(C) Fig. 5. Comparison of excess path loss from measurements by Okumura et al. [ l ] with theoretical predictions for (a)f, = 1317 and , 1430 MHz; (b) f = 922 MHz; and (c) fc = 453 MHz. Broken curves represent measurements and the solid curves are theory.

V . CONCLUSION A theoretical model of UHF propagation in urban environments has been developed to predict the range dependence of the average signal strength, and is found to be in good agreement with measurements. The model shows how the building height and separation influence signal strength. Because the excess attenuation is found to depend on h and D only through the grazing angle CY,variations of ground elevation and slope can be accounted for through the definition of CY. When finding the local tangent to the terrain, one need only take into account the terrain in the abbreviated settling distance 0.1 (Nod) in front of the mobile.

REFERENCES
[l] Y. Okumura, E. Ohmori, T. Kawano, and K. Fukuda, Field strength and its variability in VHF and UHF land-mobile radio service, Rev. Elec. Commun.Lab., vol. 16, pp. 825-873, 1968.

121 M. Hata, Empirical formula for propagation loss in mobile radio services, JEEE Trans. Veh. Technol., vol. VT-29, pp. 317-325, 1980. [31 K. K. Kelly, Flat suburban area propagation at 820 MHz, IEEE Trans. Veh. Technol., vol. VT-27, pp. 189-197, 1978. [dl G. D. Ott and A. Plitkins, Urban path-loss characteristics at 820 MHz, IEEE Trans. Veh. Technol., vol. VT-27, pp. 189-197, 1978. PI V. Graziano, Propagation correlations at 900 MHz, IEEE Trans. Veh. Technol., vol. VT-27, pp. 182-189, 1978. 161 K. Allsebrook and J . D. Parsons, Mobile radio propagation in British cities at frequencies in the VHF and UHF bands, IEEE Trans. Veh. Techno/., vol. VT-26, pp. 313-322, 1977. [71 H. L. Bertoni and J. Walfisch, A diffraction based theoretical model for prediction of UHF path loss in cities, in Proc. A G A R D Conf. Terrestrial Propagat. Characteristics in Modern Syst., Ottawa, 1986, pp. 8-1-8-9. VI J . Walfisch and H. L. Bertoni, A theoretical model of UHF propagation in urban environments, IEEE Trans. Antennas Propagat., to be published. 191 W. C. Y . Lee, Studies of base-station antenna height effects on mobile radio, JEEE Trans. Veh. Techno/., vol. VT-29, pp. 252260, 1980. f [lo] L. B. Felsen and N. Marcuvitz, Radiation and Scattering o Waves. Englewood Cliffs, NJ: Prentice-Hall, 1973, pp. 630-659.

68

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 31, NO. I , FEBRUARY 1988

Appendix VI Lees Model

IV. POINT-TO-POINT PREDICTION OBSTRUCTED [2], FOR PATHS ~31 This model is a point-to-point median path-loss prediction model consisting of two components. The first component is In the presence of a single terrain feature obstructing the an area-to-area path-loss prediction model, which serves as a line of sight from the base station antenna to the mobile, the base for the second component that gives point-to-point shadow loss is predicted by well-known knife-edge diffraction prediction. theory. A parameter U is introduced for evaluating shadow loss, and is defined as 11. AREA-TO-AREA COMPONENT [l] Area-to-area median path-loss prediction employs the three parameters: (3) Here h is the wavelength and r l , r2 are the distances from the equivalent knife edge to the base station and mobile, respectively. The height h is that of the shadowing terrain feature above the line connecting base station and mobile antennas. The shadow loss L, can be expressed in terms of Fresnel integrals with argument U . For different ranges of U the loss can be approximated by the simple functions listed below I. INTRODUCTION

1) the median transmission loss at 1 km, Lo 2) path-loss slope y in dB/decade 3) adjustment factor Fo
which must be specified for a frequency of 900 MHz, base station antenna height of 30 m, and a mobile antenna height of 3 m. At a distance D in kilometers the area-to-area path loss L, is then given by

Lf=Lo+y log D+Fo. (1) Values of Lo and y obtained experimentally for different environments are listed in Table I. The median path loss predicted by (1) for each set of values is plotted in Fig. 1 for Fo = 0. When predicting path loss, one choses values from Table I for the environment having man-made structures most nearly similar to the area in question.

L,=O dB
I<u

L,= -20 log ( 0 . 5 + 0 . 6 2 ~ )

O<u<l

L,= -20 log (0.5e0,g2u) 111. POINT-TO-POINT PREDICTIONNONOBSTRUCTED FOR PATHS -l<u<O [21, [31 When the terrain does not obstruct the path between the base L,= -20 log ( 0 . 4 - ~ 0 . l 1 8 4 - ( 0 . 1 ~ + 0 . 3 8 ) ~ ) station and the mobile, which implies path lengths less than - 2 . 4 < ~ <- 1 that to the radio horizon, the received signal consists of direct and reflected waves. In hilly terrain there can be two reflected L,= - 20 log ( - 0 . 2 2 5 / ~ ) waves as suggested in Fig. 2 for two types of transmission U < -2.4. (4) links A and B. One effective reflection point is closest to the mobile. By extending the ground plane in the vicinity of this In the presence of an obstructing terrain feature, the point-topoint back to the base station, as shown in Fig. 2, it is possible point path loss L; is given by to define an effective base station antenna height he. The effective antenna height may be greater than the actual L; = L f +L , (5) antenna height above the curb, as in Fig. 2, or less than the actual height. This height may be used to correct the area-to- and Lf is a given by (1). area path loss via the expression V. MERIT LEES OF MODEL L; = Lf+ 20 log (he/30) (2) Incorporating terrain information, as done in the point-towhere L; is the point-to-point path loss and L f is as given by point model, gives substantially different results from those (I) for a 30 m high antenna. obtained from the area-to-area model. This difference is An example of how the point-to-point model can be applied shown in Fig. 3 . Because of its more realistic accounting of to hilly terrain is illustrated in Fig. 3. The terrain profile is terrain features, the point-to-point model can be expected to shown in Fig. 3(a), together with the terrain slope used to find give more accurate predictions. Fairly good agreement is seen the effective antenna height. The point-to-point path loss from Fig. 4, which compares measured data with the predicted at each location is plotted in Fig. 3(b). For predictions of the point-to-point model [4] for nonobstructed comparison, the path loss in a flat suburban area is also shown paths. in Fig. 3(b) assuming y = 38.6 dB/decade. The actual The figures in this Appendix are reproduced with permission correction procedure for each location is indicated in Table 11. of the publisher, Howard W. Sams and C o . , Indianapolis,

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION TABLE I PROPAGATIONPARAMETERS

Environment Lo(dB at 1 km)
91.3 91.3 104.0 112.8 106.3 128.0
T

69

Free Space Open Area Suburban Philadelphia Newark Tokyo

20 43.5 38.5 36.8 43.1 30

Values originally given for English units in references 11-31 have been converted for metric units in this Table.

(b)
Fig. 2. T w o types of hilly terrain.

idtm)

-E4

E
Y
3 .

-I

\:
'8\D

-70

4 0 -90

* Fa/:-. .
*%

*
%

""&-

-IW I
I
I

slG;(,

'6-;3&-

-\.c
I

1
Fig. 1 .

5 E 7

10

\----_ ;\.
,
,
I

D
Transmission loss in different areas.

(KM)

TABLE I1 POINT-TO-POINT PREDICTION USING EFFECTIVE ANTENNA HEIGHT OF FIG. 3

Locations Effective Antenna he Gain (Loss) = 2 0 log he/lOO ( d B ) Distance (Miles)

B
100'

C
100'

D
200'

E
330'

F
620'

H
60'

I
60'

250'

0
0

0
1.15

6
3.75

10.37 5.5

15.85 6.5

8 7.5

-4
9

-4

10 150

Area-to-Area 0 Transmission Loss (Surburban area, dB) Point-to-point 0 Transmission Loss (dB)

113

134

140

143

145

148

113

128

129.6

127.2

137

152

154

70

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY. VOL. 37, NO. 1 , FEBRUARY 198R

10

PREDICTION

0 , 8 DB
2 a = 2 DB
/

research has dealt with theoretical aspects of wave phenomena in electromagnetics, ultrasonics, and optics. He has authored or co-authored over 65 articles on these topics, one of which received the 1984 Best Paper Award of the IEEE Sonics and Ultrasonics Group. During 1982-1983 he was on sabbatical leave at IJniversity College, London, UK, as a Guest Research Fellow of the Royal Society, and in the summer of 1983 held a Faculty Research Fellowship at USAF Rome Air Development Center, Hanscom AFB. His current research deals with UHF radio propagation in urban environments, scattering from periodic structures, and acoustic microscopy. Dr. Bertoni is a member of the International Union of Radio Science and a member of the Acoustical Society of America.

4 1

[,3
~WHIPPANY AREA
X CAMDEN, PHILADELPHIA

Joseph R. Child (M73) received the B S E E degree from Drexel University, Philadelphia, PA, 0 2 4 6 8 10 and has taken graduate studies for the M B A degree at George Washington University, WashingP R E D I C T E D(DB) ton, DC He is a system engineer with over 20 years of Fig 4 Indication of errors in point-to-point predictions under nonobstructed experience in the computer, mobile radio, telecomconditions munications, and receiver systems used in electronic warfare As Vice President of Engineering at Airfone, Inc , he designed and developed the first from Mobile Communications Design Fundamentals, by public telephone system used in commercial aircraft William C . Y . Lee, copyright 1986. in the United States Prior to this. he held a number of manaeement and engineering positions at Computer Sciences Corporation, ITT, MCI, and Page REFERENCES Communications Engineers, Inc. Mr. Child served as; Chairman of the Washington Telecommunications ~ ~ ~ N~~ iYork:~ ~ ~ ~ ~ ~ , [ I ] w, C. y , b e , Mobile Communjcations Society from 1980 through 1984, and was a member of United Nations McGraw-Hill, 1982. conunittee, CCIR Study Group 8B, responsible for developing the U.S.s of prediction, [2] ~, Path loss over hilly terrain and general ~ . ; ~ ~ ~ ~ position on the subject of public air-to-ground correspondence. in Mobile Communications E ~ ~ ; ~ ~ N~~, York: M , ~ G ~technical . Hill, 1982, ch. 4. [3] ~~, Prediction of propagation loss, in Mobile Communications Design Fundamentals. Indianapolis, IN: Howard W . Sams Book Co., 1986, ch. 2. Studies of base station antenna height effects on mobile radio, [4] -, IEEE Trans. Veh. Techno/., vol. VT-29, pp. 250-260, May 1980.

Nadia S. Adawi was born in Princeton, NJ, on August 29, 1958 She received the B S E E degree from the University of Missouri, Columbia, in 1980 From 1980 to 1981 she was an Electronics Engineer with the Federal Communications Commissions AM Broadcast Division In 1981 \he joined the law firm of Becker, Gurman, Lukas, Meyers, OBrien as an Engineering Consultant in cellular and common carrier system design In 1983 she became Director of Engineering for Maxcell Telecom Plus, then an applicant in third tier and lower cellular markets Frcm 1984 to 1986 she was Director of Engineering for Lunayach Communications Consultants, engineering consultants specializing in real-world cellular engineering In 1986 she became Director of Technical Support for Cellular Radio Corporation, a manufacturer of a PBX-based cellular system designed for small markets She is currently Director of Systems Engineering for Vision Systems, Inc , a telecommunications system design company offering complete installation services dnd management consulting in Arlington, VA

William A. Daniel received the B S. degree in electrical engineering from George Washington University, Washington, DC, in 1965 He is currently Chief of the Propagation Analysis Branch in the Office of Engineering and Technology at the Federal Communications Commission He is active in CCIR Study Group 5 and serve\ as Chairman of U S Ad Hoc Committee 5D which deals with the propagation aspects of the terrestrial mobile and broadcasting Fervices.

Henry L. Bertoni (M67-SM79-F87) was born in Chicago, IL, on November 15, 1938 He received the B S degree in electrical engineering from Northwestern University, Evanston. IL, i n 1960, the M S degree in electrical engineering in 1962, and the Ph D degree in electrophysics in 1967, both from the Polytechnic Institute of Brooklyn (now Polytechnic University), Brooklyn, NY. After graduation he stayed on at the Polytechnic and I \ now a Professor in the Department of Electrical Engineering and Computer Science His

John E. Dettra (S55-M57) received the B.S.E.T. degree from Capitol Institute of Technology, Washington, DC, in 1968. From 1956 to 1970, he was associated with Adair Engineering Company designing antenna systems for broadcast, medium, and short-wave stations including propagation studies and antenna measurements. From 1970 to the present he has been President of Dettra Communications, Inc., continuing the consulting engineering practice and specializing in land mobile applications. He has been qualified to give expert testimony on radio engineering matters before the FCC and the public utilities commission in ten states. He holds FCC First Class and Amateur Licenses and owns and operates radio station WQRA. blr. Dettra is an active member of AFCCE, the IEEE Vehicular Technology Society, and has been made a Fellow of the Capitol Institute of Technology and The Radio Club of America.

SPECIAL ISSUE ON MOBILE RADIO PROPAGATION Robert P. Eckert received the B.Sc degree in physics from Georgetown University, Washington, DC, in 1951, and did graduate work in electrical engineering at the Johns Hopkins University, Baltimore, MD. He was a communications systems engineer for Philco-Ford from 1962 to 1967. From 1967 to 1980, he worked for the Kelly Scientific Corporation designing communications systems for state and local governments. Since 1980, he has worked for the Federal Communications Commission specializing in matters relating to radio propagation prediction, spectrum efficiency, and frequency-assignment algorithms.

71
model used for the mobile telephone system design. He joined ITT Defense Communications Division in 1979, where he worked on military communication systems and adaptive antenna array theory. He is now Vice President of Corporate Technology for PacTel Mobile Communications, Imine, CA. He has written more than 70 technical papers, has been granted six patents, and is the author of two books: Mobile Communications Engineering (New York: McGraw-Hill, 1982) and Mobile Communication Design Fundamentals (Indianapolis, IN: Howard W. Sams Book Co., 1986). A third book, Mobile Cellular Telecommunication Systems, will be published in June 1988 by McGraw-Hill. He presented two talks on spectrum efficiency issues sponsored by the FCC in 1985 and 1987, respectively, and also conducts seminars on cellular mobile systems, domestically and internationally, sponsored by George Washington University. Dr. Lee is an Associate Editor of IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY. He is a Fellow of the Radio Club of America, a member of CCIR Group 8, an affiliate member of University of California at Irvines School of Engineering. and co-chairman of CTIAs subcommittee for Advanced Radio Technologies.

Earl H. Flath, Jr. (A44-M44-SMSl-LF87) received the B S degree in electrical engineering (with honors) from Southern Methodist University, Dallas, TX, in 1943, and the M S degree in applied science from the University of Cincinnati, Cincinnati, OH, in 1954 He was with the Naval Research Laboratory from 1945 to 1951, the Ling-Temco-Vought and Processor Companies from 1951 to 1962, with Continental Electronics Manufacturing Co., from 1962 to 1973, with E-Systems, Inc., Commercial Division, from 1973 to 1983, with Electrocom Automation, from 1983 to 1985, and is now with Consulting Engineering. He has been involved in system engineering and technical and administrative management for the last 42 years. Currently, he is involved in consulting on the design and implementation of land-mobile communication systems. He is a joint holder of Patent 2 980 909-Retarded Surface Wave Antenna. Mr. Flath is a member of Sigma Tau, Eta Kappa Nu, the Radio Club of America, APCO, and TSPE. He is a Registered Professional Engineer in the District of Columbia and the State of Texas.

Robert T. Forrest (S61-M64) received the B S E E degree from Remselaer Polytechnic Institute, Troy, NY, in 1964, and the M S degree in physics from Lynchburg College, Lynchburg, VA in 1973 His early experience in radio propagation was in the design and testing of microwave and troposcatter system as a civilian Systems Engineer working for the US Air Force i n Rome, NY and Wiesbaden, Germdny In 1969 he loined the General Electric Company i n Lynchburg, VA, where, a5 a Mobile Systems Engineer, he developed their propagation analysis facility In 1973 he was tramferred to the systems group where he wds responsible for the design and testing of mobile communications systems In 1985 he left GE to become a founding principal of Communications Technology Associates in Lynchburg, VA, a broad-based comniunicdtions engineering firm He 15 presently in charge of the Two-way Communications group at CTA He has presented several papers at IEEE conferences on the subject of radio propdgation Mr Forrest is a member of the IEEE Vehicular Technology and Communication\ Societies, and has \erved as chairman of the Central Virginia Section He 15 a Regi\tered Professional Engineer i n the States of Virginia and North Carolina

Samuel R. McConoughey (S48-A50-M56M67-SM77-F88) received the B.S.E.E. degree from Iowa State College (University), Iowa City. He held various engineering and marketing positions with Northrop-Page Communications Engineers, LTV-Continental Electronics Manufacturing Co., Prodelin, General Electric Co. (Mobile Communications Department), Michigan Wisconsin Pipeline Co., and AT&T Long Lines. He joined the Federal Communications Commission, Washington, DC, in 1971. He held various positions there including Senior Engineer, Hearing Division (1984-1985), and since 1985 has been Senior Engineer, Accounting and Audit Division, Depreciation Engineering. He is co-editor of an IEEE Press book, Land-Mobile Communications Engineering , and author of several papers in IEEE publications. Mr. McConoughey has been elected Member of the IEEE Vehicular Technology Board of Directors in 1976. He was founder and first Chairman, Professional Group on Communication Systems (PGCS forerunner to ComSOC) Syracuse Chapter, 1955. He was a member of PGCS National Administrative Committee, 1955-1958; Vice-chairman, 1960; Chairman, 1961. He received several IEEE and other professional awards. He is Secretary, ITU CCIR US Ad Hoc Study Group BA-Land Mobile, Member of the Washington Telecommunications Society, and a Registered Professional Engineer in the States of Iowa and Texas.

William C. Y . Lee (M@-SM80-F82) received the B.S. degree from the Chinese Naval Academy, Taiwan, ROC, in 1953, and the M.S. and Ph.D. degrees from The Ohio State University, Columbus, both in electrical engineering, in 1960 and 1963, respectively. From 1964 to 1979 he worked for Bell Telephone Laboratories. He created both fading and propagation path loss models, searched effective schemes in reducing mukipdth fading, studied mobile antennas and polarization effects, and participated in designing and analyzing the Bell System Advanced Mobile Phone Service (AMPS) system. He developed a software program based on his UHF propagation

John P. Murray (S54-M61-SM84) received the B.S.E.E. degree from Rensselaer Polytechnic Institute, Troy, NY, in 1957 and the masters of engineering administration degree from George Washington University, Washington, DC, in 1967. He has been involved in the development and application of radio propagation methods and models since 1959 in the U.S. Air Force, the Electromagnetic Compatibility Analysis Center, Sachs-Freeman Associates, the MITRE Corporation, and the Institute for Telecommunications Sciences, where he served as Associate Director for Spectrum Utilization. Since 1983, he has been president of John Murray Associates, a radio systems planning and design group which develops computer techniques for radio design and planning and for radio frequency management. Major emphasis of his research has been on application to the design of mobile and tactical systems, cellular radio, radio networks, and frequency planning. He has recently been involved in planning for new radio services and frequency management in developing countries. Mr. Murray was chairman of the 1986 IEEE Vehicular Technology Conference. He is a member of CCIR Study Groups I and V.

72

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 37, NO. 1 , FEBRUARY 1988

Herbert Sachs received the B.S. and M.S. degrees from the Illinois Institute of Technology, Chicago, in 1948 and 1958, respectively. He was employed by Stewart-Warner Electric for three years, Hallicrafters Corporation for three years, the Armour Research Foundation for 13 years, and the Vertex Corporation for two years. In these positions, he managed national TV and radio service operations, designed television receivers, performed communications system research, and provided electromagnetic compatibility consulting services. He designed his first mobile radio system in 1953 for the Chicago Fire Department. In 1969, he became a principal and Executive Vice President of Sachs Freeman Associates. The company specializes in Federal Government physics and engineering research. His major contribution was the development of the firms land-mobile radio capability which supported corporate offices in Maryland, Illinois, New Jersey, and Texas. It is estimated that about 300 radio system studies were performed for State, County, and City governments in connection with police, fire, emergency medical, transit, public works, and other services. He retired from the company in December 1986. Since then, he has provided consulting services to a number of jurisdictions. His other interests include microwave system design, computer-aided dispatch systems, basic and enhanced 91 1 emergency telephone systems, dispatch center design, telephone switching systems, and mobile data communications.

George L. Schrenk was born in Seymour, IN. He received the B.S., M.S., and Ph.D. degrees in theoretical physics from Indiana University, Bloomington, in 1958, 1959, and 1963, respectively, and an honorary M.A. degree from the University of Pennsylvania, Philadelphia, in 1971. He joined the Engineering Faculty of the University of Pennsylvania in 1963. From 1968 through 1980 he held the tenured rank of Associate Professor, and since 1980 he has held the rank of Adjunct Professor of Engineering. He is the President and Founder of Comp Comm, Inc., a telecommunications consulting firm with offices in Collingswood, NJ and Washington, DC. Comp Comm, Inc. was founded in the early 1970s and provides both engineering and information services to the land mobile telecommunications industry. He has published over 50 technical papers and numerous technical reports. Since the founding of Comp Comm, Inc. he has done extensive engineering work in the design and analysis of all types of mobile communication systems. Dr. Schrenk has participated extensively in the work of the IEEE VTS 8001 900 MHz Propagation Committee. He conducted much of the analysis and comparison of various propagation models for land-mobile systems presented in Section V. He also developed and implemented the field strength measurement techniques described in detail in Section VI.

Neal H. Shepherd (SM54-F70-LF88) received the B.S.E.E. degree from Texas A&M College, College Station, TX, in 1942 After graduation, he served four years as communication officer in the U.S. Signal Corps and the U S . Air Force One of his assignments while in the military service was the design, installation, and maintenance of a communication network for fighter control installations in the San Diego and Los Angeles areas. In 1946, he joined the General Electric Co., where he held various positions until his retirement from that company in 1982 His earlier work included development of mobile communication equipment, design of L-band radar transmitters, and the design of an antenna system for the Titan I1 weapons system sites Twenty years prior to retirement, he was a consulting engineer engaged in the development of advance mobile communication systems, propagation studies, noise measurements, standards committee work, and international radio regulations. One of his later assignments was to assist in the development of alignment and check out procedures for the Los Angeles Police Department simulcast radio system He has had several technical papers published Mr Shepherd was elected to the grade of Fellow in the Radio Club of America in 1978 He received the Avant Garde Award of the IEEE Vehicular Technology Society in 1980, the Founders Award of the IEEE Electromagnetic Compatability Society in 1983, and the IEEE Centennial Medal in 1984. He chairs a 900 MHz Radio Propagation Ad Hoc Committee of the IEEE Vehicular Technology Society

Floyd D. Shipley (M66) was born in Chico, TX, on June 3, 1926. He received the B.S. and M.S. degrees in physics from North Texas State University, Denton, in 1949 and 1951, respectively. He has held positions in electronic engineering in the field of antenna design and propagation since 1952 at the National Bureau of Standards (DOFL), Chance Vought Aircraft, Inc., Texas Instruments, Inc., and Decibel Products, Inc., where he was responsible for many original designs for aircraft, base and mobile antennas. He has been in private practice as a Consulting Engineer for the past 12 years, primarily in the field of Land Mobile Communications Systems Engineering with emphasis on antenna system, propagation, and interference problems. He has also conducted seminars on antenna systems and RFI for land mobile systems nationwide and overseas. Mr. Shipley is a member of the IEEE Vehicular Technology Society and a Registered Professional Engineer in the State of Texas.