## Propagation and Radio System Design Issues in Mobile Radio Systems for the GloMo Project

### Theodore S. Rappaport, Keith Blankenship, Hao Xu Mobile and Portable Radio Research Group

### Bradley Department of Electrical and Computer Engineering Virginia Polytechnic Institute and State University

### Revised January 31, 1997

This tutorial was developed for the benet of the commercial and military wireless community working on the DARPA GloMo program and related activities, and was sponsored by DARPA/ETO under the GloMo program.

I. Introduction

An accurate quantitative understanding of the radio propagation channel is imperative for reliable wireless system design. For instance, radio propagation conditions limit the area which can be covered by a transmitter or the maximum data rate in a system. Radio propagation conditions also directly aect the battery power requirements for mobile transceivers, since receiver circuitry has a battery drain that is proportional to its complexity, and transmitter circuitry has a battery drain that is directly related to RF power output.

This tutorial introduces the reader to the fundamental concepts of propagation in mobile radio. The reader will be provided with the tools necessary to make simple link budget calculations and, given data on the time dispersive nature of the mobile radio channel, determine the maximum unequalized data rate allowed by the channel. An extensive bibliography points the reader to important papers on various topics, which can be consulted for further study.

Mobile radio propagation is usually studied in terms of large-scale eects and small-scale eects. Large- scale eects involve the variation of the mean received signal strength over large distances or long time intervals, whereas small-scale eects involve the uctuations of the received signal strength about a local mean, where these uctuations occur over small distances or short time intervals.

This tutorial begins in Section II by discussing the large-scale eects and presents the useful d^{n} model
for calculating path loss. A table of typical path loss exponents, n, is provided. After introducing
receiver noise, a practical link budget is carried out using the d^{n} model. Other more sophisticated
models for calculating large-scale path loss are mentioned and referenced.

Section III discusses the small-scale eects in mobile radio propagation. After introducing the chan- nel impulse response, a discussion of the various parameters used to characterize the mobile radio propagation channel is embarked upon. The various types of fading, including \fast," \slow," \ at,"

and \frequency selective" are dened. The time dispersive nature of the mobile radio propagation channel, which limits the maximum unequalized data rate that can be attained, will be discussed. The statistical methods for modeling the mobile radio propagation channel are mentioned and referenced, as are the various methods by which the channel parameters are measured. Section IV summarizes and concludes the tutorial.

II. Large-Scale Propagation, Battery, and Range Issues

Large-scale fading analysis is concerned with predicting the mean signal strength as a function of
transmitter-receiver (T-R) separation distance (d) over T-R separations of hundreds, thousands, or
millions of meters. In this section, the widely-employed and easy-to-use d^{n} model is presented, where
d represents the T-R separation distance and relates to path loss in terms of an empirical path loss
exponent, n, for whicht ypical values are tabulated. Thed^{n} model has been shown historically to be a
very good rst-cut model for prediction of the distance-dependent received power in a wireless system
[1]. More sophisticated models which take into account more specic information are also referenced.

A. d^{n} Path Loss Model- Range vs. Battery/Power Drain

The d^{n} path loss model is generally used to predict the power transfer between a transmitter and a
receiver. This model takes into account the decrease in energy density suered by the electromagnetic
wave due to spreading, as well as the energy loss due to the interaction of the electromagnetic wave
with the propagation environment. \Path loss" is the term used to quantify the dierence (in dB)
between the transmitted power, Pt (in dBm), and received power, Pr (in dBm). (The gains of the
transmitting and receiving antennas may be implicitly included or excluded in these power quantities).

The d^{n} model predicts that the mean path loss,PL(d), measured in dB, at a T-R separationd will be
PL(d)= PL(d^{0})+10 nlog^{10} d

d^{0}

!

(dB) (1)

where PL(d^{0}) is the mean path loss in dB at close-in reference distance d^{0}, and n is the empirical
quantity { the \path loss exponent." Note that when n=2, the path loss is the same as free space {
received signals fall o by 20 dB per decade increase in distance. The reference distance, d^{0},isc hosen
to be in the far-eld of the antenna, at a distance at which the propagation can be considered to
be close enough to the transmitter such that multipath and diraction are negligible and the link is
approximately that of free-space. Typically,d^{0} is chosen to be 1 m for indoor environments and 100 m

or 1 km in outdoor environments. The free space distance must be in the far-eld of the antenna, which
is related to the physical size and frequency of the antenna. Without explicit measured information
on the close-in receive distance PL(d^{0}), it can be measured or estimated by the following formula:

PL(d^{0}) = 20log^{10} 4d^{0}

!

(2)
where =c=f is the wavelength of the transmitted signal (cis the speed of light, 3^{}10^{8} m/s and f
is the frequency of the transmitted signal in Hz).

The path losses at dierent geographical locations at the same distance d (for d > d^{0}) from a xed
transmitter exhibit a natural variability due to dierences in local surroundings, blockage or terrain
over which the signals travel. This variabilityo ver a large number of independent measured locations
the same distance away from the transmitter results in \log-normal shadowing" and is usually found
to follow a Gaussian distribution (with values in dB) about the distance-dependent mean path loss,
PL(d), with standard deviation dB about the mean path lossPL(d).

The path loss exponent, n, is an empirical constant that is often measured, but can also be derived theoretically in some environments. It varies depending upon the radio propagation environment.

Table 1, taken from [1, p. 104], gives typical values forn. Typical values for the log-normal shadowing in outdoor environments range between 8 and 14 dB. Path loss exponents for indoor environments are presented in Table 2, which also presents measured values of .

TABLE 1

Typical Path Loss Exponents

### Environment Path Loss Exponent,

nFree Space 2

Urban area cellular/PCS 2.7 to 4.0 Shadowed urban cellular/PCS 3to5 In building line-of-sight 1.6 to 1.8

Obstructed in building 4to6

Obstructed in factories 2to3

TABLE 2

Path loss exponents and ^{} values for specific indoor environments.

### Environment Freq.(MHz)

n### (dB) Source

Indoor { Retail Store 914 2.2 8.7 [2]

Indoor { Grocery Store 914 1.8 5.2 [2]

Indoor { Hard Partition Oce 1500 3.0 7.0 [2]

Indoor { Soft Partition Oce 900 2.4 9.6 [2]

Indoor { Soft Partition Oce 1900 2.6 14.1 [2]

Indoor {F actory (LOS) 1300 1.6{2.0 3.0{5.8 [2]

Indoor {F actory (LOS) 4000 2.1 7.0 [2]

Indoor { Suburban Home 900 3.0 7.0 [2]

Indoor {F actory (Obstructed) 1300 3.3 6.8 [2]

Indoor {F actory (Obstructed) 4000 2.1 9.7 [2]

Indoor { Oce Same Floor 914 2.76{3.27 5.2{12.9 [3]

Indoor { Oce Entire Building 914 3.54{4.33 12.8{13.3 [3]

Indoor { Oce Wing 914 2.68{4.01 4.4{8.1 [3]

Indoor {Av erage 914 3.14 16.3 [3]

Indoor { Through One Floor 914 4.19 5.1 [3]

Indoor { Through Two Floors 914 5.04 6.5 [3]

Indoor { Through Three Floors 914 5.22 6.7 [3]

### Example One. Path loss calculations for free space and standard urban channels.

Evaluate the path loss at a distance of 10 km for a radio signal with a carrier frequency of 900 MHz for free space and standard urban channels.

### Solution to Example One

= c
f ^{= 3}

10^{8}

9^{}10^{8} =1=3m (3)

a) For free space propagation:

Let d^{0} =1 km, n=2, then the path loss at the refence distance of1km will be

PL(d^{0} = 1 km) = 20log^{10} 4d^{0}

!

= 20log^{10}^{}41000

=91 :5dB (4)

PL(d)= PL(d^{0})+10 nlog^{10} d
d^{0}

!

=91 :5 + (10)(2)log^{10}^{}10000
1000

= 111:5dB (5) b) For standard urban:

Let d^{0} =1 km, n=4, then the path loss will be
PL(d)= PL(d^{0})+10 nlog^{10} d

d^{0}

!

=91 :5 + (10)(4)log^{10}^{}10000
1000

= 131:5dB (6) Notice that in the urban case there is 20 dB more loss over the 10 km path.

### Example Two. The eect of talk time on battery life.

Assume a 1 Amp-hour battery is used on a wireless portable communications device. Assume that the radio receiver draws 35 mA on receive and 250 mA during both transmission and reception. How long would the phone work (i.e. what is the battery life), if the user has one 3-minute call every hour?

What is the maximum talk time available if the transmitter is operating continuously?

### Solution to Example Two

a) If the user has one 3-minute call every hour, battery lifetime T will be
T = 1000^{}60 (mA-minute)

((60 3)^{}35+3 ^{}250) (mA-minute)/hour =21 :86 hours (7)
b) The battery lifetime T for continuous transceiver operation is

T = (1000^{}60) (mA-minute)

250 mA = 240 minutes = 4 hours (8)

Notice that the battery lasts at least 5 times longer when the transmitter is used for only 3 minutes per hour compared to when the transmitter is on continuously.

B. Other Models

Although extremely useful for quick estimations of link performance, the d^{n} model combines all prop-
agation eects into a single parameter { the path loss exponent n. More sophisticated models have
been developed to take into account other important factors that may vary from site to site, such as
terrain, urban clutter, antenna heights, and diraction. For outdoor propagation, some of the most
widely used models are as follows:

Longley-Rice ([4], [5], [6]),

Durkin ([7], [8]),

Okumura ([9]),

Hata ([10]),

COST-231 ([11]),

Walsch and Bertoni ([12]),

Wideband PCS microcell ([13]).

For indoor propagation, most models in wide use rely on the basicd^{n}model for free space propagation,
but also account for signal losses suered in traversing each inner partition or
oor. The world's rst
indoor propagation planning tool, SMT Plus, is described in [14] and provides system planning using
blue prints. Similar tools for wireless planning and simulation are being developed at VAT ech as part
of the DARPA GloMo project.

C. Signal Penetration into Buildings

Building penetration issues will become very important as urban wireless systems, which seek to provide ubiquitous coverage, are widely deployed. Measurements of signal penetration losses are reported in [15], [16], [17], [18], [19], and [20], which cite average penetration losses of between 7.6 and 16.4 dB into a building, depending upon building materials and frequency of operation.

D. Link Budget Calculations - Bandwidth, Power, Distance Tradeos

The limiting factor on a wireless link is the signal-to-noise ratio (SNR) required by the receiver for useful reception.

SNR =Ps=N (9)

or when powers are measured in dBm units

SNR(dB)= Ps(dBm) N(dBm) (10) The audio or video quality of a receiver is directly linked to the SNR; the greater the SNR, the better is the reception quality.

The power received from a transmitter at a separation distance of (d) directly impacts the SNR since the desired signal level is represented in the received power. The received power can be evaluated using the next equation:

Ps=Pr(d)= PtGtGr

PL ^{(11)}

or when the gains are measured in dB units and the power is measured in dBm

Ps(dBm)=( Pt)dBm+(Gt)dB+(Gr)dB (PL(d))dB (12) wherePr(d) is the received power in dBm, which is a function of the T-R separation distance in meters, Ptis the transmitted power in dBm,(Gt)dB is the gain of the transmitter antenna in dB, (Gr)dB is the gain of the receiver antenna in dB, and PL(d)dB is the path loss of the channel in dB.

The noise might consist of thermal noise generated in the receiver, co-channel or adjacent-channel in- terference in frequency division or time division multiple access systems, or multiple access interference in code division multiple access spread spectrum systems. If only the thermal noise is considered, the noise power N in dBm is given by

N =KT^{0}BF (13)

or

N(dBm)= 174 (dBm) + 10log^{10}B+F (dB) (14)

where K =1 :38^{}10 ^{23} J/K is the Boltzmann's constant, T^{0} = 290 K is standard temperature, B is
the receiver bandwidth in Hz, and F is the noise gure of the receiver in dB (see [1, p. 565]). Typical
values for F range from 5 to 10 dB for commercial receivers.

### Example Three. Noise power vs. BW.

Typical values for the bandwidth in a wireless system are from 10 kHz to several MHz. If the noise gure is assumed to be 10 dB, what is the noise level at the receiver for a 30 kHz system? Plot the noise level vs. BW, where the BW is allowed to vary from 10kHz to 100 MHz.

### Solution to Example Three

The noise level at the receiver is :

N = 174 (dBm) + 10log^{10}B+F (dB) = 174 + 10log^{10}(30000) + 10 = 119 dBm (15)
The plot of noise level vs. BW is presented in Figure 1.

### Example Four. Received power vs. distance for free space and a shad- owed urban area.

A typical cellular subscriber unit transmits 0.6 watts of power. If the transmitter output is applied to a unity gain antenna with a 900 MHz carrier frequency, what is the received power in dBm at a

10^{1} 10^{2} 10^{3} 10^{4} 10^{5}

−125

−120

−115

−110

−105

−100

−95

−90

−85

−80

Thermal noise level vs. receiver BW, when F=10dB

BW (kHz)

Noise level (dBm)

Fig. 1. Thermal noise level vs. receiver BW, when^{F} = 10 dB.

free space distance of 5 km from the antenna? What isPr(5 km) in a shadowed urban area where free space is assumed within 1 km from the antenna, and n = 4 holds for d > 1 km ? Plot the received power vs. distance. Assume unity gain for the receiver antenna.

### Solution to Example Four

Given:

Transmitter power, Pt= 0:6 W. Carrier frequency, fc= 900 MHz.

The wave length and path loss can be determined as:

= c

f ^{= 3}^{}^{10}^{8}=(9^{}10^{8}) = 1=3 m (16)

PL(d^{0}) = (4)^{2}d^{0}^{2}

^{2} ^{= (4})^{2}1000^{2}

(1=3)^{2} = 1:42^{}10^{9} = 91:53 dB (17)
a) For free space n= 2, at d= 5km

PL(d) =PL(d^{0}) + 10nlog^{10} d
d^{0}

!

= 91:53 + (10)(2)log^{10}^{}5000
1000

= 105:53 dB = 105:5 dB (18) Pr = (Pt)dBm+ (Gt)dB + (Gr)dB (PL(d))dB = 27:8 + 0 + 0 105:5 = 77:7 dBm (19) where the antenna gains are 0 dB.

b) For a shadowed urban area where n= 4, atd= 5 km the path loss and the received power will be:

PL(d) =PL(d^{0}) + 10nlog^{10} d
d^{0}

!

= 91:53 + (10)(4)log^{10}^{}5000
1000

= 119:5 dB (20)

Pr = (Pt)dB+ (Gt)dB+ (Gr)dB (PL(d))dB = 27:8 + 0 + 0 119:5 = 91:7 dBm (21) c) The plot of received power vs. separation distance is shown in Figure 2. Notice that, as the distance increases, the signal becomes markedly weaker in the urban environment.

### Example Five. Maximum separation distance vs. transmitted power (with xed BW).

In Example Four, if the signal to noise ratio is required to be at least 25 dB for the receiver to properly distinguish the signal from noise, what will be the maximum separation distance? Assume the receiver has BW of B = 30 kHz and noise gure of F = 10 dB. If the transmitted power is allowed to vary from 0.1 W to 3 W, plot the maximum separation distance vs. the transmitted power.

### Solution to Example Five

If the receiver has a BW of B = 30 kHz and a noise gure of F = 10 dB, then the noise level will be N = 119 dBm (See Example Three). A required SNR of 25 dB means that the received signal power, Pr, must be such that Pr > ( 119 + 25) dBm or Pr > 94 dBm. Assuming that the transmitted signal power is 0.6 W, or 27.78 dBm, this allows a path loss of up to PL=Pt Pr, or 122 dB.

Assuming an outdoor cellular environment using a reference distance d^{0} of 1 km and an operating
frequency of 900 MHz ( = 1=3 m), the path loss at the reference distance, PL(d^{0}) = 91:5dB (See
Example Four).

a) For free space channel:

1 2 3 4 5 6 7 8 9 10

−120

−110

−100

−90

−80

−70

−60

Pt=0.6W, d0=1km, fc=900 MHz,(Assume unity gain antennas).

Separation distance (km)

Received signal power (dBm)

In shadowed urban (n=4) In free space (n=2)

Fig. 2. Received power vs. separation distance, when^{P}^{t}= 0^{:}6 W^{;}^{d}^{0}= 1 km^{;}^{f}^{c}= 900 MHz^{;}^{n}= 4^{:}(Assume unity gain
antennas at both ends).

The path loss exponent is 2 in free space, so the d^{n} path loss model predicts that a path loss of 122
dB occurs at

122 = 91:5 + 10(2)log^{10} d
1 km

!

(22) Solving for d, we nd d = 33:5 km. Thus, the link could operate at SNR of 25 dB or greater for T-R separations of up to 33.5 km in free space.

b) For a shadowed urban channel:

Assuming a path loss exponent of 4, typical for a shadowed urban area cellular radio, the d^{n} path loss

model predicts that a path loss of 117 dB occurs at

122 = 91:5 + 10(4)log^{10} d
1 km

!

(23) or d= 5:8 km. Thus, the link could operate at SNR of 25 dB or greater for T-R separations of up to 5.8 km in a typical shadowed urban area.

c) The plot of maximum separation distance vs. the transmitted power is shown in Figure 3. Notice that for the same amount of transmitted power, the maximum separation distance decreases by a factor of its square root in the urban environment as compared to the free space environment.

0 0.5 1 1.5 2 2.5 3

10^{0}
10^{1}
10^{2}

B=30 kHz, F=10dB, SNR=25dB, Pt=0.6W,(Assume unity gain antennas).

Transmitted power (W)

Maximum separation distance (km) In shadowed urban (n=4)

In free space (n=2)

Fig. 3. Maximum separation distance vs. transmitted power, when ^{B} = 30 kHz^{;}^{F} = 10 dB^{;}^{SN}^{R} = 25 dB^{;}^{P}^{t} =
0^{:}6 W^{:}(Assume unity gain antennas at both ends).

### Example Six. Maximum BW vs. transmitted power (with xed separa-

### tion distance).

If a maximum separation distance of 5 km is required and if the transmitted power is 0.6 W, what is the maximum BW allowed for a mobile communication system operating over free space? Assume a 10 dB receiver noise gure and a required SNR of 25 dB. If the transmitted power is allowed to vary from 0.1 W to 3 W, plot the maximum BW vs. transmitted power. Assume unity gain antennas at both ends.

### Solution to Example Six

a) From Example Four we know that the pass loss is: PL(d) = 105:5 dB at a distance of d= 5kmfor a free space channel. The received power is:

Pr = (Pt)dBm+ (Gt)dB + (Gr)dB (PL(d))dB = 27:8 + 0 + 0 105:5 = 77:7 dBm (24) Assuming that the signal to noise ratio must be 25 dB, then the maximum noise power should be N = 77:7 25 = 102:7 dB. From

N = 174 dBm + 10log^{10}B+F (dB) = 174 + 10log^{10}(B) + 10 = 102:7 dBm (25)
Solving for B, we nd

B = 1:349MHz (26)

b) For a shadowed urban area, let d= 5 km, n= 4, the path loss will be:

PL(d) = 119:5 dB (27)

The received power is:

Pr= (Pt)dB + (Gt)dB + (Gr)dB (PL(d))dB = 27:78 + 0 + 0 119:5 = 91:7 dBm (28) The maximum noise level should be N = 116:7 dBm for SNR= 25 dB, and solving for B, we nd

N = 174 dBm + 10log^{10}B+F (dB) = 174 + 10log^{10}(B) + 10 = 116:7 dBm (29)

B = 53:7 kHz: (30)

Notice that in the urban environment the channel bandwidth is 20 times smaller than that in a free space channel for the same quality of reception on the 5 km link.

c) The plot of maximum BW vs. transmitted power is shown in Figure 4.

0 0.5 1 1.5 2 2.5 3
10^{0}

10^{1}
10^{2}
10^{3}
10^{4}

d=5km, F=10dB, SNR=25dB,(Assume unity gain antennas).

Transmitted power (W)

Maximum receiver BW (kHz) In shadowed urban (n=4)

In free space (n=2)

Fig. 4. Maximum BW vs. transmitted power, when ^{d}= 5 km^{;}^{F} = 10 dB^{;}^{SNR} = 25 dB^{;}^{n}= 4^{:}(Assume unity gain
antennas at both ends).

E. Choice of Modulation and its Eect on Eciency

Various modulation techniques are used in mobile communication systems. In the rst generation mo- bile radio systems, analog modulation schemes are employed. Since digital modulation oers numerous benets, it is being used to replace conventional analog systems.

The most popular analog modulation technique used in mobile radio systems is frequency modulation (FM). FM oers many advantages over amplitude modulation (AM). FM has better noise immunity and superior qualitative performance in fading when compared to amplitude modulation because the information in FM signals is represented as frequency variations rather than as amplitude variations.

In FM systems it is possible to tradeo bandwidth occupancy for improved noise performance by varying the modulation index(i.e. the RF bandwidth). FM is more power ecient than AM. Since an FM signal is a constant envelope signal, the transmitted power of an FM signal is constant regardless of the amplitude of the message signal. This allows Class C power ampliers, which have power eciencies on the order of 70%, to be used for RF power amplication. In AM, however, linear Class A or AB ampliers, which have power eciencies of 30 40%, must be used to maintain the linearity between the applied message and the amplitude of the transmitted signal. Some of the disadvantages of the FM systems are ineciency in bandwidth, complexity of the receiver and transmitter, and the requirement that the received signal power must be above a threshold for correct detection, etc. [1, p.198]

Modern mobile communication systems use digital modulation techniques. Digital modulation has many advantages over analog modulation. Some advantages include greater noise immunity and ro- bustness to channel impairments, easier multiplexing of various forms of information (e.g. voice, data, and video), and greater security. Several factors in uence the choice of a digital modulation scheme.

A desirable modulation scheme provides low bit error rates at low received signal-to-noise ratios, per- forms well in multipath and fading conditions, occupies a minimum of bandwidth, and is easy and cost-eective to implement. Existing modulation schemes do not simultaneously satisfy all of these requirements. Some are better in terms of the bit error rate performance, while others are better in terms of bandwidth eciency. Depending on the demands of the particular application, trade-os are made when selecting a digital modulation. For example, higher level modulation schemes (M-ary key- ing) decrease bandwidth occupancy but increase the required received power, and hence trade power eciency for bandwidth eciency. In general, the modulation, interference, and implementation of the time-varying eects of the channel as well as the performance of the specic demodulator are analyzed as a complete system using simulation to determine relative performance and ultimate selection. [1, p.220]

### Example Seven. Battery life vs. transmitted power.

If a 1 Amp-hour battery is used for a mobile wireless terminal, and the continuous transmitted power of the terminal is 0:6 W, what is the battery life? Plot the battery life vs. transmitted power, if the transmitted power is allowed to vary from 0.1W to 3 W. Assume the power supply voltage is 12 V, and the transmitted power is 60% of the overall power consumed by the terminal.

### Solution of Example Seven

The consumed power of the subscriber unit is 0:6=60% = 1 W, and the consumed current is 1 W=12 V =

0:0833 amp. The lifetime of the battery is 1 Amp-hour=0:0833 amp = 12 hour.

The plot of the battery life vs. transmitted power is shown in Figure 5. This curve shows the important relationship between talk time and transmitted power. Notice that in the example we used 60% power eciency, considering a 70% power eciency for the Class C amplier and some eciency loss in the rest of the circuitry.

0 0.5 1 1.5 2 2.5 3

0 10 20 30 40 50 60 70 80

Assume 1 Amp−hour battery and 60% overall efficiency

Transmitted power (W)

Battery life (hours)

Fig. 5. The battery life vs. transmitted power, when using 1 Amp-hour battery and assuming 60% overall eciency.

A combined plot of the maximum separation distance and battery life vs. transmitted power is shown in Figure 6. The combined plot of the maximum BW and battery life vs. transmitted power is shown in Figure 7. These plots show the relationship among the maximum separation distance, maximum

bandwidth, transmitted power and battery life. The plots can be used to design the propagation system or tradeo the maximum separation distance and the maximum bandwidth against the battery life.

For instance, when the bandwidth is given to be 30 kHz, from Figure 6, one can tradeo the maximum separation distance against battery lifetime by varying the transmitted power. If we change the transmitted power fromW = 0:6W to W = 1W, then from the plot we can nd that the value of the battery life decreases approximately fromt= 12 hours to t= 9 hours, while the maximum separation distance in standard urban increases from d = 5:5 km to d = 6:5 km . Similar estimation can be undertaken for the tradeo of the maximum bandwidth against the battery life using Figure 7, where the maximum separation distance is known to be 5 km. In Figures 6 and 7 we assumed 60% overall power eciency as we did in Example Seven.

Fig. 6. Separation distance and battery life vs. transmitted power.(Assume unity gain antennas and 60% eciency).

Fig. 7. BW and battery life vs. transmitted power.(Assume unity gain antennas and 60% eciency).

III. Small-Scale Fading

Whereas in large-scale propagation the average local signal strength is studied over large spatial dis- tances, small-scale fading is concerned with the more rapid uctuations of the signal over short time periods or over short distances. Multiple versions of the signal arrive at the receiver at dierent times and are subjected to constructive and destructive interference, which results in fading. Fading mani- fests itself in three main ways [1, p. 139]: (1) variation of the signal strength over short distances or short time intervals, (2) undesired frequency modulation on the signal due to Doppler shifts, and (3) time dispersion due to multipath propagation delays.

In this section, the channel impulse response is rst presented. Various useful gross-level parameters which have been developed to characterize the channel are presented next, followed by a discussion of the types of small-scale fading which can occur. The models that have been developed to simulate small-scale fading are useful for system design and simulation and are mentioned. The section concludes with a description of the methods used to obtain real-world data on the small-scale characteristics of a wireless channel.

A. Impulse Response

In mobile radio, signal propagation between the transmitter and receiver can be conceptualized by introducing the concept of the mobile radio channel impulse response, which introduces a ltering action on the signal. The channel impulse response is assumed to be time-invariant (which is usually not the case in mobile radio). The baseband impulse response, h(t), can be expressed as

h(t) =^{X}^{N}

i^{=1}aie^{}^{i}(t i) (31)

where ai is the voltage amplitude of the i^{th} arriving signal, i is the phase shift of the i^{th} arriving
signal, and i is the time delay of the i^{th} arriving signal. This shows that, in general, the received
signal is a series of time-delayed, phase-shifted, attenuated versions of the transmitted signal. If the
channel is not time-invariant, thenai,i, andi are also functions of time. The parameters ofh(t) can
be directly measured using wideband channel sounding techniques (see Section III-E). The parameters
may be used to create realistic small scale models of the propagation channel for system design and
simulation. Software packages such as SIRCIM [21] and SMRCIM [22] use this concept.

B. Parameters of the Mobile Radio Channel

In order to characterize the mobile radio channel, several parameters have been developed which provide insight into the eect of the channel upon the transmitted signal. The most important of these are the RMS delay spread, the coherence bandwidth, and the coherence time of the channel.

TABLE 3

Typical RMS delay spreads in various environments.

### Environment Freq. (MHz)

### (ns) Notes Source

Urban { New York City 910 1300 Average [23]

Urban { New York City 910 600 Standard Deviation [23]

Urban { New York City 910 3500 Maximum [23]

Urban { San Francisco 892 1000{2500 Worst Case [24]

Suburban 910 200{310 Averaged Typical Case [23]

Suburban 910 1960{2110 Averaged Extreme Case [23]

Indoor { Oce Building 1500 10{50 [25]

Indoor { Oce Building 1500 25 Median [25]

Indoor { Oce Building 850 270 Maximum [26]

Indoor { Oce Buildings 1900 70{94 Average [27]

Indoor { Oce Buildings 1900 1470 Maximum [27]

The RMS delay spread of a mobile radio channel characterizes the time dispersive nature of the channel.

The RMS delay spread, , is found from the impulse response function of the channel according to the formula

=^{q}^{2} ()^{2} (32)

where

=

Pka^{2}_{k}k

Pka^{2}_{k} ^{(33)}

(often referred to as the mean excess delay), and

^{2} =

Pka^{2}_{k}_{k}^{2}

Pka^{2}_{k} ^{(34)}

where ak is the voltage amplitude of the k^{th} multipath component and k is the delay of the k^{th}
multipath component. Table 3 provides some typical values for . The value of describes the
time delay spread in a multipath channel, beyond what would be expected for free space line-of-sight
transmission.

Coherence bandwidth is a statistical measure of the range of the frequencies over which the channel can be considered \ at" (i.e., a channel which passes all spectral components with approximately equal

gain and linear phase). In other words, coherence bandwidth is the range of frequencies over which two frequency components have a strong potential for amplitude correlation [1]. In [28], it is shown that if any two arbitrary spectral components over the bandwidth Bc exhibit a cross-correlation greater than 0.9, then Bc is related to the RMS delay spread according to the approximate formula

Bc^{} 1

50 (35)

Similarly, if the cross-correlation of any two spectral components is reduced to 0.5, then Bc is related to the RMS delay spread according to the approximate formula

Bc^{} 1

5 (36)

While approximate, the essence of these two equations is that the coherence bandwidth bears an inverse relationship to the RMS delay spread, .

Relative motion between the transmitter and receiver impresses a Doppler shift upon the frequency of
the transmitted signal, which means that the received signal will be a frequency shifted version of the
transmitted signal. The frequency shift, fd, is given by fd = (v=)cos , where v is the magnitude of
the relative velocity between the transmitter and receiver and is the wavelength of the transmitted
signal. The angle describes the angle of the received signal in relation to the direction of motion. In
mobile radio, where the relative transmitter-receiver velocity varies with time, fd will also vary with
time, thus the frequency of the received signal will appear to vary with time. As a result, the Doppler
shift tends to introduce a FM modulation into the signal. For instance, a continuous-wave (CW) signal
of frequency fc, which is transmitted over a mobile radio channel, will spread out over the bandwidth
fc^{}fm, where fm =max(fd) is the maximum Doppler shift suered by the signal. The direction of
arrival of energy determines the exact Doppler shift.

The coherence time of a mobile radio channel is the time over which the channel impulse response can be considered stationary, thus the same signal received at dierent points in time is likely to be highly correlated in amplitude. As a rule of thumb (see [1, p. 164]), the coherence time,Tc, is inversely proportional to the maximum Doppler shift suered by the signal, fm, according to the formula

Tc= 0:423

fm (37)

C. Types of Fading

There are two types of fading due to time dispersion, and two types of fading due to Doppler spread.

Time dispersion is due to the multipath delays in the channel, and has nothing to do with the motion of a mobile or the channel. Doppler spread has to do with the motion of the mobile or the channel, and has nothing to do with the multipath time delay spread of the channel.

To describe fading due to time dispersion, the terms \ at" and \frequency selective" are used. \Flat"

implies that the channel has a constant amplitude response and linear phase response over a bandwidth which is greater than the bandwidth of the transmitted signal. In the time domain, this implies that the channel impulse response is like a delta function as compared to the signal modulation symbol duration. \Frequency selective" implies that the channel possesses a constant gain and linear phase response over a bandwidth which is smaller than the bandwidth of the transmitted signal. And in the time domain, the impulse response of the channel has a time duration that is equal to or greater than the width of the modulation signal. That is, when the channel time dispersion is greater than the time it takes to send the signal, the channel has memory and induces distortion that can only be undone by an equalizer. This is why high data rate digital mobile radio standards like GSM and IS-136 require an equalizer, while older, lower data rate mobile systems like AMPS do not require an equalizer.

The data rate which can be supported by a radio channel is a function of the multipath delays which occur due to re ection and diraction in the radio channel, as well as the complexity of the receiver. In simple receivers, where an equalization circuit is not used, the maximum data rate R (bits/second) that may be supported by the channel is inversely proportional to the RMS delay spread that exists in the channel. The eects of the time delay spread on portable radio communications channels with digital modulation were studied extensively by Chuang in [29], Glance and Greenstein in [30], and Thoma, Fung and Rappaport in [31]. For nonequalized channels, the maximum data rate which may be sent before the time dispersion produces signicant errors from intersymbol interference into channel is related to the RMS delay spread by

max(Rb) = d

(38)

where factor d is dependent on the specic channel, interference level, and modulation type being used. A common rule-of-thumb based on the extensive work in [29], [30] and [31] is that when the RMS delay spread of the channel, , exceeds one-tenth of the data rate of the transmitted signal, an equalizer is required to mitigate the time-dispersive eects introduced by the channel. Thus, for d= 0:1, the maximum data rate through a channel without equalization is

max(Rb) = 0:1

(39)

When the receiver uses an equalizer, it is possible to transmit data rates that greatly exceed the inverse of the RMS delay spread. In such a case, the update rate of the equalizer is dependent on the rate of change of the channel in time, which is described by the Doppler spread. The Doppler spread describes the time rate of change of the state of the channel. For example, if a channel has a 100 Hz

Doppler spread, then the radio channel changes its \state" at a rate of 100 times per second, and a particular trained state of an equalizer may be assumed to be static for 1/100 or 0.01 seconds. The Doppler spread, when used to describe the \staticness" or short term \stationarity" of the channel, yields insight into the proper length for block codes that may be used to protect data on a wireless link.

### Example Eight. Nonequalized data rates for dierent environments.

Using the data in Table 3 calculate the maximum unequalized data rates for dierent environments.

Assume d= 0:1.

### Solution to Example Eight

For nonequalized channels, the maximum data rate is given by:

max(Rb) = d

(40)

For Urban-New York City, = 1300(ns) max(Rb) = d

= 0:1

1300^{}10 ^{9} = 76;923 bps = 76:9 kbps (41)
For suburban, = 200(ns)

max(Rb) = d

= 0:1

200^{}10 ^{9} = 500;000 bps = 500 kbps (42)
For indoor oce building, = 70(ns)

max(Rb) = d

= 0:1

70^{}10 ^{9} = 1:429 Mbps (43)

Notice how a small value of yields a larger bit rate. Of course, equalization may be used to increase data rates at the expense of more complexity and power drain.

\Fast" and \slow" fading relate the coherence time of the channel, Tc, to the symbol period of the transmitted signal, Ts. Fast fading occurs when Tc < Ts; therefore, the channel changes faster than the transmitted signal. On the other hand, slow fading occurs when Tc > Ts; therefore, the channel changes slower than the transmitted signal. Generally speaking, in today's mobile radio, the fading is almost always slow, since Doppler spreads are usually less than 100 Hz, whereas symbol rates are on the order of 30 kHz or more.

The envelopes of signals in at fading channels can often be described as being distributed according to either Rayleigh or Ricean distributions. The Rayleigh distribution describes the distribution of the envelope about its RMS value when the line-of-sight between the transmitter and receiver is obstructed.

The Rayleigh distribution describes the envelope because, due to the arrival of numerous out-of-phase multipath components, the in-phase (I) and quadrature (Q) components of the signal are Gaussian in nature. Hence, the signal envelope, which is the square root of the sum of the squares of the I and Q signals, follows a Rayleigh distribution. The Ricean distribution, on the other hand, describes the envelope of the signal when there is a dominant line-of-sight component in the received signal. As the dominant signal becomes weak, the Ricean distribution degenerates into a Rayleigh distribution.

D. Models for Small-Scale Propagation Phenomena

For detailed wireless systems design, the small-scale eects in the mobile radio channel are often simulated, and a simulated transmitted signal is subjected to the eects of the simulated channel.

Demodulation of the corrupted signal at the receiver may result in bit errors. The amount of the distortion the channel has upon the signal will directly inpact the quality of the demodulated signal.

In some channel models, an impulse response is generated, with which the transmitted bit stream can be convolved to obtain the received signal. Generally speaking, the models are statistical in nature, representing typical channels in various propagation environments. However, progress is being made into ray tracing and other methods which use site-specic information to derive the channel impulse response function for specic geographical regions. It should be noted that small-scale simulation is vital for synchronization, equalization, error control coding, or diversity design in practical wireless systems.

Clarke [32] and Gans [33] developed methods by which the Rayleigh fading envelope can be simulated.

The added complexity of multipath time delay is rectied in such models as the two-ray Rayleigh fading model [1, p. 188] which builds upon the model of Clarke and Gans. Software packages such as SIRCIM [21] and SMRCIM [22] are statistical in nature, based on measured data. These programs generate channel impulse responses for typical propagation environments based upon environmental factors and mobile velocity. Recently, models have been introduced to statistically model the angle- of-arrival of the received signal ([34], [35]). Such models are needed for research into adaptive arrays for mobile communications.

E. Measurements of Small-Scale Propagation Phenomena

The small-scale phenomena of at fading channels can be measured with narrowband CW signals.

However, in order to measure the time dispersive eects in small-scale propagation as discussed above, wideband techniques are required. Historically, three wideband techniques have been used [1, pp.153{

159]. These include the following:

the swept frequency technique, in which a network analyzer is used to sweep a frequency range,
measuring the s-parameter, s^{21}, over the bandwidth. The measurement can be converted to the
time domain by means of a fast Fourier transform. This method suers from the fact that the
time required to sweep the channel is often much longer than the coherence time Tc.

pulsed techniques, in which a pulse train with inter-pulse times greater than the maximum measur- able channel delay is transmitted. The echoes of the received pulses are recorded by the receiver.

The measure of the channel impulse response is determined from the strength and time delays of the pulses. This method suers from the need for wideband RF ltering, and therefore the dynamic range is limited.

spread spectrum sliding correlator techniques, in which a spread spectrum signal is transmitted.

The receiver's pseudonoise (PN) clock runs at a slightly slower rate than the PN clock rate of the transmitted signal. This allows the receiver's PN code to gradually slide relative to the transmitted PN code and to eventually become correlated with all multipath components in the channel. The method benets from the ability to use narrowband processing, thus rejecting much of the passband noise otherwise admitted by the pulsed techniques.

Recently, there has been growing interest in measurements at higher frequencies, including the 28 GHz, 37.2 GHz, and 60 GHz frequency bands. References [36] through [50] contain literature on higher frequency measurements.

F. Impact of Antenna Gain and Fading

As discussed in section III-C, in a radio communication system, multipath can limit performance either by introducing fading in narrowband systems or causing intersymbol interference in wideband systems [51]. One way to mitigate the eect of multipath interference is to use the spatial ltering technique.

The basic idea of spatial ltering is to use directional antennas instead of omnidirectional antennas to emphasize signals received from one direction and attenuate signals from other directions. Spatial ltering can be employed in one of three forms: sectorized systems, switched beam systems, or adaptive antennas.

In sectorized systems, the cell is divided into three or six angular regions. The base station typically uses directional antennas with 60 or 120 degree beamwidths to cover these regions. The system capacity is increased due to the decrease of the amount of the co-channel interference from the other users within its own channel. The proposed IS-95 CDMA standard incorporates a degree of spatial processing through the use of simple sectored antennas at the cell site. It employs three 120 degree wide receive and transmit beams to cover the azimuth. Sectoring nearly triples system

capacity in CDMA.[52]

In switched beam systems, each sector in a cell is divided into smaller angular regions, each of which is covered by a narrow beamwidth antenna. A simple switched beam technique is to select the antenna which provides the best signal for a particular mobile unit[53].

In adaptive antenna systems, a steerable adaptive antenna is used on the base station receiver.

The adaptive array is capable of steering a directional antenna beam in order to maximize the signals from a desired user while attenuating signals from the other users. Systems which form a dierent beam for each user using switched beam or adaptive antenna technology are also referred to as intelligent antennas or smart antennas.[53]

A list of references is included for the further study of spatial ltering. Some of the recent work at MPRG in this area is presented in [34], [35] and [54], including the analysis of CDMA cellular systems employing adaptive antennas in multipath environments, descriptions of the geometrically based statistical channel model for macrocellular mobile environments and geometrically based model for light-of-sight multipath radio channels. For example, [35] uses the geometrically based single bounce macrocell (GBSBM) channel model to analyze multipath and fading. When the antenna beamwidth is narrowed, the radius of the scattering circle is decreased. With a decrease in the radius of the scattering circle, the GBSBM model predicts that the range of angles of arrival of multipath components and the Doppler spread will both decrease. Consequently, the fading will be reduced. The plot of the fading envelopes with scattering circles of dierent radii is presented in [35].

IV. Summary and Conclusion

This tutorial has given the reader a brief introduction to the concepts relevant to the mobile radio
propagation channel and practical system design, bandwidth, and power issues. Large-scale eects
were dealt with in Section II, focusing on the widely-used d^{n} path loss model. An example of the use
of this model was given Section II-D. The reader is encouraged to employ the list of typical path loss
exponents in Table 1 for quick link budget calculations. Section III dealt with the small-scale eects
in mobile radio propagation. The parameters which characterize the mobile radio propagation channel
{ , Bc, and Tc { were introduced. The relation of these parameters to those of the transmitted
signal determine the type of fading the signal experiences and some important aspects of wireless
communication system design, including whether or not an equalization stage is required in the receiver.

Some of the statistical methods for modeling the mobile radio propagation were touched upon, before the tutorial concluded with a cursory description of the techniques by which the parameters of the mobile radio propagation channel are measured in practice. The extensive bibliography should provide

the reader with a starting point by which to explore the topics covered in this tutorial in greater depth.

V. Acknowledgement

Thanks to Rob Ruth and Ken Gabriel of DARPA, and Jim Schaner at Hughes, who are program managers supporting our propagation work. Special thanks to Richard Roy of ArrayComm and Bruce Fette of Motorola who have provided suggestions for improvement for this document.