LESS SYSTEMS AND lRELESS

OP

Cordless Systems
Time Division Duplex
DECT Operation
Adaptive Differential Pulse Code Modulation
Wireless Local Loop
The Role of WLL
Propagation Considerations for WLL
Orthogonal Frequency Division Multiplexing
Multichannel Multipoint Distribution Service
Local Multipoint Distribution Service
WiMAX and IEEE 802.16 Broadband Wireless Access Standards
IEEE 802.16 Architecture
Services
IEEE 802.16 MAC Layer
IEEE 802.16 Physical Layer
IEEE 802.16a
Recommended Readings and Web Sites
11.5 Key Terms, Review Questions, and Problems
Key Terms
Review Questions
Problems
Appendix ItA Linear Predictive Filters
317
318 CHAPTER 11 ! CORDLESS SYSTEMS AND \VIRELESS LOCAL LOOP

In this chapter, we look at two technologies that bring wireless access into the resi-
dence and office: cordless systems and wireless local loop (WLL).

Standardized cordless systems have evolved from cordless telephone technology.

Originally, cordless telephones were developed to provide users with mobility
within a residence or small office by separating the handset from the rest of the
telephone (called the base station) and providing a simple analog wireless link. As
technology improved, digital cordless telephones were developed. The products on
the market used proprietary wireless interfaces. Because the same manufacturer
sold the base station and the handset as a unit, there was no need for standards.
Standards-making bodies became interested in standardizing cordless technol-
ogy to widen its range of applicability, in two directions. First, cordless systems can
support multiple users from the same base station, which could include either
multiple telephone handsets or both voice and data devices (e.g., fax or printer).
Second, cordless systems can operate in a number of environments:
• Residential: Within a residence a single base station can provide voice and
data support, enabling in-house communications as well as providing a con-
nection to the public telephone network.
• Office: A small office can be supported by a single base station that provides
service for a number of telephone handsets and data devices. In a larger office,
multiple base stations can be used in a cellular configuration, with the base
stations connected to a PBX (private branch exchange) switch. Such a config-
uration can serve hundreds or even thousands of users.
• Telepoint: Telepoint refers to the provision of a base station in a public place,
such as a shopping mall or airport. This application has not succeeded in the
marketplace.
A number of design considerations have driven the development of cordless
standards. [WEBBOO] lists the following:
1. The range of the handset from the base station is modest, up to about 200 m.
Thus, low-power designs are used. Typically, the power output is one or two
orders of magnitude lower than for cellular systems.
2. The handset and the base station need to be inexpensive. This dictates the use of
simple technical approaches, such as in the area of speech coding and channel
equalization.
3. Frequency flexibility is limited, because the user owns the base station as well as
the mobile portion and can install these in a variety of environments. Hence, the
system needs to be able to seek a low-interference channel wherever it is used.
Although a number of different standards have been proposed for cordless sys-
tems, the most prominent is DECT (digital enhanced cordless telecommunications),l

IDECT formerly stood for digital European cordless telephone.

 11.1 / CORDLESS SYSTEMS 319

Table 11.1 DECT and PWT Parameters

developed in Europe. The U.S. equivalent is known as PWT (personal wireless

telecommunications). Table 11.1 shows some of the key parameters for DECT and
PWT. These systems use an approach referred to as time division duplex (TDD). We
begin with a general discussion ofTDD and then turn to the details of DECT.

Time Division Duplex

In the technique ofTDD, also known as time-compression multiplexing (TCM) , data
are transmitted in one direction at a time, with transmission alternating between the
two directions. We first look at simple TDD and then discuss TDMNTDD.
Simple TDD To achieve the desired subscriber data rate with simple TDD, the
transmitter's bit stream is divided into equal segments, compressed in time to
a higher transmission rate, and transmitted in bursts, which are expanded at the
other end to the original rate. A short quiescent period is used between bursts
going in opposite directions to allow the channel to settle down. Thus, the actual
data rate on the channel must be greater than twice the data rate required by the
two end systems.
The timing implications are shown in Figure 11.1. The two sides alternate in
the transmission of data. Each side sends blocks of some fixed length, which take
a time T b to transmit; this time is a linear function of the number of bits in
a block. In addition, a time Tp is required for the propagation of a signal
from transmitter to receiver; this time is a linear function of the distance
between transmitter and receiver. Finally, a guard time Tg is introduced to turn
the channel around. Thus, the time to send one block is (Tp + T b + T g ). How-
ever, because the two sides must alternate transmissions, the rate at which blocks
 320 CHAPTER II/CORDLESS SYSTEMS AND WIRELESS LOCAL LOOP

Base
station -+--+--t-+---+------'\---+--+-----
I
I
I

Handset _-----'_--'-.L.---1.-_ _--'-----'----'-_'-- _

-l H f- Time~
Tp Tb Tg

Tp = Propagation delay
Tb = Burst transmission time
Tg = Guard time
Figure 11.1 Transmission Using Time Division Duplex

can be transmitted by either side is only 1I[2(Tp + T b + Tg )]. We can relate this
tothe effective data rate, R, as seen by the two end points as follows. Let B be the
size of a block in bits. Then the effective number of bits transmitted per second,
or effective data rate, is

The actual data rate, A, on the medium can easily be seen to be

Combining the two, we have

A = ( TT+ T)
2R 1 + p
b
g

The choice of block size, B, is a compromise between competing requirements. For a

larger block size, the value of T b becomes larger compared to the values of Tp and
Tg . Now consider that we have a fixed value of R, which is the data rate required for
the link, and we need to determine the value of A. If B is increased, there is a
decrease in the actual data rate,A. This makes the task of implementation easier. On
the other hand, this is accompanied by an increase in the signal delay due to buffer-
ing, which is undesirable for voice traffic.

L
 11.1 ! CORDLESS SYSTEMS 321

TDlVIA/TDD TDD was developed for use on wired subscriber systems but has
now found application in wireless configurations. In a wireless configuration, TDD
involves transmitting and receiving on the same carrier frequency but at different
times. Wireless TDD is typically used with TDMA, in which a number of users
receive forward (base to handset) channel signals in turn and then transmit reverse
(handset to base) channel signals in turn, all on the same carrier frequency.
Two important advantages of TDD over using a TDMA system with two
different carriers, one in each direction, are improved ability to cope with fast fading
and improved capacity allocation.
With respect to fast fading, the pattern of fading changes as the frequency
changes. 2 It is possible for a mobile unit receiving on one frequency and trans-
mitting on another to experience strong fade in one direction only. One way to
cope with fast fading is spatial diversity of the antennas. In this scheme, a receiver
employs two antennas, spaced ideally at least one wavelength apart. With such a
configuration there is a good chance that if one of the antennas is experiencing
significant fade, the other is not. The receiver can then select the strongest incom-
ing signal for demodulation. This approach is fine for the base station but is not
practical for small, inexpensive handsets. Further, we wish to avoid complex
adaptive equalization algorithms or forward error control algorithms. With TDD
however, spatial diversity can still be used. For each forward channel burst arriv-
ing at the base station, the base station selects the antenna with the stronger sig-
nal for reception. The base station then uses that same antenna for the next
transmission on the corresponding forward channel. Because the same frequency
is used in both directions, this antenna should provide the strongest signal back at
the handset.
The other advantage of TDMA/TDD over TDMAIFDMA is in the area of
capacity allocation. For many data-oriented (as opposed to voice oriented) applica-
tions, it is likely that the volume of traffic will be greater in the forward direction
than in the reverse direction. If a conventional FDMA scheme is used, with equal
bandwidth on forward and reverse channels, then the system must be sized for the
forward channel and capacity on the reverse channel is wasted. On a TDD system, it
is possible for a controller to allocate dynamically the number of TDMA slots in
each direction so that more slots are available on the forward channel when needed.

DEeT Operation
Figure 11.2 indicates the protocol architecture that supports DECT operations.
At the physical layer, data are transmitted in the TDMA-TDD frames over one of
10 RF carriers. The modulation method is Gaussian-filtered FSK, with a nominal
deviation of 288 kHz. This is essentially the same as the GMSK technique used for
GSM (Chapter 10). The ratio of data rate to transmission bandwidth is 2 (compare
Table 6.2), and the data rate is 1.152 Mbps.
The medium access control (MAC) layer selects the physical channels and
then establishes or releases connections on those channels. It also multiplexes

2See Section 5.4 for a discussion of fast fading.

322 CHAPTER 11 / CORDLESS SYSTEMS /I,NI) WIRELESS LOCAL LOOP

Mobile unit Base station

~ ~ ~ ~ r-"'\ r-"'\ r-"'\ r-"'\ ~
~ ~
C.J C.J
.S: .S:
en
~
C.J ""
~
en "0
....c: en
~
C.J
~""
en "0
....c:
.S: ....c: .S: ....c:
~ ~

.-
~ ~

'0
....""c:
""
~

....en
~
OIl
<::
en
en
~
.- :>
.-
~
C.J E
~
OIl '0
....""c:
""
~

....en
~
OIl
<::
en
en
~
.- :>
o""
~
C.J E
~
OIl
~
""o ""en~
<::
c: ~ ""en
~
<::
c:
0 ....""<::c: E <::
DECT messages 0 ....""<::c: E <::

--; ....= OIl --; ....= OIl<::

~ ~
C.J en .2 E C.J en .2 E
~
en
~
<::
en .... ~
en
~ C.J en ....
U E~
"20
C.J
~
c:
c:0 E
en
~ :=:E u E~
"20 ~
c:
c:0 E
en
~ :5
:E0
Q.. ;:
0
Q.. ;:
=
Q, C.J
~
c: u :E
Q,
=
C.J
~
c:
u :E
00 00
c:
0
c:
0
U U

Data link control Data link control

Link control
frames
Mediu:rnaccesscontrol Medium access control

Physical Physical

Radio link

Figure 11.1 DECT Protocol Architecture

information into the TDMA-TDD frame format. The MAC layer supports three
serVIces:
4> Broadcast: Broadcast messages sent in the A field
4& Connection oriented: Transfer of user data in the B field

• Connectionless: Supports individual DECT messages sent in the A field

The data link control layer provides for the reliable transmission of messages
using traditional data link control procedures, including error detection and auto-
matic repeat request. 3
Above the data link control layer are a set of services:
• Call control: Manages circuit-switched calls, including connection setup
and release.
@ Supplementary services: Services independent of any call that support
operations.
4& Connectionless message service: Support of connectionless messages. This ser-

vice will segment longer messages into smaller blocks for transmission and
reassemble at reception, if necessary.
.. Connection-oriented message service: Support of connection-oriented messages.

3See Appendix C for a general discussion of data link control protocols.

 ILl/CORDLESS SYSTEMS 323

• Mobility management: Handles functions necessary for the secure provision of

DECf services. Mobility management is organized into seven groups of services:
- Identity procedures: Used for the mobile unit to identify itself to the base
station
- Authentication procedure: Establishes that the mobile unit is a valid net-
work user
- Location procedure: Used in systems with multiple base stations to track
the location of a mobile unit
- Access rights procedure: Establishes that the mobile unit has the right to
gain access to a specific type of local or global network
- Key allocation procedure: Distributes encryption keys for protecting
network control information and user information
- Parameter retrieval procedure: Used to exchange information about the
parameters of the mobile unit and network operation
- Ciphering-related procedure: Encryption and decryption operations

Adaptive Differential Pulse Code Modulation

For voice digitization, DECT makes use of a procedure known as ADPCM. ADPCM
has been standardized by ITU-T and is used in a number of applications, including
digital circuit-switched networks such as ISDN, and in a number of wireless local
loop implementations. We first look at the basic principles involved, then describe the
simpler differential PCM (DPCM) scheme, and finally examine ADPCM. 4
Differential Quantization Differential quantization is based on the principle
that speech signals tend not to change much between two samples. Thus, with ordinary
PCM, the quantized values of adjacent samples will generally be near to one another
and the transmitted PCM values contain considerable redundancy. Hence, it might
make sense to transmit only the difference value between adjacent samples rather
than the actual value, leading to the use of fewer bits per sample (Figure 11.3). Thus, if
m(k) is the kth sample, we transmit the difference d(k) = m(k) - m(k - 1). If the
receiver starts with the correct initial value, then on the basis of the sequence of
difference samples d(k), it can reconstruct the sequence of samples m(k).
However, if we merely transmit the difference between the current and the
immediately preceding samples, using fewer bits, there is the danger that the output at
the receiver will begin to drift away from the true value at an increasing amount. If the
difference value between two samples exceeds that which can be represented by the
bits transmitted, then the receiver will err in reproducing the input and there is no way
to make a correction subsequently. As a way to address this problem, the encoder could
not only transmit a difference value; it could also replicate the decoding function that
takes place at the receiver. The encoder could then transmit the difference between the
current sample and what the encoder knows is the receiver's output from the previous
transmission. In effect, the encoder is instructing the decoder to make incremental
adjustments to its previous output, thus automatically providing a correction.

4Before proceeding, the reader may wish to review the discussion of PCM in Section 6.4.
324 CHAPTER 11 ! CORDLESS SYSTEMSAND\X7IRELESS LOCAL LOOP

Current Next
sample sample

I n leveb

Coding
region

m levels

f'igure 11.3 Normal and Differential Quantization

Differential PCM The scheme that has just been described is based on the
assumption that difference values will be less than sample values. We can do even
better than that by recognizing that voice signals change relatively slowly and there-
fore we can fairly accurately estimate, or predict, the value of the kth sample m(k)
from the values of the preceding samples. For an estimate of the kth sample, m(k),
we then transmit the difference d(k) = m(k) - m(k). If we use an accurate esti-
mator, this difference value should be less than the difference between successive
samples. At the receiver, the same estimation function is used and the incoming
difference value is added to the estimate of the current sample. This is the principle
behind differential PCM (DPCM). Note that the simple difference scheme of the
preceding paragraph is just a special case of DPCM in which we use the estimator
m(k) = m(k - 1).
Figure 11.4a is a block diagram of the transmitter. An understanding of this
diagram takes some discussion. To begin, an analog sample m(k) is taken of the voice
signal get); this is a pulse amplitude modulation (PAM) sample. We are now faced
with a difficulty. We could construct a predicted value m(k), take the difference,
and transmit that. However, at the receiving end, instead of the past analog samples
m (k - 1), m (k - 2), and so on, what we have is the sequence of quantized differences
from which we could construct a sequence of quantized samples, mq(k). So we cannot
reconstruct m(k) but can only determine mq(k), the estimate of the quantized sample
mq(k), with the estimate based on the preceding quantized samples mq(k - 1),
mq(k - 2), and so on. If the transmitter is basing its prediction on values of m(k) and
the receiver is basing its prediction on values of mq(k), there is an error introduced in
reconstruction. Instead, the transmitter should also determine mq(k) and transmit the
difference d(k) = m(k) - mq(k). As we have just seen, the receiver can also generate
mq(k), and from the received dq(k), the receiver can reconstruct mq(k).
 11.1 I CORDLESS SYSTEMS 325

g(t) dik )
Sample I---~ Quantizer 1---....:....----.---..

r---- ------------------------1
I
I
1-----------......., 1
1
I I
I
: mik) I
I 1
I
I
1
1- _

(a) DPCM encoder

1----------------------------- 1
: dik ) + mq(k) : Smoothing g(t)
--+----+1 I 1---------.....-....-+-1
+ : OOu
I
I
I
L...-------l Predictor 1
1
I I
I -----------------------~

(b) DPCM decoder

Figure 11.4 DPCM Transmitter and Receiver

It remains to show that Figure llAa produces the desired quantized differ-
ences. The output of the predictor is designated mq(k), the predicted value of m(k).
The difference is taken to produce
d(k) = m(k) - mq(k)
This difference is then quantized to produce dq(k). We can express this as
dq(k) = d(k) + e(k)
where e(k) is the quantization error that results from approximating the analog
d(k) by the digital dq(k). The input to the predictor consists of this quantized dif-
ference plus the predictor output, which is fed back to become part of the input:

mq(k) = mq(k) + dq(k)

= [m(k) - d(k)] + [d(k) + q(k)]
= m(k) + q(k)
This demonstrates that mq(k) is a quantized version of m(k). Therefore, the
input to the predictor is the desired mq(k), which is what is needed for operation at
the receiver. Figure 11.4b shows the logic at the receiver. The portion of the receiver
in the dashed box is the same as the portion of the transmitter in the dashed box.
The two dashed boxes have the same input, dq ( k), and therefore produce the same
output, mq(k). At the receiver, this output is passed through a filter to produce an
analog output that is an estimate of the original analog source.
326 CHAPTER II/CORDLESS SYSTEMS AND WIRELESS LOCAL LOOP

Figure 11.5 Linear Predictor Circuit

_The predictor is often the linear weighted sum of previous samples, imple-
mented using shift registers, and a delay equal to the sample interval (Figure 11.5).
The predictor implements the equation
N
mik) = ~ Aimq(k - i) (11.1)
i=l

For a further discussion of linear predictors, see Appendix llA.

Adaptive Differential PCM The performance of DPCM can be improved
by using adaptive prediction and quantization, so that the predictor and differ-
ence quantize adapt to the changing characteristics of the speech being coded.
In this subsection, we describe the ADPCM scheme standardized by ITU-T in
Recommendation G.726,5 which is used in DECT and in a number of wireless
local loop schemes.
Figure 11.6 shows the basic ADPCM logic. The transmitter converts a 64 kbps
digitized voice stream into a compressed 32-kbps stream. The overall block
diagram is the same as that for DPCM. In this case, the adaptive predictor makes
use not only of previous predictor output values but also previous quantized
difference values. In both the predictor and the quantizer, parameter values of the
algorithms are dynamically adjusted based on the statistical properties of previous
samples. As with DPCM, the heart of the decoder is a replication of part of the
logic of the encoder.
The adaptive quantizer takes as input the difference d(k) between the PCM
signal s(k) and the signal estimate se(k) and produces a 4-bit (3 for the magnitude
and 1 for the sign) output ADPCM I(k). The first two columns of Table 11.2 show
the mapping. Prior to quantization, d(k) is converted to a base 2 logarithmic repre-
sentation and scaled by y(k), which is an adaptation factor based on the rate of
change of the signal over recent samples. The result is then mapped into 4 bits for
transmission as the ADPCM signal.

50eneral Aspects of Digital Transmission Systems; Terminal Equipments: 40,32,20,16 kbit/s ADPCM.
1990.
 11.1 / CORDLESS SYSTEMS 327

32 kbps
Difference ADPCM
64 kbps + signal d(k) I Adaptive I I(k)
PCM
s(k) -
I I quantizer I
1---- ----------------------------------- - - - - - - - 1
I I
I I
I I
1 Inverse I
I I
I adaptive 1
I I
I quantizer I
I Signal
I estimate
+ I

1 I Adaptive I Quantized:
I s.(k)
I I predictor I Reconstructed
I + difference I
I signal dik) :
I signal I
I I
I I
Il JI

(a) ADPCM encoder

r----------------------------------------
I
I
I
I
Quantized
32kbps 1 don
I erence Reconstructed
ADPC M I Inverse signal + signal
I 64 kbps
adaptive I PCM
I
I quantizer +
I
I
I Adaptive I

predictor f+-
I I
I Signal I
I 1
I estimate 1
I I
I I
I I
I
IL ~

(b) ADPCM decoder

Figure 11.6 Simplified ADPCM Transmitter and Receiver

Table 11.2 ADPCM Quantizer Normalized Input/Output Characteristics

328 CHAPTER 11/ CORDLESS SYSTEMS AND WIRELESS LOCAL LOOP

Table 11.3 Mean Opinion Score (MOS) Values

Number Score Quality Scale Impairment Scale

5 Excellent Imperceptible
4 Good (Just) perceptible but not annoying
3 Fair (Perceptible and) slightly annoying
2 Poor Annoying (but not objectionable)
1 Unsatisfactory Very annoying (objectionable)

The inverse adaptive quantizer takes the 4-bit signal I(k) and produces a dif-
ference signal dq(k), which serves as input to the adaptive predictor. The second and
third columns of Table 11.3 show the mapping.
The adaptive predictor has the structure of a linear predictor (Figure 11.5) but
is more complex than for DPCM. For ADPCM, the structure is a combination of a
second-order recursive filter and a sixth-order nonrecursive filter. The equation is
[compare Equation (11.1)]
2 6
seek) = 2: Ai(k - i)se(k - i) + 2: Bi(k - i)dq(k - i) (11.2)
i=1 i=1

where the coefficients A and B are updated using algorithms that adapt to the rate
of change of the Se and dq samples. Thus, the adaptive predictor is adjusted to adapt
to the rate of change of the difference samples as well as the rate of change of the
signal estimates.

Subjective Measureluent of Coder Perfonnance The goal of speech

coding is to exploit signal redundancies, signal structure, and knowledge of human
perception to code the signal to various bit rates while maintaining as high a quality
as possible. In assessing the performance of a speech coder, subjective measurements
of quality are more relevant than objective measures such as SNR or the mean
square error. However, reliable values of subjective measurements are in general
more difficult to obtain.
The most widely used subjective measurement as a standard for determin-
ing quality of service is the mean opinion score (MOS), obtained from formal
subjective testing [DAUM82]. For the MOS measurement, a group of subjects
listens to a sample of coded speech, and each classifies the output on a 5-point
scale. To enhance the reliability of the test, it is repeated with various speech
samples and with various groups of subjects. The MOS scale is used in a number
of specifications as a standard for quality (e.g., IEEE 802.16; described in
Section 11.3).
Table 11.3 shows the scale that is used, with two alternative but equivalent
interpretations. Figure 11.7 illustrates the MOS results reported in [RABI95] for
various speech coders, including ADPCM. Note that even uncoded speech does not
get a full rating of 5. This is because subjects may sometimes award a score of 4 to a
 11.2 / WIRELESS LOCAL LOOP 329

5r-----------------------,

4
----
V'l
0
~
'-"
Q,l
;..
Q
~
[IJ

3
....=
Q
....=
c:.
Q

=
eo=
Q,l

~
2

1
Coder Uncoded PCM ADPCM LD-CELP CELP CELP LPC
speech
Bit rate (kbps) 64 32 16 8 4.2 2.4

ADPCM: adaptive differential PCM

CELP: code-excited linear prediction
LD-CELP: low-delay CELP
LPC: linear predictive coding

Figure 11.7 Subjective Speech Quality for Various Speech Coders

sample that ideally deserves a 5 [JAYA84]. With this phenomenon in mind, a score
of 4.0 to 4.5 is a very useful indicator of high-quality digitization.

11.2 WIRELESS LOCAL LOOP

Traditionally, the provision of voice and data communications to the end user, over
the local loop, or subscriber loop, has been provided by wired systems. For residen-
tial subscribers, twisted pair has been and continues to be the standard means of
connection. For business and government subscribers, twisted pair, coaxial cable,
and optical fiber are in use.
As subscribers have demanded greater capacity, particularly to support Internet
use, traditional twisted pair technology has become inadequate. Telecommunications
providers have developed a number of technologies to meet the need, including ISDN
330 CHAPTER 11 / CORDLESS SYSTEMS AND WIRELESS LOCAL LOOP

(integrated services digital network), and a family of digital subscriber loop technolo-
gies, known as xDSL. In addition, cable operators have introduced two-way high-
speed service using cable modem technology. Thus, wired technologies are responding
to the need for reliable, high-speed access by residential, business, and government
subscribers.
However, increasing interest is being shown in competing wireless tech-
nologies for subscriber access. These approaches are generally referred to as
wireless local loop (WLL), or fixed wireless access. WLL alternatives are narrow-
band, which offer a replacement for existing telephony services, and broadband,6
which provide high-speed two-way voice and data service. Table 11.4, based
on one in [WEBBOO] , compares WLL and other alternatives for fixed sub-
scriber support.
In this section, we provide an overview of WLL, then look at some technical
issues, and finally examine two specific WLL approaches. Section 11.3 is devoted to
an evolving standard for WLL known as IEEE 802.16.

Table 11.4 Alternative Providers for End User Access

6The term broadband is not precise. Generally, it refers to systems that provide user data rates of greater
than 2 Mbps, up to 100s of Mbps.
 1l.2 / WIRELESS LOCi\L LOOP 331

The Role ofWLL

Figure 11.8 illustrates a simple WLL configuration. A WLL provider services one or
more cells. Each cell includes a base station antenna, mounted on top of a tall build-
ing or tower. Individual subscribers have a fixed antenna mounted on a building or
pole that has an unobstructed line of sight to the base station antenna. From the
base station, there is a link, which may either be wired or wireless, to a switching
center. The switching center is typically a telephone company local office, which
provides connections to the local and long-distance telephone networks. An Inter-
net service provider (ISP) may be collocated at the switch or connected to the
switch by a high-speed link.
Figure 11.8 shows what amounts to a two-level hierarchy. More complex config-
urations have also been implemented, in which a base station may serve a number of
subordinate base station antennas, each of which supports a number of subscribers.
The WLL has a number of advantages over a wired approach to subscriber
loop support:
• Cost: Wireless systems are less expensive than wired systems. Although the
electronics of the wireless transmitter/receiver may be more expensive than
those used for wired communications, with WLL the cost of installing kilome-
ters of cable, either underground or on poles, is avoided, as well as the cost of
maintaining the wired infrastructure.
• Installation time: WLL systems typically can be installed rapidly. The key
stumbling blocks are obtaining permission to use a given frequency band and
finding a suitable elevated site for the base station antennas. Once these
hurdles are cleared, a WLL system can be installed in a small fraction of the
time required for a new wired system.

Wireless
links

Base station
antenna _------- _--. _~..
. .. I Residence
'"------- III ,m I

"", "
,, , ' " ,
,, " ,
,, '"
Wire ,, '"
link ,,, '"
,, Office
Switching ,, building
center ,,
,,
,,
,,
,
~~===~-, (;overnunent
agency

Figure 11.8 WLL Configuration

332 CHAPTER 11 / CORDLESS SYSTE1V1S AND WIR.ELESS LOCAL LOOP

• Selective installation: Radio units are installed only for those subscribers who
want the service at a given time. With a wired system, typically cable is laid out
in anticipation of serving every subscriber in a local area.
WLL needs to be evaluated with respect to two alternatives:
41Wired scheme using existing installed cable: A large fraction of the earth's
inhabitants do not have a telephone line. For high-speed applications, many
subscribers with telephone lines do not have a line of sufficient quality or are
too far from the central office to effectively use xDSL. Many of these same
subscribers also do not have cable TV or their cable provider does not offer
two-way data services. Finally, because WLL has become cost-competitive
with wired schemes, new installations face a genuine choice between the wired
and wireless approaches.
• Mobile cellular technology: Current cellular systems are too expensive and do
not provide sufficient facilities to act as a realistic alternative to WLL. Even
when 3G systems become available, they are likely to be more expensive and
less functional than broadband WLL alternatives. A major advantage of WLL
over mobile cellular is that, because the subscriber unit is fixed, the subscriber
can use a directional antenna pointed at the base station antenna, providing
improved signal quality in both directions.
In the United States, the Federal Communications Commission has set aside
15 frequency bands for use in commercial fixed wireless service, at frequencies of 2 to
40 GHz. In other countries, similar frequency bands have been allocated. Note that
these frequencies are considerably higher than those used for cellular systems. At
these frequencies, often referred to as millimeter wave frequencies, propagation
characteristics are quite different from those in the MHz ranges. We look at propaga-
tion considerations next. Then we introduce a technique that is used in a number of
WLL systems, known as orthogonal FDM (OFDM). Finally, we examine the two
approaches of most interest for the WLL application: local multipoint distribution
service (LMDS) and multichannel multipoint distribution service (MMDS).

Propagation Considerations for WLL

For most high-speed WLL schemes, frequencies in what is referred to as the milli-
meter wave region are used. Although the term millimeter wave is not precisely
defined, a common boundary is 10 GHz; that is, frequencies above 10 GHz, up to
about 300 GHz, are considered to be in the millimeter wave region'? The reasons for
using frequencies in this range for WLL include the following:
1. There are wide unused frequency bands available above 25 GHz.
2. At these high frequencies, wide channel bandwidths can be used, providing high
data rates.
3. Small size transceivers and adaptive antenna arrays can be used.

7The free space wavelength at 10 GHz is 30 mm, and at 300 GHz is 1 mm.
 11.2 I WIRELESS LOCAL LOOP 333

However, millimeter wave systems have some undesirable propagation

characteristics:
Free space loss increases with the square of the frequency [Equation (5.2)];
thus losses are much higher in this range than in the ranges used for traditional
microwave systems.
2. Generally, below 10 GHz, we can ignore attenuation due to rainfall and atmos-
pheric or gaseous absorption. Above 10 GHz, these attenuation effects are large.
3. Multipath losses can be quite high. As was pointed out in Chapter 5, reflection
occurs when an electromagnetic signal encounters a surface that is large relative
to the wavelength of the signal; scattering occurs if the size of an obstacle is on
the order of the wavelength of the signal or less; diffraction occurs when the
wavefront encounters the edge of an obstacle that is large compared to the
wavelength.
Because of these negative propagation characteristics, WLL systems can only
serve cells of a limited radius, usually just a few kilometers. Also, obstructions,
including foliage, must be avoided along or near the line of sight. Finally, rainfall and
humidity effects limit the range and availability of WLL systems.
Fresnel Zone For effective communication at millimeter wavelengths, there should
be an unobstructed line of sight between transmitter and receiver. The question then
arises as to how much space around the direct path between transmitter and receiver
should be clear of obstacles. A useful criterion relates to the concept of the Fresnel zone.
The definition of Fresnel zones is based on the theory that any small element
of space in the path of an electromagnetic wave may be considered the source
of secondary wavelet, and that the radiated field can be built up by the super-
position of all these wavelets. On the basis of this theory, it can be shown that
objects lying within a series of concentric circles around the direct line of sight
between two transceivers have constructive or destructive effects on commu-
nication. Those that fall within the first circle, the first Fresnel zone, have the most
serious negative effects.
Consider a point along the direct path between a transmitter and receiver, that
is, a distance S from the transmitter and a distance D from the receiver, with the
total distance along the path equal to S + D (Figure 11.9). Then the radius of the
first Fresnel zone at that point is

i
Transmitter f - - - - - - - - - - - - - - - + - - - L - + - - - - - - - - - - - - - - - i Receiver

1 + - - - - - - S - - - - - - - 0 . _ <- - - - - - D - - - - - - - - + - 1

Figure 11.9 The First Fresnel Zone

'I 334 CHAPTER 11 / CORDLESS SYSTEMS AND \VIRELESS LOCAL LOOP

R = (TSD
\jS+i5
where R, S, D, and A are in the same units, and A is the wavelength of the signal aloI
the path. For convenience, this can be restated as

R
m
= 17.3 1_1_ SkmDkm
\j fGHz Skm + D km
where R is expressed in meters, the two distances are in kilometers, and the sign;
frequency is in gigahertz.

It has been found that if there is no obstruction within about 0.6 times the ractin:
of the first Fresnel zone at any point between the two transceivers, then attenuation duc
to obstructions is negligible [FREE97]. One noteworthy example of an obstruction i1
the ground. Hence, the height of the two antennas must be such that there is no POinl
along the path at which the ground is within 0.6 times the radius of the first Fresnel zone
Atmospheric Absorption At frequencies above 10 GHz, radio waves propa-
gating through the atmosphere are subject to molecular absorption. The absorption
as a function of frequency is very uneven, as shown in Figure 11.10. There is a peak
of water vapor absorption at around 22 GHz and a peak of oxygen absorption near
60 GHz. This figure suggests that there is a favorable window for communication
roughly from 28 GHz to 42 GHz, where the attenuation is on the order of
0.13 dB/km, and another favorable window from 75 GHz to 95 GHz, where the
attenuation is on the order of 0.4 dB/km.
However, Figure 11.10 only shows the absorption effects at a particular tem-
perature, relative humidity, and atmospheric pressure. Although the shapes of these
two curves remain the same, the actual values vary, especially with temperature and
relative humidity. Table 11.5, taken from [DALK96], shows the effects of tempera-
ture and humidity.

Effect of Rain One of the most serious concerns for millimeter wave propaga-
tion is attenuation due to rain. The presence of raindrops can severely degrade the
 11.2 / WIRELESS LOCAL LOOP 335

100

50

20
,
02
11
10

2
e
~
'C 1
'-'

::::=
0
ell
=
=
0.5
,
t1
-<
H2O I I~
0.2

~ ~
V
0.1
1 1/1
1
I II
0.05
I
I /
0.02 / /
0.01 /
..,
/
0.005
1
---- 2 5
II

10 20 50 100
Frequency (GHz)
Figure 11.10 Absorption Due to Atmospheric Gas (atmospheric
pressure: 1013 mb; temperature: 15°C; water vapor:
7.5 glm 3 )

reliability and performance of communication links and, during periods of heavy

rain, may outweigh all other factors. The effect of rain on millimeter wave propaga-
tion is complex, depending on drop shape, drop size, rain rate, and frequency.
A widely used formula for estimating attenuation due to rain is
(11.3)

where the attenuation is in units of dB/km, the rain rate R is measured in mm/hr,
and the parameters a and b depend on the distribution of drop sizes and on
336 CHAPTER II/CORDLESS SYSTEMS ANDWTIULESS LOCAL LOOP

Table 11.5 Clear Air Absorption at

lI
28 GHz (dB/km)

Relative Humidity

---
U
° 0°
0%
·0.02
50%

0.05
100%

·0:08
~
...
-=...
= 10°
20°
0.02
• 0.02
0.08
0.12 I
0.14
0.25
e
~

30° ·0.02 0.20 0.44

~ 40° 0.01 0.33 0.79

frequency. Further, attenuation is affected by whether the electromagnetic wave is

horizontally or vertically polarized. Table 11.6, based on [FREE97], gives typical
values for a and b as a function of frequency for horizontal and vertical waves.
Having determined a and b, attenuation then depends on the rain rate R. The
principal concern is the percentage of time the rain rate exceeds a certain threshold.
This depends on the climate zone. Table 11.7, developed by lTD, divides the earth
into 15 climate zones based on precipitation patterns and shows the value of R that
is exceeded for various percentages of time over the course of a year. Such informa-
tion can be used to determine availability of a radio link.

Example 11.4 At a frequency of 30 GHz, with a vertically polarized wave, in climate

zone P, 0.01 % of the time, rainfall exceeds 145 mm/hr. At that rain rate, the attenuation
that must be added to free space loss is A =:= 0.167 X 145 = 24.2 dB/km.

Effects ofVegetation Through part of its path, a WLL link may run through
vegetation, particularly foliage of tall trees. In some suburban and small town
areas, such obstacles may be unavoidable for some subscribers, even with
rooftop subscriber antennas. A study reported in [PAPA97] reached the follow-
ing conclusions:

Table 11.6 Coefficients for Estimating Attenuation Due to Rain

0.880
0.923
1.265
1.264
1.065
1.000
0.929
0.868

j
 11.2 / \-VIRELESS LOCAL LOOP 337

Table 11.7 Rainfall Intensity Exceeded (mm/hr) for Various Rain Regions

Rain Climate Zone

ABC D E FGHJ K LMN P Q

1 0.1 0.5 2.1 0.6 3 2 8 1.5 2 4 5 12 24

0.3 0.8 2 3 4.5. .2.4. 5 7 4 13 4.2 7 11 15 34 49
0.1 2 3 5 8 6 8 12 10 20 12 15 22 35 64 72
0.01 8 12 15 19 22 28 30 32 35 42 60 63 95 145 115

1. The presence of trees near subscriber sites can lead to multipath fading.
2. The principal multipath effects from the tree canopy are diffraction and scattering.
3. Measurements in regularly planted orchards have found attenuation values
between 12 and 20 dB per tree for deciduous trees and up to 40 dB for 1 to 3
coniferous trees, when the foliage is within 60% of the first Fresnel zone.
4. The multipath effects are highly variable due to wind.
Thus, when installing a WLL system, efforts should be made to avoid the pres-
ence of tree foliage within 60% of the first Fresnel zone for each subscriber. How-
ever, the presence of trees does not preclude communications, but means that
adequate countermeasures, such as forward error correction, are required.

Orthogonal Frequency Division Multiplexing

OFDM, also called multicarrier modulation, uses multiple carrier signals at differ-
ent frequencies, sending some of the bits on each channel. This is similar to FDM.
However, in the case of OFDM, all of the subchannels are dedicated to a single
data source.
Figure 11.11 illustrates OFDM. Suppose we have a data stream operating at
R bps and an available bandwidth of Nib, centered at fa. The entire bandwidth could be
used to send the data stream, in which case each bit duration would be 1/R. The alter-
native is to split the data stream into N substreams, using a serial-to-parallel converter.
Each substream has a data rate of RIN bps and is transmitted on a separate subcarrier,
with a spacing between adjacent subcarriers of lb' Now the bit duration is NIR.
To gain a clearer understanding of OFDM, let us consider the scheme in terms
of its base frequency, lb' This is the lowest-frequency subcarrier. All of the other
subcarriers are integer multiples of the base frequency, namely 21b, 31b' and so on, as
shown in Figure 11.12a. The OFDM scheme uses advanced digital signal processing
techniques to distribute the data over multiple carriers at precise frequencies. The
precise relationship among the subcarriers is referred to as orthogonality. The result,
as shown in Figure l1.12b, is that the peaks of the power spectral density of each
subcarrier occurs at a point at which the power of other subcarriers is zero. With
OFDM, the subcarriers can be packed tightly together because there is minimal
interference between adjacent subcarriers.
Note that Figure 11.12 depicts the set of OFDM subcarriers in a frequency
band beginning with the base frequency. For transmission, the set of OFDM
338 CHAPTER 11 / CORDLESS SYSTEMS AND \-\TIRELESS LOCAL LOOP

R/Nbps

10 + (N - 1)1b/2

R/Nbps
Rbps Serial-to-
Parallel R/Nbps
Convertor

R/N bps ,.------,

10 - (N - l)jb 12
Figure 11.11 Orthogonal Frequency Division Multiplexing

subcarriers is further modulated to a higher frequency band. For example, for the
IEEE 802.11a LAN standard, discussed in Chapter 14, the OFDM scheme con-
sists of a set of 52 subca~riers with a base frequency of 0.3125 MHz. This set of
subcarriers is then translated to the 5-GHz range for transmission.
OFDM has several advantages. First, frequency selective fading only affects
some subchannels and not the whole signal. If the data stream is protected by a
forward error-correcting code, this type of fading is easily handled. More impor-
tant, OFDM overcome intersymbol interference (lSI) in a multipath environment.
As discussed in Chapter 5, lSI has a greater impact at higher bit rates, because the
distance between bits, or symbols, is smaller. With OFDM, the data rate is reduced by
a factor of N, which increases the symbol time by a factor of N. Thus, if the symbol
period is T; for the source stream, the period for the OFDM signals is NT;. This
 11.2/ \vIRElESS LOCAL LOOP 339

A
,, "'-,,
I " \

-
I
.-- I
'"
Q I
I

~ I
Q,/
'Q 0
=
.<::
Q..
e
<

-A

t-------1I(2tb) - - - - -

period = T = lItb
(a) Three subcarriers in time domain

S(n

(b) Three subcarriers in frequency domain

Figure 11.12 Illustration of Orthogonality of OFDM

dramatically reduces the effect of lSI. As a design criterion, N is chosen so that NTs is
significantly greater than the root-mean-square delay spread of the channel.
As a result of these considerations, with the use of OFDM, it may not be
necessary to deploy equalizers, which are complex devices whose complexity
increases with the number of symbols over which lSI is present.
A common modulation scheme used with OFDM is quadrature phase shift key-
ing (QPSK). In this case, each transmitted symbol represents two bits. An example of an
OFDMlQPSK scheme, used in a MMDS system [CISCOO], occupies 6 MHz made up of
512 individual carriers, with a carrier separation of a little under 12 kHz. To minimize
340 CHAPTER 11 / CORDLESS SYSTEMS AND \VIRELESS LOCAL LOOP

lSI, data are transmitted in bursts, with each burst consisting of a cyclic prefix followed
by data symbols. The cyclic prefix is used to absorb transients from previous bursts
caused by multipath. For this system, 64 symbols constitute the cyclic prefix, followed by
512 QPSK symbols per burst. On each subchannel, therefore, QPSK symbols are sepa-
rated by a prefix of duration 64/512 symbol times. In general, by the time the prefix is
over, the resulting waveform created by the combined multipath signals is not a func-
tion of any samples from the previous burst. Hence there is no lSI.

Multichannel Multipoint Distribution Service

Table 11.8 shows five frequency bands in the range 2.15 GHz to 2.68 GHz that have
been allocated in the United States for fixed wireless access using MMDS. The first
two bands were licensed in the 1970s when they were called multipoint distributions
services (MDSs), for broadcast of 6 MHz TV channels. In 1996, the FCC increased
the allocation to its present range and allowed for multichannel services, called
MMDS. MMDS has been used to compete with cable TV providers and to provide
service in rural areas not reached by broadcast TV or cable. For this reason, MMDS
is also referred to as wireless cable.
The transmitted power allowed by the FCC enables an MMDS base station to
service an area with a radius of 50 km, but subscriber antennas must be in the line of

Table 11.8 Fixed Wireless Communications Bands, FCC Allocation

ISM = industrial, scientific, and medical

LMDS = local multipoint distribution service
MDS = multipoint distribution service
MMDS = multichannel multipointatribution service
UNII = unlicensed national information infrastructure
 11.2 / WIRELESS LOCAL LOOP 341

sight. MMDS can be used to support two-way services. MMDS is also used in other
countries for two-way access. Thus, MMDS is an alternative for broadband data
services, such as Internet access.
The principal disadvantage of MMDS, compared to LMDS, which we discuss
next, is that at the lower frequency of MMDS, it offers much less bandwidth than
LMDS. With current technology, a single MMDS channel can offer upstream (sub-
scriber to base station) transfer rates of 27 Mbps, with individual subscriber rates of 300
kbps to 3 Mbps [ORTIOO]. However, developments in OFDM may allow higher data
rates or more subscriber support for upstream MMDS. In any case, because of its lesser
bandwidth, MMDS is likely to be used mainly by residential subscribers and small busi-
nesses, whereas LMDS appeals to larger companies with greater bandwidth demands.
The advantages of MMDS over LMDS include the following:
1. MMDS signals have larger wavelengths (greater than 10 cm) and can travel
farther without losing significant power. Hence MMDS can operate in consid-
erably larger cells, thereby lowering base station equipment costs.
2. Equipment at lower frequencies is less expensive, yielding cost savings at both
the subscriber and base station.
3. MMDS signals don't get blocked as easily by objects and are less susceptible
to rain absorption.

Local Multipoint Distribution Service

LMDS is a relatively new WLL service to deliver TV signals and two-way broad-
band communications, operating at millimeter frequencies. In the United States,
LMDS will be offered at frequencies near 30 GHz; in Europe and some other areas,
frequencies near 40 GHz will be used. Table 11.8 shows frequency bands that have
been allocated in the United States for fixed wireless access using LMDS.
LMDS has the following advantages:
• Relatively high data rates, in the Mbps range
• Capable of providing video, telephony, and data
• Relatively low cost in comparison with cable alternatives
The principal disadvantage of LMDS is the short range from the base station,
requiring a relatively large number of base stations to service a given area.
In a typical system, the base station antenna is located on top of a tall building
or high pole overlooking the service area, with line of sight to subscribers with the
possible exception of tree canopies. The base station antenna covers a sector 60° to
90° wide. Thus, full coverage requires 4 to 6 antennas. A radius of 2 to 4 km is typical.
With the bandwidths available, an upstream data rate from the subscriber of up to
1 Mbps is reasonable, with a downstream capacity of 36 Mbps.
An important consideration is that LMDS's short-wavelength signals cannot
pass around or penetrate objects like buildings, walls, or thick foliage. A certain
amount of foliage can be tolerated, as is discussed earlier in this section, but dense
foliage may require other countermeasures. By overlapping cell coverage, it may be
possible to obtain coverage in shielded areas in one cell from the base station of a
neighboring cell. Use of repeaters or reflectors is another possibility.
342 CHAPTER 11 / CORDLESS SYSTEMS AND vVIRELESS LOCAL LOOP

." »

11.3 ~~'1lm15 SO~ ..16 BRCi}1IDB~~I1U.. J

lIC~I1SS S'J?1INB1IR.IlS .:,

With the growing interest in LMDS WLL services, a need was recognized within the
industry to develop standards for this service. In response to this need the IEEE 802
committee set up the 802.16 working group in 1999 to develop broadband wireless
standards. The charter for the group is to develop standards that [MARK99]
• Use wireless links with microwave or millimeter wave radios
• Use licensed spectrum (typically)
• Are metropolitan in scale
• Provide public network service to fee-paying customers (typically)
• Use point-to-multipoint architecture with stationary rooftop or tower-mounted
antennas
• Provide efficient transport of heterogeneous traffic supporting quality of
service (QoS)
• Are capable of broadband transmissions (>2 Mbps)
In essence, IEEE 802.16 standardizes the air interface and related functions
associated with LMDS. As of this writing, standards have been developed in the
areas indicated in Table 11.9. In addition, an industry group, the WiMAX (World-
wide Interoperability for Microwave Access) Forum, has been formed to promote
the 802.16 standards and to develop interoperability specifications.
The work on 802.16 in Table 11.9 is the farthest along and is the one most likely
to generate the most industry interest, as it is targeted at available LMDS frequency
bands. In this section, we provide an overview of 802.16 architecture and services and
then look in more detail at the 802.16 specification. Finally, a summary of the 802.16a
standard is presented.

IEEE 802.16 Architecture

Systenl Reference Architecture The 802.16 standards are designed with respect
to the abstract system reference model shown in Figure 11.13. An 802.16 wireless

Table 11.9 IEEE 802.16 Standards

IEEE802.16c
IEEE 802.16e Physical and Medium Access Control Layers for Combined Fixed and
Mobile Operation in Licensed Bands
IEEE 802.16.2 Coexistence of Fixed Broadband Wireless Access Systems
 11.3 /WIMAXAND IEEE 802.16 BROADBAND WIRELESS ACCESS STANDARDS 343

- - - - - - - - - - - - - End-to-end delay - - - - - - - - - - - -.....

_ - - - - - - Transit delay ------~.

MAC delay

TBNI-l
E

Provi~ed
SNI~;n::;a<e
Subscriber Core I
terminal network I serVIce

Repeater
(optional)

SNI: STS network interface

STS: Subscriber transceiver station
BTS: Base transceiver station
BNI: BTS network interface
Figure 11.13 IEEE 802.16 System Reference Points

service provides a communications path between a subscriber site, which may be either
a single subscriber device or a network on the subscriber's premises (e.g., a LAN, PBX,
IP-based network) and a core network (the network to which 802.16 is providing
access). Examples of a core network are the public telephone network and the Internet.
Three interfaces are defined in this model. IEEE 802.16 standards are concerned with
the air interface between the subscriber's transceiver station and the base transceiver
station. The standards specify all the details of that interface, as discussed subsequently
in this subsection. The system reference model also shows interfaces between the trans-
ceiver stations and the networks behind them (SNI and BNI). The details of these inter-
faces are beyond the scope of the 802.16 standards. The reason for showing these
interfaces in the system reference model is that the subscriber and core network tech-
nologies (such as voice,ATM, etc.) have an impact on the technologies used in the air
interface and the services provided by the transceiver stations over the air interface.
Finally, the system reference model includes the optional use of some sort of
repeater. The air interface specification allows for the possibility of repeaters or
reflectors to bypass obstructions and extend cell coverage.
Protocol Architecture Protocols defined specifically for wireless transmission
address issues relating to the transmission of blocks of data over the network. In
OSI terms, higher-layer protocols (layer 3 or 4 and above; see Figure 4.3) are inde-
pendent of network architecture and are applicable to a variety of networks and
communications interfaces. Thus, a discussion of 802.16 protocols is concerned with
lowest two layers of the OSI model.
Figure 11.14 relates the four protocol layers defined in the 802.16 protocol
architecture to the OSI model. Working from the bottom up, the lowest two layers
of the 802.16 protocol model correspond to the physical layer of the OSI model and
include such functions as
• Encoding/decoding of signals
• Preamble generation/removal (for synchronization)
• Bit transmission/reception
344 CHAPTER 11 / CORDLESS SYSTEMS AND WIRELESS LOCAL LOOP

Digital audio/ Digital Bridged Back Virtual Frame

video multicast telephony ATM IP LAN haul point-to-point relay

Convergence
OSIdata
link layer
Medium access control

Transmission
OSI physical
layer
Physical

Figure 11.14 IEEE 802.16 Protocol Architecture

In addition, the physical layer of the 802 model includes a specification of the trans-
mission medium and the frequency band. Generally, this is considered "below" the
lowest layer of the OSI model. However, the choice of transmission medium and
frequency band is critical in wireless link design, and so a specification of the
medium is included. In general, the 802.16 physical layer is concerned with these
medium-dependent issues, and the transmission layer is concerned with the bulleted
items listed previously.
Above the physical and transmission layers are the functions associated with
providing service to subscribers. These include

• On transmission, assemble data into a frame with address and error detec-
tion fields.
• On reception, disassemble frame, and perform address recognition and error
detection.
• Govern access to the wireless transmission medium.

These functions are grouped into a medium access control (MAC) layer. The
protocol at this layer, between the base station and the subscriber station, is
responsible for sharing access to the radio channel. Specifically, the MAC protocol
defines how and when a base station or subscriber station may initiate transmis-
sion on the channel. Because some of the layers above the MAC layer, such
as ATM, require specified service levels (QoS), the MAC protocol must be able
to allocate radio channel capacity so as to satisfy service demands. In the
downstream direction (base station to subscriber stations), there is only one trans-
mitter and the MAC protocol is relatively simple. In the upstream direction, mul-
tiple subscriber stations are competing for access, resulting in a more complex
MAC protocol.
 J 1.3 /WIMAX AND IEEE 802. J 6 BRO.WBANDWIRELESS ACCESS STANDARDS 345

Above the MAC layer is a convergence layer that provides functions specific
to the service being provided. A convergence layer protocol may do the following:
• Encapsulate PDU (protocol data unit) framing of upper layers into the native
802.16 MACIPHY frames.
• Map an upper layer's addresses into 802.16 addresses.
• Translate upper layer QoS parameters into native 802.16 MAC format.
• Adapt the time dependencies of the upper layer traffic into the equivalent
MAC service.
In some cases, such as digital audio and video, no convergence layer is needed
and the stream of digital data is presented to the transmission layer. Upper-layer
services that make use of a PDU structure do require a convergence layer.
An example of the protocol structure supported by the convergence layer is
the handling ofTCP/IP based traffic, as shown in Figure 11.15 (compare Figure 4.2).
Higher-level data are passed down to LLC (logical link control), which appends
control information as a header, creating an LLC protocol data unit (PDU). This
control information is used in the operation of the LLC protocol, which is a form of
data link control protocol (see Appendix C). The entire LLC PDU is then passed
down to the MAC layer, which appends control information at the front and back of
the packet, forming a MAC frame. Again, the control information in the frame is
needed for the operation of the MAC protocol. The figure shows the use of TCP/IP
and an application layer above the 802.16 protocols.

Application data Application layer

TCP layer
I
J

I1PI
~

LLC layer
I
I

I h~;r I MAC layer

I I

I 1-_ _'------- _
I I

- - - - - - TCP segment - - - - - -
~.-------IP datagram - - - - - - - + -
..

- - - - - - - LLC protocol data unit -------~

..
• MAC frame - - - - - - - - - - - - -
Figure 11.15 IEEE 802.16 Protocols in Context
346 CHAPTER 11 / CORDLESS SYSTEiVIS AND WIRELESS LOCAL LOOP

Services
Requirements for the IEEE 802.16 standards are defined in terms of bearer services
that the 802.16 systems must support. A bearer service refers to the type of traffic
generated by a subscriber network or core network in Figure 11.13. For example, an
802.16 interface must be able to support the data rate and QoS required by an ATM
network or an IP-based network, or support the data rate and delay requirements of
voice or video transmissions.
IEEE 802.16 is designed to support the following bearer services:

• Digital audio/video multicast: Transports one-way digital audio/video streams

to subscribers. The principal example of this service is a broadcast radio and
video similar to digital broadcast cable TV and digital satellite TV. A special
case of this service is two-way video such as in teleconferencing. In this latter
case, delay requirements are stringent because of the interactivity involved.
• Digital telephony: Supports multiplexed digital telephony streams. This service
is a classic WLL service that provides a replacement for wired access to the
public telephone network.
• ATM: Provides a communications link that supports the transfer of ATM cells
as part of an overall ATM network. The 802.16 link must support the various
QoS services defined for ATM.
• Internet protocol: Supports the transfer of IP datagrams. The 802.16 link must
provide efficient timely service. In addition, a variety of QoS services are now
defined for IP-based networks, and 802.16 should support these.
• Bridged LAN: Similar to the IP-based support. A bridge LAN service enables
transfer of data between two LANs with switching at the MAC layer.
• Back-haul: For cellular or digital wireless telephone networks. An 802.16 system
may be a convenient means to provide wireless trunks for wireless telephony
base stations.
• Frame relay: Similar to ATM. Frame relay uses variable-length frames in con-
trast to the fixed-length cells of ATM.

Another way of viewing the service requirements for 802.16 is shown in

Table 11.10, which is taken from the 802.16 functional requirements document.
Bearer services are grouped in three broad categories:

• Circuit based: These services provide a circuit-switching capability, in which

connections are set up to subscribers across a core network.
• Variable packet: IP and frame relay are examples of services that make use of
variable-length PDUs. Another example is MPEG video, which is a video
compression scheme in which successive blocks of digital video information
may be of varying sizes.
• Fixed-length celUpacket: This service is for ATM.

Table 11.10 summarizes requirements in three categories. The first category is the
data rate that must be supported. The second category refers to error performance. For
 11.3 /\VIMAXAND IEEE 802.16 BROADBAND WIRELESS ACCESS STANDARDS 347

Table 11.10 IEEE 802.16 Services and QoS Requirements

most services an upper limit on the bit error ration (BER) is defined. For ATM, various
specific QoS error parameters are also used.
The final category is maximum one-way delay. To place this delay in context,
Figure 11.13 shows three categories of delay defined in the 802.16 standards:
• Medium access delay: Once a transceiver station is ready to transmit, the
medium access delay measures the amount of time that the station must wait
before it can transmit.
• Transit delay: This is the delay from SNI to BNI or BNI to SNI. It includes the
medium access delay plus the processing at the MAC layer for preparing
transmission (from the STS or BTS) and at the MAC layer for reception (at
the BTS or STS).
• End-to-end delay: The total delay between a terminal in the subscriber network,
to the ultimate service beyond the core network. This includes the transit delay.
The maximum one-way delay category specified in Table 11.10 refers to
transit delay.
348 CHAPTER II/CORDLESS SYSTEMS AND WIRELESS LOCAL LOOP

IEEE 802.16 MAC Layer

Data transmitted over the 802.16 air interface from or to a given subscriber are struc-
tured as a sequence of MAC frames. The term MAC frame as used in this context
refers to the PDU that includes MAC protocol control information and higher-level
data. This is not to be confused with a TDMA frame, which consists of a sequence of
time slots, each dedicated to a given subscriber. A TDMA time slot may contain
exactly one MAC frame, a fraction of a MAC frame, or multiple MAC frames. The
sequence of time slots across multiple TDMA frames that is dedicated to one sub-
scriber forms a logical channel, and MAC frames are transmitted over that logical
channel.
Connections and Service Flow The 802.16 MAC protocol is connection
oriented. That is, a logical connection is set up between peer entities (MAC users)
prior to the exchange of data between those entities. Each MAC frame includes a
connection ID, which is used by the MAC protocol to deliver incoming data to the
correct MAC user. In addition, there is a one-to-one correspondence between a con-
nection ID and service flow. The service flow defines the QoS parameters for the
PDUs that are exchanged on the connection.
The concept of a service flow on a connection is central to the operation of the
MAC protocol. Service flows provide a mechanism for upstream and downstream
QoS management. In particular, they are integral to the bandwidth allocation
process. The base station allocates both upstream and downstream bandwidth on
the basis of the service flow for each active connection. Examples of service flow
parameters are latency (maximum acceptable delay), jitter (maximum acceptable
delay variation), and throughput (minimum acceptable bit rate).
FraIne Fonnat A good way to get a grasp of the MAC protocol is to ~xamine the
frame format (Figure 11.16). The frame consists of three sections:
• Header: Contains protocol control information needed for the functioning of
the MAC protocol.
• Payload: The payload may be either higher-level data (e.g., an ATM cell, an IP
packet, a block of digital speech) or a MAC control message.
• CRC: The cyclic redundancy check field contains an error-detecting code (dis-
cussed in Section 8.1).
Three header formats are defined. There is a generic header format in both
the uplink (toward the base station) and downlink (toward the subscriber)
directions. These formats are used for frames that contain either higher-level
data or a MAC control message. The third format is used for a bandwidth
request frame.

48 or 56 bits 2: 0 bits 32 bits

_MAC
_ _____
header ~
_ _ _ _ _.....CRC 1
Payload (optional)

Figure 11.16 IEEE 802.16.1 Frame Format

11.3/ \VIMAX AND IEEE 802.16 BROADBAND WIRELESS ACCESS STANDARDS 349

EC

Connection identifier

HCS

(a) Generic downlink header

I I I

EC I EKS
I Length

I I
I C0rn;tection ide~tifier, I I I I I
I I I I I I

~~IARQI
I I I I I

I I
FC
I FSN Grant management

HCS
I I I

Grant management
.
Unsolicited grant service
~
~,-_-,-_G...L:r_a_n_ts.....:p_e_r_inL-:t_er_v......~_l_-'-_
Unsolicited grant service
with activity detection

All others

(b) Generic uplink header

1 10

Connection identifier
HT
=1 Bandwidth request

(c) Bandwidth request header

ARQ: Automatic repeat request HCS: Header check sequence

EC: Encryption control HT: Header type
EKS: Encryption key sequence PM: Poll me bit
FC: Fragment control SI: Slip indicator
FSN: Fragment sequence number

Figure 11.17 IEEE 802.16 MAC Header Formats

The downlink header format is shown III Figure 11.17a. It consists of the
following fields:
• Encryption control (1 bit): Indicates whether the payload is encrypted.
• Encryption key sequence (4 bits): An index into a vector of encryption key
information, to be used if the payload is encrypted.
• Length (11 bits): Length in bytes of the entire MAC frame.
350 CHAPTER II/CORDLESS SYSTEMS AND \,VIRELESS LOCAL LOOP

• Connection identifier (16 bits): A unidirectional, MAC-layer address that iden-

tifies a connection to equivalent peers in the subscriber and base station MAC.
1
A CID maps to a SFID, which defines the QoS parameters to the service flow
associated with that connection.
• Header type (1 bit): Indicates whether this is a generic or bandwidth
request header.
• ARQ indicator (1 bit): Indicates whether the frame belongs to an ARQ enabled
connection. If so, the ARQ mechanism described in Section 8.4 is used, and a
2-byte control field is prepended at the beginning of the frame. The control bit
structure contains a 4-bit retry number and a 12 -bit sequence number. The retry
number field is reset when a packet is first sent and is incremented whenever it
is retransmitted (up to the terminal value of 15). The sequence number field is
assigned to each packet on its first transmission and then incremented.
• Fragment control (2 bits): Used in fragmentation and reassembly, as explained
subsequently.
• Fragment sequence number (4 bits): Sequence number of the current fragment.
• Header check sequence (8 bits): An 8-bit CRC used to detect errors in the header.

Fragmentation is used to divide a higher-level block of data into two or more

fragments in order to reduce the size of MAC frames. This is done to allow efficient
use of available bandwidth relative to the QoS requirements of a connection's ser-
vice flow. If fragmentation is not used, then the fragment control (FC) field is set to
00. If fragmentation is used, then the all of the fragments are assigned the same frag-
ment sequence number (FSN) and the FC field has the following interpretation: first
fragment (10); intermediate fragment (11); last fragment (01). The MAC user at the
destination is responsible for reassembling all of the fragments with the same FSN.
The uplink header format is shown in Figure 11.17b. It contains all of the fields
of the downlink header, plus an 8-bit grant management field. This field is used by
the subscriber to convey bandwidth management needs to the base station. There
are three different encodings of this field, depending on the type of connection. The
subfields within the GM field include
• Slip indicator (1 bit): If set, indicates a slip of uplink grants relative to the
uplink queue depth.
• Poll-me (1 bit): If set, requests a poll by the base station.
• Grants per interval (7 bits): The number of grants required by a connection.
• Piggyback request (8 bits): The number of bytes of uplink capacity requested
by the subscriber for this connection.
The first two formats for the GM field are associated with the unsolicited
grant service (UGS). This service is designed to support real-time service flows. In
essence, the base station periodically, using MAC management messages, grants an
allocation of bytes to the subscriber on a given connection. The allocation is
designed to keep up with the real time demands. If a subscriber finds that its queue
of data to send has exceeded a threshold, the subscriber sends a GM field with the
SI bit set and either requests a poll for bandwidth by setting the PM bit or requests I
J
11.3 / \VIMAX AND IEEE 802.16 BROADBAND WIRELESS ACCESS STANDARDS 351

that a given number of bandwidth grants be executed in the next time interval. The
latter technique is used if this is a UGS with activity detection; this simply means
that the flow may become inactive for substantial periods of time. For other types of
service, the GM field may be used to make a request for capacity. This is referred to
as a piggyback request because the request is made as part of a MAC frame carrying
user data rather than in a separate bandwidth request management MAC frame.
Finally, the bandwidth request header is used by the subscriber to request
additional bandwidth. This header is for a MAC frame with no payload. The IS-bit
bandwidth request field indicates the number of bytes of capacity requested for
uplink transmission.

MAC Managell1ent Messages IEEE 802.16 defines a number of control mes-

sages that are used by the base station and the subscriber to manage the air interface
and manage the exchange of data over the various connections. Messages are used to
exchange operating parameters and status and encryption-related information and
for capacity management. The following is a brief summary of the messages currently
defined:

o Uplink and downlink channel descriptor: Transmits characteristics of the

physical channel.
41 Uplink and downlink access definition: Allocates access to the uplink and

downlink channels.
e Ranging request and response: The request is used by the subscriber to deter-
mine network delay and to request power and/or modulation adjustment. The
response from the BS contains the requested parameters.
• Registration request, response, and acknowledge: The request is transmitted by
the subscriber at initialization time and contains various relevant parameters. The
base station replies with a response, and the subscriber sends an acknowledge to
complete the handshake.
• Privacy key management request and response: Used to exchange information
about encryption keys.
41 Dynamic service addition request, response, and acknowledge: The request is

sent by a subscriber to request a new service flow.

e Dynamic service change request, response, and acknowledge: The request is
sent by a subscriber or the base station to dynamically change the parameters
of an existing service flow.
e Dynamic service deletion request and response: The request is sent by the
base station or subscriber to delete an existing service flow.
til Multicast polling assignment request and response: Sent by a subscriber to

request inclusion in a multicast polling group.

o Downlink data grant type request: Sent by the base station to the subscriber to
dictate the modulation method and FEe technique to be used on a particular
connection, to enhance performance.
.. AR Q acknowledgment: Used to acknowledge successful receipt of one or
more packets from the peer MAC entity.
352 CHAPTER 11 / CORDLESS SYSTEMS AND WIRELESS LOCAL LOOP

Table 11.11 IEEE 801.16 Physical Layer Modes

DAMA-TDMA: Demand assignement multiple access, time division multiple access

FDD: Frequency division duplexing
FSDD: Frequency shift division duplexing
TDD: Time division duplexing
TDM: Time division multiplexing

IEEE 802.16 Physical Layer

The 802.16 physical layer supports a different structure for the point-to-
multipoint downstream channels and the multipoint-to-point upstream channels
(Table 11.11). These structures reflect the differing requirements in the two direc-
tions. In general, most systems will require greater downstream capacity to
individual subscribers to support asymmetric data connections such as Web appli-
cations over the Internet. For the upstream direction, the issue of medium access
needs to be addressed, because there are a number of subscribers competing
for the available capacity. These requirements are reflected in the physical layer
specification.
Upstreal11. Transluission Upstream transmission uses a DAMA-TDMA (demand
assignment multiple access-time division multiple access) technique. As was defined in
Chapter 9, DAMA is a capacity assignment technique that adapts as needed to opti-
mally respond to demand changes among the multiple stations. TDMA is simply the
technique of dividing time on a channel into a sequence of frames, each consisting of a
number of slots, and allocating one or more slots per frame to form a logical channeL
With DAMA-TDMA, the assignment of slots to channels varies dynamically.
Upstream transmission makes use of a Reed-Solomon code for error correc-
tion and a modulation scheme based on QPSK.
Downstreal11. Transulission In the downstream direction, the standard speci-
fies two modes of operation, one targeted to support a continuous transmission
stream (mode A), such as audio or video, and one targeted to support a burst trans-
mission stream (mode B), such as IP-based traffic.
For the continuous downstream mode, a simple TDM scheme is used for chan-
nel access. The duplexing technique that is used for allocating capacity between
upstream and downstream traffic is known as FDD (frequency division duplex).
FDD simply means that a different frequency band is used for transmission in each
direction. This is equivalent to a FAMA-FDMA (fixed assignment multiple access-
frequency division multiple access) scheme. FDD implies that all subscribers can
transmit and receive simultaneously, each on their own assigned frequencies.
11.3 I \VIMAX AND IEEE 802.16 BROADBAND \VIRELESS ACCESS STANDARDS 353

For the burst downstream mode, the DAMA-TDMA scheme is used for chan-
nel access. Three alternative techniques are available for duplexing traffic between
upstream and downstream:
• FDD with adaptive modulation: This is the same FDD scheme used in the
upstream mode, but with a dynamic capability to change the modulation and
forward error correction schemes.
• Frequency shift division duplexing (FSDD): This is similar to FDD, but
some or all of the subscribers are not capable of transmitting and receiving
simultaneously.
• Time division duplexing (TDD): This technique is discussed in Section 11.1.
A TDMA frame is used, with part of the time allocated for upstream transmis-
sion and part for downstream transmission.
The availability of these alternative techniques provides considerable flexibil-
ity in designing a system that optimizes the use of capacity.

IEEE 802.16a
The initial version of the 802.16 standard, ratified in 2002 and referred to simply as
IEEE 802.16, operates in the 10-to-66-GHz frequency band and requires line-of-
sight antennas. The IEEE 802.16a standard, ratified in 2003, does not require line-
of-sight transmission and allows use of lower frequencies (2 to 11 GHz), many of
which are unregulated. IEEE 802.16a can achieve a range of 50 km and data rates of
over 70 Mbps.
IEEE 802.16a provides a new physical layer standard with some modifica-
tions to the 802.16 MAC layer. Compared to the higher frequencies, the 2-to-
ll-GHz range offers the opportunity to reach many more customers less
expensively, though at generally lower data rates. This suggests that such services
will be oriented toward individual homes or small- to medium-sized enterprises.
Design of the 2-to-11-GHz physical layer is driven by the need for non-line-of-
sight (NLOS) operation. This is essential to support residential applications since
rooftops may be too low for a clear sight line to a BS antenna, possibly due to
obstruction by trees. Therefore, significant multipath propagation must be
expected. Furthermore, outdoor mounted antennas are expensive due to both
hardware and installation costs. The 802.16a standard specifies that systems
implement one of three air interface specifications, each of which provides for
interoperability:
• WirelessMAN-SCa: This uses a single-carrier modulation format.
e WirelessMAN-OFDM: This uses orthogonal frequency division multiplexing
with a 256-point transform. Access is by TDMA.
• WirelessMAN-OFDMA: This uses orthogonal frequency division multiple
access with a 2048-point transform. In this system, multiple access is provided
by addressing a subset of the multiple carriers to individual receivers
Because of the propagation requirements, the use of advanced antenna sys-
tems is supported. To accommodate the more demanding physical environment and
354 CHAPTER 11 / COIU)LESS SYSTEMS AND WIRELESS LOCAL LOOP

Table 11.12 Data Rates Achieved at Various 802.16a Bandwidths

Modulation QPSK QPSK 16QAM 16QAM 64QAM 64QAM

Code Rate 112 3/4 112 3/4 2/3 3/4
1.75 MHz 1.04 2.18 2.91 4.36 5.94 6.55
3.5 MHz 2.08 4.37 5.82 8.73 11.88 13.09
7.0 MHz. 4.15 8.73 11.64 17.45 23.75 26.18
10.0 MHz 8.31 12.47 16.63 24.94 33.25 37.40
20.0 MHz 16,62 24.94 33.25 49.87 66.49 74.81

different service requirements found at frequencies between 2 and 11 GHz, the

802.16a project upgrades the MAC to provide automatic repeat request (ARQ).
Also, an optional mesh topology is defined to expand the basic point-to-multipoint
architecture.
IEEE 802.16a supports both TDD and FDD modes of operation, along with a
range of channel bandwidths. Table 11.12 shows the data rates that can be achieved
at various 802.16a channel bandwidths.

[PHIL98] is a detailed treatment of DECT and PWT.

A thorough analysis of propagation effects at millimeter wavelengths is found in
[FREE97]. A briefer but useful survey of the same topic is [DALK96].
[NORDOO] is a general overview of LMDS technology and applications. [FUNG98]
focuses on some detailed transmitter/receiver design issues for MMDS. [WEBBOO] is a good
overall survey of wireless local loop technology. [KELLOO] provides a thorough analysis of
OFDM. [BOLC01] surveys design issues related to broadband wireless access. [MORA04]
provides a detailed technical treatment of broadband wireless access.
[EKLU02] provides an overview of IEEE 802.16. [KOFF02] and [JOHN04] cover the
OFDM portion of 802.16a.

BOLCOI Bolcskei, H., et al. "Fixed Broadband Wireless Access: State of the Art,
Challenges, and Future Directions." IEEE Communications Magazine, January
2001.
DALK96 Dalke, R.; Hufford, G.; and Ketchum, R. Radio Propagation Considerations for
Local Multipoint Distribution Systems. National Telecommunications and Informa-
tion Administration Publication PB97116511, August 1996.
EKLU02 Eklund, c., et al. "IEEE Standard 802.16: A Technical Overview of the Wire-
lessMAN Air Interface for Broadband Wireless Access." IEEE Communications
Magazine, June 2002.
FREE97 Freeman, R. Radio System Design for Telecommunications. New York: Wiley, 1997.
FUNG98 Fung, P. "A Primer on MMDS Technology." Communication Systems Design,
April 1998. Available at www.commsdesign.com.
JOHN04 Johnston, D., and Yaghoobi, H. "Peering into the WiMAX Spec." Communica~
tions System Design, January 2004. Available at www.commsdesign.com.
 11.5/ KEY TERMS, REVIEW QUESTIONS,AND PROBLEMS 355

KELLOO Keller, T., and H. "Adaptive Multicarrier Modulation: A Convenient Frame-

work for Time-Frequency Processing in Wireless Communication." Proceedings of
the IEEE, May 2000.
KOFF02 Koffman, I., and Roman, V. "Broadband Wireless Access Solutions Based on
OFDMAccess in IEEE 802.16." IEEE Communications Magazine, April 2002.
MORA04 Morais, D. Fixed Broadband Wireless Communications: Principles and Practi-
cal Applications. Upper Saddle River, NJ: Prentice Hall, 2004.
NORDOO Nordbotten, A. "LMDS Systems and Their Application." IEEE Communica-
tions Magazine,June 2000.
PHIL98 Phillips, J.; and Namee, G. Personal Wireless Communications with DECT and
PWT. Boston: Artech House, 1998.
WEBBOO Webb, W. Introduction to Wireless Local Loop: Broadband and Narrowband
Systems. Boston: Artech House, 2000.

Recomlnended Web Sites:

• Wireless Communications Association International: Represents the fixed broadband
wireless access industry worldwide. Good set of links.
• The IEEE 802.16 Working Group on Broadband Wireless Access Standards: Contains
working group documents plus discussion archives.
• WiMAX Forum: An industry group promoting the interoperability of 802.16 products
with each other.
• NIST Radio Frequency Technology Division: The U.S. National Institute of Stan-
dards and Technology. Good source of information on fixed wireless access and
other topics.
• OFDM Forum: Industry consortium to promote OFDM.

Key TenTIs

adaptive differential PCM fixed wireless access multicarrier modulation

(ADPCM) Fresnel zone multichannel multipoint distri-
cordless linear predictive filter bution service (MMDS)
differential quantization local multipoint distribution orthogonal frequency division
digital enhanced cordless service (LMDS) multiplexing (OFDM)
telecommunications logical link control (LLC) time division duplex (TDD)
(DECT) MAC frame wireless cable
fixed broadband wireless medium access control (MAC) wireless local loop (WLL)
356 CHAPTER 11 / CORDLESS SYSTEMS AND\VIRELESS LOCAL LOOP

Revie-w Questions
11.1 In what ways has traditional cordless telephony been extended by standardization?
11.2 What is the difference between TDD and TDM?
11.3 List and briefly define the DECT logical channels.
11.4 What are some key advantages of WLL over a wired subscriber loop?
11.5 What are the principal advantages of millimeter wave transmission compared to
transmission at longer wavelengths?
11.6 What are the principal disadvantages of millimeter wave transmission compared to
transmission at longer wavelengths?
11.7 What are the principal advantages of OFDM?
11.8 What are the relative advantages of MMDS compared to LMDS?
11.9 What are the key differences among IEEE 802.16, IEEE 802.16.2, and IEEE 802.16.3?

ProbleU1S
11.1 GSM, described in Chapter 10, makes use of a TDMA scheme in which there are a
number of duplex channels, with each channel consisting of two dedicated frequency
bands, one for transmission in each direction. Suppose that we wished to design a
comparable system, using the same total amount of bandwidth, but using TDD. List
the key parameters of a TDD-GSM system and compares these to the real GSM
parameters.
11.2 A LOS microwave link 12 km long is planned for installation in Germany, which is in
climate zone H (Table 11.7). The operational frequency is 38 GHz and the desired
time availability is 99.99%. Local rain rate statistics are not available. The antenna
polarization is horizontal.
a. What amount of attenuation is to be anticipated?
b. Repeat for 99.9% and 99% and comment on the practicality of the requirement
for various availabilities.

The linear predictive filter is based on using an approximation of the Taylor series. For a funCtion
get) that has derivatives of all orders at t, the function can be represented by its Taylor series:

g(x) = :L 00 (x -

n=O n.
, tt
g(n)(t) (11.4)

where g(n)(t) is the nth derivative of get). Now, suppose we are going to sample get) with a
sampling interval of T s. Then, defining T s = x - t, we can restate Equation (11.4) as
oo(TY
get + T s ) = :L _ s , g(n)(t) (11.5)
n.
n=O

Equation (11.5) shows that the value of the signal at time t + Ts can be predicted from
the value of its signal and its derivatives at time t. In fact, for small values of Ts (much less
than one), we can produce a good approximation with just the first derivative:
(11.6)

8The development in this appendix is based on material in [LATH98].

)
 APPENDIX l1A I UNEAR PREDICTIVE FILTERS 357

Let us denote the kth sample of get) taken at time t = kTs as m(k); then
m(k ± 1) = g(kTs ± Ts). Further, we can approximateg'(kTs) by [g(kTs) - g(kTs - Ts)]/Ts'
Therefore, from Equation (11.6), we have

m (k + 1) ~ m (k) + T s [
m(k) - m(k -
T
1)]
s
= 2m(k) - m(k - 1)
Thus, we can compute an approximate predication of the sample at (k + 1) from the
preceding two samples. The approximation improves as we add more samples. If N samples
are used, the approximation formula becomes:
(11.7)
The right hand side of Equation (11.7) is m(k), the predicted value of m(k). Thus,
(11.8)
This is the linear predictor illustrated in Figure 11.5. The prediction coefficients Ai are
determined from the statistical properties of the samples.