n Figure 14. Average MSDU length effects on data throughput.

Data offered load

0 0.2

0

0.1

Data throughput

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.4 0.6 0.8 1

MSDU length 200

MSDU length 300

MSDU length 500

MSDU length 1000

MSDU length 2000

IEEE Communications Magazine • September 1997 125

must be satisfied. It is obvious from the figure that an echo

canceler must be used since a large proportion of the voice

traffic exceeds the 25-ms requirement in delay. Thus, it is

assumed that an echo canceler is employed, and that voice

packets delayed by more than 0.5 s at the receiver become

useless and have to be discarded. Thus, the performance measure

of interest for voice traffic is the probability that a voice

packet will be discarded due to its late arrival at the receiver.

For clean voice quality communications, a packet loss rate of

1 percent should be maintained [13]. Shorter voice payloads

incur larger overheads, translating into longer delays. At the

other extreme, longer payloads imply longer packetization

delays. Thus, these two parameters must be traded off. As is

seen from Fig. 15, the best operating points appear to be

around 100–400 octets long for voice payload. When the CFP

repetition interval is 5, the recommended voice payload

lengths have been shown to be 100–200 octets long [9]. Note

that the average delay calculated when the voice payload is

50, 100, 200, 400, and 800 octets is 200, 186, 205, 233, and 284

ms, respectively.

Figure 16 shows the impact of voice payload length on data

throughput over a range of offered loads. It is shown that data

traffic will suffer more as the voice payload length is

decreased. Given a fixed amount of voice information to be

transmitted during the CFP, shortening the voice payload

length will result in more frames (i.e., overhead) transmitted.

Shortening the payload length will therefore lengthen the

duration of CFP operation, leaving less available bandwidth

for the transmission of data if the CFP is foreshortened. Thus,

from the point of view of data traffic, the voice payload

should be made relatively long. However, beyond 200 octets,

the data throughput improvement is marginal.

THE EFFECT OF POLLING SCHEME ON PERFORMANCE

As mentioned previously, an AP drops a station from the

polling list if the station does not transmit and receive any

data for k consecutive polls in the current CFP interval. To

see the appropriate values of k, the effect of k on data

throughput and voice delay is plotted, as illustrated in Figs. 17

and 18. Figure 17 shows throughput plotted against offered

load. For the PCF, five voice station pairs are used with voice

payload fixed at 200 octets.

The curves indicate that a higher value of k tends to reduce

the aggregate data throughput. When k increases, there is a

higher probability that a voice station will receive or have traffic

to transmit, which tends to prolong the duration of the

CFP. Prolonging the CFP corresponds to a reduction in the

amount of time that data stations have access to the channel.

In Fig. 18, the value of k has very little impact on the voice

packet loss rate, mainly due to the fact that voice stations

operate on an ON/OFF basis. That is, when a voice station

does not have any data to send during an OFF period, it is

likely that it will not have any data to send in the near future.

Thus, when a communicating pair of voice buffers are empty,

the best policy is to drop the stations from the polling list

immediately (k = 1). If the CFP is foreshortened due to light

traffic at that particular instant in time, the wait until the next

polling cycle is still well under the acceptable delay specifications

levied by the echo canceler. Therefore, from a data

throughput perspective, it is best to select k = 1 and have a

foreshortened CFP period.

n Figure 15. Complementary cumulative distribution for voice

delay.

x (s)

0 0.1

0.0001

0.001

Probability{X > x}

0.01

0.1

1

0.2 0.3 0.4 0.5 0.6 07

Voice payload = 50

Voice payload=100

Voice payload=200

Voice payload=400

Voice payload=800

n Figure 16. Effect of 1voice on data throughput.

Data offered load

0 0.2

0.05

0.1

Data throughput

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5 0.6 0.8 1

Voice payload = 50

Voice payload = 90

Voice payload=130

Voice payload=170

Voice payload=210

n Figure 17. Data throughput vs. offered load for several values of k.

Data offered load

0.30 0.2

0.32

Data throughput

0.34

0.36

0.38

0.4

0.42

0.44

0.4 0.6 0.8 1

k = 1

k = 2

k = 3

k = 4

k = 5

n Figure 18. Effect of k on voice delay.

x (s)

0 0.1

0.0001

0.001

Probability {X > x}

0.01

0.1

1

0.2 0.3 0.4 0.5 0.6 07

k = 1

k = 2

k = 3

k = 4

k = 5

126 IEEE Communications Magazine • September 1997

IMPACT OF THE NUMBER OF

VOICE STATIONS ON PERFORMANCE

The results of this last section concentrate on the effect the

number of voice stations has on voice delay in the infrastructure

network. The PCF is simulated using a fixed voice

payload of 200 octets, k = 1, and the number of voice stations

varies between 4 and 18.

In Fig. 19, the complementary cumulative distribution is

plotted for voice delay. As the number of voice stations

increases, so does the amount of packet loss. This is due to the

fact that more voice packets will be competing as the number

of voice stations increases. When the CFP interval is set to 4,

approximately up to 16 voice stations can be supported.

CONCLUSION

The primary contributions of this work include a detailed

investigation of both the DCF and the PCF operating over a

common CFP repetition interval. The simulation model includes

asynchronous data being transmitted over the DCF, which is not

delay-sensitive, and packetized voice traffic transmitted over

the PCF, which requires bounded delay. The model includes

the effect of a bursty error channel, which is typical of a wireless

radio environment where multipath fading is commonly

experienced. The final contribution includes a scheme to drop

voice stations from the CFP if they are idle for a specific period

of time. Dropping idle voice stations frees available bandwidth

for stations with packets queued for transmission.

The general conclusions derived from the study are:

• The efficiency delivered by the DCF is reasonably high if

the average MSDU length is longer than 500 octets, the

Fragmentation_Threshold is set to 800 octets, the

RTS_Threshold is set to 250 octets, and the medium is

relatively clean (BER better than 10–6).

• Based on our assumptions and simulation model, realtime

services such as packet voice can be transported by

the PCF. However, packet voice systems must employ an

echo canceler since the end-to-end delay cannot be

bounded under 25 ms.

• Compromised performance for both data and voice traffic

is achieved when the voice payload length is approximately

200 octets long.

• When a voice station does not have any data to receive

and transmit during a poll, the station should be dropped

from the list immediately (i.e., k = 1) so that the remaining

bandwidth can be allocated to other stations.

REFERENCES

[1] R.O. LaMaire et al., “Wireless LANs and Mobile Networking: Standards

and Future Directions,” IEEE Commun. Mag., vol. 34, no. 8, Aug. 1996,

pp. 86–94.

[2] ETSI TC-RES, “Radio Equipment and Systems (RES); High Performance

Radio Local Area Network (HIPERLAN); Functional Specification,” ETSI,

06921 Sophia Antipolis Cedex, France, draft prETS 300 652, July 1995.

[3] Wireless Medium Access Control and Physical Layer WG, IEEE Draft

Standard P802.11, “Wireless LAN,” IEEE Stds. Dept, D3, Jan. 1996.

[4] K. C. Chen, “Medium Access Control of Wireless LANs for Mobile Computing,”

IEEE Network, vol. 8, no. 5, Sept. 1994, pp. 50–63.

[5] H. S. Chhaya and S. Gupta, “Throughput and Fairness Properties of

Asynchronous Data Transfer Methods in the IEEE 802.11 MAC Protocol,”

PIMRC ’95, 1995, pp. 613–17.

[6] W. Diepstraten, “A Wireless MAC Protocol Comparison,” IEEE P802.11-92/51.

[7] J. Weinmiller, H. Woesner, and A. Wolisz, “Analyzing and Improving the

IEEE 802.11-MAC Protocol for Wireless LANs,” Proc. MASCOTS ’96, San

Jose, CA, Feb. 1996, pp. 200–6.

[8] D. Bantz and F. Bauchot, “Wireless LAN Design Alternatives,” IEEE Network,

vol. 8, no. 2, Apr. 1994, pp. 43–53.

[9] B. Crow et al., “Investigation of the IEEE 802.11 Medium Access Control

(MAC) Sublayer Functions,” Proc. INFOCOM 97, Kobe, Japan, Apr. 1997,

pp. 126–33.

[10] E. Gilbert, “Capacity of a Burst Noise Channel,” Bell Sys. Tech. J., vol.

39, Sept. 1960, pp. 1253–66.

[11] P. Brady, “A Model for Generating On-Off Speech Patterns in Two-Way

Conversation,” Bell Sys. Tech. J., vol. 48, no. 7, Sept. 1969, pp. 2445–72.

[12] M. de Prycker, Asynchronous Transfer Mode: Solution for Broadband

ISDN, 3rd ed., Englewood Cliffs, NJ: Prentice Hall, 1995.

[13] L. Hanzo et al., “A Packet Reservation Multiple Access Assisted Cordless

Telecommunications Scheme,” IEEE Trans. Vehic. Tech., vol. 43, no.

2, May 1994, pp. 234–44.

BIOGRAPHIES

BRIAN P. CROW (bcrow@mitre.org) received a B.S. from Arizona State University

in 1987 and an M.S. from the University of Arizona in 1996. From

1988 to 1992, he was a signal officer in the U.S. Army. He is currently a

lead engineer with the MITRE Corporation. His research interests include

wireless and broadband networks, and network and systems management.

INDRA WIDJAJA received a B.A.Sc. degree from the University of British

Columbia, an M.S. from Columbia University, and a Ph.D. from the University

of Toronto, all in electrical engineering. From 1994 to 1997, he was

assistant professor of ECE at the University of Arizona. Since July 1997, he

has been with Fujitsu Network Communications. His research interests

include mobile and wireless networks, switching architectures, and traffic

management.

JEONG GEUN KIM received B.S. and M.S. degrees, both in electrical engineering,

from Yonsei University, Seoul, Korea, in 1990 and 1992, respectively.

He is pursing a Ph.D. at the University of Arizona, studying issues of wireless

ATM and performance evaluation of broadband networks.

PRESCOTT T. SAKAI received an M.S.E.E. from the University of Arizona in Tucson.

He is currently a new product planning and applications engineer at

Cypress Semiconductor, where he is involved with defining next-generation

data communication products.

n Figure 19. Effect of voice stations on voice delay.

x (s)

0 0.1

0.0001

0.001

Probability {X > x}

0.01

0.1

1

0.2 0.3 0.4 0.5 0.6 0.7

Voice stations = 4

Voice stations = 6

Voice stations=10

Voice stations=14

Voice stations=18

Data stations 10

Average MSDU length 1000 octets

Channel rate 1 Mb/s

BERgood 10-10

a 30 s--1

b 10 s-1

RTS_Threshold 250 octets

Fragmentation_Threshold 800 octets

Short_Retry_Limit 5

Long_Retry_Limit 7

DSSS preamble 144 bits

DSSS header 48 bits

Station buffer size 300 frames

Slot_Time 20 ms

SIFS_Time 10 ms

DIFS_Time 50 ms

Attribute Typical value

n Table 2. Default attribute values for the

infrastructure network unless otherwise

specified.

BERbad 10-5

Number of voice stations 10

Voice transmission rate 64 kb/s

Voice station buffer size 100 frames

CFP_Max_Duration 0.39 s

CFP_Repetition_Interval 0.41 s

PIFS_Time 30 ms

Attribute Typical value

124 IEEE Communications Magazine • September 1997

EFFECT OF FRAGMENTATION_THRESHOLD ON

MAXIMUM DATA THROUGHPUT

The Fragmentation_Threshold is used to combat the effects

of poor channel quality. By reducing the size of the packets

transmitted, there is a better probability of successful transmission,

especially under poor channel conditions. However,

under good channel conditions, fragmentation is a hindrance

because the associated overhead tends to reduce the aggregate

throughput. Figure 13 shows the maximum data throughput

plotted against Fragmentation_Threshold for various

values of BERbad, when the average MSDU length is 1000

octets. When the channel is in a good condition (i.e., BERbad

less than 10–5), fragmentation only hinders the maximum

throughput because of the additional overhead. However,

when BERbad is high, the benefits of using fragmentation

become apparent. In the figure, the difference between the

peak and smallest values of maximum throughput for the

BERbad 10–4 curve is almost 140 kb/s.

Since a typical WLAN terminal will experience the whole

spectrum of channel qualities, the optimum

Fragmentation_Threshold should be set between 500 and 800

octets. This range of values is ideal for neither a clean channel

nor a degraded channel, but offers acceptable performance

across the entire spectrum of channel qualities.

EFFECT OF MSDU LENGTH ON DATA THROUGHPUT

Figure 14 shows the effects of average MSDU length, ldata, on

throughput performance. The curves are obtained for a

BERbad of 10–5. The IEEE 802.11 MAC and PHY layers add

a total of 58 octets for overhead. Given a clean channel like

that shown in Fig. 14, the longer the MSDU is, the more efficient

the system becomes. When the channel is operating in a

degraded mode, we have observed that the benefits of a large

MSDU length become less pronounced.

INFRASTRUCTURE NETWORK

The infrastructure network supporting voice and data traffic

is now considered. Data traffic is transported through the

CFP and voice traffic through the CP. All results below are

shown using a CFP repetition interval of four beacon periods.

With five simultaneous voice conversations in progress (10

stations total), the aggregate voice throughput is approximately

272 kb/s. This is calculated by considering that each voice

station is transmitting at 64 kb/s and the channel is in the ON

state for 42.5 percent of the time (based on the ON/OFF

model described above).

THE EFFECT OF VOICE PAYLOAD

LENGTH ON PERFORMANCE

The effect of voice payload length, lvoice, on both data and

voice performances is investigated first. The number of voice

stations, Nvoice, is set to five pairs. The first five voice stations

are located in the BSS; the others are located elsewhere and

are generated through the AP. This voice scheme is used

because it is assumed that voice traffic would not occur

between stations within the same BSS, due to their close

proximity to each other. For voice traffic, only the delay

between an AP and another mobile station in the same BSS is

considered. All measurements are done at the MSAP (MAC

service access point). Figure 15 displays the influence of the

voice payload length on data traffic performance.

In Fig. 15, the random variable X denotes the end-to-end

delay between an AP and a mobile station. Here the delay is

measured from the time the first bit is generated at the transmitter

until the time the last bit is received at the receiver.

Since voice packets, unlike data packets, are bounded by a

specified delay (e.g., 0.5 s), any packets exceeding the delay

requirements must be discarded. A complementary cumulative

distribution plot is used to determine the percentage of

voice packets which are discarded because they are not transmitted

within the delay bounds. The figures illustrate the complementary

cumulative distribution, Pr{X > x}, for voice delay

in seconds. As discussed previously, voice delay can tolerate as

much as 0.5 s if an echo canceler is used. Without an echo

canceler, a much more stringent voice delay (under 25 ms)

n Figure 12. RTS_Threshold effects on data throughput.

RTS_Threshold

0 500

0.3

0.35

Maximum data throughput

0.4

0.45

0.5

0.55

0.6

0.65

0.7

0.75

0.8

1000 1500 2000 2500

MSDU length 200

MSDU length 500

MSDU length 1000

MSDU length 2000

n Figure 13. Fragmentation_Threshold effects on data throughput.

Fragmentation_Threshold

0 500

0.1

0.2

Maximum data throughput

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1000 1500 2000 2500

BERbad 10-3

BERbad 10-4

BERbad 10-5

BERbad 10-6

where it stopped, and resumes polling

at that same point when the CFP starts.

Since all stations operating during the

CFP are packetized voice users, they all

have the same QoS requirements;

therefore, priority polling mechanisms

are not required. A simple polling

scheme is proposed to allocate unused

bandwidth to currently active voice

users. When the AP is prepared to poll

a station during the CFP, if the AP has

an MPDU queued for transmission to the

station, the poll and MPDU can be combined

and transmitted as a single frame.

Otherwise, the AP sends a sole CF-Poll

(no Data) to the station. If the AP sends

k consecutive CF-Polls to a station, and

the station responds each time without

any payload to transmit (i.e., Null Function),

the station is dropped from the

polling list for that CFP_Repetition_Interval.

During the next interval, the station

will be added back into the polling list

and the process will start over. The

polling scheme will drop stations that

are not active transmitting and receiving

voice packets. When all voice stations

have been dropped from the

polling list, the AP will send a CF_END

frame indicating that the asynchronous

users can start using the channel until

the start of the next CFP interval.

The voice stream is modeled using

an ON/OFF process, where stations are

either transmitting (ON) or listening

(OFF). The amount of time sitting in

the OFF or ON state is exponentially

distributed, where the mean value of

the silence (OFF) period is 1.35 s, and

the mean value of the talk spurt (ON)

period is 1 s. The voice transmission

rate in the ON state is 64 kb/s. The

transition rates are representative of

real telephonic speech patterns that

were obtained from measurement [11].

The length of the voice payload

should be chosen so that voice packetization

delay is minimized and header

overhead is not large, which is a conflicting

goal. No retransmissions will be performed

for voice frames since this traffic

is delay-sensitive. QoS parameters for voice typically limit maximum

delay to 25 ms without echo canceling, and 500 ms using

echo canceling [12]. Asynchronous data frames are transmitted

in the CP portion of the repetition interval using the DCF

protocol described above. Table 2 lists the additional default

values used for simulation of the infrastructure network.

SIMULATION RESULTS

Simulation results are shown for an ad hoc network and an

infrastructure network. The results below are presented in

the form of plots and, where applicable, with 95 percent confidence

intervals. The throughput plots shown below represent

aggregate throughput. Approximate throughput per station

can be calculated by dividing the aggregate throughput by the

total number of data stations in the BSS.

AD HOC NETWORK

For the ad hoc network, we assume all

mobile stations generate asynchronous

data traffic with the same intensity.

Figure 11 shows the aggregate data

throughput in megabits per second versus

the offered load in megabits per

second for several BERs (i.e., the

BERbad). The offered load is defined

to be the average number of bits per

second passed down to the MAC sublayer

at the source. The throughput is

the average number of bits per second

passed up from the MAC sublayer at

the destination.

Note that the burst error transition

rates for this model indicate that more

time will be spent in the “bad” state

than in the “good” state. When the

medium is relatively clean, BERbad is

less than 10–6, and a maximum

throughput of approximately 77 percent

is possible. However, the maximum

throughput can drop to

approximately 20 percent under harsh

fading. Thus, it is clear that the channel

condition can adversely affect the

throughput performance of the IEEE

802.11 system. It is also noted that the

throughput saturates around 90 percent

under ideal channel conditions

due to overhead, collisions, IFS, and

backoff intervals.

THE EFFECT OF RTS ON MAXIMUM

DATA THROUGHPUT

As stated previously, the RTS/CTS

handshaking mechanism is used to combat

the effects of collisions. The

RTS/CTS reserves the channel for transmission

of a larger data packet, with the

desired effect that if a collision occurs

with the RTS/CTS handshake, less

bandwidth will have been wasted than if

the larger data packet had been transmitted

and corrupted. RTS_Threshold

is a manageable parameter used to

determine when to precede a data

packet with an RTS/CTS handshake.

In the plots below, the maximum data

throughput is plotted against RTS_Threshold for various values

of data MSDU length. The maximum data throughput is

defined as the maximum value of throughput obtained over

all offered loads when RTS_Threshold is held constant at a

specified value. A bursty channel error model is used with the

transition rates given in Table 1.

Figure 12 shows the impact of RTS_Threshold on maximum

data throughput when there is no fragmentation of data

packets. As shown in the figure, under all of the MSDU values

the peak throughput occurs when the RTS_Threshold is

set at approximately 250 octets. The maximum throughput

values begin to taper off considerably when the RTS_Threshold

begins to exceed 400 octets, indicating that collisions are

having an adverse impact on system throughput. We have

also varied the number of data stations and observed the

same conclusion.

n Table 1. Default attribute values for the

ad hoc network unless otherwise specified.

Two different simulation models are presented in this article.

The first model represents an ad hoc network, where

all stations in the BSS are capable of directly communicating

with all other stations in the BSS. All stations in the ad hoc

network are assumed to be asynchronous data users. The second

model represents an infrastructure network which characterizes

a single BSS with an AP. The infrastructure network

operates with asynchronous data users in the CP and packetized

voice terminals operating in the CFP. Both simulation

models are implemented using the physical-layer parameters

specified in the standard for the DSSS implementation. More

detailed explanation of the simulation model is found in [9].

Several assumptions have been made to reduce the complexity

of the model. A short description of each of the

assumptions is provided below:

• The effects of propagation delay on the model are neglected.

This is a fairly realistic assumption if transmission distances

are on the order of 100 ft between stations.

• The “hidden terminal” problem is not addressed in the

simulation models.

• The basic rate of 1 Mb/s was simulated for the DSSS. This

decision was made because the enhanced rate, 2 Mb/s,

would add additional complexity since control, multicast,

and broadcast frames are required to be transmitted at the

basic rate (to ensure that all stations in the BSS can be

properly received), while management and data frames

are transmitted at any available rate (1 Mb/s or 2 Mb/s).

• No stations operate in the “power-saving” mode (PSMode).

By requiring all stations to be “awake” at all

times, transmitted MPDUs can be received immediately

by the destination station without buffering at the AP.

• No interference is considered from nearby BSSs reusing

the same DSSS spreading sequence.

When the PCF and DCF coexist together in the infrastructure

network, all stations operating during the CP are

asynchronous data users, and all users operating during the

CFP are packetized voice users.

A finite transmit buffer is maintained for each station. If

the finite buffer fills, all newly generated MSDUs will be considered

dropped without returning.

For the ad hoc and infrastructure network simulations, a

burst error model is introduced to characterize fading in the

communications channel [10]. A two-state continuous-time

Markov chain is used to represent the burst error model.

State G represents the channel in a “good” state. This indicates

that the channel is operating with a very low bit error

rate (denoted by BERgood). State B indicates the channel is

operating in a fading condition with a higher error rate,

denoted by BERbad. The transition rate from state G to state

B is denoted by a, while the transition rate from state B to

state G is denoted by b. A frame is considered to be corrupt if

it contains one or more bit errors.

The simulation model uses the error model above to determine

whether each transmitted frame or MPDU was transmitted

successfully. When the frame is transmitted, a portion of

the frame can be sent over the communications medium when

the channel is in state G, and a portion can be transmitted

when the channel is in state B. The number of bits transmitted

in the frame during state B is denoted by n1, and the number

transmitted during state G is n2. The probability that the

frame is transmitted successfully is then calculated as

Pr{success} = (1 – BERbad)n1 · (1 – BERgood)n2.

AD HOC NETWORK MODEL

With the ad hoc network model, all users are assumed to be

asynchronous data users, and they shall operate in a selfcontained

BSS. The arrival of frames from a station’s higherlayer

protocol to the MAC sublayer is modeled with exponential

interarrival times and a truncated geometric distribution for

the frame lengths. The truncated geometric distribution is

used to ensure that the MSDU does not exceed the maximum

length established by the specification (i.e., 2312 octets). However,

the simulation model can easily accommodate other

arrival processes and frame length distributions.

During the simulation, if collisions or bit errors affect the

transmission of a frame, retransmission will occur according to

the backoff procedure described previously. The number of

retransmissions is limited before the frame is dropped from the

station’s transmit queue. In the case of MSDUs smaller than

RTS_Threshold, the number of retransmissions is limited to

Short_Retry_Limit. For MSDUs larger than RTS_Threshold, the

maximum number of retransmissions is set by Long_Retry_Limit.

The number of retransmissions is extended in this case since

short RTS frames are not as wasteful of bandwidth as larger

data payloads. Typical default values used in the simulation of

the ad hoc network are illustrated in Table 1.

INFRASTRUCTURE NETWORK MODEL

The effect of a single BSS with an AP is simulated, where

asynchronous data users transmit during the CP and packetized

voice users transmit during the CFP. The coexistence

of the DCF and PCF is illustrated in Fig. 8, where, for the

purposes of this simulation, the value of CFP_Max_Duration

is provided in Table 2. The duration of the

CFP_Repetition_Interval is approximately 0.4096 s; therefore,

approximately 94 percent of the repetition interval

can be allocated by the AP for contention-free services.

During the CFP, if a station is polled by the AP to transmit,

the station can transmit directly to another station in the BSS

(Fig. 10) or to a station in another BSS. When the transmission

is directed to a station in another BSS, the source station

transmits the frame to the AP, who is responsible for forwarding

the frame through the DS to the remote AP servicing the destination

station. Since the size of the BSS is relatively small, all

packetized voice activity is assumed to occur between stations

in different BSSs. Therefore, the simulation model directs all

voice traffic from a station through the AP. All voice traffic

destined for a mobile station is also delivered via the AP.

The polling scheme during the CFP uses a cyclical scheduling

algorithm, where each station is polled sequentially in the

order in which it is placed in the polling list. When the CFP

ends, the AP keeps track of the location in the polling list

n Figure 11. Burst error effects on data throughput.

Data offered load

0 0.2

0

0.1

Data throughput

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.4 0.6 0.8 1

BERbad 10-3

BERbad 10-4

BERbad 10-5

BERbad 10-6

IEEE Communications Magazine • September 1997 123

known as CF-aware stations. The method by which

polling tables are maintained and the polling sequence

is determined, is left to the implementor.

The PCF is required to coexist with the DCF and

logically sits on top of the DCF (Fig. 4). The CFP repetition

interval (CFP_Rate) is used to determine the

frequency with which the PCF occurs. Within a repetition

interval, a portion of the time is allotted to contention-free

traffic, and the remainder is provided for contention-based

traffic. The CFP repetition interval is initiated by a beacon

frame, where the beacon frame is transmitted by the AP. One

of its primary functions is synchronization and timing. The

duration of the CFP repetition interval is a manageable

parameter that is always an integral number of beacon frames.

Once the CFP_Rate is established, the duration of the CFP is

determined. The maximum size of the CFP is determined by

the manageable parameter CFP_Max_Duration. The minimum

value of CFP_Max_Duration is the time required to

transmit two maximum-size MPDUs, including overhead, the

initial beacon frame, and a CF-End frame. The maximum

value of CFP_Max_Duration is the CFP repetition interval

minus the time required to successfully transmit a maximumsize

MPDU during the CP (which includes the time for

RTS/CTS handshaking and the ACK). Therefore, time must

be allotted for at least one MPDU to be transmitted during

the CP. It is up to the AP to determine how long to operate

the CFP during any given repetition interval. If traffic is very

light, the AP may shorten the CFP and provide the remainder

of the repetition interval for the DCF. The CFP may also be

shortened if DCF traffic from the previous repetition interval

carries over into the current interval. The maximum amount

of delay that can be incurred is the time it takes to transmit

an RTS/CTS handshake, maximum MPDU, and ACK. Figure

8 is a sketch of the CFP repetition interval, illustrating the

coexistence of the PCF and DCF.

At the nominal beginning of each CFP repetition interval,

all stations in the BSS update their NAV to the maximum

length of the CFP (i.e., CFP_Max_Duration). During the

CFP, the only time stations are permitted to transmit is in

response to a poll from the PC or for transmission of an ACK

a SIFS interval after receipt of an MPDU. At the nominal

start of the CFP, the PC senses the medium. If the medium

remains idle for a PIFS interval, the PC transmits a beacon

frame to initiate the CFP. The PC starts CF transmission a

SIFS interval after the beacon frame is transmitted by sending

a CF-Poll (no data), Data, or Data+CF-Poll frame. The PC

can immediately terminate the CFP by transmitting a CF-End

frame, which is common if the network is lightly loaded and

the PC has no traffic buffered. If a CF-aware station receives

a CF-Poll (no data) frame from the PC, the STA can respond

to the PC after a SIFS idle period, with a CF-ACK (no data)

or a Data + CF-ACK frame. If the PC receives a Data + CFAck

frame from a station, the PC can send a Data + CFACK

+ CF-Poll frame to a different station, where the

CF-ACK portion of the frame is used to acknowledge receipt

of the previous data frame. The ability to combine polling and

acknowledgment frames with data frames, transmitted

between stations and the PC, was designed to improve efficiency.

If the PC transmits a CF-Poll (no data) frame and the

destination station does not have a data frame to transmit, the

station sends a Null Function (no data) frame back to the PC.

Figure 9 illustrates the transmission of frames between the PC

and a station, and vice versa. If the PC fails to receive an

ACK for a transmitted data frame, the PC waits a PIFS interval

and continues transmitting to

the next station in the polling list.

After receiving the poll from

the PC, as described above, the

station may choose to transmit a

frame to another station in the

BSS. When the destination station

receives the frame, a DCF ACK is

returned to the source station, and

the PC waits a PIFS interval following

the ACK frame before

transmitting any additional frames.

Figure 10 illustrates station-to-station

frame transmission during the

CFP. The PC may also choose to

transmit a frame to a non-CFaware

station. Upon successful

receipt of the frame, the station

would wait a SIFS interval and

reply to the PC with a standard

ACK frame. Fragmentation and

reassembly are also accommodated

with the Fragmentation_Threshold

value used to determine whether

MSDUs are fragmented prior to

transmission. It is the responsibility

of the destination station to

reassemble the fragments to form

the original MSDU.

n Figure 8. Coexistence of the PCF and DCF.

CFP CP CFP

PCF DCF

NAV NAV

CP

CFP repetition interval CFP repetition interval

B B PCF DCF

n Figure 9. PC-to-station transmission.

SIFS

PIFS

B

U1+ACK

D1+Poll

SIFS

Contention free period

D2+ACK+Poll

SIFS

D3+ACK+Poll

PIFS

D4+Poll

SIFS

CF-End

SIFS

U2+ACK

SIFS

NAV

U4+ACK

SIFS

CP

n Figure 10. Station-to-station transmissions.

SIFS

PIFS

B

Sta-to-sta

D1+Poll

PIFS

CFP repetition interval

Contention period

Contention free period

D2+Poll

SIFS

CF-End

SIFS

ACK

SIFS

U2+ACK

SIFS

NAV

122 IEEE Communications Magazine • September 1997

SIMULATION MODEL

Defer access

120 IEEE Communications Magazine • September 1997

updating their NAVs based on the RTS from the source station

and CTS from the destination station, which helps to

combat the “hidden terminal” problem. Figure 6 illustrates

the transmission of an MPDU using the RTS/CTS mechanism.

Stations can choose to never use RTS/CTS, use

RTS/CTS whenever the MSDU exceeds the value of

RTS_Threshold (manageable parameter), or always use

RTS/CTS. If a collision occurs with an RTS or CTS MPDU,

far less bandwidth is wasted when compared to a large data

MPDU. However, for a lightly loaded medium, additional

delay is imposed by the overhead of the RTS/CTS frames.

Large MSDUs handed down from the LLC to the MAC may

require fragmentation to increase transmission reliability. To

determine whether to perform fragmentation, MPDUs are compared

to the manageable parameter Fragmentation_Threshold.

If the MPDU size exceeds the value of Fragmentation_Threshold,

the MSDU is broken into multiple fragments. The resulting

MPDUs are of size Fragmentation_Threshold, with exception of

the last MPDU, which is of variable size not to exceed Fragmentation_

Threshold. When an MSDU is fragmented, all fragments

are transmitted sequentially (Fig. 7). The channel is not

released until the complete MSDU has been transmitted successfully,

or the source station fails to receive an acknowledgment

for a transmitted fragment. The destination station

positively acknowledges each successfully received fragment by

sending a DCF ACK back to the source station. The source

station maintains control of the channel throughout the transmission

of the MSDU by waiting only an SIFS period after

receiving an ACK and transmitting the next fragment. When an

ACK is not received for a previously transmitted frame, the

source station halts transmission and recontends for the channel.

Upon gaining access to the

channel, the source starts transmitting

with the last unacknowledged

fragment.

If RTS and CTS are used, only

the first fragment is sent using the

handshaking mechanism. The duration

value of RTS and CTS only

accounts for the transmission of

the first fragment through the

receipt of its ACK. Stations in the

BSS thereafter maintain their NAV

by extracting the duration information

from all subsequent fragments.

The collision avoidance portion

of CSMA/CA is performed through

a random backoff procedure. If a

station with a frame to transmit

initially senses the channel to be busy;

then the station waits until the channel

becomes idle for a DIFS period, and

then computes a random backoff time.

For IEEE 802.11, time is slotted in

time periods that correspond to a

Slot_Time. Unlike slotted Aloha,

where the slot time is equal to the

transmission time of one packet, the

Slot_Time used in IEEE 802.11 is

much smaller than an MPDU and is

used to define the IFS intervals and

determine the backoff time for stations

in the CP. The Slot_Time is different

for each physical layer implementation.

The random backoff time is an integer

value that corresponds to a number of

time slots. Initially, the station computes

a backoff time in the range 0–7. After the medium becomes

idle after a DIFS period, stations decrement their backoff

timer until the medium becomes busy again or the timer

reaches zero. If the timer has not reached zero and the medium

becomes busy, the station freezes its timer. When the timer

is finally decremented to zero, the station transmits its frame. If

two or more stations decrement to zero at the same time, a collision

will occur, and each station will have to generate a new

backoff time in the range 0–15. For each retransmission attempt,

the backoff time grows as ë22 + i · ranf()û · Slot_Time, where i is

the number of consecutive times a station attempts to send an

MPDU, ranf() is a uniform random variate in (0,1), and ëxû

represents the largest integer less than or equal to x. The idle

period after a DIFS period is referred to as the contention

window (CW). The advantage of this channel access method is

that it promotes fairness among stations, but its weakness is

that it probably could not support DTBS. Fairness is maintained

because each station must recontend for the channel

after every transmission of an MSDU. All stations have equal

probability of gaining access to the channel after each DIFS

interval. Time-bounded services typically support applications

such as packetized voice or video that must be maintained

with a specified minimum delay. With DCF, there is no

mechanism to guarantee minimum delay to stations supporting

time-bounded services.

POINT COORDINATION FUNCTION (PCF)

The PCF is an optional capability, which is connection-oriented,

and provides contention-free (CF) frame transfer. The

PCF relies on the point coordinator (PC) to perform polling,

enabling polled stations to transmit without contending for

n Figure 6. Transmission of an MPDU using RTS/CTS.

DIFS

RTS

Source

SIFS

Destination

Other

SIFS SIFS

DIFS

NAV (RTS)

NAV (CTS)

NAV (data)

Defer access Backoff started

CW

Data

CTS ACK

n Figure 7. Transmission of a fragmented MPDU.

SIFS SIFS SIFS

Fragment burst

SIFS

Src

Dest.

Other

Other

SIFS

ACK 0 ACK 1 ACK 2

NAV (CTS) NAV (fragment 1)

NAV (fragment 0)

NAV (ACK 0)

NAV (ACK 11)

NAV (frag 2)

SIFS

DIFS

Fragment 0 Fragment 1 Fragment 2 CW

IEEE Communications Magazine • September 1997 121

the channel. The function of the PC is performed by

the AP within each BSS. Stations within the BSS that

are capable of operating in the CF period (CFP) are

is illustrated in Fig. 3. Note that

the frame body (MSDU) is a variable-

length field consisting of the

data payload and 7 octets for

encryption/decryption if the

optional Wired Equivalent Privacy

(WEP) protocol is implemented.

The IEEE standard 48-bit

MAC addressing is used to identify

a station. The 2 duration

octets indicate the time (in

microseconds) the channel will be

allocated for successful transmission

of a MAC protocol data unit (MPDU). The type bits

identify the frame as either control, management, or data.

The subtype bits further identify the type of frame (e.g., Clear

to Send control frame). A 32-bit cyclic redundancy check

(CRC) is used for error detection.

DISTRIBUTED COORDINATION FUNCTION

The DCF is the fundamental access method used to support

asynchronous data transfer on a best effort basis. As identified

in the specification, all stations must support the DCF. The

DCF operates solely in the ad hoc network, and either operates

solely or coexists with the PCF in an infrastructure network.

The MAC architecture is depicted in Fig. 4, where it is

shown that the DCF sits directly on top of the physical layer

and supports contention services. Contention services imply

that each station with an MSDU queued for transmission

must contend for access to the channel and, once the MSDU

is transmitted, must recontend for access to the channel for all

subsequent frames. Contention services promote fair access to

the channel for all stations.

The DCF is based on carrier sense multiple access with

collision avoidance (CSMA/CA). CSMA/CD (collision detection)

is not used because a station is unable to listen to the

channel for collisions while transmitting. In IEEE 802.11, carrier

sensing is performed at both the air interface, referred to

as physical carrier sensing, and at the MAC sublayer, referred

to as virtual carrier sensing. Physical carrier sensing detects the

presence of other IEEE 802.11 WLAN users by analyzing all

detected packets, and also detects activity in the channel via

relative signal strength from other sources.

A source station performs virtual carrier sensing by sending

MPDU duration information in the header of request to

send (RTS), clear to send (CTS), and data frames. An MPDU

is a complete data unit that is passed from the MAC sublayer

to the physical layer. The MPDU contains header information,

payload, and a 32-bit CRC.

The duration field indicates the

amount of time (in microseconds)

after the end of the present frame

the channel will be utilized to

complete the successful transmission

of the data or management

frame. Stations in the BSS use

the information in the duration

field to adjust their network allocation

vector (NAV), which indicates

the amount of time that

must elapse until the current

transmission session is complete

and the channel can be sampled again for idle status. The

channel is marked busy if either the physical or virtual carrier

sensing mechanisms indicate the channel is busy.

Priority access to the wireless medium is controlled through

the use of interframe space (IFS) time intervals between the

transmission of frames. The IFS intervals are mandatory periods

of idle time on the transmission medium. Three IFS intervals

are specified in the standard: short IFS (SIFS), point

coordination function IFS (PIFS), and DCF-IFS (DIFS). The

SIFS interval is the smallest IFS, followed by PIFS and DIFS,

respectively. Stations only required to wait a SIFS have priority

access over those stations required to wait a PIFS or DIFS

before transmitting; therefore, SIFS has the highest-priority

access to the communications medium. For the basic access

method, when a station senses the channel is idle, the station

waits for a DIFS period and samples the channel again. If the

channel is still idle, the station transmits an MPDU. The

receiving station calculates the checksum and determines

whether the packet was received correctly. Upon receipt of a

correct packet, the receiving station waits a SIFS interval and

transmits a positive acknowledgment frame (ACK) back to

the source station, indicating that the transmission was successful.

Figure 5 is a timing diagram illustrating the successful

transmission of a data frame. When the data frame is transmitted,

the duration field of the frame is used to let all stations

in the BSS know how long the medium will be busy. All

stations hearing the data frame adjust their NAV based on

the duration field value, which includes the SIFS interval and

the ACK following the data frame.

Since a source station in a BSS cannot hear its own transmissions,

when a collision occurs, the source continues transmitting

the complete MPDU. If the MPDU is large (e.g., 2300

octets), a lot of channel bandwidth is wasted due to a corrupt

MPDU. RTS and CTS control frames can be used by a station

to reserve channel bandwidth prior to the transmission of

an MPDU and to minimize the amount of bandwidth

wasted when collisions occur. RTS and CTS control

frames are relatively small (RTS is 20 octets and CTS

is 14 octets) when compared to the maximum data

frame size (2346 octets). The RTS control frame is

first transmitted by the source station (after successfully

contending for the channel) with a data or management

frame queued for transmission to a specified

destination station. All stations in the BSS, hearing the

RTS packet, read the duration field (Fig. 3) and set

their NAVs accordingly. The destination station

responds to the RTS packet with a CTS packet after

an SIFS idle period has elapsed. Stations hearing the

CTS packet look at the duration field and again

update their NAV. Upon successful reception of the

CTS, the source station is virtually assured that the

medium is stable and reserved for successful transmission

of the MPDU. Note that stations are capable of

n Figure 4. MAC architecture.

Point coordination

function (PCF)

MAC

extent

Distributed coordination function (DCF)

Used for contention services

and basis for PCF

Required for contention-free

services

n Figure 5. Transmission of an MPDU without RTS/CTS.

DIFS

Data

Source

Destination

Other

SIFS

DIFS

CW

ACK

NAV

Backoff after defer

and IR. The FHSS utilizes the 2.4 GHz Industrial, Scientific,

and Medical (ISM) band (i.e., 2.4000–2.4835 GHz). In the

United States, a maximum of 79 channels are specified in the

hopping set. The first channel has a center frequency of 2.402

GHz, and all subsequent channels are spaced 1 MHz apart.

The 1 MHz separation is mandated by the FCC for the 2.4

GHz ISM band. The channel separation corresponds to 1

Mb/s of instantaneous bandwidth. Three different hopping

sequence sets are established with 26 hopping sequences per

set. Different hopping sequences enable multiple BSSs to

coexist in the same geographical area, which may become

important to alleviate congestion and

maximize the total throughput in a single

BSS. The reason for having three different

sets is to avoid prolonged collision periods

between different hopping sequences in a

set [3]. The minimum hop rate permitted

is 2.5 hops/s. The basic access rate of 1

Mb/s uses two-level Gaussian frequency

shift keying (GFSK), where a logical 1 is

encoded using frequency Fc + f and a logical

0 using frequency Fc – f. The enhanced

access rate of 2 Mb/s uses four-level

GFSK, where 2 bits are encoded at a time

using four frequencies.

The DSSS also uses the 2.4 GHz ISM

frequency band, where the 1 Mb/s

basic rate is encoded using differential

binary phase shift keying (DBPSK),

and a 2 Mb/s enhanced rate uses differential

quadrature phase shift

keying (DQPSK). The spreading is

done by dividing the available bandwidth

into 11 subchannels, each 11

MHz wide, and using an 11-chip

Barker sequence to spread each

data symbol. The maximum channel

capacity is therefore (11 chips/symbol)/(

11 MHz) = 1 Mb/s if DBPSK

is used [8]. Overlapping and adjacent

BSSs can be accommodated by

ensuring that the center frequencies

of each BSS are separated by at

least 30 MHz [3]. This rigid requirement

will enable only two overlapping

or adjacent BSSs to operate without interference.

The IR specification identifies a wavelength range from

850 to 950 nm. The IR band is designed for indoor use only

and operates with nondirected transmissions. The IR specification

was designed to enable stations to receive line-of-site

and reflected transmissions. Encoding of the basic access rate

of 1 Mb/s is performed using 16-pulse position modulation

(PPM), where 4 data bits are mapped to 16 coded bits for

transmission. The enhanced access rate (2 Mb/s) is performed

using 4-PPM modulation, where 2 data bits are mapped to 4

coded bits for transmission.

MEDIUM ACCESS CONTROL SUBLAYER

The MAC sublayer is responsible for the channel allocation

procedures, protocol data unit (PDU) addressing, frame

formatting, error checking, and fragmentation and reassembly.

The transmission medium can operate in the contention mode

exclusively, requiring all stations to contend for access to the

channel for each packet transmitted. The medium can also

alternate between the contention mode, known as the contention

period (CP), and a contention-free period (CFP). During

the CFP, medium usage is controlled (or mediated) by the

AP, thereby eliminating the need for stations to contend for

channel access. IEEE 802.11 supports three different types of

frames: management, control, and data. The management

frames are used for station association and disassociation with

the AP, timing and synchronization, and authentication and

deauthentication. Control frames are used for handshaking

during the CP, for positive acknowledgments during the CP,

and to end the CFP. Data frames are used for the transmission

of data during the CP and CFP, and can be combined

with polling and acknowledgments during the CFP. The stan-

n Figure 2. Sketch of an infrastructure network.

STA

BSS STA

BSS

Portal

IEEE 802.X

DS: Distribution system

STA STA

AP

AP

n Figure 3. Standard IEEE 802.11 frame format.

Frame

control

2

Duration

conn. ID

2

Address

6

Address

6

Address

6

Address

6

Frame

body

0–2312

CRC

4

Sequence

control

2

Octets:

Protocol

version

2

Type

2

Subtype

4

To DS

1

From DS

1

Last

fragment

1

Retry

1

Power

mgt

1

More

data

1

EP

1

EP

Bits: 1

IEEE Communications Magazine • September 1997 119

dard IEEE 802.11 frame format

identified as IEEE 802.11 [3]. This project was initiated in 1990,

and several draft standards have been published for review.

The scope of the standard is “to develop a Medium Access

Control (MAC) and Physical Layer (PHY) specification for

wireless connectivity for fixed, portable and moving stations

within a local area.” The purpose of the standard is twofold:

• “To provide wireless connectivity to automatic machinery,

equipment, or stations that require rapid deployment,

which may be portable, or hand-held or which may

be mounted on moving vehicles within a local area”

• “To offer a standard for use by regulatory bodies to standardize

access to one or more frequency bands for the

purpose of local area communication” [3].

The IEEE 802.11 draft standard describes mandatory support

for a 1 Mb/s WLAN with optional support for a 2 Mb/s

data transmission rate. Mandatory support for asynchronous

data transfer is specified as well as optional support for distributed

time-bounded services (DTBS). Asynchronous data

transfer refers to traffic that is relatively insensitive to time

delay. Examples of asynchronous data are available bit rate

traffic like electronic mail and file transfers. Time-bounded

traffic, on the other hand, is traffic that is bounded by specified

time delays to achieve an acceptable quality of service

(QoS) (e.g., packetized voice and video).

Of particular interest in the specification

is the support for two fundamentally

different MAC schemes to

transport asynchronous and timebounded

services. The first scheme,

distributed coordination function

(DCF), is similar to traditional legacy

packet networks supporting besteffort

delivery of the data. The DCF

is designed for asynchronous data

transport, where all users with data

to transmit have an equally fair

chance of accessing the network. The point coordination function

(PCF) is the second MAC scheme. The PCF is based on

polling that is controlled by an access point (AP). The PCF is

primarily designed for the transmission of delay-sensitive traffic.

While the DCF has been studied by several researchers

[4–7], the combined performance of the DCF and PCF operating

in a common repetition interval is much less understood.

In this article, the performance of an ad hoc network (DCFonly)

and an infrastructure network (DCF and PCF) are

investigated by means of simulation. We also investigate the

effect of channel errors on the performances of PCF and

DCF, which is absent in all previous studies. Channel degradation,

in terms of burst errors due to multipath fading, will

be factored into the simulations, and the effects on throughput

and delay will be determined. We also develop an efficient

polling scheme used during the PCF to drop inactive

stations from the polling list for a polling cycle, thereby providing

more bandwidth to currently active stations.

In the remainder of the article, we will summarize the

IEEE 802.11 WLAN specification (emphasis on the MAC

sublayer), briefly describe the simulation model which supports

asynchronous data and packetized voice traffic, and provide

performance results from the simulation.

DESCRIPTION OF THE

IEEE 802.11 DRAFT STANDARD

ARCHITECTURE

The basic service set (BSS) is the fundamental building block

of the IEEE 802.11 architecture. A BSS is defined as a group

of stations that are under the direct control of a single coordination

function (i.e., a DCF or PCF) which is defined below.

The geographical area covered by the BSS is known as the

basic service area (BSA), which is analogous to a cell in a cellular

communications network. Conceptually, all stations in a

BSS can communicate directly with all other stations in a BSS.

However, transmission medium degradations due to multipath

fading, or interference from nearby BSSs reusing the same

physical-layer characteristics (e.g., frequency and spreading

code, or hopping pattern), can cause some stations to appear

“hidden” from other stations.

An ad hoc network is a deliberate grouping of stations into

a single BSS for the purposes of internetworked communications

without the aid of an infrastructure network. Figure 1 is

an illustration of an independent BSS (IBSS), which is the formal

name of an ad hoc network in the IEEE 802.11 standard.

Any station can establish a direct communications session

with any other station in the BSS, without the requirement of

channeling all traffic through a centralized access point (AP).

In contrast to the ad hoc network, infrastructure networks

are established to provide wireless users with specific services

and range extension. Infrastructure networks in the context of

IEEE 802.11 are established using APs. The AP is analogous

to the base station in a cellular communications network. The

n Figure 1. Sketch of an ad hoc network.

STA

Independent BSS

STA STA

118 IEEE Communications Magazine • September 1997

AP supports range extension by providing

the integration points necessary

for network connectivity

between multiple BSSs, thus forming

an extended service set (ESS).

The ESS has the appearance of one

large BSS to the logical link control

(LLC) sublayer of each station

(STA). The ESS consists of multiple

BSSs that are integrated together

using a common distribution

system (DS). The DS can be thought

of as a backbone network that is

responsible for MAC-level transport

of MAC service data units

(MSDUs). The DS, as specified by

IEEE 802.11, is implementationindependent.

Therefore, the DS

could be a wired IEEE 802.3 Ethernet

LAN, IEEE 802.4 token bus LAN, IEEE 802.5 token ring

LAN, fiber distributed data interface (FDDI) metropolitan

area network (MAN), or another IEEE 802.11 wireless medium.

Note that while the DS could physically be the same

transmission medium as the BSS, they are logically different,

because the DS is solely used as a transport backbone to

transfer packets between different BSSs in the ESS.

An ESS can also provide gateway access for wireless users

into a wired network such as the Internet. This is accomplished

via a device known as a portal. The portal is a logical

entity that specifies the integration point on the DS where the

IEEE 802.11 network integrates with a non-IEEE 802.11 network.

If the network is an IEEE 802.X, the portal incorporates

functions which are analogous to a bridge; that is, it

provides range extension and the translation between different

frame formats. Figure 2 illustrates a simple ESS developed

with two BSSs, a DS, and a portal access to a wired LAN.

PHYSICAL LAYER

The IEEE 802.11 draft specification calls for three different

physical-layer implementations: frequency hopping spread

spectrum (FHSS), direct sequence spread spectrum (DSSS),

IEEE 802.11

Wireless Local Area Networks

0163-6804/97/$10.00 © 1997 IEEE

ireless computing is a rapidly emerging technology

providing users with network connectivity without

being tethered off of a wired network. Wireless local area networks

(WLANs), like their wired counterparts, are being

developed to provide high bandwidth to users in a limited

geographical area. WLANs are being studied as an alternative

to the high installation and maintenance costs incurred by traditional

additions, deletions, and changes experienced in

wired LAN infrastructures. Physical and environmental necessity

is another driving factor in favor of WLANs. Typically,

new building architectures are planned with network connectivity

factored into the building requirements. However, users

inhabiting existing buildings may find it infeasible to retrofit

existing structures for wired network access. Examples of

structures that are very difficult to wire include concrete

buildings, trading floors, manufacturing facilities, warehouses,

and historical buildings. Lastly, the operational environment

may not accommodate a wired network, or the network may

be temporary and operational for a very short time, making

the installation of a wired network impractical. Examples

where this is true include ad hoc networking needs such as

conference registration centers, campus classrooms, emergency

relief centers, and tactical military environments.

Ideally, users of wireless networks will want the same services

and capabilities that they have commonly come to expect

with wired networks. However, to meet these objectives, the

wireless community faces certain challenges and constraints

that are not imposed on their wired counterparts.

Frequency Allocation — Operation of a wireless network

requires that all users operate on a common frequency band.

Frequency bands for particular uses must typically be approved

and licensed in each country, which is a time-consuming process

due to the high demand for available radio spectrum.

Interference and Reliability — Interference in wireless communications

can be caused by simultaneous transmissions (i.e.,

collisions) by two or more sources sharing the same frequency

band. Collisions are typically the result of multiple stations waiting

for the channel to become idle and then beginning transmission

at the same time. Collisions are also caused by the “hidden

terminal” problem, where a station, believing the channel is

idle, begins transmission without successfully detecting the

presence of a transmission already in progress. Interference is

also caused by multipath fading, which is characterized by random

amplitude and phase fluctuations at the receiver. The

reliability of the communications channel is typically measured

by the average bit error rate (BER). For packetized

voice, packet loss rates on the order of 10–2 are generally

acceptable; for uncoded data, a BER of 10–5 is regarded as

acceptable. Automatic repeat request (ARQ) and forward

error correction (FEC) are used to increase reliability.

Security — In a wired network, the transmission medium can

be physically secured, and access to the network is easily controlled.

A wireless network is more difficult to secure, since

the transmission medium is open to anyone within the geographical

range of a transmitter. Data privacy is usually

accomplished over a radio medium using encryption. While

encryption of wireless traffic can be achieved, it is usually at

the expense of increased cost and decreased performance.

Power Consumption — Typically, devices connected to a

wired network are powered by the local 110 V commercial power

provided in a building. Wireless devices, however, are meant to

be portable and/or mobile, and are typically battery powered.

Therefore, devices must be designed to be very energy-effi-

Brian P. Crow, The MITRE Corporation

Indra Widjaja, Fujitsu Network Communications

Jeong Geun Kim, University of Arizona

Prescott T. Sakai, Cypress Semiconductor