Performance comparison of the IEEE 802.

11 and AIr
infrared wireless MAC protocols
P. Barker, A.C. Boucouvalas
Multimedia Communications Research Group
School of Design, Engineering & Computing
Bournemouth University
Fern Barrow, Poole, BH12 5BB, UK
Email: {pbarker, toucouv}@bournemouth.ac.uk

Abstract
A performance comparison is made between the IEEE
802.11 Infrared W-LAN and the IrDA Advanced
Infrared (AIr) MAC protocols. Both protocols are
CSMA/CA based medium access schemes that operate
on a similar basis but with certain significant
differences. The study uses an existing analytical
model of the 802.11 MAC protocol with suitable
modifications for the AIr MAC protocol. Verification
of the results is provided from simulation modelling
using the OPNET Modeler package.

1. Introduction
The Infrared (IR) medium has been shown to be an
attractive alternative to RF for short range indoor
wireless data communications, with the principle
benefit being the use of low-cost components with low
power consumption [1]. The Infrared Data Association
(IrDA) has produced a set of standards for directed
point-to-point links with data rates up to 4 Mbps which
have been widely adopted. Devices such as digital
cameras, mobile phones and PDAs can communicate
using IR in addition to the traditional applications of
file transfer and printing [2]. At the heart of the IrDA
protocol is the IrLAP data link layer which is HDLCNRM based. However there are limitations in the IrDA
protocol from the point-to-point nature of the links and
the IrLAP protocol [3]. Recently IrDA and IBM have
produced a draft protocol specification called
Advanced Infrared (AIr). The basis of AIr is to provide
a robust non-directed multiple access IR medium with
a suitable MAC (Medium Access Control) layer [4].
The AIr protocol can be seen as being comparable with
that of the IEEE 802.11 protocol. The 802.11 is an
international standard for wireless local area networks
(W-LANs) which was approved in 1997. The standard
allows a number of physical layer options, one of
which is Infrared [5].
Both systems use a CSMA/CA (Carrier Sensing
Multiple Access with Collision Avoidance) based
MAC protocol with RTS/CTS medium reservation that
operate in a very similar fashion. However there are
subtle but significant differences between the two
systems in their collision avoidance and media access

methods. This paper compares the performance of the

two MAC protocols by examining the throughput
performance in relation to network size and collision
avoidance parameters.
2. The Advanced Infrared (AIr) protocol
The AIr physical layer uses wide angle transceivers
with a robust modulation scheme. The modulation used
is 4PPM with variable repetition encoding (4PPM/VR)
with a base data rate of 4 Mbps. Variable repetition
encoding means repeating each 4PPM symbol a
number (1, 2, 4, 8, or 16) of times and using majority
voting on valid symbols. There is therefore a trade-off
between effective data rate and link quality. By
doubling the symbol repetition, the data rate is halved,
but the signal-to-noise ratio is improved by a factor of
3dB. This also enables variable data rates without
physical changes in transceiver circuitry [6].
The AIr frame structure is as shown in figure 1. The
frame consists of preamble and synchronisation fields
that indicate the start of the frame, a robust header, and
a variable length main body (with PDU data 0 to 2K
bytes) which if present is protected by a 32 bit CRC.
The robust header contains essential information for
MAC operation and is always encoded with the
maximum 16 repetition rate (RR). The main body
(including the CRC) has a variable RR.

Pream

Sync

Robust Header

256 bits
RR=1

160 bits
RR=1

32 bits
RR=16

Main Body

CRC*

variable length
32 bits
RR=1variable RRRR=1
RR=1
* only present with main body
RR=1

Figure 1. AIr Frame Structure

In the AIr protocol the IrLAP layer is split into three
sub-layers of the AIr MAC, the AIr LM (Link
Manager) and the AIr LC (Link Controller). The LM
layer provides multiplexing for client protocols and the
LC layer is a HDLC-ABM (Asynchronous Balanced
Mode) based data link layer. The AIr MAC protocol is
responsible for establishing access to the IR medium
and avoiding packet collision. Access to the medium
can be both reserved (using RTS/CTS exchange) or
unreserved.

2.1 The AIr MAC protocol

Collision avoidance in the AIr MAC protocol involves
waiting a randomly chosen set of Collision Avoidance
Slots (CAS) each of 800secs before transmitting an
RTS frame. If another contending station (i.e. with
collision avoidance timer running) receives the start of
an RTS frame, the collision avoidance (CA) timer is
paused, and resumed (from the beginning of the next
time slot) when the media becomes free again. The
target station replies to the RTS frame with a CTS
frame to confirm establishment of the reservation after
which the sending station transmits a burst of data
packets. The number of packets transmitted is
determined by the packets-per-burst (ppb) parameter
which must be such that the burst time does not exceed
the reservation time, which has an upper limit of
500ms. After the final packet is transmitted, the
sending station sends an EOB (End Of Burst) frame.
The target station replies with an EOBC (End Of Burst
Confirm) frame to complete termination of the
reservation. Stations then synchronise before medium
contention resumes. Also, each time a station requires

to transmit a frame following reception of a frame, the

station must wait a turn-around delay time(TAT). This
is to cover receiver latency and for AIr has a fixed
value of 200 secs. The number of CAS slots used is
chosen randomly from a 'CAS window'. This normally
has a value of 8 by which a CAS value is chosen
randomly from 0 to 7. If two or more stations choose
the same CAS value, then an RTS collision will occur.
This will be detected by the expiration of a Wait-forCTS timer, after which a new CAS value is chosen and
the contention restarted. If the CAS window is small in
relation to the number of contending stations then RTS
collision can be quite likely [7].
The AIr MAC has a CAS window adjustment
algorithm which can be employed to improve
contention. If an RTS collision occurs, the CAS
window for the RTS retry is increased by an
adjustment value. Following a successful transmission,
the CAS window is reduced by the adjustment value.
The adjustment has a recommended value of 4. There
is also an upper and lower limit to the CAS window
[8].

CA slots
TAT

ppb
RTS

DATA

DATA

DATA

EOB

EXIT1

CTS

EOBC

EXIT2

TAT

TAT

Source

Destination
CA slots
defer contention

Other

Figure 2. AIr MAC reserved mode data transfer process

3. The IEEE 802.11 W-LAN protocol

The IEEE 802.11 is an international standard for
wireless local area networks (W-LANs) that was
approved by the IEEE in 1997. The standard specifies a
common MAC layer to a number of possible physical
layers. The physical layer options include 2.4 GHz RF
with direct-sequence or frequency-hopping spread
spectrum, or L-PPM diffuse Infrared. All mediums can
operate at 1 or 2 Mbps data rates.
The frame structure for the IR physical layer is shown
in figure 3. The 'header' consists a number of fields that
indicate the start of the frame and the data rate being
used. The header portions are encoding using on-off-

keying (OOK) modulation whereas the remainder of

the frame uses L-PPM. Specifically, 4-PPM (each
symbol is 2 bits) is used for the 2 Mbps data rate and
16-PPM (each symbol is 4 bits) for the 1 Mbps, with
the same 'slot' time used for each scheme.
Header

Length

CRC

MPDU

96 to 112 slots

16 bits

16 bits

0 to 1500 bytes

Figure 3. 802.11 IR physical layer frame structure

3.1 The 802.11 MAC protocol

DIFS

SIFS

The media access function of the 802.11 MAC

protocol is known as the Distributed Co-ordination
Function (DCF) but is essentially just a CSMA/CA
process. With the DCF, when a station has data to
send, it first listens to the media to see if it is busy. If
the media is detected free for a period equal to a time
period called the Distributed Inter-Frame Space (DIFS)
then a packet may be transmitted. If the channel is
detected busy within the DIFS, then channel access is
deferred until the media is detected free again. The
station then uses a contention window procedure in
which a random number of contention slots are chosen.
If the media is detected busy during the contention
backoff period, the contention timer is paused and
resumed when the media is detected free again. When
the contention timer reaches zero the packet is
transmitted. The media can be accessed using either the
basic access method or with RTS/CTS media
reservation. With the basic access method, a data
packet is transmitted and is acknowledged with an
ACK frame from the receiving station. With the
RTS/CTS method, an RTS/CTS exchange is used to
reserve the media, a single data frame is transmitted
and is acknowledged with an ACK. Also with the
RTS/CTS method, when the link is turned-around, the
station must wait a period called the Short Inter-Frame
Space (SIFS) before transmitting. Any other station
that hears the RTS/CTS exchange updates a Network
Allocation Vector (NAV) to determine the period of
contention deferral. The process is shown in figure 4.

RTS

As with the AIr MAC protocol, the 802.11 DCF uses a

contention window (CW) adjustment algorithm.
However in this case the adjustment is exponential
instead of linear. For an ith successive retry for access
to the medium, the contention window becomes
2i CW min. Upon a successful transmission, the
contention window is returned to CW min. The
maximum contention window CW max is given by
2m CW min, where m is the maximum backoff stage. For
the IR medium, the specified values are CW min = 64
and CW max = 1024 [9].
The principle differences between the AIr MAC and
the 802.11 MAC can therefore be summarised as
follows.

The maximum data rate of AIr is 4 Mbps. The

maximum data rate of 802.11 is 2 Mbps.

AIr uses non-directed line-of-sight links. 802.11IR

uses non-directed non-line-of-sight (diffuse) links.

A burst of data packets can be transmitted with AIr

in each reservation. Only one data packet at a time
transmitted with the 802.11 protocol.

The AIr MAC uses a linear adjustment (both up

and down) of the contention window. The 802.11
MAC uses an exponential (up only) adjustment of
the contention window.

DATA

SIFS

contention slots
SIFS

Source

CTS

ACK

DIFS

RTS

Destination

defer contention (NAV)

DIFS

Other

Figure 4. IEEE 802.11 MAC RTS/CTS channel access process

4. Performance Modelling of MAC processes

The analysis used here uses the performance model of
the 802.11 MAC process by Bianchi [10] with a
modification of the model for the AIr MAC process. In
both cases, if we define a transmission slot period in
which data is transmitted, the normalised throughput S
is determined by:

S=

E[data transmitted in slot]

E[length of slot]

(1)

This can therefore be written as:

S=

PS Ptr L
(2)
(1 Ptr ) + Ptr PS TS + Ptr (1 PS )TC

where TS is the average time the channel is sensed busy

for a successful transmission, and TC is the average
time the channel is busy if a collision occurs. L is the
packet data payload transmission time (we assume
constant packet sizes) and is the contention slot time.
Ptr is the probability of at least one transmission in the
chosen time slot and PS is the probability of the
transmission being successful.

where W is the contention window size. However if the

contention window adjustment algorithm is used then
we require a Markov model analysis of the process.
The Markov model used is different for AIr and 802.11
as AIr uses a linear adjustment and 802.11 uses an
exponential backoff adjustment. The analysis relates
to the probability of collision p. This is assumed to be
constant and independent for the Markov analysis.
However, p is actually dependent on and the network
size n:

p = 1 (1 ) n1

(10)

For the 802.11 MAC process, the times TS and TC are

given by:

This produces a non-linear system from which results

can be obtained by numerical techniques (e.g.
successive approximation).

T S (802 .11) = RTS + SIFS + CTS + SIFS

5. Model verification by simulation

+ ( L + H ) + SIFS + ACK + DIFS

T C (802 .11) = RTS + DIFS

(3)

(4)

where H is data packet overhead transmission time.

For the AIr MAC process, they are given by:

TS ( AIr ) = RTS + TTA + CTS + TTA

+ ppb( L + H ) + EOB + TAT

(5)

+ EOBC + EXIT 2

TC ( AIr ) =

(6)

For both systems, the transmission probability Ptr is

given by:

Ptr = 1 (1 ) n

(7)

where n is the number of stations in the network, and

is probability of transmitting in a chosen time slot. The
probability of a successful transmission PS is the
probability of exactly one transmission on the network
in the chosen time slot, and is therefore given by:

PS =

n (1 ) n 1
Ptr

(8)

The probability is determined by examining the

collision avoidance process for both systems. If there is
no adjustment of the contention window, then it can be
seen that is simply given by:

2
W +1

(9)

Results from the mathematical analysis are verified by

comparison with those from simulation models using
the OPNET Modeler package. OPNET uses a set of
graphical hierarchical domains that represent the
structure of a communications network from the
network topology down to specific processes which are
implemented as C/C++ coded finite state machines.
The IEEE 802.11 OPNET model was initially
produced by Baldwin [11] and is now part of the
standard OPNET model library.
An OPNET model of the AIr protocol has been
produced by the authors [12]. This was based on the
state transition tables provided in the AIr specification
documentation. As draft documents, these contain
certain errors and inconstancies which also needed to
be addressed in creating the model. The models use the
'radio' extension to the Modeler package with
modifications to emulate the behaviour of the IR
medium, which is not inherently supported.
The AIr OPNET model network domain for 5-stations
is shown in figure 5. Each 'node' has a set of attributes
representing the protocol parameters that can be set at
the network level or promoted to the simulation level
where a set of values can be assigned. The model also
includes a 'monitor' node which collects simulation
statistics and compiles results.

Figure 5. shows a normalised throughput comparison

against fixed contention window size (without any
window adjustment), for a network size of 5 stations.
The AIr MAC plots are shown for ppb ( packets-perburst) values of 1 (to compare with the single packet
802.11) and with a ppb of 20. This shows that the AIr
protocol suffers from a much larger contention
overhead from the larger contention slot size.

Figure 5. AIr OPNET 5-station model network

6. Analysis Results
The following fixed parameter settings are used for the
comparison analysis. To have a valid comparison
between the systems, unless otherwise stated, the same
data payload size of 2k bytes (16384 bits) is used. One
of the more obvious differences between the two
systems is the contention slot size. For the AIr MAC,
the CA slot is used in carrier sensing, and so must be
large enough to cover the RTS transmission and the
beginning of the CTS reply to cope with 'hidden nodes'.
The 802.11 uses the DIFS for carriers sensing before
employing the contention slots. These can therefore be
much smaller.
Parameter
Data payload
Data Packet Overhead
Data Rate
RTS
CTS
EOB
TAT
CA slot
EXIT2

Value
16384 bits
250 s
4 Mbps
244 s
232s
232 ms
200 s
800 s
200 s

Figure 5. throughput Vs fixed contention window

size, n = 5
The remaining figures use the contention adjustment
process. For the AIr MAC process, an initial contention
window of 8 slots is used with a adjustment of 4. For
the 802.11 MAC process, an minimum window of 64
slots is used, with a maximum backoff stage of 4.
Figure 6 shows normalised throughput against data
payload size (bits). For AIr MAC plots are shown for a
ppb of 1, 7 and 30 packets. The throughput
performance of 802.11 MAC in fact matches almost
exactly with the AIr MAC with a ppb of around 20.
Again it can be seen that the AIr MAC performance is
much worse for a small bust size.

Table 1. AIr MAC parameter values

Parameter
Data payload
Data Packet Overhead
Data Rate
DIFS
SIFS
RTS
CTS
ACK
CA slot

Value
16384 bits
200 s
2 Mbps
128 s
28 s
144 s
120 s
120 s
8 s

Table 2. 802.11 MAC parameter values

Figure 6. throughput Vs data payload size, n = 5

Figure 7 shows throughput against network size. In
general it can be seen that the performance is
independent of network size above 2 or 3 stations.

Again the 802.11 performance (relative) matches that

of the AIr MAC for a ppb around 20 packets.

Figure 7. throughput Vs network size using

contention window adjustment processes.

7. Conclusions
The IrDA Advanced Infrared (AIr) and IEEE 802.11
W-LAN protocols can be seen as operating on very
similar principles and can therefore be favourably
compared in their performances. The principle
differences are in the maximum available data rate (2
Mbps for 802.11, 4 Mbps for AIr), the number of
packets sent after reservation of the medium (1 for
802.11, multiple for AIr) and the method of contention
window adjustment (exponential backoff for 802.11,
linear adjustment for AIr). However a comparison
analysis of the protocols has revealed that a further
significant difference is the size of the contention
window used. For 802.11 the slot size is 8 secs for the
infrared medium, while AIr has a specified slot size of
800 secs. The 802.11 slot can be much smaller
because it uses the DIFS in packet sensing. This causes
the AIr protocol to suffer when only transmitting 1
packet per reservation making it only slightly better
than the 802.11 despite transmitting at twice the data
rate. The AIr protocol is therefore only takes advantage
of the higher data rate when transmitting a burst of
packets in a reservation which reduces the relative
effect of the larger contention overhead.

References

1.

Kahn, J. M and Barry, J. R. Wireless Infrared

Communications. Proceedings of the IEEE. 1997;
85(2):265-298.

2.

Williams, S. IrDA: Past, Present and Future. IEEE

Personal Communications. 2000; 7(1):11-19.

3.

Barker, P.; Boucouvalas, A. C., and Vitsas, V.

Performance modelling of the IrDA infrared

wireless communications protocol. International

Journal of Communications Systems. 2000; 13(78):589-604.
4.

Gfeller, F. and Hirt, W. Advanced infrared (AIr):

physical layer for reliable transmission and
medium access . International Zurich Seminar on
Broadband Communications ; Feb 15-17 2000;
Zurich, Switzerland: 77-84.

5.

Crow, B.; Widjaja, I.; Kim, J. G., and Sakai, P.

Performance of IEEE 802.11 wireless local area
networks. Proceedings of the SPIE. 1996;
2917:480-491.

6.

IrDA.
Advanced Infrared Physical Layer
Specification, 1.0. Infrared Data Association;
1998.

7.

IBM Corporation. Advanced Infrared (AIr) MAC

Draft Specification . 1999.

8. IBM Corporation.
AIr Link Manager Draft
Specification. 1999.
9. IEEE. Wireless LAN Medium Access Control
(MAC) and Physical Layer Specification 802.11:
IEEE; 1997.
10. Bianchi, G. Performance Analysis of the IEEE
802.11 Distributed Coordination Function. IEEE
Journal on Selected Areas in Communications.
2000; 18(3):535-547.
11. Baldwin, R. O. and Davis, N. J. Implementation
of an IEEE 802.11 Wireless LAN model using
OPNET. Opnetworks '98; 1998; 1998.
12. Barker, P. and Boucouvalas, A.C. A Simulation
Model of the Advanced Infrared (AIr) MAC
protocol using OPNET. CSDSPN 2000,
Bournemouth University, UK, 18-20 July 2000:
153-156