7 Medium access control

The medium access control (MAC) layer provides, among other things, addressing
and channel access control that makes it possible for multiple stations on a network
to communicate. IEEE 802.11 is often referred to as wireless Ethernet and, in terms
of addressing and channel access, 802.11 is indeed similar to Ethernet, which was
standardized as IEEE 802.3. As a member of the IEEE 802 LAN family, IEEE 802.11
makes use of the IEEE 802 48-bit global address space, making it compatible with
Ethernet at the link layer. The 802.11 MAC also supports shared access to the wireless
medium through a technique called carrier sense multiple access with collision avoidance
(CSMA/CA), which is similar to the original (shared medium) Ethernet’s carrier sense
multiple access with collision detect (CSMA/CD). With both techniques, if the channel
is sensed to be “idle,” the station is permitted to transmit, but if the channel is sensed
to be “busy” then the station defers its transmission. However, the very different media
over which Ethernet and 802.11 operate mean that there are some differences.
The Ethernet channel access protocol is essentially to wait for the medium to go
“idle,” begin transmitting and, if a collision is detected while transmitting, to stop
transmitting and begin a random backoff period. It is not feasible for a transmitter to
detect a collision while transmitting in a wireless medium; thus the 802.11 channel
access protocol attempts to avoid collisions. Once the medium goes “idle,” the station
waits a random period during which it continues to sense the medium, and if at the end of
that period the medium is still “idle,” it begins transmitting. The random period reduces
the chances of a collision since another station waiting to access the medium would
likely choose a different period, hence the collision avoidance aspect of CSMA/CA.
The simple distributed, contention-based access protocol supported by the CSMA/CA
technique is the basis for the 802.11 MAC protocol and also where the similarity to
Ethernet ends. The wireless medium, being very different from the wired medium,
necessitates a number of additional features:

r The wireless medium is prone to errors and benefits significantly from having a low
latency, link level error recovery mechanism.
r In a wireless medium not all stations can “hear” all other stations. Some stations may
“hear” the station on one end of an exchange but not the station at the far end (the
hidden node problem).
r The data rate that a channel can support is affected greatly by distance and other
environmental effects. Also, channel conditions may change with time due to station
182 Next Generation Wireless LANs

LLC LLC

MSDU MSDU

MPDU
MAC MAC

PSDU PSDU

PPDU
PHY PHY

STA 1 STA 2

Figure 7.1 Protocol layering and messaging.

mobility or environmental changes. Stations need to continually adjust the data rate
at which they exchange information to optimize throughput.
r Stations, often being mobile, need management mechanisms for associating with and
disassociating from WLANs as they change location.

This chapter provides an overview of the 802.11 MAC prior to the enhancements
introduced in 802.11n. After some background information on protocol layering, there
is a brief overview of 802.11 management functions. This is followed by a more detailed
overview of the channel access and data transfer aspects.

7.1 Protocol layering

Some basic concepts regarding protocol layering and messaging and illustrated in
Figure 7.1 are needed in order to understand the MAC functionality. In this layered
model, each entity, PHY and MAC, offers services to the entity in the layer immediately
above it and user data is transferred between the layers as a service data unit (SDU).
The MAC receives data from the logical link control (LLC) layer, and delivers data to
the LLC layer through the MAC SDU (MSDU). The PHY receives data from the MAC
and delivers data to the MAC in a PHY SDU1 (PSDU).
A protocol is the means by which entities in the layered model exchange data and
control information with their peer entities. This exchange takes place through protocol
data units (PDUs). The MAC exchanges MAC PDUs (MPDUs) with its peer and the
PHY exchanges PHY PDUs with its peer.
Another commonly used term in the 802.11 standard is station or STA, which refers
to the MAC and PHY in the context of the device that incorporates these entities. In

1 The IEEE 802.11 standard uses the term PLCP SDU instead of PHY SDU, where PLCP is the physical
layer convergence procedure, a sublayer at the top of the PHY. The terms are equivalent.
 Medium access control 183

Beacon Interval

TBTT TBTT TBTT TBTT

Medium busy
Beacon

Beacon

Beacon

Beacon
Figure 7.2 Beacon transmission on a busy network.

practice this is the network adaptor in a laptop or the communication subsystem in a

mobile phone. An AP (access point) is a station with additional functions related to
managing an infrastructure BSS and providing access to the distribution system.

7.2 Management functions

As described in Chapter 1, the BSS is the basic building block of an 802.11 WLAN.
There are two types of BSS, the independent BSS (IBSS), which is an ad-hoc association
of stations that communicate with one another directly, and the infrastructure BSS, which
is anchored by an AP that may be connected to a distribution system (DS) and through
which the majority of the data transfer takes place, both station to station and station
to DS. In this book we are primarily concerned with the infrastructure BSS; however,
much of the discussion applies equally to IBSSs.
A station becomes aware of the existence of a BSS through scanning, that is passively
seeking Beacon transmissions or actively probing for the existence of an AP through a
Probe Request/Response exchange.
A station’s membership of a BSS is dynamic. The station may turn on and off, or the
station may be mobile and move in or out of the area covered by the BSS. A station
becomes a member of a BSS by becoming “associated” with the BSS. On leaving the
BSS, a station becomes “disassociated.” In an ESS comprised of multiple infrastructure
BSSs, a station may migrate from one BSS to another BSS within the ESS through
“reassociation.”

7.2.1 Beacons
The AP in an infrastructure BSS periodically broadcasts Beacon frames. The Beacon
period defines a fixed schedule of target beacon transmission time (TBTTs) and the
Beacon frame itself is transmitted on or as close to the TBTT as possible subject to the
medium being idle (Figure 7.2).
The Beacon frame carries regulatory information, capability information and infor-
mation for managing the BSS.

7.2.2 Scanning
Scanning is the process by which a station discovers a BSS and the attributes associated
with that BSS. Two forms of scanning are possible: passive scanning and active scanning.
184 Next Generation Wireless LANs

Passive scanning is a receive only operation that is compatible with all regulatory
domains. With passive scanning the station looks for Beacon transmissions and may
switch channels to find these transmissions. Beacon frames include, among other things,
information on the country code, maximum allowable transmit power, and the channels
to be used for the regulatory domain. Once the station has discovered the AP through its
Beacon transmission and has this regulatory information it may probe the AP directly
for additional information using a Probe Request/Response exchange if that additional
information is not present in the Beacon frame itself.
Active scanning may be used when it is permitted by the regulatory domain in which
the station operates. With active scanning a station transmits Probe Request frames on
each of the channels where it is seeking a BSS. Depending on the generality of the
search, the Probe Request frame includes the following addressing information:
r SSID (service set identifier). The SSID in the Probe Request may be the SSID of the
specific ESS for which the station is seeking BSSs or it may be the wildcard SSID.
r BSSID (BSS identifier). The BSSID in the Probe Request may be the BSSID of a
specific BSS or it may be the wildcard BSSID.
r DA (destination address). The DA of the Probe Request frame is the broadcast
address or the specific MAC address of the AP station.

An AP that receives a broadcast Probe Request sends a Probe Response to the station
making the request if the following conditions are true:(a) the SSID is the wildcard SSID
or matches the SSID of the ESS and (b) the BSSID is the wildcard BSSID or the AP’s
BSSID. Multiple APs may respond to a Probe Request using normal channel access
procedures to avoid collisions.

7.2.3 Authentication
Authentication is the process by which two stations that wish to communicate establish
their identity to a mutually acceptable level. The original 802.11 specification supported
two authentication methods operating at the link level: open system authentication and
shared key authentication. With the former any station may be admitted as a member of
a BSS. With the latter, stations rely on the wired equivalent privacy (WEP) protocol to
demonstrate knowledge of a shared encryption key.
WEP has been shown to be insecure and the newer security techniques in 802.11
support authentication through the IEEE 802.1X standard (IEEE, 2004). Authentication
is performed prior to association; however, a detailed discussion on how this is performed
is beyond the scope of this book.

7.2.4 Association
Before a station is allowed to send data via an AP it must become associated with the
AP. Association provides a mapping between the station and AP that allows messages
within the DS (distribution system) to reach the AP with which the station is associated
 Medium access control 185

and ultimately to the station itself. At any given instant a station may only be associated
with a single AP.
Association is initiated by the station with the station sending an Association Request
to the AP. If the station is admitted, the AP responds with an Association Response.
With the Association Request and Response exchange, the station and AP exchange
capability information (support for optional features) and the AP informs the station of
specific operating parameters within the BSS.

7.2.5 Reassociation
Reassociation supports BSS-transition mobility, allowing a station to move from a
current association with one AP to another within the same ESS. This keeps the DS
informed of the current mapping between AP and station. Reassociation may also be
performed to change attributes of the station association such as station capability
information.
Reassociation is initiated by the station with the station sending a Reassociation
Request to the AP. The AP responds with a Reassociation Response.

7.2.6 Disassociation
Disassociation terminates an existing association and may be performed by either the
station or the AP. Stations should attempt to disassociate when they leave the network.
However, because loss of communication may prevent this, a timeout mechanism allows
the AP to disassociate the station without a message exchange should the station become
unreachable.
To disassociate a station from the BSS, the AP or station sends a Disassociation frame.
Disassociation is not a request, thus the other party merely acknowledges reception of
the frame.

7.3 Distributed channel access

The specific CSMA/CA mechanism used in the 802.11 MAC is referred to as the
distributed coordination function (DCF). A station that wishes to transmit first performs
a clear channel assessment (CCA) by sensing the medium for a fixed duration, the DCF
inter-frame space (DIFS). If the medium is idle then the station assumes that it may
take ownership of the medium and begin a frame exchange sequence. If the medium
is busy, the station waits for the medium to go idle, defers for DIFS, and waits for a
further random backoff period. If the medium remains idle for the DIFS deferral and the
backoff period, the station assumes that it may take ownership of the medium and begin
a frame exchange sequence.
The random backoff period provides the collision avoidance aspect. When the network
is loaded, multiple stations may be waiting for the medium to go idle having accumulated
packets to send while the medium was busy. Since each station probabilistically selects
186 Next Generation Wireless LANs

a different backoff interval, collisions where more than one station begins transmission
at the same time are unlikely.
Once a station has gained access to the medium, it maintains control of the medium
by keeping a minimum gap, the short inter-frame space (SIFS), between frames in a
sequence. Another station will not gain access to the medium during that sequence since
it must defer for a fixed duration that is longer than SIFS. Rules limit the types of frame
exchange sequences that are allowed and the duration of those sequences to prevent one
station from monopolizing the medium.
Fundamental to CSMA/CA is the carrier sense. The DCF uses both physical and
virtual carrier sense functions to determine the state of the medium. The physical carrier
sense resides in the PHY and uses energy detect and preamble detect with frame length
deferral to determine when the medium is busy. The virtual carrier sense resides in the
MAC and uses reservation information carried in the Duration field of the MAC headers
announcing impeding use of the medium. The virtual carrier sense mechanism is called
the network allocation vector (NAV). The medium is determined to be idle only when
both the physical and virtual carrier sense mechanisms indicate it to be so.
The DCF also makes use of the immediate feedback provided by the basic acknowl-
edgement mechanism that has the responder send an ACK frame in response to the
initiator’s data or management frame. Not receiving the ACK response frame is a likely
indication that the initiator’s transmission was not correctly received, either due to
collision or poor channel conditions at the time of the data transmission.
To further minimize the chance of collisions, and as a more robust collision detect
mechanism, the initiating station may begin a sequence with a short control frame
exchange using robustly modulated RTS and CTS frames. This sets the NAV in the
stations surrounding both the initiator and responder, some of which may be hidden
nodes unable to detect the more remote station’s transmissions and thus only able to
defer for frame transmissions from nearby nodes.
The DCF provides a distributed contention-based channel access function. Stations
compete for channel access without the need for a central coordinator or arbiter. This
mechanism is remarkable efficient and fairly apportions bandwidth among the active
stations.

7.3.1 Basic channel access timing

Basic channel access timing from the original 802.11 specification is illustrated in
Figure 7.3. The different inter-frame space (IFS) durations effectively provide access to
the wireless medium at different priority levels.

7.3.1.1 SIFS
The short inter-frame space (SIFS) is used to separate a response frame from the
frame that solicited the response, for example between a data frame and the ACK
response. SIFS is designed to be as short as possible but still accommodate the latencies
incurred in a reasonable implementation. These latencies include the decode latency in
the PHY for demodulating the received frame, the MAC processing time for the received
 Medium access control 187

SIFS Slot
Medium Busy ...

SIFS Burst and response access: e.g. ACK, CTS

PIFS Priority access: e.g. Beacon

DIFS Backoff Data and Management frame access

Figure 7.3 Basic channel access priorities with associated timing.

MAC Processing
PHY Rx Latency

PHY Tx Latency

MAC

PHY

Data ACK Air Interface

SIFS

Figure 7.4 PHY and MAC latencies generating a response frame.

frame and building the response, and the transmitter startup time to send the response
(Figure 7.4).
SIFS is also used to separate individual frames in a back-to-back data burst. Stations
accessing the medium using SIFS timing do not check if the medium is busy, but simply
switch to transmit mode (if not already in transmit mode) and begin transmission at the
SIFS boundary.
The SIFS duration for a particular PHY is defined by the aSIFSTime parameter. For
the 802.11a, 802.11g, and 802.11n PHYs the value is 16 µs.

7.3.1.2 Slot time

Timing for the other IFS durations is SIFS plus an integral number of slot times with
transmission beginning on the slot boundary. In practice, propagation delays and, to a
small extent, implementation inaccuracies mean that each station sees a slightly different
boundary. The slot duration is designed to accommodate this variability and provide
enough time for a transmitting station’s preamble to be detected by neighboring stations
before the next slot boundary. During each slot time, stations not yet transmitting remain
in receive mode and check that the medium remains idle.
The slot time for a particular PHY is defined by the aSlotTime parameter. For the
802.11a, 802.11g, and 802.11n PHYs the value is 9 µs.
188 Next Generation Wireless LANs

7.3.1.3 PIFS
The PCF inter-frame space (PIFS) defer provides the next highest access priority fol-
lowing SIFS and is used to gain priority access to the medium. PIFS is defined by the
following equation:

PIFS = aSIFSTime + aSlotTime (7.1)

The AP uses the PIFS defer to gain access to the medium to send a Beacon, start
a contention free period, or to regain access to the medium if an expected response
frame is not received during a contention free period. Despite its name, which reflects its
original use for the point coordination function (see Section 9.1), PIFS is now also used
for other priority operations, such as by a station that needs to send a Channel Switch
Announcement frame (802.11h).

7.3.1.4 DIFS
The DCF inter-frame space (DIFS) is used by stations operating under the DCF to
transmit data frames and management frames and is defined by the following equation:

DIFS = aSIFSTime + 2 × aSlotTime (7.2)

A station using the DCF is allowed to transmit if it determines that the medium is idle
for the duration of the DIFS, or if it determines that the medium is idle for the duration of
the DIFS plus the remaining backoff time following the reception of a correctly received
frame.

7.3.1.5 Random backoff time

When the medium transitions from busy to idle, multiple stations may be ready to send
data. To minimize collisions, stations wishing to initiate transfer select a random backoff
count and defer for that number of slot times. The random backoff count is selected as
a pseudo-random integer drawn from a uniform distribution over the interval [0, CW],
where CW, an integer value, is the contention window.
The contention window (CW) parameter takes the initial value CWmin and effectively
doubles on each unsuccessful MPDU transmit, for example each time an ACK response
is not received for a data frame. If the CW reaches CWmax it remains at that value until
it is reset. The CW is reset to CWmin after every successful MPDU transmit.
CW, CWmin, and CWmax may take values that are a power of 2 less 1. For the
DCF, CWmin and CWmax are specified according to the particular PHY used. For
802.11a, 802.11g, and 802.11n PHYs, CWmin is 15 and CWmax is 1023. CW would
thus start with the value 15 and when “doubled” take on the next higher power of 2 less
1 until it reaches 1023, i.e. 15, 31, 63, . . . , 1023. The CW was defined this way for easy
implementation: using binary notation, “doubling” is effectively a left shift operation
with a lower order one inserted and the backoff value is obtained by using CW to mask
a full word random number.
 Medium access control 189

Initial backoff

DIFS
STA 1 Frame Defer Frame

Initial backoff Remaining backoff Remaining backoff

DIFS
STA 2 Defer Frame

DIFS
STA 3 Defer Frame

DIFS
STA 4 Defer Frame

Figure 7.5 Backoff procedure.

7.3.1.6 Random backoff procedure

To begin the random backoff procedure, the station selects a random backoff count in
the range [0, CW]. All backoff slots occur following a DIFS during which the medium
is determined to be idle. During each backoff slot the station continues to monitor the
medium. If the medium goes busy during a backoff slot then the backoff procedure is
suspended. The backoff count is resumed when the medium goes idle again for a DIFS
period.
The effect of this procedure is illustrated in Figure 7.5. When multiple stations are
deferring and go into random backoff, then the station selecting the smallest backoff
count (STA 3) will win the contention and transmit first. The remaining stations suspend
their backoff and resume DIFS after the medium goes idle again. The station with the
next largest backoff count will win next (STA 4) and then eventually the station with the
longest backoff count (STA 2). A station that begins a new access (STA 1 again) will
select a random backoff from the full contention window and will thus tend to select a
larger count than the remaining backoff for stations (such as STA 2) that have already
suspended their backoff from a previous access attempt.

7.4 Data/ACK frame exchange

Transmission over a wireless medium is error prone. Data transfer benefits from a low
latency, link level repeat mechanism that allows for the retransmission of frames that
have not been successfully demodulated at the receiver. The basic mechanism by which
this is achieved is to have the station that correctly receives a data frame addressed to
it send an immediate, positive acknowledgement in the form of an ACK frame. If the
station sending the data frame does not receive the ACK frame, it assumes the data frame
was not received and may retransmit it.
Not all data frames can be acknowledged in this way. Broadcast and multicast data
frames are directed to all or a subset of the stations in a WLAN and cannot be acknowl-
edged this way. In an 802.11 WLAN, broadcast and multicast frames do not benefit from
the additional reliability that the acknowledgement mechanism provides.
Data transfer using the Data/ACK exchange is illustrated in Figure 7.6. Here STA 1 is
transferring data to STA 2. STA 1 accesses the medium after a contention period during
190 Next Generation Wireless LANs

Contention

Contention

Contention
Period

Period

Period
STA 1 Data Data Data

ACK

ACK
STA 2

Figure 7.6 Basic data/ACK frame exchange sequence.

which it defers for DIFS followed by a random backoff period. If the medium remains
idle, STA 1 transmits a data frame addressed to STA 2. If STA 2 detects and correctly
demodulates the frame then it responds with an ACK. When STA 1 receives the ACK
it knows that the frame was correctly received and begins channel access again in order
to transmit the next frame. If, as with the second data frame in the figure, STA 2 fails to
successfully demodulate the frame then STA 1 will not receive an ACK and will then
begin channel access again to retransmit the data frame.
The number of retransmission attempts on a particular MSDU is limited. The trans-
mitting station maintains a count of the number of retransmission attempts on an MSDU
and when that count exceeds a configured retry limit the MSDU is discarded.
To enhance the reliability with which acknowledgement feedback is provided, the
ACK frame is modulated robustly, i.e. it is sent using a lower PHY data rate than data
frames sent to the same station. The additional overhead incurred with robust modulation
is relatively small since the ACK frame itself is very short.

7.4.1 Fragmentation
Fragmentation is used to break up large MSDUs to improve the chance that the MSDU
will be received correctly and to reduce the overhead of retransmission. At low data
rates, an unfragmented MSDU can occupy a large amount of air time. For example,
a 1500 byte data frame sent using the 1 Mbps 802.11b rate takes 12 ms to transmit,
making it susceptible to changing channel conditions. A bit error in the frame would
result in the entire frame being retransmitted. With fragmentation the MSDU would be
broken into smaller sections and each section encapsulated in an MPDU. Each MPDU
is sent in a separate PPDU with the preamble of each PPDU providing a new channel
estimate. A bit error would result in only the MPDU carrying the errored segment being
retransmitted.
The fragments making up an MSDU are sent as individual MPDUs. A station may
send each fragment in a separate channel access, or the fragmented MSDU may be sent
as a burst of data MPDUs following a single channel access, as illustrated in Figure 7.7.
An MSDU is fragmented when its length exceeds a threshold specified by the
dot11FragmentationThreshold attribute. Each fragment contains an even number of
bytes and all fragments are the same size, except the last fragment, which may be
smaller. The fragments are delivered in sequence.
 Medium access control 191

MSDU

Contention

MPDU
Period

SIFS

SIFS
Data Data Data
STA 1 (frag 1) (frag 2) (frag 3)

SIFS

SIFS

SIFS
ACK

ACK

ACK
STA 2

Figure 7.7 Fragment burst.

The transmitter of a fragmented MSDU maintains a timer. The attribute

dot11MaxTransmitMSDULifetime specifies the maximum amount of time allowed to
transmit a MSDU. The source station starts the timer on the first attempt to transmit
the first fragment of the MSDU. If the timer exceeds dot11MaxTransmitMSDULifetime
then all remaining fragments are discarded.
The receiver of a fragmented MSDU also maintains a timer. The attribute aMaxRe-
ceiveLifetime specifies the maximum amount of time allowed to receive an MSDU.
The receive MSDU timer starts on reception of the first fragment of the MSDU. If the
timer exceeds aMaxReceiveLifetime then all the fragments of the MSDU are discarded.
Additional fragments which may be received later are also discarded.

7.4.2 Duplicate detection

With retransmission there is the possibility that a frame that was correctly received
may be received again, for example if the transmitter retransmits a frame because the
ACK response itself was not correctly demodulated. To detect duplicate frames the
Data frame includes a Retry bit and a Sequence Control field consisting of a sequence
number and a fragment number. The Retry bit is set on any frame that is retransmitted.
The sequence number is generated as an incrementing sequence of integers assigned to
MSDUs and management frames. If the MSDU or management frame is fragmented
then each fragment receives the same sequence number with an incrementing fragment
number.
To detect duplicate frames, the receiving station keeps track of the sequence number
and fragment numbers of the last MSDU or management frame that it received from
each station communicating with it. In other words, it maintains a cache of <transmit
address, sequence number, fragment number> tuples for each fragment received for the
last sequence number seen from each transmit address other than broadcast or multicast
192 Next Generation Wireless LANs

Contention

Contention
Period

Period
STA 1 Data Data

Contention

Contention
Period

Period
STA 2 Data Data
ACK

ACK

ACK

ACK
STA 3

Figure 7.8 Two stations competing for access.

addresses. If the station receives an MPDU with the Retry bit set that matches an entry
in the cache then it rejects the MPDU as a duplicate.

7.4.3 Data/ACK sequence overhead and fairness

The basic Data/ACK frame exchange has a fixed overhead associated with it. Since most
data frames are successfully transmitted, this overhead includes the contention period
during which the medium is essentially idle, the overhead associated with transmitting
the data frame itself, the radio turnaround time at the receiver (SIFS), and the transmission
of the ACK frame. While this overhead is essentially fixed, the duration of the data frame
is not fixed since it depends on the modulation and coding scheme (data rate) used by
the PHY. The higher the data rate the shorter the data frame duration and the greater the
fixed overhead relative to the overall duration of the transfer.
The distributed channel access mechanism promotes fairness in the sense that all
stations on the network with data to send will, on average, each send the same number of
data frames. If they are all using the same packet size they will see the same throughput
irrespective of their individual PHY data rates. So, for example, suppose there are two
stations on the network, STA 1 and STA 2, both competing for access to send data to STA
3 (Figure 7.8). Suppose also that STA 1 is using a high data rate while STA 2 is using
a low data rate. Each station competes for channel access to send one data frame and
each station will on average get the same number of transmit opportunities. However,
because STA 2 is using a lower data rate it will use proportionately more air time to send
its data frames than STA 1.

7.5 Hidden node problem

The distributed nature of channel access in 802.11 WLANs makes the carrier sense
mechanism critical for collision free operation. The physical carrier sense, which logi-
cally resides in the PHY, is responsible for detecting the transmissions of other stations.
However, in some situations it may not be possible for the physical carrier sense to
detect the transmissions of all stations. Consider the situation in Figure 7.9 where there
 Medium access control 193

Figure 7.9 Hidden node problem.

is data transfer between STA 1 and the AP. Transmissions from STA 1 can be detected
by the AP and STA 2. A distant node, STA 3, can detect transmissions from the AP but
not from STA 1. STA 3 is a hidden node with respect to communication between STA 1
and the AP. When STA 1 transmits a frame to the AP there is a chance that STA 3 would
still see the medium as idle and also begin a frame transmission.

7.5.1 Network allocation vector

One mechanism defined to overcome the hidden node problem is the network allocation
vector (NAV). The NAV is a function that logically resides in the MAC and provides
a virtual carrier sense to augment the physical carrier sense. Each MAC frame carries
a Duration field that is used to update the NAV in any station other than the addressed
station that successfully demodulates the frame. The Duration field holds a time value
that indicates the duration for which the sending station expects the medium to be busy
referenced from the end of the last symbol of the PPDU carrying the MAC frame.
All frames2 include the Duration field and may set the NAV in neighboring stations.
However, to do so the frame must be successfully demodulated by the neighboring sta-
tions. The NAV is most effectively set in neighboring stations using robustly modulated
control frames, such as the RTS/CTS exchange, rather than data frames.

7.5.1.1 RTS/CTS frame exchange

To protect a station’s transmissions from hidden nodes, a station may begin a sequence
with an RTS/CTS exchange as illustrated in Figure 7.10. The RTS (request to send) is
sent by the initiator (STA 1) and the station addressed by the RTS (STA 2) responds with
a CTS (clear to send). The RTS frame occupies less air time than the data frame and
is thus less susceptible to collision than the longer data frame transmitted alone. Also,
loss of the RTS to collision would be quickly detected. The RTS and CTS are robustly
modulating so that they are broadly received.

2 Except the PS-Poll frame, which uses this field for other purposes.
194 Next Generation Wireless LANs

STA 1 RTS Data

AP CTS ACK

NAV (RTS)
STA 2 NAV (CTS)

STA 3 NAV (CTS)

Figure 7.10 RTS/CTS exchange for hidden node protection.

The Duration field of the RTS frame carries a NAV setting to cover the CTS response
plus the time needed for the subsequent frame exchange. The CTS response has its
Duration field set to the Duration field value seen in the RTS less SIFS and the duration
of the CTS response itself. In the diagram, the hidden node (STA 3) would receive the
CTS frame and set its NAV to defer for the subsequent frame exchange. STA 2 sees both
the RTS and CTS.
The RTS/CTS exchange is required when the length of a data or management frame
exceeds the threshold set by the dot11RTSThreshold attribute. The dot11RTSThreshold
is a local management attribute and may be set to 0 so that all MPDUs are delivered with
an RTS/CTS exchange, to the maximum allowed MPDU length so that the RTS/CTS
need not be used at all, or any value in between.

7.5.2 EIFS
Another mechanism used to protect against hidden nodes is the extended inter-frame
space (EIFS). A station uses EIFS instead of DIFS to defer if a frame is detected but
not correctly received, i.e. the MAC determines that the frame check sequence (FCS) is
invalid. EIFS is defined as:

EIFS = aSIFSTime + ACKTxTime + DIFS (7.3)

where ACKTxTime is the time required to transmit an ACK frame at the lowest manda-
tory PHY data rate. EIFS is intended to prevent a station from transmitting over the
ACK of a hidden node when a station is unable to demodulate the data frame and thus
correctly set its NAV. If during the EIFS defer a valid frame is received (for example,
the ACK) then a DIFS defer is used following the actual frame instead of continuing
with EIFS. EIFS usage is illustrated in Figure 7.11.

7.6 Enhanced distributed channel access

Enhanced distributed channel access (EDCA) is an extension of the basic DCF intro-
duced in the 802.11e amendment to support prioritized quality of service (QoS). The
EDCA mechanism defines four access categories (ACs). Each AC is characterized by
 Medium access control 195

Table 7.1 AC relative priorities and mapping from 802.1D user priorities

802.1D 802.1D
Priority User priority Designation AC Designation
Lowest 1 BK
AC BK Background
2 –
0 BE
AC BE Best effort
3 EE
4 CL
AC VI Video
5 VI
6 VO
AC VO Voice
Highest 7 NC

If frame is not correctly received then

NAV is not set Following a frame which is not correctly
received a station uses EIFS since ACK
NAV may not be detected
Data
SIFS

If ACK or any valid frame is received

ACK then station continues with DIFS defer

EIFS

DIFS Backoff

Data

Figure 7.11 EIFS usage.

specific values for a set of access parameters that statistically prioritize channel access
for one AC over another. The relative access priorities of the four ACs and the mapping
of 802.1D (MAC bridging) user priorities to ACs are give in Table 7.1. An MSDU
with a particular user priority is said to belong to a traffic category (TC) with that user
priority.
Under EDCA, egress traffic (traffic leaving the system) is sorted logically into four
queues, one for each AC (Figure 7.12). An instance of the EDCA access function
operates for each queue contending for access with that AC’s access parameters when
the queue is non-empty. The EDCA access functions, like DCF, compete for access to
the medium by deferring for a fixed period, the arbitration inter-frame space (AIFS),
when the medium goes idle and then for a random backoff period. The parameters for
EDCA access are similar to the parameters that are used for the DCF, but defined per
AC. The AIFS value for each AC is referenced as AIFS[AC]. The contention window
from which the random backoff count is selected is referenced as CW[AC].
The contention window for a particular AC, CW[AC], starts with the value
CWmin[AC]. If a frame transmission for a particular AC is not successful CW[AC]
is effectively doubled as described in Section 7.3.1.5. If CW[AC] reaches CWmax[AC]
196 Next Generation Wireless LANs

MSDU
User Priority

Mapping to
AC_BK AC_BE AC_VI AC_VO Access Category

Transmit queues
for ACs

Per-queue EDCA
functions with
internal collision
resolution

Figure 7.12 EDCA reference implementation. Reproduced with permission from IEEE (2007)


c IEEE.

it remains at that value until it is reset. CW[AC] is reset to CWmin[AC] after a successful
MPDU transmit in that AC.
If two (or more) instances of the EDCA access function gain access simultaneously,
the internal collision is resolved by the highest priority AC gaining access and the other
AC behaving as if an external collision occurred by doubling its contention window and
re-arming for another access attempt.

7.6.1 Transmit opportunity

A key concept introduced in the 802.11e amendment is the transmit opportunity (TXOP).
A TXOP is a bounded period during which a station may transfer data of a particular
traffic class. Under EDCA, a TXOP is obtained by the station through the channel access
procedure using access parameters for the particular class of traffic for which the TXOP
will be used. Once the TXOP has been obtained, the station may continue to transmit
data, control, and management frames and receive response frames, provided the frame
sequence duration does not exceed the TXOP limit set for that AC. A TXOP limit of zero
means that only one MSDU or management frame can be transmitted before competing
again for access.
The TXOP concept promotes resource fairness rather than throughput fairness in that
all stations accessing the network with traffic of the same class will on average receive
the same amount of air time. Suppose two stations are competing for access, one with
a high PHY data rate and the other with a low PHY data rate as shown in Figure 7.13.
Both stations will on average receive the same amount of air time, but the station with
the higher PHY data rate will see higher throughput than the station with the lower PHY
 Medium access control 197

Contention TXOP TXOP

Period

STA 1 Data Data Data

Contention
Period
STA 2 Data Data
ACK

ACK

ACK

ACK

ACK
STA 2

Figure 7.13 TXOP usage with different PHY data rates.

SIFS Slot
Medium Busy 16 µs 9 µs
...

SIFS Burst and response access : e.g. ACK, BA, CTS

PIFS Priority access: e.g. Beacon

Non-QoS STA
(legacy) data DIFS Backoff Legacy data /mgmt access
access

AIFS [AC_VO] Backoff[AC_VO] Voice TXOP

QoS STA
AIFS [AC_VI] Backoff[AC_VI] Video TXOP
data access
(802.11e)
AIFS [AC_BE] Backoff[AC_BE] Best Effort TXOP

AIFS [AC_BK] Backoff[AC_BK] Background TXOP

Figure 7.14 Channel access priorities including EDCA with associated timing.

data rate. Contrast this with the situation in Figure 7.8 where both stations see the same
effective throughput but with unequal resource utilization.

7.6.2 Channel access timing with EDCA

Extending Figure 7.3 to include EDCA access timing we arrive at Figure 7.14. This
figure shows the access priorities for the four ACs in relation to each other and to the
DCF.
The arbitration inter-frame space (AIFS) for a particular AC is defined by the
equation

AIFS[AC] = aSIFSTime + AIFSN[AC] × aSlotTime (7.4)

where AIFSN[AC] is the slot count.

198 Next Generation Wireless LANs

Table 7.2 Default EDCA access parameters for 802.11a, 802.11g, and
802.11n PHYs

AC CWmin CWmax AIFSN TXOP limit

AC BK 31 1023 7 0
AC BE 31 1023 3 0
AC VI 15 31 2 3.008 ms
AC VO 7 15 2 1.504 ms
legacy 15 1023 2 0

7.6.3 EDCA access parameters

The EDCA access parameters are provided in the EDCA Parameter Set information
element that is present in Beacon and Probe Response frames. Stations on the BSS use
the last seen version of the parameters and the AP may adjust the parameters over time,
for example based on network load or the number of associated stations.
The default EDCA access parameters for the 802.11a, 802,11b, and 802.11n PHYs are
given in Table 7.2. The default EDCA parameters are used if the AP does not broadcast
a different set of parameters. While not an access category, the table also shows for
comparison the equivalent parameters for the DCF (labeled legacy).
The EDCA access parameters determine the degree by which one AC is prioritized
over another. The AIFSN parameter provides weak differentiation. ACs with lower
values gain access more frequently than ACs with higher values, other parameters
being equal. The CWmin parameter provides much stronger differentiation. The random
backoff count is selected from the range [0, CW], where CW is typically equal to CWmin.
Increasing the range from which the backoff count is selected has a bigger effect on
the overall defer period than differences in AIFSN. The TXOP limit is also a strong
differentiator. An AC with a large TXOP limit will receive more air time than an AC
with a small TXOP limit assuming equal allocation of TXOPs.

7.6.4 EIFS revisited

Under DCF, a station must defer for EIFS instead of DIFS following a frame that is
detected but not successfully demodulated. The intent is to protect a possible ACK
response to the unsuccessfully demodulated frame from a more distant node that may
not be detected by the station. To effect equivalent behavior under EDCA, a station
must defer for EIFS − DIFS + AIFS[AC] following a frame that is detected but not
successfully demodulated.
The EIFS and EIFS − DIFS + AIFS[AC] defer are a convoluted way of saying that
following an unsuccessfully demodulated frame a station must defer for SIFS plus the
duration of an ACK frame before performing the usual DIFS or AIFS defer.
 Medium access control 199

7.6.5 Collision detect

When a station obtains a TXOP it may transmit for the duration of the TXOP, either
as a single transmission or as a burst of back-to-back data frames. There is, however,
a chance that two stations will obtain channel access simultaneously and the resulting
colliding transmissions are likely to be unintelligible to the receiving peers.
To minimize the loss of air time due to these collisions, stations must perform a
short frame exchange at the beginning of the TXOP to detect a collision. The short
frame exchange may be either an RTS/CTS exchange or a short single Data/ACK
exchange.
Collision detect is also necessary for the correct operation of the backoff algorithm,
which must have its contention window size doubled should a collision occur. Doubling
the contention window reduces the chance that a collision will occur the next time the
two stations attempt channel access.
The short frame exchange performed at the beginning of the TXOP also allows the
two stations involved in the frame exchange to set the NAV of their neighboring stations.

7.6.6 QoS Data frame

To support the QoS features and block acknowledgement (discussed below), the 802.11e
amendment introduced a new data frame, the QoS Data frame. The QoS Data frame has
the same fields as the regular Data frame, but includes an additional QoS Control field
(see Section 11.1.5). The QoS Control field carries various subfields for managing QoS
and other features introduced in the amendment. The TID or traffic identifier identifies
the TC to which the frame belongs. Under EDCA, the TID field carries the user priority,
which is mapped to the AC through Table 7.1.
The Ack Policy subfield determines how the data frame is acknowledged by the
receiving peer and carries one of the following values:
r Normal Ack – if correctly received, the recipient responds to the QoS Data frame
with an ACK response.
r No Ack – the recipient does not respond to the QoS Data frame. This may be useful
for traffic that has a low tolerance for jitter or delay and does not benefit from
retransmission.
r No Explicit Ack – there may be a response frame, but it is not an ACK. This policy
is used when polling under centrally coordinated channel access (see Section 9.2).
r Block Ack – the recipient takes no action on the received frame except to record its
reception. This policy is used under the block acknowledgement protocol.

7.7 Block acknowledgement

The block acknowledgement protocol, introduced with the 802.11e amendment,

improves efficiency by allowing for the transfer of a block of data frames that are
acknowledged with a single Block Acknowledgement (BA) frame instead of an ACK
200 Next Generation Wireless LANs

frame for each of the individual data frames. Unlike the normal acknowledgement
mechanism, however, the block acknowledgement mechanism is session oriented and a
station must establish a block acknowledgement session with its peer station for each
traffic identifier (TID) for which block data transfer is to take place. A particular block
acknowledgment session is thus identified by the <transmit address, receive address,
TID> tupple.
The 802.11e amendment introduced two flavors of the block acknowledgement pro-
tocol: immediate block ack and delayed block ack. The two flavors differ in the manner
in which the block acknowledgement control frames are exchanged. Under immediate
block ack the Block Ack Request (BAR) frame solicits an immediate Block Ack (BA)
frame response, i.e. the BA is returned within SIFS of receiving the BAR and thus
within the same TXOP. With delayed block ack the BAR is sent in one TXOP and the
BA response is returned in a separate, subsequent TXOP. Immediate block ack provides
lower latency and improved performance over delayed block ack, which was primarily
defined for ease of implementation.
The block acknowledgement protocol is described in more detail in Section 8.3, but
a brief overview is provided here to illustrate the basic concept. Block acknowledge-
ment is enabled in one direction for a particular TID with the exchange of an ADDBA
Request and ADDBA Response. The station with data to send – the originator – sends
an ADDBA Request to the station that will receive the data – the recipient. The recip-
ient acknowledges the correct receipt of the ADDBA Request with an ACK and then
responds some time later with an ADDBA Response to which the originator responds
with an ACK. The ADDBA exchange allows the originator and responder to exchange
parameters such as the responder reorder buffer size. To tear down a block ack session,
the originator or the recipient sends a DELBA request which, if correctly received, is
acknowledged with an ACK.
Block data transfer occurs as follows. The originator transmits one or more QoS Data
frames addressed to the recipient and containing the TID of the block acknowledgement
session. The Ack Policy field is set to Block Ack. The block of data frames need
not be transmitted in sequence and may include retransmitted frames. The recipient is
responsible for reordering the frames and delivering them in sequence to the higher layer
and performs this function using a reorder buffer. The recipient will hold frames in the
reorder buffer until gaps in the sequence number space are filled. The originator limits
the sequence number range for which acknowledgements are outstanding so as not to
overrun the recipient reorder buffer.
After sending a block of data frames, the originator sends a Block Ack Request (BAR)
frame. The BAR frame performs two functions; it flushes the recipient’s reorder buffer
and it solicits a Block Ack (BA) frame. The recipient’s reorder buffer may need to be
flushed to advance passed holes in the sequence number space resulting from MSDUs
that did not make it through after exhausting their retransmission count or lifetime
limit. Flushing the reorder buffer releases MSDUs that may be held up behind these
holes.
The BAR frame includes a Starting Sequence Control field that contains the sequence
number of the oldest MSDU in the block for which an acknowledgement is expected.
 Medium access control 201

TXOP TXOP
Contention

Contention
Period

Period

BAR
RTS

RTS
STA 1 Data Data Data Data Data Data

(a) CTS

CTS

BA
STA 2

TXOP TXOP
Contention

Contention
Period

Period

BAR
STA 1 Data Data Data Data Data Data Data

(b)
ACK

ACK

BA
STA 2

Figure 7.15 Block data frame exchange sequence.

MSDUs in the recipient’s buffer with sequence numbers that precede the starting
sequence number are either forwarded to the LLC layer (if complete) or discarded
(if one or more fragments are missing). The solicited BA frame contains a bitmap that
represents the acknowledgement state of the data frames received beginning with the
starting sequence number from the BAR frame.
On receiving the BA frame the originator discards acknowledged data frames and
re-queues data frames not acknowledged for retransmission. The originator may also
discard data frames that have reached their retransmission count or lifetime limit.
With a block ack session in place, the originator may still solicit a regular ACK for
QoS Data frames by setting the Ack Policy field to Normal Ack.

7.7.1 Block data frame exchange

The block data frame exchange using the immediate block ack protocol is illustrated in
Figure 7.15(a) with STA 1 transferring data to STA 2. After a contention period, STA
1 gains a TXOP. As a collision detect mechanism and to set the NAV in neighboring
stations, STA 1 performs a short frame exchange, in this case RTS/CTS. STA 1 then sends
a back-to-back burst of data frames with SIFS separating the individual transmissions
until the TXOP limit is reached. Since it has more data to send, STA 1 again accesses
the wireless medium and gains a TXOP. An RTS/CTS exchange is again performed
followed by the remaining frames in the block sent as a back-to-back burst. STA 1
then sends a BAR frame, which solicits a BA response from STA 2. The BA response
indicates which of the data frames in the block were correctly received.
As an alternative to performing the RTS/CTS exchange, STA 1 may use a Data/ACK
exchange with STA 2 to effect collision detect, as illustrated in Figure 7.15(b). Collision
detect is necessary through one of these mechanisms as it reduces the degradation in
network throughput that would otherwise occur if two stations collided for the full
202 Next Generation Wireless LANs

duration of the TXOP. The Data/ACK exchange provides more limited protection near
the transmitter (due to the higher order modulation used for data frames), but is more
efficient than the RTS/CTS exchange with which no information transfer takes place.
It should be noted that the block transfer is independent of the TXOP. The block
transfer may occur over multiple TXOPs or it may be contained within a single TXOP.

References

IEEE (2004). IEEE Std 802.1XTM -2004, IEEE Standard for Local and Metropolitan Area Net-
works: Port-Based Network Access Control.
IEEE (2007). IEEE Std 802.11TM -2007, IEEE Standard for Information Technology – Telecom-
munications and Information Exchange Between Systems – Local and Metropolitan Networks –
Specific Requirements. Part 11: Wireless LAN Medium Access Control (MAC) and Physical
Layer (PHY) Specifications.