Adjacent Channel Interference In IEEE 802.11 N

Adjacent Channel Interference in IEEE 802.11n
Anatolij Zubow and Robert Sombrutzki
Humboldt University Berlin
{zubow, sombrutz}@informatik.hu-berlin.de
Abstract—In this paper we analyze the adverse effects of
Adjacent Channel Interference (ACI) on 802.11 with a focus
on new 802.11n standard. ACI is causing problems that are
related to the carrier sensing mechanism in 802.11. On the
one hand, the carrier sensing is sometimes too restrictive thus
preventing concurrent transmissions which leads to a variant of
the exposed terminal problem. On the other hand, the carrier
sensing is sometimes too optimistic thus causing packet collisions
which is a form of the hidden node problem. Both problems are
especially severe in multi-radio systems, where the radios are very
closely spaced. Such problems already investigated in 802.11a/b/g
still remain with 802.11n. Our results show that the number
of available orthogonal channels in IEEE 802.11n depends on
the spatial spacing between the radios, channel width (20 MHz
vs. 40 MHz), RF band (2.4 vs. 5 GHz) and traffic pattern. In
a multi-radio system the situation is worst, e.g. in the 2.4 GHz
we were not able to find 2 orthogonal channels. The adverse
effect of ACI can be reduced in two ways. First, by increasing
the spatial separation between the radios; a spacing of less than
1 meter already improves the situation significantly, e.g. 40 cm are
sufficient to get 2-3 orthogonal 20 MHz channels in the 2.4 GHz
band with reduced transmission power. Furthermore, a distance
of 90 cm is also sufficient so that a 40 and a 20 MHz channel can
be used simultaneously without any interference. However, in a
multi-radio system the spatial spacing between the radios cannot
be increased due to space limitations. The only option to overcome
ACI related problems is to reduce the transmit power making
power control essential. Finally, our analysis revealed that 802.11
is an inappropriate protocol for multi-channel MAC/routing
protocols based on multi-radio systems where an explicit MAC
layer link-scheduling is more promising.
Index Terms—Wireless Networks, IEEE 802.11n, Multi Chan-
nel, Multi Radio, Adjacent Channel Interference, Orthogonal
Channels, Measurements
I. INTRODUCTION
Wireless networks based on standards like IEEE 802.11
are an important research topic in industry and academia.
To increase the network capacity lots of work was done on
multi-channel MAC and routing protocols that simultaneously
use the multiple channels available in IEEE 802.11 [1]. The
majority of multi-channel protocol designers assume the exis-
tence of several non-overlapping and therefore non-interfering
(orthogonal) channels, e.g. 3 for 802.11b/g and 12 for 802.11a,
when evaluating their protocols. While implementing real-
world prototypes of their multi-channel protocols, some au-
thors realized that Adjacent Channel Interference (ACI) be-
tween supposable non-overlapping channels causes serious
problems when used with 802.11. The impact from ACI was
much higher in multi-radio systems where network devices
are equipped with multiple radios, since the spacing between
antennas of different radios is small due to space constraints.
The majority of multi-channel protocol research is based
on the outdated 802.11a/b/g standard. However, the updated
802.11n standard [2] offers lots of improvements like the
use of wider channels (channel bonding) and the use of less
guard carriers which have an effect on channel orthogonality.
Moreover, the signal filtering was improved, i.e. less energy is
bleeding over to adjacent channels. Therefore, it is necessary
to examine the influence of ACI in 802.11n again using state-
of-the-art hardware and software.
The main contributions of this paper are as follows. First,
we describe the adverse effects of ACI on 802.11 like the
increased probability for hidden and exposed terminal prob-
lems. Second, we give a brief overview of the radio spectrum
usage in 802.11 with the focus on 802.11n. Third, we present
experimental results showing the impact of ACI on 802.11n
and compare them with results for 802.11b/g. In contrast
to other studies we take a holistic view on the impact of
ACI. Thus we are able to separate the impact of ACI on the
individual components of 802.11 - Clear Channel Assessment
(CCA) at the transmitter and error correction at the receiver.
Fourth, we discuss the impact of our results on current research
fields. We identify promising research areas as well as research
directions where we think that have only little prospects.
II. RELATED WORK
The impact of ACI on 802.11b/g/a was extensively stud-
ied [3]–[6]. The studies can be classified in whether multi-
radio systems, i.e. a network node is equipped with multiple
802.11 radios, or single-radio systems were analyzed. For
multi-radio systems, Draves et al. [4] could not find any
non-interfering channels within 802.11b/g and 802.11a. The
disillusioning result was that they had to operate one radio in
the 2.4 GHz and the other in the 5 GHz band. Of the expected
15 non-interfering frequency channels only two remained. The
first systematical measurement was conducted by Robinson et
al. [5]. They observed that merely plugging an additional wire-
less card into a PC workstation and operating it in a passive
monitor mode can reduce the throughput. They accounted this
to board crosstalk and radiation leakage of the passive cards.
Regarding single-radio systems, Adya et al. [3] found out
that separating two radios by at least 30 cm 3 non-interfering
channels became available for 802.11b/g. In our previous
study we evaluated 802.11b/g/a for both multi-radio as well
as single-radio systems [6]. The results can be summarized
as follows: The number of available non-interfering channels
depends on the spatial spacing between radios, PHY modu-
lation, RF band (2.4 vs. 5 GHz), traffic pattern and whether
2012 IEEE Wireless Communications and Networking Conference: PHY and Fundamentals
978-1-4673-0437-5/12/$31.00 ©2012 IEEE 1163

single- or multi-radio systems are used. A general statement
about channel orthogonality in 802.11 cannot be made. ACI
was identified as main problem. For nearby transceivers (multi-
radio systems) at most 2 non-interfering channels, one within
2.4 GHz and the other within the 5 GHz band was identified.
Moreover, we observed asymmetric packet flows depending on
the considered traffic pattern. Furthermore, we also observed
hardware and software related problems. The most recent work
on ACI together with the upcoming 802.11n standard can be
found in [7]. The authors analyzed the impact of channel
bonding in 802.11n. However, the used experimental setup
allowed them to analyze the impact of ACI on CCA only.
The main observation was that using 40MHz channels in an
unplanned fashion can lead to serious throughput degradation.
III. BACKGROUND
The objective of this section is twofold. First, we describe
the effects of ACI on 802.11 namely a variant of the hidden
and the exposed terminal problem. Finally, the radio spectrum
usage of the different 802.11 PHY modes is presented. The
focus here is to address the particularities of 802.11n.
A. Hidden Terminal Problem
The Hidden Terminal Problem (HTP) is a well known
problem. It happens when transmissions from two nodes,
that cannot hear each other, collide at the receiver for one
of the nodes. Various solutions to solve this problem have
been proposed. The 802.11 standard recommends the use of
RTS/CTS exchange to avoid HTP.
B. Exposed Terminal Problem
The Exposed Terminal Problem (ETP) occurs when a node
is prevented from sending due to the presence of another
transmitter nearby. This occurs because the carrier sense
mechanism (CCA) used in 802.11 is conservative, and prevents
a node from transmitting when another node is transmitting,
for the fear of causing a collision. Several solutions to this
problem have been proposed as well, but most proposals
require modifications of the carrier sensing mechanism.
C. Adjacent Channel Interference
ACI is a form of interference that is caused by nearby
transmitters on distinct frequency channels ”bleeding over”
to another channel [8, p. 74]. When using 802.11 ACI has the
following consequences: For the case of two nearby transmit-
ters the overlapping ACI of one transmitter causes a spurious
carrier sensing at the other thus preventing two concurrent
transmissions. Remember that 802.11 is a protocol based on
Carrier Sense Multiple Access (CSMA) which follows the
listen-before-talk paradigm. That means that a station is only
allowed to transmit if the medium is idle. ACI may trigger the
carrier sensing mechanisms to report that the medium is busy.
In this case the station will misleadingly defer its transmission.
However, at both receivers there is a sufficient high signal-to-
interference ratio so that it would be possible to successfully
decode both signals. Note, the effect of ACI is smaller than
that of co-channel interference because only a small amount
of energy is ”bleeding over” to another channel. Thus ACI
causes a variant of the ETP which significantly reduces the
spatial reuse in the network and thus wasting radio resources.
In addition to the ETP a variant of the HTP can be caused
by ACI. Here we distinguish between two cases. For the
case of a receiver and a transmitter on adjacent channels in
close proximity the weak incoming signal at the receiver gets
corrupted by the ACI from the strong outgoing signal of the
nearby transmitter. The reason for that is that the interfering
sender is unable to sense an ongoing transmission via CCA. In
contrast to the above ETP problem here the CCA mechanism
is not sensitive enough.
We also observed a problem for two receiving nodes in close
vicinity. Here ACI corrupts the weak signal at both receivers.
In contrast to the above HTP problem here both transmissions
are suffering. Again, ACI causes a variant of the HTP that
like the one before cannot be tackled with RTS/CTS since the
two links are on distinct channels and therefore the receiver is
unable to decode the network allocation vector (NAV) value
from the RTS/CTS packets. We will later show that even for
two nodes in very close proximity it is not possible to receive
a packet send on an adjacent channel (crosstalk)1
.
D. Radio Spectrum Usage
1) 802.11b/g/a: The Direct Sequence Spread Spectrum
(DSSS) PHY (802.11b) has at most 13 channels2
in the
2.4GHz band with channel spacing of 5 MHz. Channel 1 is
centered at 2.412GHz, channel 2 at 2.417GHz, and so on.
Within a channel, most of the signal energy is spread across
a 22 MHz band. To prevent interference to adjacent channels,
the first side lobe is filtered to 30 dB below the power at the
channel center frequency. Additional lobes are filtered to 50 dB
below the power at the channel center [9]. ACI influences
the number of channels that can be used simultaneously. The
IEEE 802.11b specifies that 25 MHz spacing is sufficient. The
multi-carrier (OFDM) PHY in the 5 GHz band (802.11a)
offers 8 channels for indoor and 11 for outdoor use, each
20 MHz wide. In comparison to the 2.4GHz band the channel
spacing’s are larger – 20 instead of 5 MHz3
[9]. The 802.11a
transmission spectrum mask is described in the specification
as: ”The transmitted spectrum shall have a 0 dBr bandwidth
not exceeding 18 MHz, -20dBr at 11 MHz frequency offset,
-28dBr at 20 MHz frequency offset and -40 dBr at 30 MHz
frequency offset and above.” [10]. According to 802.11g the
multi-carrier OFDM PHY is also available in the 2.4GHz
band. The transmit mask is the same as for 802.11a.
2) 802.11n: The most important modification in 802.11n
having an impact on ACI is the use of wider channels (channel
bonding). Channels having a bandwidth of 40 MHz (HT40)4
can be used which effectively doubles throughput. Moreover
1Crosstalk is partially working for 802.11b, whereas it is not working for
the OFDM physical layer (802.11a/g/n) at all.
2In Japan there is an additional channel 14.
3Channel 36 is centered at 5.180 GHz, channel 40 at 5.200 GHz, and so on.
4HT stands for “High Throughput”.
1164

to improve the spectral efficiency the number of OFDM data
subcarriers was increased from 48 to 52 which reduces the
number of guard (null) carriers and thus might increase ACI
on neighboring channels. The 802.11n transmission spectrum
mask is described in the 802.11n specification as: ”When
transmitting in a 20 MHz channel, the transmitted spectrum
shall have a 0 dBr bandwidth not exceeding 18 MHz, -20dBr at
11 MHz frequency offset, -28 dBr at 20 MHz frequency offset,
and the maximum of -45dBr and -53dBm/MHz at 30 MHz
frequency offset and above. When transmitting in a 40 MHz
channel, the transmitted spectrum shall have a 0 dBr bandwidth
not exceeding 38 MHz, -20dBr at 21MHz frequency offset,
-28 dBr at 40 MHz offset, and the maximum of -45dBr and
-56 dBm/MHz at 60 MHz frequency offset and above.” [2]
3) Discussion: By comparing the transmission spectrum
masks of the different 802.11 PHY modes with each other
we observe the following. The signal in 802.11b is best
filtered. Starting at a frequency offset of 22 MHz, the signal
is already attenuated by 50 dB. Thus the ACI impact should
be the lowest. In 802.11n (20 MHz) the filtering of the OFDM
signal was improved compared to 802.11a/g - the signal at a
frequency offset of 30 MHz and more must be attenuated by
45 instead of 40 dB. Therefore it is interesting to analyze if
this has an impact on the number of orthogonal channels.
IV. EVALUATION
The objective of this section is to evaluate the impact from
ACI on the performance of 802.11 with a major focus on
802.11n. At first we present the experimental methodology.
Thereafter the results are presented and discussed.
A. Methodology
The experimental setup is given in Fig. 1. We considered the
following three scenarios termed as: (i) TX-TX, (ii) RX-RX
and (iii) TX-RX. In the TX-TX scenario both middle nodes
(M0 and M1) transmit at the same time. Two packet flows
were setup as follows: M0 → L and M1 → R. To obtain
strictly directive flows MAC layer broadcasts were used, i.e.
no acknowledgment packets were sent out. For non-interfering
channels one would expect that both radios in the middle
can transmit in parallel. So the total throughput should equal
the sum of the single packet flows. In the RX-RX scenario
both nodes in the middle are the destinations of two flows
originated in L and R respectively. Since two flows can be
received simultaneously, one would again expect a significant
throughput increase when using two non-interfering channels.
Finally, in the TX-RX scenario the node L sends packets to
M0 while at the same time M1 transmits to the right node.
This scenario mimics a forwarding operation of a relay in a
multi-hop mesh network.
Our objective in considering these three different scenarios
is to analyze problems in 802.11 connected to ACI in an
isolated way. With the help of the TX-TX scenario we intend
to analyze the adverse effects from ETP, whereas the other
two scenarios are used to show the impact from HTP.
To systematically analyze the impact from ACI we used
a moving robot (Fig. 1, right). We automatically varied the
distance between M0 and M1 (d1) from 10cm to 140 cm in
10 cm steps, of which a separation of 10 cm mimics a multi-
radio device, i.e. a device equipped with two or more radios.
Distances d2 and d3 were around 400 and 500cm respectively.
The link between nodes L and R is obstructed by a thin wall.
The goal was to mimic a link with medium SNR. All nodes
were placed 80cm above the ground and had clear line of
sight to each other with the exception of the link between L
and R. Moreover we evaluated different TX power level to
differentiate the impact on weak and strong links respectively.
We used Netgears WNDR3700v2 equipped with 2 wireless
interfaces (Atheros AR9220/9223)5
and 8 internal metamate-
rial antennas from Rayspan6
. On the software side we chose
OpenWrt as operating system and ath9k7
as WiFi driver. The
packet generation and capturing was done with the Click
Modular Router8
software and additional elements for 802.11n
support (see [11]). All nodes were running in WiFi monitor
mode. During the measurements only one radio was active.
During transmission and reception we monitored that the CPU
load remained within safe grounds and did not become the
bottleneck. The measurements took place at night.
M1
R
d1
M0
L
d2
d3
wall
Fig. 1. Measurement setup consists of 3 fixed (M1, L and R) and a single
moving node (M0).
Besides 802.11n bitrates9
we considered the DSSS
(802.11b) and the OFDM physical layers (802.11g/a) with
a bitrate of 5.5 and 6 Mbps, respectively. Broadcast MAC
frames with a size of 2300 Bytes (802.11a/b/g) and 3832 Bytes
(802.11n) were sent out as fast as possible (backlogged
queues). As mentioned since MAC layer acknowledgments
were not used, we were able to obtain strictly directive
flows. For all experiments the link between L and M0 was
fixed at channel 1 for 802.11b/g/n and 149 for 802.11a/n,
respectively, while the channel for the link between M1 and R
was varied from 1-11 and 149-165 respectively (Fig. 1). Note,
that in the 5 GHz band channels 149 to 165 was the largest
working contiguous frequency range (5 channels), which can
be used without support of Dynamic Frequency Selection. If
not otherwise stated for 40 MHz channels in 802.11n we used
5More details on the used hardware can be found in [11]
6see http://www.commnexus.org/assets/011/9474.pdf
7Linux-Wireless: http://http://linuxwireless.org
8The Click Modular Router: http://www.read.cs.ucla.edu/click/
9With bitrate we refer to a specific MCS available in 802.11n [12].
1165

Parameter Value
Location Indoors
Systems Netgear WNDR3700v2
Scenarios TX-TX (M0 + M1 transmitting),
RX-RX (M0 + M1 receiving),
RX-TX (M0 receiving, M1 transmitting)
Antenna separations 10-140 cm (10 cm steps)
Physical layer 802.11a/b/g/n
Transmission power 16 & 27 dBm (2.4 GHz)
12 & 17 dBm (5 GHz)
Bitrates 1 Mbps (.11b), 6 Mbps (.11g) , 6.5 Mbps
(.11n, MCS=0, 20 MHz), 13 Mbps (.11n,
MCS=0, 40 MHz)
Radio frequencies 2.4 GHz (ch. 1–11),
5 GHz (ch. 149–165)
Transmission mode MAC Broadcast
WiFi frame size 3832, 2300 Bytes
RTS/CTS Disabled
Flow duration 30 sec
WiFi country code United States (US)
TABLE I
MEASUREMENTS PARAMETERS
HT40+, i.e. the depicted channel is the center frequency of the
lower 20 MHz band. E.g. for a HT40+ transmission on channel
1 the upper 20MHz band is centered on channel 5. Note, that
using HT40+ the channel 165 becomes unavailable. For each
scenario and channel assignment the experiment lasted 30 s.
Beforehand, the links were independently measured to ensure
that the signal was strong enough and that the Packet Error
Rate was close to zero for all links and the channel utilization10
was low. The remaining parameters we used throughout our
measurements are summarized in Table I.
B. Results
The results section is divided in two parts. At first we
present results for the 2.4 GHz band, afterwards results for the
5 GHz band are presented. For the analysis the most important
statistics were: (i) sending rate at the transmitting nodes and
(ii) receiving rate at the receiving nodes. Both statistics were
computed on MAC layer as well as hardware based. For the
later one the performance registers of the Atheros chip were
read-out [11]. These results leading to the same conclusions
were kept out due to space limitations (ref. to [12]).
1) 2.4GHz Band: Table II summarizes the results for the
2.4 GHz band. The depicted channel spacing was estimated as
follows: (i) in the TX-TX scenario it is the required channel
spacing so that both transmitters M0 and M1 can send with the
maximum rate. We do not take care whether the packets were
also correctly received at both receivers L and R. What counts
is the sending rate at the transmitters measured at the MAC
layer. (ii) in the RX-RX scenario it is the required channel
spacing so that both receivers M0 and M1 can receive with
the maximum rate. (iii) in the TX-RX scenario we present the
required channel spacing so that the receiver M1 receives at
full rate (flow: L → M1) while the transmitter M0 is able
to send at the maximum rate (flow: M0 → R). Note, that
we ignore both the send and the receive rate at L and R
respectively. In general the sending node requires less channel
spacing to be able to send at full rate. Therefore, we examine
10As reported by Atheros performance registers (see [11]).
both cases separately in the TX-RX scenario, i.e. TX-RX (TX)
and TX-RX (RX) represent the required channel spacing so
that the transmitter M0 and the receiver M1 can send/receive
at maximum rate respectively.
We evaluated two different TX powers: 27 and 16dBm
respectively. Assuming the free-space path loss model the
distance of 50 cm at 27 dBm TX power is roughly equivalent
to the distance of 10 cm at 16 dBm TX power. Note, that the
full TX power of 27dBm with a distance of 10 cm mimics a
multi-radio system, i.e. two radios in one network node11
. For
our conclusions we assume that 11 channels are available in
the 2.4 GHz band.
The entries in Table II and III are read as follows. Consider
as an example the TX-TX scenario in the 2.4GHz band where
both flows use HT20 and transmit at 16 dBm. Here a channel
spacing of 9 (equates to 9×5 MHz = 45MHz) is required when
the spatial separation between both radios is 10 to 20 cm.
In the TX-TX scenario the spatial spacing between both
transmitters plays a crucial role. In a multi-radio system the
spacing between the radios is small which makes it impossible
to use more than 1 channel in the 2.4 GHz band. However,
by reducing the TX power from 27 to 16 dBm 2 orthogo-
nal channels become available even for a very close spatial
spacing. When using a high TX power a spatial spacing of at
least 20 cm is necessary so that 2 orthogonal channels become
available. When using a 40 and a 20 MHz channel a spatial
spacing of at least 30 cm or a reduced TX power of 16dBm
is required to get 2 orthogonal channels.
For the RX-RX scenario the spatial spacing between both
receivers plays no role. For 802.11g as well as 802.11n a
channel spacing of 5 channels is sufficient which results in
3 orthogonal channels. When using 802.11b together with
802.11g/n a channel spacing of only 4 is required12
. This is
mainly due to the better signal filtering in 802.11b. When using
a 40MHz (HT40+) together with a 20MHz channel a channel
spacing of 10 is required so that both streams will not interfere
with each other resulting in 2 orthogonal channels.
In the TX-RX scenario the situation is more complicated.
To make our multi-channel relaying scheme possible, the
transmitter M0 and the receiver M1 need channel spacing so
that they both can send/receive at the maximum rate. From our
results we see that the sending node M0 requires less spatial
spacing to be able to send at a full rate. In a multi-radio system
even with a reduced TX power it is impossible to use more
than 1 channel in the 2.4 GHz band. Only by increasing the
spatial spacing beyond 20 cm permits 2 orthogonal channels.
By reducing the TX power to 16 dBm together with a spatial
separation of at least 40 cm it is possible to use 3 orthogonal
channels in 2.4GHz band. When using a 40 and a 20 MHz
channel a reduced power as well as a spatial separation of at
least 90 cm is required to get 2 orthogonal channels.
2) 5 GHz Band: The results for the 5 GHz band are sum-
marized in Table III. As with 2.4GHz in the TX-TX scenario
11We are aware of the fact that our setup does not consider any other
sources of interference like board crosstalk.
12The same is true if only 802.11b is used.
1166

Scenario
802.11b vs. 802.11g 802.11g vs. 802.11g 802.11n: HT20 vs. HT20 802.11n (HT40) vs. 802.11g
16dBm 27dBm 16dBm 27dBm 16dBm 27dBm 16dBm 27dBm
TX-TX
10 - 20 cm : 9 10 - 20 cm : n/a 10 - 20 cm : 9 10 - 20 cm : n/a 10 - 20 cm : 9 10 - 20 cm : n/a 10 cm : 10 - 30 cm : n/a
20 - 40 cm : 5 20 - 40 cm : 10 20 - 40 cm : 5 20 - 40 cm : 10 20 - 50 cm : 5 20 - 40 cm : 10 20 - 40 cm : 9 30 cm - : 10
40 cm - : 4 40 cm - : 8 40 cm - : 4 40 cm - : 8 50 cm - : 4 40 cm - : 8 40 cm - : 8
RX-RX
4 (802.11b)
10 cm - : 5
10 (802.11n)
5 (802.11g) 9 (802.11g)
TX-RX (RX)
10 cm : n/a 10 cm : n/a 10 - 20 cm : n/a 10 - 20 cm : n/a 10 -20 cm : n/a 10 - 20 cm : n/a 10 - 90 cm : n/a
20 - 30 cm : 5 20 - 30 cm : 10 20 - 40 cm : 7 20 - 40 cm : 10 20 -40 cm : 7 20 - 30 cm : 10 90 cm : 10 n/a
40 cm - : 4 40 cm : 8 40 cm - : 6 40 cm : 9 40 cm - : 6 40 cm : 9
TX-RX (TX) 10 cm - : 4 10 cm - : 8
TABLE II
IMPACT OF ACI FOR DIFFERENT SCENARIOS AND CONFIGURATIONS IN THE 2.4 GHZ BAND. SHOWN IS THE MINIMUM CHANNEL SPACING BETWEEN TWO
NON-INTERFERING CHANNELS REQUIRED FOR A GIVEN SPATIAL SPACING BETWEEN THE RADIOS. CHANNEL SPACING OF 1 IS EQUIVALENT TO 5 MHZ.
the spatial separation between both transmitters (M0 +M1) is
very important. For the multi-radio system case we were not
able to find 2 orthogonal channels, i.e. in a frequency chunk of
100MHz it is not possible to use two channels independently.
Again, increasing the spatial separation between the two
transmitting nodes helps: when using two 20 MHz channels
a spatial separation of 20 cm is required so that a channel
spacing of 4 is sufficient. Furthermore the required channel
spacing can be halved when the spatial separation is increased
to 90 cm or beyond. Again reducing the TX power has a
similar effect like increasing the spatial separation.
In the RX-RX scenario the spatial spacing between both
receivers plays no role. A channel spacing of 2 is sufficient
when using two 20 MHz channels. In case of a 40 and a
20 MHz or two 40 MHz channels a spacing of 3 channels is
required.
In the TX-RX scenario we again see that the reception at
M1 requires the most channel spacing. The relaying case is
not possible with only 5 available channels. Even reducing
the TX power does not help which is bad news for multi-
radio systems. Only by increasing the spatial separation to
30 cm allows us to use two orthogonal 20 MHz channels. By
further increasing the spatial separation to 70 cm we can use
three 20 MHz channels simultaneously. A spatial separation of
80 cm is sufficient to simultaneously use two 40MHz chan-
nels. Here we see the decisive advantage of wider channels;
in a frequency chunk of 100MHz it is possible to use two
40 MHz but only three 20 MHz channels independently.
V. IMPLICATIONS
Our results have significant implications on what areas of
research are promising and which have only small prospects.
A. Research Areas with little Prospects
The research of multi-channel protocols on the
MAC/Routing layer is related to the availability of multiple
orthogonal channels in 802.11 which can simultaneously
be used to improve the network capacity. Lots of multi-
channel protocols residing on the MAC and routing layer
were proposed [1]. The majority of them use 802.11 as
physical/MAC layer. It is important to note that the benefit
of a multi-channel protocol heavily depends on the number
of available orthogonal channels; the greater the number the
higher the benefit from multi-channel protocols. However,
our results showed the existence of only a few orthogonal
channels. Moreover, the majority of multi-channel proposals
require multi-radio systems. However, the small spatial
separation between radios results in a high effect from
ACI which means that even fewer orthogonal channels
are available. Furthermore, the current development trend
is to combine even more channels already on the PHY
layer in order to increase link capacity (e.g. 160 MHz in
802.11ac [13]) thus leaving no room for multi-channel
protocols at the MAC and routing layer.
B. Promising Research Areas
In the following we present 5 research directions that we
believe that they are promising.
a) Adjacent Channel Interference: Efforts must be made
to reduce the energy bleeding over to neighboring channel and
thus reducing the adverse effect from ACI. This is especially
vital for multi-radio systems where a node is equipped with
multiple radios. Because of the small spatial separation of
the radios the influence from ACI is very high. The problem
can be solved by two approaches: (i) the use of better signal
filtering or (ii) by using novel adjacent channel interference-
cancellation techniques like the one proposed in [14].
b) Coexistence with Legacy Devices: The current trend
is the continued merging of channels already at the physical
layer. In 802.11n and 802.11ac up to 2 and 8 channels can be
merged together, respectively. Our results show that ACI leads
to an increase of hidden and exposed terminal problems. Very
wide channels (e.g. 160 MHz in 802.11ac) are very vulnerable.
The impact should therefore be examined in great detail.
c) Hidden and Exposed Terminal Problem: The number
of hidden and exposed terminal problem increases signifi-
cantly due to ACI when multi-channel protocols are used
on overlapping channels. We have to think about solutions
how to solve these problems. Note, that the hidden terminal
problem cannot be tackled by the use of RTS/CTS exchange.
This is because in general an RTS or CTS packet cannot be
received on a neighboring channel which is however necessary
in order to decode the included NAV. So a different type of
channel reservation scheme which respects the impact from
1167

Scenario
802.11n: HT20 vs. HT20 802.11n: HT20 vs. HT40 802.11n: HT40 vs. HT20 802.11n: HT40 vs. HT40
12dBm 17dBm 12dBm 17dBm 12dBm 17dBm 12dBm 17dBm
TX-TX
10 cm : 4 10 cm : n/a 10 - 20 cm : n/a 10 - 50 cm : n/a 10 - 20 cm : n/a 10 - 60 cm : n/a 10 - 60 cm : n/a 10 - 120 cm : n/a
20 - 30 cm : 3 20 - 30 cm : 4 20 - 60 cm : 3 60 - 80 cm : 3 30 - 60 cm : 4 70 - 80 cm : 4 70 - 120 cm : 3 130 cm - : 3
40 cm - : 2 40 - 80 cm : 3 70 cm - : 2 90 cm - : 2 70 cm - : 3 90 cm - : 3 130 cm - : 2
90 cm - : 2
RX-RX 2 3
TX-RX (RX)
10 cm : n/a 10 - 20 cm : n/a 10 - 30 cm : n/a 10 - 30 cm : n/a 10 - 20 cm : n/a 10 - 30 cm : n/a 10 - 60 cm : n/a 10 - 70 cm : n/a
20 cm : 4 30 - 60 cm : 3 40 cm - : 2 40 cm - : 2 30 - 40 cm : 4 40 - 70 cm : 4 70 cm - : 3 80 cm - : 3
30 cm - : 2 70 cm - : 2 50 cm - : 3 80 cm - : 3
TX-RX (TX) 1 2
TABLE III
IMPACT OF ACI FOR DIFFERENT SCENARIOS AND CONFIGURATIONS IN THE 5 GHZ BAND. SHOWN IS THE MINIMUM CHANNEL SPACING BETWEEN TWO
NON-INTERFERING CHANNELS REQUIRED FOR A GIVEN SPATIAL SPACING BETWEEN THE RADIOS. CHANNEL SPACING OF 1 IS EQUIVALENT TO 20 MHZ.
ACI explicitly is required.
d) Power Control: Through an intelligent adaptation of
transmit power (power control) the number of orthogonal
channels can be increased significantly. This is particularly
important for the case of multi-radio systems were power
control is the only available simple solution to increase the
number of orthogonal channels.
e) Explicit Link Scheduling: Given our results that ACI
increases the possibility for hidden an exposed terminal prob-
lems it appears that an explicit MAC layer link-scheduling [15]
is more promising than a CSMA protocol like 802.11. Both
problems can be easily solved with such a scheme. This is
particularly of great interest for multi-radio-systems where the
ETP results from the carrier sensing on neighboring channels.
VI. CONCLUSION
We have analyzed the adverse effects of ACI on 802.11 with
a focus on the new 802.11n standard. When using overlapping
channels ACI causes problems that are related to the Carrier
Sensing (CS) mechanism in 802.11 namely variants of the
well known hidden as well as exposed terminal problem. For
the case of two nearby transmitters the CS is too restrictive
preventing concurrent transmissions and thus reducing the
spatial reuse in the network. In the event that a receiver and
a transmitter are located close to each other the CS is too
optimistic causing the weak incoming signal at the receiver
to get corrupted by the ACI from the strong outgoing signal
and thus wasting radio resources. Both problems are especially
severe in multi-radio systems, where the radios are separated
by only a few centimeters.
In the main part we presented results from extensive mea-
surements showing the impact of ACI on three different
scenarios for different spatial spacings between the radios,
channel width and RF band. Our objective was to determine
the number of orthogonal channels. In a multi-radio system
with a small spatial spacing between the radios the results were
poor. From the 11 channels available in 2.4GHz as well as the
5 analyzed channels in 5 GHz we were not able to find more
than 1 orthogonal channel. The adverse effect of ACI can be
reduced in two ways. First, by increasing the spatial separation
between the radios; a spacing of less than 1 meter already
improves the situation significantly, e.g. 40 cm are sufficient
to get 2-3 orthogonal 20 MHz channels in the 2.4GHz band
depending on the transmission power. However, increasing the
spatial separation between the radios is not an option for multi-
radio systems where the spatial spacing between the radios
cannot be increased due to space limitations. Another option
to overcome ACI is the reduction of the transmit power.
Furthermore, we discussed the impact of our results on
current research fields. Here we identified promising research
areas as well as research directions where we think that they
have only little prospects.
REFERENCES
[1] J. Mo, H.-S. W. So, and J. Walrand, “Comparison of Multichannel MAC
Protocols,” IEEE Transactions on Mobile Computing, 2008.
[2] “IEEE Standard for Information technology–Telecommunications and
information exchange between systems–Local and metropolitan area
networks–Specific requirements Part 11: Wireless LAN Medium Access
Control (MAC) and Physical Layer (PHY) Specifications Amendment
5: Enhancements for Higher Throughput,” pp. c1–502, Oct. 2009.
[3] A. Adya, P. Bahl, J. Padhye, A. Wolman, and L. Zhou, “A Multi-
Radio Unification Protocol for IEEE 802.11 Wireless Networks,” in
BROADNETS’04: Proceedings of the First International Conference on
Broadband Networks, Washington, DC, USA, 2004.
[4] R. Draves, J. Padhye, and B. Zill, “Routing in multi-radio, multi-hop
wireless mesh networks,” in MobiCom’04: International Conference on
Mobile Computing and Networking, 2004.
[5] J. Robinson, K. Papagiannaki, C. Diot, X. Guo, and L. Krishnamurthy,
“Experimenting with a Multi-Radio Mesh Networking Testbed,” in
WiNMee’05: Proceedings of the First Workshop on Wireless Network
Measurements, Riva del Garda, Italy, April 2005.
[6] J. Nachtigall, A. Zubow, and J.-P. Redlich, “The impact of adjacent chan-
nel interference in multi-radio systems using IEEE 802.11,” IWCMC
’08., 2008.
[7] V. Shrivastava, S. Rayanchu, J. Yoon, and S. Banerjee, “802.11n under
the microscope,” Internet Measurement Conference ’08, 2008.
[8] T. S. Rappaport, Wireless Communications: Principles and Practice,
2nd ed. Prentice Hall PTR, December 2002.
[9] M. Gast, 802.11 Wireless Networks: The Definitive Guide, Second
Edition. O’Reilly Media, Inc., 2005.
[10] IEEE 802 LAN/MAN Standards Committee, “Wireless LAN Medium
Access Control (MAC) and Physical Layer (PHY) specificationsAmend-
ment 1: High-speed Physical Layer in the 5 GHz band. IEEE Standard
802.11a,” 1999.
[11] A. Zubow and R. Sombrutzki, “Evaluation of a Low-cost IEEE802.11n
MIMO Testbed,” Tech. Rep., 2011.
[12] ——, “Reinvestigating Channel Orthogonality - Adjacent Channel In-
terference in IEEE 802.11n Networks,” Tech. Rep., 2011.
[13] “IEEE P802.11 Wireless LANs, Proposed TGac Draft Amendment,”
Tech. Rep., 2011. [Online]. Available: http://mentor.ieee.org/802.11/
dcn/10/11-10-1361-03-00ac-proposed-tgac-draft-amendment.docx
[14] R. H. Morelos-Zaragoza and S. Kuruba, “An Adaptive Adjacent Channel
Interference Cancellation Technique,” Faculty Publications, 2009.
[15] L. Bao and J. J. Garcia-Luna-Aceves, “Distributed dynamic channel
access scheduling for ad hoc networks,” J. Parallel Distrib. Comput.,
January 2003.
1168

Adjacent Channel Interference In IEEE 802.11 N

More Related Content

Similar to Adjacent Channel Interference In IEEE 802.11 N (20)

More from Asia Smith (20)

Recently uploaded (20)

Adjacent Channel Interference In IEEE 802.11 N