Achieving Single Channel, Full Duplex Wireless
Communication
Jung Il Choi
†
, Mayank Jain
†
, Kannan Srinivasan
†
, Philip Levis, Sachin Katti
Stanford University
California, USA
{jungilchoi,mayjain,srikank}@stanford.edu, pal@cs.stanford.edu, skatti@stanford.edu
†
Co-primary authors
Abstract
This paper discusses the design of a single channel full-duplex
wireless transceiver. The design uses a combination of RF and
baseband techniques to achieve full-duplexing with minimal ef-
fect on link reliability. Experiments on real nodes show the full-
duplex prototype achieves median performance that is within 8%
of an ideal full-duplexing system.
This paper presents Antenna Cancellation, a novel technique for
self-interference cancellation. In conjunction with existing RF in-
terference cancellation and digital baseband interference cancella-
tion, antenna cancellation achieves the amount of self-interference
cancellation required for full-duplex operation.
The paper also discusses potential MAC and network gains with
full-duplexing. It suggests ways in which a full-duplex system can
solve some important problems with existing wireless systems in-
cluding hidden terminals, loss of throughput due to congestion, and
large end-to-end delays.
Categories and Subject Descriptors
C.2.1 [Computer-Communication Networks]: Network Archi-
tecture and Design—Wireless communication
General Terms
Design, Performance
1. INTRODUCTION
A basic precept of wireless communication is that a radio can-
not transmit and receive on the same frequency at the same time,
i.e. operate in a full duplex fashion. As wireless signals attenuate
quickly over distance, the signal from a local transmitting antenna
is hundreds of thousands of times stronger than transmissions from
other nodes. Hence it has been generally assumed that one can-
not decode a received signal at a radio while it is simultaneously
transmitting.
This paper challenges that assumption, and shows via analysis
and practical implementations on 802.15.4 radios that it is possible
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. To copy otherwise, to
republish, to post on servers or to redistribute to lists, requires prior specific
permission and/or a fee.
MobiCom’10, September 20–24, 2010, Chicago, Illinois, USA.
Copyright 2010 ACM 978-1-4503-0181-7/10/09 ...$10.00.
to build full duplex radios. The implementation is fairly simple,
and can be built using off-the-shelf hardware with software radios.
In theory, it is possible to build a full duplex, single channel ra-
dio using existing techniques. For a system with an antenna each
for transmit and receive, since the system knows the transmit an-
tenna’s signal, it can subtract it from the receive antenna’s signal
and decode the remainder using standard techniques. For example,
for 802.15.4 systems, which use 0dBm transmit power, the power
of the transmit antenna’s signal at a receive antenna placed 6 inches
away is ∼-40dBm. The noise floor is ∼-100dBm, hence if we can
remove 60dB of self-interference by cancellation, we can decode
the receive antenna’s signal.
One can envision implementing the above interference cancella-
tion idea completely in the analog domain using noise cancellation
circuits [17]. But practical noise cancellation circuits can only han-
dle a dynamic range of at most 30dB [18], leaving us far off from
our 60dB goal. Similarly, we could implement interference cancel-
lation after ADC sampling in the digital domain using techniques
such as ZigZag decoding [8]. But existing ADCs do not have the
resolution to let the received signal through (which is 60dB below
the noise floor due to the transmit signal’s interference). Even when
combined, these techniques cannot subtract 60dB of interference
necessary to decode signal from the receive antenna.
This paper presents antenna cancellation, a novel technique for
signal cancellation that allows us to implement practical full du-
plex radios. Antenna cancellation by itself provides ∼30dB of sig-
nal cancellation, and in combination with noise cancellation and
digital interference cancellation, provides around 60dB reduction,
allowing a node to simultaneously transmit and receive.
The basic idea behind antenna cancellation is to use two trans-
mit and one receive antenna. For a wavelength λ, the two transmit
antennas are placed at distances d and d +
λ
2
away from the re-
ceive antenna. Offsetting the two transmitters by half a wavelength
causes their signals to add destructively and cancel one another.
This creates a null position where the receive antenna hears a much
weaker signal. We can then apply noise cancellation and digital in-
terference cancellation on the weaker signal to remove any residue.
The evaluation presented in this paper explores how antenna place-
ment affects cancellation and the signal profile at the transmit an-
tenna’s intended receiver. Also, since antenna placement is dictated
by a single carrier frequency while wireless transmission uses a
band of frequencies, we study the impact of bandwidth on antenna
cancellation. We show that for narrowband systems, the technique
is sufficiently robust.
This paper combines three self-interference cancellation schemes,
antenna cancellation, RF interference cancellation, and digital can-
cellation, to implement a practical 802.15.4 full-duplex radio. It
provides results from real world experiments showing the feasibil-
ity of a full-duplex design. The full-duplex prototype comes within
8% of the performance of an ideal full-duplex system. The ideal
full-duplex system would double the aggregate throughput com-
pared to a half-duplex system, while the prototype achieves 84%
median physical layer throughput gain compared to half-duplex op-
eration.
There are three basic limitations to our design: transmit power,
size and bandwidth. Because the combination of techniques have a
limited potential to cancel up to ∼80dB of signal, very strong trans-
mitters cannot be canceled. For example, it cannot completely can-
cel transmitters that are higher than 20dBm: WiFi is just within the
realm of possibility. This limitation can be overcome with the use
of more precise components for implementing antenna and RF in-
terference cancellation. In terms of size, the design requires at least
λ
2
in addition to regular antenna spacing. Our current prototype,
for example, uses the 2.4GHz band and approximately 7 inches of
space for antenna placement (in 5.1GHz, the antenna placement
may be closer). This means that while such an antenna design can
be part of an access point or laptop body, it cannot easily fit in a
PCI-Express wireless card.
Antenna cancellation, as described in this paper, has a funda-
mental limit in performance for any given bandwidth. This makes
antenna cancellation less effective for signals with bandwidth >
100MHz. Many current and planned future wireless technologies
do not use much more bandwidth than 100MHz. Some components
used in this paper are also limited in their operation over larger
bandwidths. The noise cancellation circuit, for example, shows
degraded performance when used with 20MHz 802.11 signals as
compared to 5MHz 802.15.4 signals.
The full-duplexing scheme uses two RF chains per node to achieve
nearly twice the throughput of a half-duplex system. A natural
follow-up question is if this method is any better than using a 2x2
MIMO system, which can also potentially double throughput us-
ing 2 RF chains per node. However, the potential gains due to
full-duplex go beyond the physical layer. With new media access
control (MAC) layer designs that support full duplex, some of the
most challenging problems in wireless networks can be mitigated,
including hidden terminals, congestion, and end-to-end delay in
multihop networks. Section 6 discusses some of these potential
implications in detail.
2. WIRELESS FULL DUPLEXING
This section examines why existing cancellation techniques, RF
and digital, are not enough to achieve full-duplex.
To understand the challenges in implementing wireless full-duplex,
we need to understand the way signals are received at wireless
nodes. The received signal from the antenna is amplified through
an automatic gain control stage (AGC) and downconverted to ei-
ther baseband or intermediate frequency, filtered and then sampled
through an Analog-to-Digital Converter (ADC) to create digital
samples.
The accuracy of digital samples depends on the resolution of the
ADC. The AGC adjusts the gain of the received signal to match
the maximum level of the ADC to get maximum resolution in the
received signal. For the receiver to decode a weaker signal using
digital cancellation, the signal needs to be strong enough to be cap-
tured within the resolution of the ADC. Typical ADCs are 8-12 bit,
representing a range of 48-72dB. For an 8-bit ADC, if the weaker
signal is 40dB lower in power than the stronger signal, it only gets
1-bit resolution.
0
40
80
120
160
200
-0 -10 -20 -30 -40 -50 -60
Throughput (Kbps)
TX Power (dBm)
No Cancellation With Digital Cancellation
Figure 1: Receive throughput using digital interference cancel-
lation with varying self-interference signal power. Digital in-
terference cancellation gives an SNR gain of only about 10dB,
while full-duplexing in this setup requires ∼46dB.
2.1 Limitation of Existing Interference
Cancellation Schemes
A small experiment shows the inefficacy of using only interfer-
ence cancellation on digital samples to implement a full-duplex
node. The “full-duplex” node used for this test has a receive RF
board trying to decode packets from a 802.15.4 transmitter placed
a few meters away. The 802.15.4 node transmits packets at 0dBm
power. The receiver has a perfect link with an SNR of >10dB to
the 802.15.4 transmitter. A second RF board on the full-duplex
node continuously transmits packets causing interference at the re-
ceiver. A digital cancellation technique is used to try and cancel the
node’s self-interference. We defer the details of this technique to
Section 4.2.
Figure 1 shows the resulting throughput for different transmit
powers of the self-interference signal. Even with digital cancella-
tion, the self-interference signal transmit power needs to be ∼36dB
lower than the transmit power of the intended transmitter for the re-
ceiver to receive any intended packets. As a comparison, the figure
also shows that the receiver can receive intended packets, without
any digital cancellation, only if the transmit power of the (self-
)interferer is ∼46dB lower than the intended transmitter. Thus,
digital cancellation gives an SNR gain of 10dB. For a true full-
duplex operation, we want the transmit powers of the intended and
interfering transmitters to be equal.
This shows the limitation of using existing digital interference
cancellation techniques for achieving full-duplex. A node’s trans-
mit signal completely overwhelms its receive ADC such that the
digital samples do not retain any information of the weaker signal
that a node is trying to receive.
Another option is to use an existing RF interference cancella-
tion chip [17] to reduce self-interference before sending the signal
through the ADC stage. An evaluation shows that this technique
can achieve a reduction in interference of ∼25dB [18]. A combi-
nation of RF and digital interference cancellation still falls short of
being able to reduce interference enough to make full-duplex fea-
sible.
This paper introduces an additional mechanism, Antenna Can-
cellation to further reduce the effect of self-interference. After
combining antenna cancellation with RF interference cancellation,
the received digital samples retain enough resolution of the desired
received signal that digital interference cancellation techniques be-
Digital Interference
Cancellation
d
d + λ/2
TX1 TX2RX
QHx220
RF
Interference
Reference
Input
Output
RF Analog
RF ! Baseband
ADC
RF Analog
Baseband ! RF
DAC
Encoder
Decoder
Digital
Interference
Reference
RF Interference Cancellation
TX Signal Path RX Signal Path
Antenna Cancellation
Power Splitter
Figure 2: Block diagram of a wireless full-duplex node.
Colored blocks correspond to different techniques for self-
interference cancellation. The power splitters introduce a 6dB
reduction in signal, thus power from TX1 is 6dB lower com-
pared to power from TX2, without the need for an additional
attenuator.
come feasible. A brief overview of the antenna cancellation scheme
follows.
2.2 Antenna Cancellation
This scheme uses the insight that transmissions from two or more
antennas result in constructive and destructive interference patterns
over space. In the most basic implementation, the transmission sig-
nal from a node is split among two transmit antennas. A separate
receive antenna is placed such that its distance from the two trans-
mit antennas differs by an odd multiple of half the wavelength of
the center frequency of transmission.
For example, if the wavelength of transmission is λ, and the dis-
tance of the receive antenna is d from one transmit antenna, then the
other transmit antenna is placed at d + λ/2 away from the receive
antenna. This causes the signal from the two transmit antennas to
add destructively, thus causing significant attenuation in the signal
received, at the receive antenna.
Destructive interference is most effective when the signal ampli-
tudes at the receiver from the two transmit antennas match. The
input signal to the closer transmit antenna is attenuated to get the
received amplitude to match the signal from the second transmit
antenna, thus achieving better cancellation. A general implemen-
tation could use differently placed or more than three antennas to
achieve better cancellation.
Antennas are optimally placed only for line-of-sight (LOS) com-
ponents. If antennas are placed in a corner, for example, the re-
flected signals from each transmit antenna will not necessarily can-
cel. While this puts a fundamental limitation on the performance of
the antenna cancellation, signal strength of the reflected signals is
typically much weaker than LOS due to longer signal path and at-
tenuation when reflected. It is possible to bring this signal into the
dynamic range of the ADC by using RF interference cancellation
after the antenna cancellation stage.
Figure 2 shows a block diagram of a system incorporating all the
techniques for full-duplex operation. While each technique has its
own limitations, this paper shows connecting all three techniques
in series can overcome the limitations.
3. ANTENNA CANCELLATION
This section analyzes the possible reduction in self-interference
by using antenna cancellation. It also evaluates its limits with re-
spect to bandwidth of the signal being transmitted and the sensitiv-
ity of antenna cancellation to engineering errors. It shows, using
actual measurements, that antenna cancellation achieves 20dB re-
duction in self-interference. This section also evaluates the effects
of using two transmit antennas for antenna cancellation on the com-
munication range. It shows that antenna cancellation degrades the
received signal at other nodes in the network by at most 6dB com-
pared to the single antenna setup.
3.1 Performance of Antenna Cancellation
In an ideal scenario, the amplitudes from the two transmit an-
tennas would be perfectly matched at the receiver and the phase
of the two signals would differ by exactly π. However, we find
that the bandwidth of the transmitted signal places a fundamental
bound on the performance of antenna cancellation. Further, real
world systems are prone to engineering errors which limit system
performance. The sensitivity of the antenna cancellation to ampli-
tude mismatch at the receive antenna and to the error in receive
antenna placement is important to consider.
To analyze the reduction in interference using antenna cancel-
lation, we look at the self-interference signal power at the receive
antenna after antenna cancellation. It is derived in Appendix A to
be:
2A
ant
“
A
ant
+
A
ant
”
|x[t]|
2
„
1 − cos
„
2π
d
ant
λ
««
+
“
A
ant
”
2
|x[t]|
2
where A
ant
is the amplitude of the baseband signal, x[t], at the
receive antenna received from a single transmit antenna.
A
ant
is
the amplitude difference between the received signals from the two
transmit antennas at the receive antenna.
d
ant
represents the error
in receiver antenna placement compared to the ideal case where the
signals from the two antennas arrive π out of phase of each other.
This equation lets us evaluate the sensitivity of antenna cancellation
to receive antenna placement, change of transmit frequency, and
amplitude matching at the receive antenna.
d
ant
also captures the effect of bandwidth on antenna cancella-
tion. Consider a 5MHz signal centered at 2.48GHz. Thus, the sig-
nal has frequency components between 2.4775GHz and 2.4825GHz.
If the receive antenna is placed perfectly for the center frequency,
there is a small error in placement for the other frequencies within
that bandwidth.
We can map the difference in wavelength to the error in receiver
placement. For example, a δ difference in wavelength is similar to
a δ/4 error in receiver placement. Thus,
d
ant
for 2.4775GHz in this
case would be ∼
1
4
`
c
2.4775∗10
6
−
c
2.48∗10
6
´
, where c is the speed
of light. This gives
d
ant
∼ 0.025mm, corresponding to 60.7dB an-
tenna cancellation for the 2.48GHz center frequency. Thus, 60.7dB
is the best antenna cancellation possible for a 5MHz signal in the
2.4GHz band using the 3 antenna scheme described in this paper.
Similarly, using 20MHz and 85MHz bandwidths give best case re-
duction of 46.9dB and 34.3dB respectively.
As can be seen from the effect of bandwidth, antenna cancel-
lation does not provide a frequency flat channel at the receiver if
there is perfect amplitude matching. This distortion in the received
signal can be a problem for the RF and digital interference cancel-
(a) Received power with distance mismatch
(b) Received power with amplitude mismatch
Figure 3: Performance of antenna cancellation with distance
and amplitude mismatch for signals with different bandwidth.
A 1mm mismatch can restrict the receive power reduction to
∼29dB. An amplitude mismatch of 10%, corresponding to 1dB
variation, can restrict the receive power reduction to ∼20dB.
lation stages, since they use the undistorted transmission signal as
reference for cancellation.
Any error in receive antenna placement adds to
d
ant
. To see the
effect of receive antenna placement error, suppose the receive an-
tenna is 1mm off from the optimal position, i.e.
d
ant
= 1mm.
With perfect amplitude matching and with a λ of 12.1cm (for a
center frequency of 2.48GHz), we see a 28.7dB reduction in power
compared to no antenna cancellation. Figure 3(a) shows the theo-
retical performance of antenna cancellation with error in receiver
placement, for different bandwidths.
Figure 3(b) shows the theoretical performance of antenna can-
cellation with error in amplitude matching, assuming perfect center
frequency receiver placement, for different bandwidths. For exam-
ple, say the amplitude of one signal is 10% higher than the other,
i.e.
A
ant
= 0.1 ∗ A
ant
. In this case, the powers of the two signals
differ by ∼ 1dB. With this
A
ant
, the reduction in received power
due to antenna cancellation is 23dB, if we ignore the effect of band-
width. For a 5MHz bandwidth, the same
A
ant
gives a 22.994dB
reduction. Thus, a small amplitude mismatch tends to dominate
the performance restrictions on antenna cancellation. Since ampli-
tude mismatch affects different frequencies equally, the resulting
frequency response is fairly flat, thus giving a less distorted input
to the later cancellation stages. Thus, amplitude mismatch may end
up helping the later stages of interference cancellation.
TX1
TX2
25
35
45
55
65
0 5 10 15 20 25
SNR (dB)
Position of RX Antenna (cm)
Only TX1 Only TX2 Both TX1 and TX2
~20dB
Figure 4: Received SNR for different receive antenna place-
ments. The received SNR is fairly monotonic with distance
when any one transmit antenna is active. With both transmit
antennas active, there is a sharp reduction in receive power at
the null point.
3.2 Antenna Cancellation in Practice
Figure 4 shows the effect of antenna cancellation with transmit-
ter TX1 attenuated by 6dB compared to TX2. Experiments show
that the received power from the two TX antennas differs by about
5.1dB when the receiver is placed at the null point. Thus, this setup
has an amplitude mismatch of ∼1dB causing the cancellation to be
restricted to ∼20dB as shown in the previous analysis. The above
analysis did not consider the multipath effect. However, results
from the measurements show that the multipath effect is not a dom-
inant component in our experimental setup.
3.3 Effect of Antenna Cancellation on Intended
Receivers
While antenna cancellation can reduce self-interference from a
node’s own transmitter, an important question is how this affects
the received signal at nodes other than the transmitter. Another
question is how does our cancellation technique compare to a sim-
ple technique such as having the signals between the two transmit
antennas phase shifted by π. Unlike our technique, the phase shift
approach does not require an attenuator and gives a null point ex-
actly at the center.
The contour map in Figure 5(a) shows received power with both
transmit antennas transmitting a single frequency tone at the same
power with a phase difference of π using a simple simulation with
a freespace propagation model. Each contour line corresponds to
a specific received power. Figure 5(b) shows the received signal
strength with different transmit powers from the transmit antennas
such that amplitudes match at the null point without any phase shift
in antenna signals. The null points achieved in the two cases are at
different locations, but both schemes are equally good in terms of
signal reduction at the null point.
The difference between these two cases becomes clearer by look-
ing at the received signal at larger distances. Figure 5(c) shows the
received signal strength profile, over space, for a single transmit
antenna over a distance of 30m from the transmitter. This is the
baseline for comparison of the two schemes with antenna cancella-
(a) Equal powers (b) Different powers (c) Single transmitter
(d) Equal powers (e) Different powers
Figure 5: Freespace signal strength profiles for equal transmit powers and different transmit powers on two transmit antennas. This
simulation uses a pathloss exponent of 2. Figures (a) and (b) correspond to a short-range study. When transmit powers are equal, the
minimum received signal is in the middle and when the transmit powers are different, the minimum is closer to the lower transmit
power antenna. Figures (c), (d) and (e) correspond to a long-range study. When transmit powers are equal, receivers equidistant
from the transmit antenna pair can see huge differences in the received signal strength. When transmit powers are different, however,
such differences are much smaller.
tion. Figure 5(d) shows the contours over larger distances for the
same setup as Figure 5(a). It is apparent that even in normal com-
munication range, there are locations with very low received power
due to the destructive interference.
Figure 5(e) shows the contours of received power when one trans-
mit signal is attenuated by 6dB compared to the other and there is
no phase shift between the two transmitted signals. The effect of
destructive interference is much lower in this case.
In case of two transmit antennas, the signals from the two anten-
nas get added constructively or destructively at the receiver. At dis-
tances much larger than the spacing between the transmit antennas,
the signals from both antennas undergo almost equal attenuation.
With equal receive power from both antennas, a perfectly destruc-
tive combining of the two signals causes the received signal to be
zero power. In case of unequal transmit powers, the received power
at these distances is different from the two transmit antennas. Even
when the signals combine perfectly out of phase, the resulting sig-
nal is not zero power.
Comparing with the single antenna case, using our antenna can-
cellation scheme leads to a maximum degradation of 6dB at any
receiver location. In a real network setting, diversity gains due to
two transmit antennas would offset this degradation. Thus, antenna
cancellation can give significant reduction at the null position with-
out having a large effect on reception at other nodes. Following
antenna cancellation, further reduction is obtained by RF and digi-
tal interference cancellation techniques.
4. INTERFERENCE CANCELLATION
This section explains two interference cancellation mechanisms
used in full-duplexing nodes after the antenna cancellation stage.
The first is RF interference cancellation using a noise canceler. The
second is digital cancellation that takes place, in software, after the
received signal is discretized.
4.1 RF Interference Cancellation
As Radunovic et al. [18] explored for 900MHz band networks,
the interference cancellation circuit based on QHx220, a noise can-
celer chip, allows removing a known analog interference signal
from a received signal. The QHx220 chip takes the known self-
interference and received signals as inputs and outputs the received
signal with the self-interference subtracted out. The chip allows
changing the amplitude and phase of the interference reference sig-
nal to match the interference in the received signal. An RF splitter
is used to give the transmit signal to the cancellation circuit as the
interference reference.
Figure 6 shows the effect of using the RF cancellation circuit. It
shows spectrum power snapshots at the receive antenna for three
