Achieving Single Channel Full-Duplex Wireless Communication
Achieving Single Channel Full-Duplex Wireless Communication
Achieving Single Channel Full-Duplex Wireless Communication
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
<strong>Achieving</strong> <strong>Single</strong> <strong>Channel</strong> <strong>Full</strong>-<strong>Duplex</strong><strong>Wireless</strong> <strong>Communication</strong>Jung Il Choi, Mayank Jain, Kannan Srinivasan,Philip Levis and Sachin Katti1
Can a wireless node transmit ANDreceive at the same time on a single band?2
Can a wireless node transmit ANDreceive at the same time on a single band?Status quo: NO3
Current wireless radios• In-band half-duplex• <strong>Full</strong>-duplex through other dimensions• E.g. different frequencies• Bandwidth is a precious resource 4
Why not full-duplex on the same band?5
Why not full-duplex on the same band?• Very strong self-interference• ~70dB stronger for 802.15.4TX RXTX RX• Analog to Digital converter (ADC) saturates6
Existing Techniques• Digital cancellation: Subtracting known interferencedigital samples from received digital samples.ZigZag [1] , Analog Network Coding [2] etc.• Hardware cancellation: RF noise cancellation circuits withtransmit signal as noise referenceRadunovic et al. [3][1] Gollakota et al. “ZigZag Decoding: Combating Hidden Terminals in <strong>Wireless</strong> Networks”, ACM SIGCOMM 2008[2] Katti et al. “Embracing <strong>Wireless</strong> Interference: Analog Network Coding”, ACM SIGCOMM 2007[3] Radunovic et al. , "Rethinking Indoor <strong>Wireless</strong>: Lower Power, Low Frequency, <strong>Full</strong>-duplex", WiMesh (SECONWorkshop),, 20107
Existing Techniques• Digital cancellation: Subtracting known interferencedigital samples from received digital samples.ZigZag [1] , Analog Network Coding [2] etc.Ineffective if ADC is saturated• Hardware cancellation: RF noise cancellation circuits withtransmit signal as noise referenceRadunovic et al. [3][1] Gollakota et al. “ZigZag Decoding: Combating Hidden Terminals in <strong>Wireless</strong> Networks”, ACM SIGCOMM 2008[2] Katti et al. “Embracing <strong>Wireless</strong> Interference: Analog Network Coding”, ACM SIGCOMM 2007[3] Radunovic et al. , "Rethinking Indoor <strong>Wireless</strong>: Lower Power, Low Frequency, <strong>Full</strong>-duplex", WiMesh (SECONWorkshop),, 20108
Existing Techniques• Digital cancellation: Subtracting known interferencedigital samples from received digital samples.ZigZag [1] , Analog Network Coding [2] etc.~15dBIneffective if ADC is saturated• Hardware cancellation: RF noise cancellation circuits withtransmit signal as noise referenceRadunovic et al. [3]~25dBThese are not enough 25dB +15dB < 70dB9
Our innovation: Antenna CancellationTX1RXd d + λ/2TX210
Our innovation: Antenna CancellationTX1RXd d + λ/2TX2~30dB self-interference cancellationEnables full-duplex when combined with Digital (15dB)and Hardware (25dB) cancellation.11
Can a wireless node transmit AND receiveat the same time on a single band?12
Can a wireless node transmit AND receiveat the same time on a single band?YES, IT CAN!<strong>Full</strong>-duplex prototype achieves 92% of thethroughput of an “ideal” full-duplex system13
Talk Outline• Design of <strong>Full</strong>-<strong>Duplex</strong> <strong>Wireless</strong>• 3 Techniques: Antenna, Hardware andDigital Cancellation• Analyzing Antenna Cancellation• Performance Results• Implications to <strong>Wireless</strong> Networks• Limitations of Design, Future Work14
Talk Outline• Design of <strong>Full</strong>-<strong>Duplex</strong> <strong>Wireless</strong>• 3 Techniques: Antenna, Hardware andDigital Cancellation• Analyzing Antenna Cancellation• Performance Results• Implications to <strong>Wireless</strong> Networks• Limitations of Design, Future Work15
Three techniques give ~70dB cancellation• Antenna Cancellation (~30dB)• Hardware Cancellation (~25dB)• Digital Cancellation (~15dB) 16
Hardware and Digital CancellationHardware Cancellation• Use existing interferencecancellation circuits(QHx220) * 18Digital Cancellation• Subtract known transmit samples from receiveddigital samples* Radunovic et al. , "Rethinking Indoor <strong>Wireless</strong>: Lower Power, Low Frequency, <strong>Full</strong>-duplex",MSR Tech Report, 2009
Bringing It TogetherAntennaCancellationRXTX SignalHardwareCancellationQHX220RFADC+BasebandDigitalCancellationTX Samples-∑Clean RX samples19
Bringing It TogetherAntennaCancellationRXTX SignalHardwareCancellationQHX220RFADC+BasebandDigitalCancellationTX Samples-∑Clean RX samples20
Bringing It TogetherAntennaCancellationRXTX SignalHardwareCancellationQHX220RFADC+BasebandDigitalCancellationTX Samples-∑Clean RX samples21
Bringing It TogetherAntennaCancellationRXTX SignalHardwareCancellationQHX220RFADC+BasebandDigitalCancellationTX Samples-∑Clean RX samples22
Our PrototypeAntennaCancellationDigitalInterferenceCancellationHardwareCancellation23
Talk Outline• Design of <strong>Full</strong>-<strong>Duplex</strong> <strong>Wireless</strong>• 3 Techniques: Antenna, Hardware andDigital Cancellation• Analyzing Antenna Cancellation• Performance Results• Implications to <strong>Wireless</strong> Networks• Limitations of Design, Future Work24
Antenna Cancellation: PerformanceTX1TX2-25-30Only TX1 ActiveRSSI (dBm)-35-40-45-50-55-600 5 10 15 20 25Position of Receive Antenna (cm)25
Antenna Cancellation: PerformanceTX1TX2RSSI (dBm)-25-30-35-40-45-50-55Only TX1 ActiveOnly TX2 Active-600 5 10 15 20 25Position of Receive Antenna (cm)26
Antenna Cancellation: PerformanceTX1TX2-25-30Only TX1 ActiveBoth TX1 &TX2 ActiveOnly TX2 Active-35RSSI (dBm)-40-45-50-55NullPosition-600 5 10 15 20 25Position of Receive Antenna (cm)27
Antenna Cancellation: PerformanceTX1TX2-25-30Only TX1 ActiveBoth TX1 &TX2 ActiveOnly TX2 Active-35RSSI (dBm)-40-45-50-55NullPosition~25-30dB-600 5 10 15 20 25Position of Receive Antenna (cm)28
Sensitivity of Antenna CancellationReduction Limit (dB)Reduction Limit (dB)dBAmplitude Mismatchbetween TX1 and TX2Error (mm)Placement Errorfor RX29
Sensitivity of Antenna CancellationReduction Limit (dB)Reduction Limit (dB)dBAmplitude Mismatchbetween TX1 and TX2Error (mm)Placement Errorfor RX30dB cancellation < 5% (~0.5dB) amplitude mismatch< 1mm distance mismatch30
Sensitivity of Antenna CancellationReduction Limit (dB)Reduction Limit (dB)dBAmplitude Mismatchbetween TX1 and TX2Error (mm)Placement Errorfor RX• Rough prototype good for 802.15.4• More precision needed for higher power systems (802.11)31
Bandwidth ConstraintA λ/2 offset is precise for one frequencyTX1RXd d + λ/2TX2fc32
Bandwidth ConstraintA λ/2 offset is precise for one frequencynot for the whole bandwidthTX1RXTX2fc -Bfcfc+Bd d + λ/233
Bandwidth ConstraintA λ/2 offset is precise for one frequencynot for the whole bandwidthTX1RXTX2d1d1 + λ-B/2TX1RXTX2d d + λ/2fc -Bfcfc+BTX1RXTX2d2d2 + λ+B/234
Bandwidth ConstraintA λ/2 offset is precise for one frequencynot for the whole bandwidthTX1RXTX2d1d1 + λ-B/2TX1RXTX2d d + λ/2fc -Bfcfc+BTX1RXTX2d2d2 + λ+B/2WiFi (2.4G, 20MHz) => ~0.26mm precision error35
Bandwidth Constraint300 MHz2.4 GHz5.1 GHz36
Bandwidth Constraint300 MHz2.4 GHz5.1 GHz• WiFi (2.4GHz, 20MHz): Max 47dB reduction• Bandwidth => Cancellation• Carrier Frequency => Cancellation37
What about attenuation at intended receivers?Destructive interference can affect this signal too!38
What about attenuation at intended receivers?Destructive interference can affect this signal too!• Different transmit powers for two TX helpsDeep Nulls at 20-30my axis (meters)3020100-10-20-30-30 -20 -10 0 10 20 30x axis (meters)Equal Transmit Powery axis (meters)3020100-10-20-30-30 -20 -10 0 10 20 30x axis (meters)Unequal Transmit Power39
What about attenuation at intended receivers?Destructive interference can affect this signal too!• Different transmit powers for two TX helpsy axis (meters)3020100-10-20-52 dBm -100 dBmy axis (meters)-30-30-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30x axis (meters)x axis (meters)Equal Transmit Power Unequal Transmit Power3020100-10-20-58 dBm-52 dBm40
What about attenuation at intended receivers?Destructive interference can affect this signal too!• Different transmit powers for two TX helps• Diversity gains in indoor environments41
Talk Outline• Design of <strong>Full</strong>-<strong>Duplex</strong> <strong>Wireless</strong>• 3 Techniques: Antenna, Hardware andDigital Cancellation• Analyzing Antenna Cancellation• Performance Results• Implications to <strong>Wireless</strong> Networks• Limitations of Design, Future Work42
Experimental Setup• 802.15.4 based signaling on USRP nodes• Two nodes at varying distances placed in anoffice building room and corridor43
Half-<strong>Duplex</strong> :- Nodes interleave transmissionsNode 1 ! 2Node 2 ! 1<strong>Full</strong>-<strong>Duplex</strong> :- Nodes transmit concurrentlyNode 1 ! 2Node 2 ! 1• <strong>Full</strong>-duplex should double aggregate throughput44
ThroughputHalf-<strong>Duplex</strong> <strong>Full</strong>-<strong>Duplex</strong> Ideal <strong>Full</strong>-<strong>Duplex</strong>1.0CDF0.80.60.40.201.84x0 50 100 150 200 250 300Throughput (Kbps)Median throughput 92% of ideal full-duplex45
ThroughputHalf-<strong>Duplex</strong> <strong>Full</strong>-<strong>Duplex</strong> Ideal <strong>Full</strong>-<strong>Duplex</strong>1.00.8CDF0.60.40.201.84x0 50 100 150 200 250 300Performance lossThroughput (Kbps)at low SNR46
Link Reception RatioHalf-<strong>Duplex</strong><strong>Full</strong>-<strong>Duplex</strong>Packet Reception Ratio1.000.750.500.2500 10 20 30 40SNR (dB)Little loss in link reliability: 88% of half-duplex on average47
Link Reception RatioHalf-<strong>Duplex</strong><strong>Full</strong>-<strong>Duplex</strong>Packet Reception Ratio1.000.750.500.2500 10 20 30 40Loss atHigh SNRSNR (dB)• Loss at High SNR: Due to spurious signal peaks in USRP48
Link Reception RatioHalf-<strong>Duplex</strong><strong>Full</strong>-<strong>Duplex</strong>Packet Reception Ratio1.000.750.500.250Loss atLow SNR0 10 20 30 40SNR (dB)• Loss at High SNR: Due to spurious signal peaks in USRP• Loss at low SNR: Due to imprecisions in prototype49
Talk Outline• Design of <strong>Full</strong>-<strong>Duplex</strong> <strong>Wireless</strong>• 3 Techniques: Antenna, Hardware andDigital Cancellation• Analyzing Antenna Cancellation• Performance Results• Implications to <strong>Wireless</strong> Networks• Limitations of Design, Future Work50
The prototype gives 1.84x throughput gainwith two radios compared to half-duplexwith a single radioSo what? PHY gains similar to 2x2 MIMO51
The prototype gives 1.84x throughput gainwith two radios compared to half-duplexwith a single radioSo what? PHY gains similar to 2x2 MIMOTrue benefit lies beyond the physical layer52
Implications to <strong>Wireless</strong> Networks• Breaks a basic assumption in wireless• Can solve some fundamental problems withwireless networks today• Hidden terminals• Primary detection in whitespaces• Network congestion and WLAN fairness• Excessive latency in multihop wireless53
Mitigating Hidden TerminalsCurrent networks havehidden terminalsN1APN2• CSMA/CA can’t solve this• Schemes like RTS/CTS introduce significant overhead54
Mitigating Hidden TerminalsCurrent networks havehidden terminalsN1APN2• CSMA/CA can’t solve this• Schemes like RTS/CTS introduce significant overhead<strong>Full</strong> <strong>Duplex</strong> solveshidden terminalsN1APN2Since both sides transmit at the same time, nohidden terminals exist55
Primary Detection in WhitespacesPrimary sensingPrimary TX(<strong>Wireless</strong> Mics)Secondary TX(Whitespace AP)Secondary transmitters should sense for primarytransmissions before channel useInterferencePrimary TX(<strong>Wireless</strong> Mics)Secondary TX(Whitespace AP)Traditional nodes may still interfere during transmissions56
Primary Detection in WhitespacesPrimary sensingPrimary TX(<strong>Wireless</strong> Mics)Secondary TX(Whitespace AP)Secondary transmitters should sense for primarytransmissions before channel usePrimary sensingPrimary TX(<strong>Wireless</strong> Mics)Secondary TX(Whitespace AP)<strong>Full</strong>-duplex nodes can sense and send at the same time57
Network Congestion and WLAN FairnessWithout full-duplex:• 1/n bandwidth for each node in network, including APDownlink Throughput = 1/n Uplink Throughput = (n-1)/n58
Network Congestion and WLAN FairnessWithout full-duplex:• 1/n bandwidth for each node in network, including APDownlink Throughput = 1/n Uplink Throughput = (n-1)/nWith full-duplex:• AP sends and receives at the same timeDownlink Throughput = 1 Uplink Throughput = 159
Reducing Round-Trip TimesLong delivery and round-trip times in multihopnetworksSolution: Wormhole routingN1 N2 N3 N4N1N2N3N4N1N2N3N4TimeHalf-duplexTime<strong>Full</strong>-duplex60
Talk Outline• Design of <strong>Full</strong>-<strong>Duplex</strong> <strong>Wireless</strong>• 3 Techniques: Antenna, Hardware andDigital Cancellation• Analyzing Antenna Cancellation• Performance Results• Implications to <strong>Wireless</strong> Networks• Limitations of Design, Future Work61
• Bandwidth ConstraintWorking on a frequency independent signal inversiontechnique• Time-varying wireless channelAuto-tuning of the hardware cancellation circuit• Multi-pathEstimate and incorporate in digital cancellation: Someexisting work does this• <strong>Single</strong> streamExtension to MIMO-like systems62
Summary• Prototype for achieving in-band full-duplex wireless• Constraints of current prototype can be overcomewith some neat ideas and careful engineering• Rethinking of wireless networks• We’ve discussed some applications like mitigatinghidden terminals and WLAN fairness• Many more possibilities63
From 3 antennas to 2 antennas! solves bandwidth problemDemo on Wednesday64
Summary• Prototype for achieving in-band full-duplex wireless• Constraints of current prototype can be overcomewith some neat ideas and careful engineering• Rethinking of wireless networks• We’ve discussed some applications like mitigatinghidden terminals and WLAN fairness• Many more possibilities65
Summary• Prototype for achieving in-band full-duplex wireless• Constraints of current prototype can be overcomewith some neat ideas and careful engineering• Rethinking of wireless networks• We’ve discussed some applications like mitigatinghidden terminals and WLAN fairness• Many more possibilitiesPS: We’re looking for jobs starting mid-2011 :)Kannan: AcademicMayank and Jung IL: Industrial Research66