Full Duplex MIMO Radios

4/25/2014 1

Dinesh Bharadia and Sachin KattiStanford University

4/25/2014 2

Refresher: Why was full duplex considered

impossible?

4/25/2014 2

TX

RX

RX

TX

Radio 1 Radio 2

Refresher: Why was full duplex considered

impossible?

4/25/2014 2

TX

RX

RX

TX

Radio 1 Radio 2

Refresher: Why was full duplex considered

impossible?

4/25/2014 2

TX

RX

RX

TX

Radio 1 Radio 2

Refresher: Why was full duplex considered

impossible?

4/25/2014 2

TX

RX

RX

TX

Radio 1 Radio 2

Self-Interference is a hundred billion times

(110dB+) stronger than the received signal

Refresher: Why was full duplex considered

impossible?

State of the Art: SISO Full duplex (Sigcomm’13)

Demonstrated that practical in-

band SISO (single antenna) full

duplex radios are possible• Self-Interference cancellation that

eliminates everything to the noise floor

• Practically achieves close to expected

theoretical 2x throughput increase

4/25/2014 3

RX RF

Frontend

TX RF

Frontend

T

+HT

+δT

Circulator

ΣAnalog Cancellation Eliminates all Tx noise and

Protects ADC from saturating

Digital Cancellation Eliminates all Linear and Non-

Linear Distortion (Harmonics)

Σ

Why care about full duplex?

Full Duplex provides a step jump in Spectral

Efficiency

4/25/2014 4

Why care about full duplex?

Full Duplex provides a step jump in Spectral

Efficiency

4/25/2014 4

LTE 2X2

1.42

Spectral Efficiency

Why care about full duplex?

Full Duplex provides a step jump in Spectral

Efficiency

4/25/2014 4

• LTE 2X2 Spectral Efficiency is 1.42 bits/sec/Hz

• Spectra Efficiency Doubled over last decade

LTE 2X2

1.42

Spectral Efficiency

Why care about full duplex?

Full Duplex provides a step jump in Spectral

Efficiency

4/25/2014 4

• LTE 2X2 Spectral Efficiency is 1.42 bits/sec/Hz

• Spectra Efficiency Doubled over last decade

LTE 2X2

1.42

Spectral Efficiency

Why care about full duplex?

Full Duplex provides a step jump in Spectral

Efficiency

4/25/2014 4

• LTE 2X2 Spectral Efficiency is 1.42 bits/sec/Hz

• Spectra Efficiency Doubled over last decade

• Full Duplex provides a step jump

LTE 2X2

1.42

Spectral Efficiency

Why care about full duplex?

Full Duplex provides a step jump in Spectral

Efficiency

4/25/2014 4

Spectral efficiency gains over last decade mostly from MIMO

For full duplex to be viable, need to work with MIMO

• LTE 2X2 Spectral Efficiency is 1.42 bits/sec/Hz

• Spectra Efficiency Doubled over last decade

• Full Duplex provides a step jump

LTE 2X2

1.42

Spectral Efficiency

This Work

4/25/2014

This Work

Designed & implemented a near-ideal MIMO in-band full

duplex radio

4/25/2014

This Work

Designed & implemented a near-ideal MIMO in-band full

duplex radio• Hybrid (analog & digital) cross and self-talk cancellation circuits &

algorithms that completely eliminate interference to the noise floor

4/25/2014

This Work

Designed & implemented a near-ideal MIMO in-band full

duplex radio• Hybrid (analog & digital) cross and self-talk cancellation circuits &

algorithms that completely eliminate interference to the noise floor

• Low complexity cancellation design

4/25/2014

This Work

Designed & implemented a near-ideal MIMO in-band full

duplex radio• Hybrid (analog & digital) cross and self-talk cancellation circuits &

algorithms that completely eliminate interference to the noise floor

• Low complexity cancellation design

4/25/2014

Prototype 3x3 MIMO full duplex radios using off-the-shelf

WiFi radios

This Work

Designed & implemented a near-ideal MIMO in-band full

duplex radio• Hybrid (analog & digital) cross and self-talk cancellation circuits &

algorithms that completely eliminate interference to the noise floor

• Low complexity cancellation design

4/25/2014

Prototype 3x3 MIMO full duplex radios using off-the-shelf

WiFi radios• Experimental indoor evaluation which demonstrates that our design

practically achieves close to the 2x theoretically expected throughput gain

Why can not we use the SISO design to

implement MIMO full duplex?

4/25/2014 6

State of the Art for SISO full duplex

4/25/2014 7

State of the Art for SISO full duplex

4/25/2014 7

Conceptual model

T

Circulator

Σ

TX RX

State of the Art for SISO full duplex

4/25/2014 7

Conceptual model

T

Circulator

Σ

TX RX

Reflector

State of the Art for SISO full duplex

4/25/2014 7

H Channel

representing environmental reflections

Conceptual model

T

Circulator

Σ

TX RX

Reflector

State of the Art for SISO full duplex

4/25/2014 7

H Channel

representing environmental reflections

Conceptual model

T

Circulator

Σ

TX RX

Reflector

R+HT

State of the Art for SISO full duplex

4/25/2014 7

Cancellation filter

H Channel

representing environmental reflections

H ≈ H

Conceptual model

R + ε

T

Circulator

Σ

TX RX

Reflector

R+HT

T ^

State of the Art for SISO full duplex

4/25/2014 7

Cancellation filter

H Channel

representing environmental reflections

H ≈ H

Conceptual model

R + ε

T

Circulator

Σ

TX RX

Reflector

R+HT

T ^

State of the Art for SISO full duplex

4/25/2014 7

Cancellation filter

H Channel

representing environmental reflections

H ≈ H

Conceptual model

R + ε

T

Circulator

Σ

TX RX

RX RF

Frontend

TX RF

Frontend

Σ

T

Σ

Circulator

TX RXPractical Realization of Cancellation filter

R

Reflector

R+HT

Reflector

H

R+HT

T ^

State of the Art for SISO full duplex

4/25/2014 7

+δT

Cancellation filter

fixed delays

RF Cancellation Circuit

tuningalgorithm

Σ

aN

a1d1

dN

H Channel

representing environmental reflections

H ≈ H

Conceptual model

R + ε

T

Circulator

Σ

TX RX

RX RF

Frontend

TX RF

Frontend

Σ

T

Σ

Circulator

TX RXPractical Realization of Cancellation filter

R

Reflector

R+HT

Reflector

H

R+HT

T

T ^

State of the Art for SISO full duplex

4/25/2014 7

Digital

Cancellation Eliminates all Linear and

Non-Linear Distortion

+δT

Cancellation filter

fixed delays

RF Cancellation Circuit

tuningalgorithm

Σ

aN

a1d1

dN

H Channel

representing environmental reflections

H ≈ H

Conceptual model

R + ε

T

Circulator

Σ

TX RX

RX RF

Frontend

TX RF

Frontend

Σ

T

Σ

Circulator

TX RXPractical Realization of Cancellation filter

R

Reflector

R+HT

Reflector

H

R + ε

R+HT

T

T ^

State of the Art for SISO full duplex

4/25/2014 7

Digital

Cancellation Eliminates all Linear and

Non-Linear Distortion

+δT

Cancellation filter

fixed delays

RF Cancellation Circuit

tuningalgorithm

Σ

aN

a1d1

dN

H Channel

representing environmental reflections

H ≈ H

Conceptual model

R + ε

T

Circulator

Σ

TX RX

RX RF

Frontend

TX RF

Frontend

Σ

T

Σ

Circulator

TX RXPractical Realization of Cancellation filter

R

Reflector

R+HT

Reflector

H

R + ε

R+HT

T

T ^

Full duplex designs are conceptually realizing an

adaptive filter that closely matches the environmental

reflection response

Key Metrics for SISO Full Duplex Design

4/25/2014 8

Key Metrics for SISO Full Duplex Design

Cancellation filters are characterized by two key metrics

4/25/2014 8

Key Metrics for SISO Full Duplex Design

Cancellation filters are characterized by two key metrics

• Complexity: The number of filter taps needed to closely match the

environmental reflection response.

– Higher number of taps More analog circuit area, more FPGA gates

– Less is better

4/25/2014 8

Key Metrics for SISO Full Duplex Design

Cancellation filters are characterized by two key metrics

• Complexity: The number of filter taps needed to closely match the

environmental reflection response.

– Higher number of taps More analog circuit area, more FPGA gates

– Less is better

• Interference residue: Amount of interference left uncancelled,

depends on the accuracy of the filter in modeling all reflections

lower is better

4/25/2014 8

Key Metrics for SISO Full Duplex Design

Cancellation filters are characterized by two key metrics

• Complexity: The number of filter taps needed to closely match the

environmental reflection response.

– Higher number of taps More analog circuit area, more FPGA gates

– Less is better

• Interference residue: Amount of interference left uncancelled,

depends on the accuracy of the filter in modeling all reflections

lower is better

4/25/2014 8

SISO Full Duplex Implications

Analog Cancellation taps 12 6x6 inches board

Digital Cancellation taps 132 295 DSP 48 Logic

Interference Residue 1dB over noise floor Almost optimal full duplex

Key Metrics for SISO Full Duplex Design

Cancellation filters are characterized by two key metrics

• Complexity: The number of filter taps needed to closely match the

environmental reflection response.

– Higher number of taps More analog circuit area, more FPGA gates

– Less is better

• Interference residue: Amount of interference left uncancelled,

depends on the accuracy of the filter in modeling all reflections

lower is better

4/25/2014 8

SISO Full Duplex Implications

Analog Cancellation taps 12 6x6 inches board

Digital Cancellation taps 132 295 DSP 48 Logic

Interference Residue 1dB over noise floor Almost optimal full duplex

Goal for any full duplex design: Minimize interference

residue to the noise floor with the lowest complexity

cancellation filter

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H11

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H11

H12

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H11

H12H13

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H11

H12H13

H11H12

H13

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H11H12

H13

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H11H12

H13

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H21

H22

H23

H11H12

H13

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H21

H22

H23

H11H12

H13,H21,H22

,H23

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H11H12

H13,H21,H22

,H23

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H11H12

H13,H21,H22

,H23

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H31

H32H33

H11H12

H13,H21,H22

,H23

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H31

H32H33

H11H12

H13,H21,H31,H22,H32

,H23 ,H33

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H11H12

H13,H21,H31,H22,H32

,H23 ,H33

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H12

H13,H21,H31,H22,H32

,H23 ,H33

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H11

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H12

H13,H21,H31,H22,H32

,H23 ,H33

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H11

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H12

H11H12

H13,H21,H31,H22,H32

,H23 ,H33

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H11

H13

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H12

H11H12

H13,H21,H31,H22,H32

,H23 ,H33

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H11

H13

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H12 H22

H23

H21

H11H12

H13,H21,H31,H22,H32

,H23 ,H33

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H11

H13

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H12 H22

H23

H21

H33

H32

H31

H11H12

H13,H21,H31,H22,H32

,H23 ,H33

How is the MIMO full

duplex problem

different?

T

Circulator

TX1 RX1

T

Circulator

TX2 RX2

T

Circulator

TX3 RX3

Reflector

H11

H13

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H12 H22

H23

H21

H33

H32

H31

H11H12

H13,H21,H31,H22,H32

,H23 ,H33

MIMO full duplex has quadratically more number of

signals to cancel because of the presence of cross

talk.

Why not replicate the SISO full duplex design to

cancel all the self-talk and cross-talk components

for MIMO?

4/25/2014 10

Naïve Solution: Replicates SISO design

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

Naïve Solution: Replicates SISO design

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11

Naïve Solution: Replicates SISO design

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11

Naïve Solution: Replicates SISO design

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11

H11^

Naïve Solution: Replicates SISO design

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H11^

Naïve Solution: Replicates SISO design

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H11^ H22

^

Naïve Solution: Replicates SISO design

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H33

H11^ H22

^

Naïve Solution: Replicates SISO design

CancellationFilter

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H33

H11^ H22

^ H33^

Naïve Solution: Replicates SISO design

CancellationFilter

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H11^ H22

^ H33^

Naïve Solution: Replicates SISO design

CancellationFilter

H21

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H11^ H22

^ H33^

Naïve Solution: Replicates SISO design

CancellationFilter

Cancellationfilter

H21

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H11^

H21^

H22^ H33

^

Naïve Solution: Replicates SISO design

CancellationFilter

Cancellationfilter

H21

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H31

H11^

H21^

H22^ H33

^

Naïve Solution: Replicates SISO design

CancellationFilter

Cancellationfilter

H21

H31

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H31

H11^

H21^

H22^ H33

^

Naïve Solution: Replicates SISO design

CancellationFilter

Cancellationfilter

H21

H31

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

Cancellationfilter

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H31

H11^

H21^

H31^

H22^ H33

^

Naïve Solution: Replicates SISO design

CancellationFilter

Cancellationfilter

H21

H31

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

Cancellationfilter

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H32

H31

H11^

H21^

H31^

H22^ H33

^

Naïve Solution: Replicates SISO design

CancellationFilter

Cancellationfilter

H21

H31

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

Cancellationfilter

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H32

H31

H32

H11^

H21^

H31^

H22^ H33

^

Naïve Solution: Replicates SISO design

CancellationFilter

CancellationFilter

Cancellationfilter

H21

H31

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

Cancellationfilter

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H32

H31

H32

H11^

H21^

H31^

H22^

H32^

H33^

Naïve Solution: Replicates SISO design

CancellationFilter

CancellationFilter

Cancellationfilter

H21

H31

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

Cancellationfilter

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H32

H31

H32

H11^

H21^

H31^

H22^

H32^

H33^

H12

Naïve Solution: Replicates SISO design

CancellationFilter

CancellationFilter

Cancellationfilter

H21

H31

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

Cancellationfilter

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H32

H31

H32

H11^

H21^

H31^

H22^

H32^

H33^

H12

H12

Naïve Solution: Replicates SISO design

CancellationFilter

CancellationFilter

Cancellationfilter

H21

H31

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

Cancellationfilter

CancellationFilter

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H32

H31

H32

H11^

H21^

H31^

H22^

H32^

H12^

H33^

H12

H12

Naïve Solution: Replicates SISO design

CancellationFilter

CancellationFilter

Cancellationfilter

H21

H31

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

Cancellationfilter

CancellationFilter

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H32

H31

H32

H11^

H21^

H31^

H22^

H32^

H12^

H33^

H13

H12

H12

Naïve Solution: Replicates SISO design

CancellationFilter

CancellationFilter

Cancellationfilter

H21

H31

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

Cancellationfilter

CancellationFilter

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H32

H31

H32

H11^

H21^

H31^

H22^

H32^

H12^

H33^

H13

H13

H12

H12

Naïve Solution: Replicates SISO design

CancellationFilter

CancellationFilter

Cancellationfilter

H21

H31

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

Cancellationfilter

CancellationFilter

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

Cancellationfilter

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H32

H31

H32

H11^

H21^

H31^

H22^

H32^

H12^

H33^

H13^

H13

H13

H12

H12

Naïve Solution: Replicates SISO design

CancellationFilter

CancellationFilter

Cancellationfilter

H21

H31

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

Cancellationfilter

CancellationFilter

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

Cancellationfilter

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H32

H31

H32

H11^

H21^

H31^

H22^

H32^

H12^

H33^

H13^

H13

H13

H12H23

H12

Naïve Solution: Replicates SISO design

CancellationFilter

CancellationFilter

Cancellationfilter

H21

H31

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

Cancellationfilter

CancellationFilter

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

Cancellationfilter

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H32

H31

H32

H11^

H21^

H31^

H22^

H32^

H12^

H33^

H13^

H23

H13

H13

H12H23

H12

Naïve Solution: Replicates SISO design

CancellationFilter

CancellationFilter

Cancellationfilter

Cancellationfilter

H21

H31

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

Cancellationfilter

CancellationFilter

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

Cancellationfilter

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H32

H31

H32

H11^

H21^

H31^

H22^

H32^

H12^

H33^

H23^

H13^

H23

H13

H13

H12H23

H12

Naïve Solution: Replicates SISO design

CancellationFilter

CancellationFilter

Cancellationfilter

Cancellationfilter

Total 9N Taps, for M=3

H21

H31

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

Cancellationfilter

CancellationFilter

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

Cancellationfilter

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H32

H31

H32

H11^

H21^

H31^

H22^

H32^

H12^

H33^

H23^

H13^

H23

H13

H13

H12H23

H12

Naïve Solution: Replicates SISO design

CancellationFilter

CancellationFilter

Cancellationfilter

Cancellationfilter

Total 9N Taps, for M=3

H21

H31

Cancellationfilter

T

Σ

Circulator

TX1

RX1

H11

Cancellationfilter

CancellationFilter

CancellationFilter

T

Σ

Circulator

TX2

RX2

H22

T

Σ

Circulator

TX3

RX3

H33

Cancellationfilter

A1 A2

3 Antenna Full Duplex MIMO Radio

A3

H11H22

H21

H33

H32

H31

H32

Cancellation filter complexity is quadratic with number of

antennas (M)

Even worse cancellation residue at each receiver

increases linearly with number of antennas (e.g. 3x3

5dB residue)H11^

H21^

H31^

H22^

H32^

H12^

H33^

H23^

H13^

H23

H13

H13

H12H23

H12

0 0.5 1 1.5 20

0.2

0.4

0.6

0.8

1

4/25/2014 12

Why does the increase in cancellation residue

matter?

Throughput performance of SISO replication design

Our Design

SISO Replication

SISO Low Complexity Replication

Gain relative to Half Duplex 3X3 MIMO

CD

F

Median throughput gain is not close to 2x because of large interference residue

0 0.5 1 1.5 20

0.2

0.4

0.6

0.8

1

4/25/2014 12

1.35x

Why does the increase in cancellation residue

matter?

Throughput performance of SISO replication design

Our Design

SISO Replication

SISO Low Complexity Replication

Gain relative to Half Duplex 3X3 MIMO

CD

F

Worse than standard Half Duplex

Median throughput gain is not close to 2x because of large interference residue

0 0.5 1 1.5 20

0.2

0.4

0.6

0.8

1

4/25/2014 12

1.35x

Why does the increase in cancellation residue

matter?

Throughput performance of SISO replication design

Our Design

SISO Replication

SISO Low Complexity Replication

Gain relative to Half Duplex 3X3 MIMO

CD

F

Worse than standard Half Duplex

Median throughput gain is not close to 2x because of large interference residue

Ideal Solution:

• Cancellation filter complexity scales linearly with

number of antennas (M)

• Cancellation Residue same as SISO and should

not be impacted by number of MIMO antennas

Technical Contributions

Designed & implemented a near-ideal MIMO in-band full

duplex radio• Hybrid (analog & digital) cross and self-talk cancellation circuits &

algorithms that completely eliminate interference to the noise floor

4/25/2014

Prototype 3x3 MIMO full duplex radios using off-the-shelf

WiFi radios• Experimental indoor evaluation which demonstrates that our design

practically achieves close to the 2x theoretically expected throughput gain

Technical Contributions

Designed & implemented a near-ideal MIMO in-band full

duplex radio• Hybrid (analog & digital) cross and self-talk cancellation circuits &

algorithms that completely eliminate interference to the noise floor

• Cancellation complexity scales linearly with M (number of antennas).

4/25/2014

Prototype 3x3 MIMO full duplex radios using off-the-shelf

WiFi radios• Experimental indoor evaluation which demonstrates that our design

practically achieves close to the 2x theoretically expected throughput gain

Technical Contributions

Designed & implemented a near-ideal MIMO in-band full

duplex radio• Hybrid (analog & digital) cross and self-talk cancellation circuits &

algorithms that completely eliminate interference to the noise floor

• Cancellation complexity scales linearly with M (number of antennas).

• Cancellation residue is the same as the SISO full duplex, i.e. it does not

degrade with increasing number of antennas.

4/25/2014

Prototype 3x3 MIMO full duplex radios using off-the-shelf

WiFi radios• Experimental indoor evaluation which demonstrates that our design

practically achieves close to the 2x theoretically expected throughput gain

T

Σ

Circulator

TX1 RX1

H11

Cancellation filter

Key Idea: Reducing Complexity

H11^

T

Σ

Circulator

TX1 RX1

H11

Cancellation filter

Circulator

TX2 RX2

Key Idea: Reducing Complexity

H11^

T

Σ

Circulator

TX1 RX1

H21H11

Cancellation filter

Circulator

TX2 RX2

Key Idea: Reducing Complexity

H11^

Can we reuse the self talk cancellation filter to also

cancel the cross talk ?

T

Σ

Circulator

TX1 RX1

H21H11

Cancellation filter

Circulator

TX2 RX2

Key Idea: Reducing Complexity

H11^

4/25/2014 15

T

Σ

Circulator

TX1 RX1

Cancellation filter

Circulator

TX2 RX2

Why can self talk cancellation filter be used to

partly model cross talk?

𝐻11 𝑓

𝐻11 𝑓𝐻21 𝑓

4/25/2014 15

T

Σ

Circulator

TX1 RX1

Cancellation filter

Circulator

TX2 RX2

Why can self talk cancellation filter be used to

partly model cross talk?

𝐻11 𝑓

𝐻11 𝑓𝐻21 𝑓

• Share Environment

• Share Reflectors

4/25/2014 15

T

Σ

Circulator

TX1 RX1

Cancellation filter

Circulator

TX2 RX2

Why can self talk cancellation filter be used to

partly model cross talk?

𝐻11 𝑓

𝐻11 𝑓𝐻21 𝑓

• Share Environment

• Share Reflectors

• Self talk and cross talk

transfer function are

related as they undergo

similar environment.

4/25/2014 15

T

Σ

Circulator

TX1 RX1

Cancellation filter

Circulator

TX2 RX2

Why can self talk cancellation filter be used to

partly model cross talk?

𝐻11 𝑓

𝐻11 𝑓𝐻21 𝑓

• Share Environment

• Share Reflectors

• Self talk and cross talk

transfer function are

related as they undergo

similar environment.

• A lot of work for modeling

self-talk is already done,

we need to create only

the difference in self talk

and cross talk.

4/25/2014 15

T

Σ

Circulator

TX1 RX1

Cancellation filter

Circulator

TX2 RX2

Why can self talk cancellation filter be used to

partly model cross talk?

𝐻11 𝑓

𝐻11 𝑓𝐻21 𝑓

• Share Environment

• Share Reflectors

• Self talk and cross talk

transfer function are

related as they undergo

similar environment.

• A lot of work for modeling

self-talk is already done,

we need to create only

the difference in self talk

and cross talk. 𝐻21 𝑓 = 𝐻𝑐𝑎𝑠𝑐𝑎𝑑𝑒 𝑓 .𝐻11 𝑓

4/25/2014 15

T

Σ

Circulator

TX1 RX1

Cancellation filter

Circulator

TX2 RX2

Why can self talk cancellation filter be used to

partly model cross talk?

𝐻11 𝑓

𝐻11 𝑓𝐻21 𝑓

• Share Environment

• Share Reflectors

• Self talk and cross talk

transfer function are

related as they undergo

similar environment.

• A lot of work for modeling

self-talk is already done,

we need to create only

the difference in self talk

and cross talk. 𝐻21 𝑓 = 𝐻𝑐𝑎𝑠𝑐𝑎𝑑𝑒 𝑓 .𝐻11 𝑓

Can we leverage this relationship to reduce the

cancellation complexity

T

Σ

Circulator

TX1 RX1

Cancellationfilter

Circulator

TX2 RX2

Empirical Observation 𝐻11 𝑓 𝐻21 𝑓

𝐻21 𝑓 = 𝐻𝑐𝑎𝑠𝑐𝑎𝑑𝑒 𝑓 . 𝐻11 𝑓

Cascade Transfer Function 𝐻11 𝑓

• Cascade Transfer

Function

• Collect cross talk and self talk for various indoor

environments

T

Σ

Circulator

TX1 RX1

Cancellationfilter

Circulator

TX2 RX2

Empirical Observation 𝐻11 𝑓 𝐻21 𝑓

𝐻21 𝑓 = 𝐻𝑐𝑎𝑠𝑐𝑎𝑑𝑒 𝑓 . 𝐻11 𝑓

Cascade Transfer Function 𝐻11 𝑓

• Cascade Transfer

Function

• Learning

Algorithm

• Complexity

Reduction

• Collect cross talk and self talk for various indoor

environments

• From all the possible cascade response, calculate

via optimization the best low complexity circuit

which achieves the cascade transfer function

(offline analysis)

• These cascade circuits are very low complexity,

thus allowing us to get close to linear complexity

T

Σ

Circulator

TX1 RX1

Cancellationfilter

Circulator

TX2 RX2

Empirical Observation 𝐻11 𝑓 𝐻21 𝑓

𝐻21 𝑓 = 𝐻𝑐𝑎𝑠𝑐𝑎𝑑𝑒 𝑓 . 𝐻11 𝑓

Cascade Transfer Function 𝐻11 𝑓

Reducing Complexity: Cascaded Cancellation

T

Σ

Circulator

TX1

RX1

Σ

H21

H31

H11

Cancellation Filter N taps

Circulator

TX2

RX2

T

Circulator

TX3

RX3

𝐻11

Reducing Complexity: Cascaded Cancellation

T

Σ

Circulator

TX1

RX1

Σ

filter C taps

Σ

H21

H31

H11

Cancellation Filter N taps

Circulator

TX2

RX2

T

Circulator

TX3

RX3

𝐻11

𝐻𝑐21

Reducing Complexity: Cascaded Cancellation

T

Σ

Circulator

TX1

RX1

Σ

filter C taps

Σ

filter D taps

H21

H31

H11

Cancellation Filter N taps

Circulator

TX2

RX2

T

Circulator

TX3

RX3

𝐻11

𝐻𝑐21 𝐻𝑐31

Reducing Complexity: Cascaded Cancellation

T

Σ

Circulator

TX1

RX1

Σ

filter C taps

Σ

Cascaded Filters

filter D taps

H21

H31

H11

Cancellation Filter N taps

Circulator

TX2

RX2

T

Circulator

TX3

RX3

𝐻11

𝐻𝑐21 𝐻𝑐31

Reducing Complexity: Cascaded Cancellation

Taps: N >> C > D

T

Σ

Circulator

TX1

RX1

Σ

filter C taps

Σ

Cascaded Filters

filter D taps

H21

H31

H11

Cancellation Filter N taps

Circulator

TX2

RX2

T

Circulator

TX3

RX3

𝐻11

𝐻𝑐21 𝐻𝑐31

Reducing Complexity: Cascaded Cancellation

Taps: N >> C > D

T

Σ

Circulator

TX1

RX1

Σ

filter C taps

Σ

Cascaded Filters

filter D taps

H21

H31

H11

Cancellation Filter N taps

Circulator

TX2

RX2

T

Circulator

TX3

RX3

𝐻11

𝐻𝑐21

Total Taps: N+ C + D, for Chain 1

General Complexity per chain: ~N << M.N

𝐻𝑐31

Taps: N >> C > D

T

Σ

Circulator

TX1

RX1

Σ

T

Σ

Circulator

TX2

RX2

Σ

T

Σ

Circulator

TX3

RX3

Σ

H11

Cancellation Filter

Cancellation Filter

Cancellation Filter

Complete Cascaded Design:

𝐻11 𝐻33 𝐻22

Taps: N >> C > D

T

Σ

Circulator

TX1

RX1

Σ

T

Σ

Circulator

TX2

RX2

Σ

T

Σ

Circulator

TX3

RX3

Σ

H21H11

Cancellation Filter

Cancellation Filter

Cancellation Filter

Complete Cascaded Design:

𝐻11 𝐻33 𝐻22

Taps: N >> C > D

T

Σ

Circulator

TX1

RX1

Σ

T

Σ

Circulator

TX2

RX2

Σ

T

Σ

Circulator

TX3

RX3

Σ

Σ

C taps

H21H11

Cancellation Filter

Cancellation Filter

Cancellation Filter

Complete Cascaded Design:

𝐻11 𝐻33 𝐻22

𝐻𝑐21

Taps: N >> C > D

T

Σ

Circulator

TX1

RX1

Σ

T

Σ

Circulator

TX2

RX2

Σ

T

Σ

Circulator

TX3

RX3

Σ

Σ

C taps

H21

H31

H11

Cancellation Filter

Cancellation Filter

Cancellation Filter

Complete Cascaded Design:

𝐻11 𝐻33 𝐻22

𝐻𝑐21

Taps: N >> C > D

T

Σ

Circulator

TX1

RX1

Σ

T

Σ

Circulator

TX2

RX2

Σ

T

Σ

Circulator

TX3

RX3

Σ

Σ

C taps D taps

H21

H31

H11

Cancellation Filter

Cancellation Filter

Cancellation Filter

Complete Cascaded Design:

𝐻11 𝐻33 𝐻22

𝐻𝑐21 𝐻𝑐31

Taps: N >> C > D

T

Σ

Circulator

TX1

RX1

Σ

T

Σ

Circulator

TX2

RX2

Σ

T

Σ

Circulator

TX3

RX3

Σ

Σ

C taps D taps

H21

H31

H11

Cancellation Filter

Cancellation Filter

Cancellation Filter

Complete Cascaded Design:

Cascaded Filter

𝐻11 𝐻33 𝐻22

𝐻𝑐21 𝐻𝑐31

Taps: N >> C > D

T

Σ

Circulator

TX1

RX1

Σ

T

Σ

Circulator

TX2

RX2

Σ

T

Σ

Circulator

TX3

RX3

Σ

Σ

C taps D taps

H21

H31

H11

Cancellation Filter

Cancellation Filter

Cancellation Filter

Complete Cascaded Design:

Cascaded Filter

𝐻11 𝐻33 𝐻22

𝐻𝑐21 𝐻𝑐31

Taps: N >> C > D

T

Σ

Circulator

TX1

RX1

Σ

T

Σ

Circulator

TX2

RX2

Σ

T

Σ

Circulator

TX3

RX3

Σ

Σ

C taps D taps

H21

H31

H11

Cancellation Filter

Cancellation Filter

Cancellation Filter

C taps

Σ

Complete Cascaded Design:

Cascaded Filter

𝐻11 𝐻33 𝐻22

𝐻𝑐21 𝐻𝑐31

𝐻𝑐12

Taps: N >> C > D

T

Σ

Circulator

TX1

RX1

Σ

T

Σ

Circulator

TX2

RX2

Σ

T

Σ

Circulator

TX3

RX3

Σ

C taps

Σ

C taps D taps

H21

H31

H11

Cancellation Filter

Cancellation Filter

Cancellation Filter

C taps

Σ

Complete Cascaded Design:

Cascaded Filter

𝐻11 𝐻33 𝐻22

𝐻𝑐21 𝐻𝑐31

𝐻𝑐12 𝐻𝑐32

Taps: N >> C > D

T

Σ

Circulator

TX1

RX1

Σ

T

Σ

Circulator

TX2

RX2

Σ

T

Σ

Circulator

TX3

RX3

Σ

C taps

Σ

C taps D taps

H21

H31

H11

Cancellation Filter

Cancellation Filter

Cancellation Filter

C taps

Σ

Complete Cascaded Design:

Cascaded Filter

𝐻11 𝐻33 𝐻22

𝐻𝑐21 𝐻𝑐31

𝐻𝑐12 𝐻𝑐32

Taps: N >> C > D

T

Σ

Circulator

TX1

RX1

Σ

T

Σ

Circulator

TX2

RX2

Σ

T

Σ

Circulator

TX3

RX3

Σ

C taps

Σ

C taps D taps

H21

H31

H11

Cancellation Filter

Cancellation Filter

Cancellation Filter

C taps

Σ

C taps

Σ

Complete Cascaded Design:

Cascaded Filter

𝐻11 𝐻33 𝐻22

𝐻𝑐21 𝐻𝑐31

𝐻𝑐12 𝐻𝑐32 𝐻𝑐32

Taps: N >> C > D

T

Σ

Circulator

TX1

RX1

Σ

T

Σ

Circulator

TX2

RX2

Σ

T

Σ

Circulator

TX3

RX3

Σ

C taps

Σ

C taps D taps

H21

H31

H11

Cancellation Filter

Cancellation Filter

Cancellation Filter

C taps

Σ

D tapsC taps

Σ

Complete Cascaded Design:

Cascaded Filter

𝐻11 𝐻33 𝐻22

𝐻𝑐21 𝐻𝑐31

𝐻𝑐12 𝐻𝑐32 𝐻𝑐32

𝐻𝑐31

Taps: N >> C > D

T

Σ

Circulator

TX1

RX1

Σ

T

Σ

Circulator

TX2

RX2

Σ

T

Σ

Circulator

TX3

RX3

Σ

C taps

Σ

C taps D taps

H21

H31

H11

Cancellation Filter

Cancellation Filter

Cancellation Filter

C taps

Σ

D tapsC taps

Σ

Complete Cascaded Design:

Cascaded Filter

𝐻11 𝐻33 𝐻22

𝐻𝑐21 𝐻𝑐31

𝐻𝑐12 𝐻𝑐32 𝐻𝑐32

𝐻𝑐31

Taps: N >> C > D

T

Σ

Circulator

TX1

RX1

Σ

T

Σ

Circulator

TX2

RX2

Σ

T

Σ

Circulator

TX3

RX3

Σ

C taps

Σ

C taps D taps

H21

H31

H11

Cancellation Filter

Cancellation Filter

Cancellation Filter

C taps

Σ

D tapsC taps

Σ

Complete Cascaded Design:

Cascaded Filter

𝐻11 𝐻33 𝐻22

𝐻𝑐21 𝐻𝑐31

𝐻𝑐12 𝐻𝑐32 𝐻𝑐32

𝐻𝑐31

Taps: N >> C > D

T

Σ

Circulator

TX1

RX1

Σ

T

Σ

Circulator

TX2

RX2

Σ

T

Σ

Circulator

TX3

RX3

Σ

C taps

Σ

C taps D taps

H21

H31

H11

Cancellation Filter

Cancellation Filter

Cancellation Filter

C taps

Σ

D tapsC taps

Σ

Complete Cascaded Design:

Cascaded Filter

Total Taps: 3N+ 4C + 2D, for M=3

General Complexity: ~ M.N << M2.N

𝐻11 𝐻33 𝐻22

𝐻𝑐21 𝐻𝑐31

𝐻𝑐12 𝐻𝑐32 𝐻𝑐32

𝐻𝑐31

Interference Residue with Cascaded Design shows no

degradation

TX1

RX1

Analog Isolation

TX2

RX2

Independent Training with SISO Replication Design

A1 A2Analog

Isolation

Txb3

RX3

A3Analog

Isolation

H11

Interference Residue with Cascaded Design shows no

degradation

TX1

RX1

Analog Isolation

TX2

RX2

Independent Training with SISO Replication Design

A1 A2Analog

Isolation

Txb3

RX3

A3Analog

Isolation

H21

H11

Interference Residue with Cascaded Design shows no

degradation

TX1

RX1

Analog Isolation

TX2

RX2

Independent Training with SISO Replication Design

A1 A2Analog

Isolation

Txb3

RX3

A3Analog

Isolation

H21

H31H11

Interference Residue with Cascaded Design shows no

degradation

TX1

RX1

Analog Isolation

TX2

RX2

Independent Training with SISO Replication Design

A1 A2Analog

Isolation

Txb3

RX3

A3Analog

Isolation

H21

H31H11

C Σ

Interference Residue with Cascaded Design shows no

degradation

TX1

RX1

Analog Isolation

TX2

RX2

Independent Training with SISO Replication Design

A1 A2Analog

Isolation

Txb3

RX3

A3Analog

Isolation

H21

H31H11

Residue=e + H21+ H31

C Σ

Interference Residue with Cascaded Design shows no

degradation

TX1

RX1

Analog Isolation

TX2

RX2

Independent Training with SISO Replication Design

A1 A2Analog

Isolation

Txb3

RX3

A3Analog

Isolation

H21

H31H11

Residue=e + H21+ H31

C Σ

Σ C

Interference Residue with Cascaded Design shows no

degradation

TX1

RX1

Analog Isolation

TX2

RX2

Independent Training with SISO Replication Design

A1 A2Analog

Isolation

Txb3

RX3

A3Analog

Isolation

H21

H31H11

Residue=e + H21+ H31

Residue=2e + H31

C Σ

Σ C

Interference Residue with Cascaded Design shows no

degradation

TX1

RX1

Analog Isolation

TX2

RX2

Independent Training with SISO Replication Design

A1 A2Analog

Isolation

Txb3

RX3

A3Analog

Isolation

H21

H31H11

Residue=e + H21+ H31

Residue=2e + H31

C Σ

Σ C

Σ C

Interference Residue with Cascaded Design shows no

degradation

TX1

RX1

Analog Isolation

TX2

RX2

Residue= 3e

Independent Training with SISO Replication Design

A1 A2Analog

Isolation

Txb3

RX3

A3Analog

Isolation

H21

H31H11

Residue=e + H21+ H31

Residue=2e + H31

C Σ

Σ C

Σ C

Interference Residue with Cascaded Design shows no

degradation

TX1

RX1

Analog Isolation

TX2

RX2

Residue= 3e

Independent Training with SISO Replication Design

A1 A2Analog

Isolation

Txb3

RX3

A3Analog

Isolation

H21

H31H11

Residue=e + H21+ H31

Residue=2e + H31

C Σ

Σ C

Σ C

Analog Isolation

TX1

RX1

Σ Cascaded

Cancel

Residue = e

A1

AnalogIsolation

TX2

RX2

A2

AnalogIsolation

TX3

RX3

A3

H21

H31H11

Improved Joint Training with Cascaded Design

Evaluation Q1: Does it work with commodity radios?

Evaluation Q1: Does it work with commodity radios?

Goal: Build 3X3 MIMO full duplex radio using commodity software radios.

Evaluation Q1: Does it work with commodity radios?

• Maxim transceiver• Challenge: Extremely high

transmitter noise and non-linearities

• 20MHz BW (transceiver limitation)• 20 dBm max TX power• WiFi 802.11n PHY

Goal: Build 3X3 MIMO full duplex radio using commodity software radios.

Evaluation Q1: Does it work with commodity radios?

• Maxim transceiver• Challenge: Extremely high

transmitter noise and non-linearities

• 20MHz BW (transceiver limitation)• 20 dBm max TX power• WiFi 802.11n PHY

Goal: Build 3X3 MIMO full duplex radio using commodity software radios.

𝐻11

𝐻22

𝐻𝑐12

𝐻𝑐21

4/25/2014 21

Evaluation Q1: Does it work with commodity radios?

• Indoor noisy

office environment

4/25/2014 21

Evaluation Q1: Does it work with commodity radios?

• Indoor noisy

office environment

4/25/2014 21

Evaluation Q1: Does it work with commodity radios?

2.43 2.44 2.45 2.46 2.47

-80

-60

-40

-20

0

20

Total TX Signal

After Self-talkCancel

After Cross-talk 1 CancelAfter Cross-talk 2 Cancel

After DigitalCancel

Noise floor

Po

we

r in

dB

m

Frequency in Ghz

• Indoor noisy

office environment

• Tunes to

environmental

changes within

8us, needs to

be re-tuned

every 60 ms

4/25/2014 21

Evaluation Q1: Does it work with commodity radios?

2.43 2.44 2.45 2.46 2.47

-80

-60

-40

-20

0

20

Total TX Signal

After Self-talkCancel

After Cross-talk 1 CancelAfter Cross-talk 2 Cancel

After DigitalCancel

Noise floor

Po

we

r in

dB

m

Frequency in GhzPrototype implementation completely cancels

interference to noise floor for a 3x3 MIMO radio

Evaluation Q2: Does it have linear Complexity

Scaling ?

SISO replicationdesign

Our design

Analog Cancellation taps (3X3)

108 56 (reduced by 1.92x)

Digital Cancellation taps (3X3)

1188 485 (reduced by 2.45x)

Tuning time (3X3) 9 ms .024 ms (reduced by 375x)

Resource Comparison between SISO replication and our design

Evaluation Q2: Does it have linear Complexity

Scaling ?

SISO replicationdesign

Our design

Analog Cancellation taps (3X3)

108 56 (reduced by 1.92x)

Digital Cancellation taps (3X3)

1188 485 (reduced by 2.45x)

Tuning time (3X3) 9 ms .024 ms (reduced by 375x)

Resource Comparison between SISO replication and our design

Complexity of both Analog and Digital

Cancellation, scales linearly as number of

antennas increases

0 0.5 1 1.5 20

0.2

0.4

0.6

0.8

1

4/25/2014 23

Evaluation Q4: Does that translate to doubling of

throughput in practice?

Our Design

SISO Replication

SISO Low Complexity Replication

Gain relative to Half Duplex 3X3 MIMO

CD

F

0 0.5 1 1.5 20

0.2

0.4

0.6

0.8

1

4/25/2014 23

1.95x

Evaluation Q4: Does that translate to doubling of

throughput in practice?

Our Design

SISO Replication

SISO Low Complexity Replication

Gain relative to Half Duplex 3X3 MIMO

CD

F

0 0.5 1 1.5 20

0.2

0.4

0.6

0.8

1

4/25/2014 23

1.95x

Evaluation Q4: Does that translate to doubling of

throughput in practice?

Our design practically achieves the theoretical

throughput doubling

Our Design

SISO Replication

SISO Low Complexity Replication

Gain relative to Half Duplex 3X3 MIMO

CD

F

Worse than standard Half Duplex

Conclusion

• Design and implementation of a near-ideal

complexity and performance full duplex MIMO

radio

– Shows that full duplex and MIMO can operate

concurrently

• Has applications to many other problems

– Radio slicing, backscatter, imaging etc

• Many implications for MAC layer design

– Feedback, beamforming, MU-MIMO4/25/2014 24