MIMO As a First-Class Citizen in 802.11

MIMO As a First-Class Citizen in 802.11

Kate C.-J. LinAcademia Sinica

Shyamnath Gollakota and Dina KatabiMIT

Wireless nodes increasingly haveheterogeneous numbers of antennas

1-antenna devices 2-antenna devices3-antenna devices

802.11 Was Designed for 1-Antenna Nodes

When a single-antenna node transmits,multi-antenna nodes refrain from transmitting

Alice

Bob

Chris

But, MIMO Nodes Can Receive Multiple Concurrent Streams

Alice

Bob

Chris

Alice

It’s Not That Simple

But, how do we transmit concurrentlywithout interfering with ongoing transmissions?

Interference!!

Interference!!Bob

Chris

Enable concurrent transmissionswithout harming ongoing transmissions

Goal

802.11n+

802.11n+

• Allows MIMO nodes to join ongoing transmissions without interfering with them

• Maintains 802.11 random access

• Implemented and shown to significantly improve the throughput

1. How to transmit without interfering with ongoing transmissions? Interference nulling Interference alignment

2. How do we achieve it in a random access manner? Multi-dimensional carrier sense



Interference Nulling

Alice

Bob nullin

g

• Signals cancel each other at Alice’s receiver• Signals don’t cancel each other at Bob’s

receiver Because channels are different

€

y

€

h1αy+h2βy=0


• Signals cancel each other at Alice’s receiver• Signals don’t cancel each other at Bob’s

receiver Because channels are different

Alice

€

βy

€

αy

€

h1

€

h2

€

≠0Bob

€

⇒ Nulling: h1α =−h2β



Q: How to transmit without interfering with ongoing

transmissions?A: Nulling

€

h1αy+h2βy=0Alice

€

βy

€

αy

€

h1

€

h2

€

≠0Bob

€

⇒ Nulling: h1α =−h2β

Alice

Bob

Chris

Is Nulling Alone Enough? NO!!

Alice

Bob

Chris

NO!


Alice

Bob

Chris

nullin

g

NO!

nullin

g

N-antenna transmitter cannot null

at N receive antennas

Chris needs to null at three antennas But, he can’t!

€

(h11α +h21β+h31γ)z = 0


Alice

Bob

€

βz

€

αz

€

gzChris

€

(h12α +h22β+h32γ)z = 0

€

(h13α +h23β+h33γ)z = 0

€

Only solution α =β=g=0

Transmit Nothing!!!

NO!null

€

z

Do we really need to null at all antennas?

No, we can use interference alignment

MIMO Basics1. N-antenna node receives in N-dimensional

space

antenna 1

antenna 2

antenna 1

antenna 1

antenna 2

antenna 3

MIMO Basics1. N-antenna node receives in N-dimensional

space2. Transmitter can rotate the received signal

To rotate received signal y to y’ = Ry, transmitter multiplies its transmitted signal by the same rotation matrix R

y’y2-antenna receiver = Ry

Interference Alignment

N-antenna node can only decode N signals

wanted signalI1I2

If I1 and I2 are aligned, appear as one interferer 2-antenna receiver can decode the wanted signal

2-antenna receiver

Interference Alignment

If I1 and I2 are aligned, appear as one interferer 2-antenna receiver can decode the wanted signal

N-antenna node can only decode N signals

2-antenna receiver I1 + I2wanted signal

align

ing

Use Nulling and Alignment

nullin

g Alice(unwanted)Bob

Chris

Alice

Bob

Chris

Null as before

align

ing

Use Nulling and Alignment

Alice

Bob

Chris

nullin

g

Alice + Chris

(unwanted)Bob

Can decode Bob’s Signal and (Alice + Chris)

All senders transmit, but the throughput is as high as if only

the 3-antenna node is transmitting all the

time

General Protocol• Each sender in a distributed way

computes where and how to null where and how to align

• Analytically proved: # concurrent streams = # max antenna per

sender





Alice

Bob

Chris

Centralized controller

But, lost the benefit of 802.11 random access

Alice

Bob

ChrisBob, Chris, both you can transmit a packet concurrently

Centralized controller

n+ maintains random access!

In 802.11, contend using carrier sense

Multi-Dimensional Carrier Sense

But, how to contend despite ongoing transmissions?

Alice

Bob

Alice

Alice

one signal

AliceBob

two signals

Say that Ben is performing carrier sense Ben

Distinguishable using simple linear algebra

Multi-Dimensional Carrier SenseAlice

BobContend

BenAlice

Contend

Alice


BobProject

BenAlice

Bob and Ben project orthogonal to Alice’s signal

Alice

Project


BobProject

orthogonal to Aliceno signal from Alice!!

Alice


BenAlice

Project

Bob and Ben project orthogonal to Alice’s signal


BobProject

Alice

Apply 802.11 contention after projection

BenAlice

Project



Alice

Alice

Bob

Bob

Detect energy after projection

Multi-Dimensional Carrier Sense

Win

LoseBen

Works for arbitrary number of antennas



Performance

Implementation

• Implemented in USRP2

• OFDM with 802.11-style modulations and convolutional codes

TestbedRandomly assign the nodes to the marked locations





Nulling Experiment

wanted signal

unwanted signal

Can Bob null his signal at Alice’s receiver?

BobAlice

010 15 20 25 27.5-32.5

0

0.5

1

1.5

2

SNR of unwanted signal [dB]

Resi

dual

inte

rfer

ence

[dB]

Nulling Experiment

802.11 SNR range

Nulling Experiment

Residual interference from Bob can reduce the SNR of wanted signal by at most ~1dB

010 15 20 25 27.5-32.5

0

0.5

1

1.5

2


Resi

dual

inte

rfer

ence

[dB]

Alignment and Nulling Experiment

010 15 20 25 27.5-32.5

0

0.5

1

1.5

2


Resi

dual

inte

rfer

ence

[dB]

Alignment and Nulling Experiment

Though alignment is harder, residual interference is still small

~1.5dB

10 15 20 25 27.5-32.50

0.5

1

1.5

2nullingalignment


Resi

dual

inte

rfer

ence

[dB]

0



Carrier Sense Experiment

tx1 tx1 + tx2 tx1 tx1 + tx2Traditional CS CS after projection

0

0.1

0.2

0.3

Time

Pow

er (m

Watt

)

0

0.005

0.01

0.015

0.02

0.025

Time

Pow

er a

fter p

roje

ction

0

0.1

0.2

0.3

Time

Pow

er (m

Watt

)


tx1 tx1 + tx2 tx1 tx1 + tx2

Can’t identify

0

0.005

0.01

0.015

0.02

0.025

Time

Pow

er a

fter p

roje

ction

Hard to distinguish

Traditional CS CS after projection

0

0.1

0.2

0.3

Time

Pow

er (m

Watt

)


0

0.005

0.01

0.015

0.02

0.025

Time

Pow

er a

fter p

roje

ction

Hard to distinguish9dB jump

tx1 tx1 + tx2 tx1 tx1 + tx2

Can identifyCan’t identify

Traditional CS CS after projection

Throughput Experiment

Compare 802.11n with 802.11n+

BobAlice Chris


0 10 20 30 40 50 60 70 800

0.2

0.4

0.6

0.8

1

Total throughput [Mb/s]

CDFs

0 10 20 30 40 50 60 70 800

0.2

0.4

0.6

0.8

1

802.11n+ 802.11n


CDFs


0 10 20 30 40 50 60 70 800

0.2

0.4

0.6

0.8

1

802.11n+ 802.11n


CDFs


~2x

n+ delivers significant throughput gain in practice

Related Works• Information theory

[CJ08], [MMK08], [JS08], …

• MIMO systems Beamforming [AASK10], SAM [TLFWZC09],

and IAC [GPK09]

First MIMO concurrent transmissions

without any central coordination

Conclusion• In today’s 802.11, MIMO is an add-on

• In 802.11n+, MIMO is a first-class citizen Higher concurrency With random access

• Shown practical via implementation and testbed evaluation

MIMO As a First-Class Citizen in 802.11

Documents

single antenna

mimo nodes refrain

antenna nodes refrain

singleantenna devices

singleantenna node

antenna nodeswhen

singleantenna systems

wireless nodes