Wireless Networking

Fundamentals and Applications

Kate C.-J. Lin ()

Network & Mobile System Lab
Research Center for IT Innovation
Academia Sinica

Agenda
Auto Rate Adaptation
Orthogonal Frequency Division Modulation
(OFDM)
Multi-Input Multi-Output Systems (MIMO)

Auto Rate Adaptation

Modulations and bit-rates
SNR and bit-error rate
Bit-rate selection algorithms

Modulations
BPSK
0 -1+0i

1 1+0i

I
Constella)on Points

Modulate digital bits

to a complex number (sample)

Modulations
Q

BPSK

QPSK

Q
11

01

I
10

16QAM

1100

1000

0101

1101

1001

0011

0111

1111

1011

0010

0110

1110

1010

0000

0100

0001

64QAM

00

Demodulation
BPSK

Received sample
I

Closet constella?on point

If Tx is actually sending 0, bit error occurs

Map the received complex number

back to digital bits

Bit-Rates in 802.11
Bitrate

802.11
Standards

DSSS
or
OFDM

Modulation

Bits
per
Symbol

Coding
Rate

1
2
5.5
11
6
9
12
18
24
36
48
54

b
b
b
b
a/g
a/g
a/g
a/g
a/g
a/g
a/g
a/g

DSSS
DSSS
DSSS
DSSS
OFDM
OFDM
OFDM
OFDM
OFDM
OFDM
OFDM
OFDM

BPSK
QPSK
CCK
CCK
BPSK
BPSK
QPSK
QPSK
QAM-16
QAM-16
QAM-64
QAM-64

1
2
1
2
1
1
2
2
4
4
6
6

1/11
1/11
4/8
4/8
1/2
3/4
1/2
3/4
1/2
3/4
2/3
3/4

MegaSymbols
per
second
11
11
11
11
12
12
12
12
12
12
12
12

Coding Rate
Avoid random errors

1/2: Add 1x redundant bits

3/4: Add 1/3x redundant bits

Havent solved the problem yet

Data input: 1, 1, 0, 1, 0, 1, 1, 0,
After encoding:
1, 1, 1, 1, 0, 0, 1, 1, 0, 0, 1, 1, 1, 1, 0, 0, .
Still one bit error Suffer from bursty errors

Interleave and De-interleave

1, -1, -1, 1, -1,
1, 1, -1, 1, 1

D/A

1, 0, 0, 1, 0,
1, 1, 0, 1, 1

Modulation

1, 1, 1, 1, 0,
0, 1, 1, 0, 0

Interleave

1, 1, 0, 1, 0

Source coding

Transmitter

channel

Receiver

1, -1, 1, -1, -1,

1, 1, -1, 1, 1

A/D

1, 0, 1, 0, 0,
1, 1, 0, 1, 1

+
De-modulation

1, 0, 1, 1, 0,
0, 1, 1, 1, 0

De-interleave

1, 1, 0, 1, 0

Decoding

Create a more uniform

distribu?on of errors

noise

Channel Quality vs. Bit-Rate

When channels are very good

Encode more bits as a sample

When channels are noisy

Encode fewer bits as a sample

Why is it affected by the channel quality?

Error Probability vs. Modulations

Given the same SNR
BPSK

QPSK
noise

Q
11

01

noise

signal
10

Decode correctly

00

Decode incorrectly

SNR = 10log10 (signal/noise)

Given the same SNR, decodable for BPSK,

but un-decodable for QPSK

SNR vs. BER (Bit Error Rate)

1

BPSK (1 megabit/s)
QPSK (2 megabits/s)
QAM-16 (4 megabits/s)
QAM-64 (6 megabits/s)

Bit Error Rate

0.1
0.01
0.001
0.0001
1e-05

10

15
20
S/N (dB)

25

30

35

802.11
operating
Figure
1-2: Theoretical
bit error region
rate (BER)5dB
versus signal-to-noise ratio for several modulation schemes assuming AGWN. The y axis is a log scale. Higher bit-rates require larger

SNR vs. PER (Packet Error Rate)

In 802.11, a packet is received correctly
if it passes the CRC check (all bits are correct)
Throughput (Megabits per Second)

Receive all or none

BPSK (1 megabit/s)
QPSK (2 megabit/s)
QAM-16 (4 megabits/s)
QAM-64 (6 megabits/s)

PER = 1-(1-BER)n
Throughput
= (1-BER)n * bit-rate

4
3

Throughput degrades
quickly even with a
little BER

2
1
0

10

15
20
S/N (dB)

25

30

Bit-Rate Selection

Throughput (Megabits per Second)

Given the SNR, select the bit-rate that

can achieve the highest throughput
6

BPSK (1 megabit/s)
QPSK (2 megabit/s)
QAM-16 (4 megabits/s)
QAM-64 (6 megabits/s)

64QAM

4
3

QPSK

2
1
0

10

15
20
S/N (dB)

25

30

Figure 1-4: Theoretical throughput in megabits per second using packets versus signal-to-

Difficulties with Rate Adaptation

Channel quality changes very quickly

Especially when the device is moving

Cant tell the difference between

poor channel quality due to noise/interference/

collision (highnoise)
poor channel quality due to distance (low

signal)
Ideally, we want to decrease the rate due to low
signal strength, but not interference/collision

Types of Auto-Rate Adaptation

Transmitter-based
SNR-based

Receiver-Based
RBAR, OAR

ACK-based

ARF, AARF

Throughput-based

SampleRate
(default in Linux)

RRAA

Selected by Tx
(Less accurate)

Selected by Rx
(Higher overhead)

Sync. ACK vs. Async ACK

backo

Tx

Data
ACK

Rx

Sync. ACK

SIFS

backo

A-ACK

DIFS

Cost the minimum overhead

Only know whether the packet is transmitted
correctly
Dont know whether the packet error is due to
incorrect rate selection or collision

Async. ACK

Cost extra overhead

Can include more detailed information

Robust Rate Adaption Algorithm (RRAA)

Dynamically enable RTS/CTS before data
transmission
Detect that the low throughput is due to
the incorrect bit-rate selection or
collision (hidden terminals)
Estimate the correct number of
transmissions to keep RTS/CTS
Disable RTS/CTS if it does not help
S. Wong, H. Yang, S. Lu, V. Bharghavan, Robust Rate Adapta?on for
802.11 Wireless Networks, ACM MOBICOM, 2006

 Remove the transmission time from the total transmission times at that bit-rat
that destination.

SampleRate

Periodically send packets at bit-rates

to that destination.
other than the current bit-rate

If the packet succeeded, decrement the number of successful packets at that bit-

r*
After remove stale results() performs
these operationsr for each stale sample, it

pkt1
pkt2
pkt and destination. rem
pkt10
calculates the pkt1
minimum
average
transmission
times
for eachpkt
bit-rate
pkt1

2
1 retry
retry

then sets the currentretry
bit-rate
for each
destination to the one with
the1smallest
average tr

Calculate the transmission time of each

To calculate
the transmission time of a n-byte unicast packet given the bit-rate b
packet

mission time.

number of retries r, SampleRate uses the following equation based on the 802.11 uni

packet length, bit-rate, number of retries,

retransmission mechanism detailed in Section 2.2:
backoff time

tx time(b, r, n) = dif s + backof f (r) + (r + 1) (sif s + ack + header + (n 8/b)

37

Sample Rates
Select the rate that has the smallest
predicted average packet transmission
time
Do not sample the rates that

have failed four successive times

are unlikely to be better than the current one

Is thought of the most efficient scheme

for static environments
J. Bicket, Bit-rate Selec?on in Wireless Networks, Ph.D Thesis, MIT, 2005

Rate Adaptation for Multicast?

Can only assign a single rate to each packet

Possible Solutions

For reliable transmission: select the rate based on

the worst node
For non-reliable transmission: provide clients
heterogeneous throughput

Recent Proposals
ZipTx
K. Lin, N. Kushman and D. Katabi, Harnessing Partial Packets in 802.11
Networks, ACM MOBICOM, 2008
Exploit partial packets with consideration of bit-rate adaptation

SoftRate
M. Vutukuru, H. Balakrishnan and K. Jamieson, Cross-Layer Wireless Bit
Rate Adaptation, ACM SIGCOMM, 2009
Exploit soft information to improve selection accuracy

FARA
H. Rahul, F. Edalat, D. Katabi and C. Sodini, Frequency-Aware Rate
Adaptation and MAC Protocols, ACM MOBICOM, 2009
Adapt the bit-rate for every OFDM subcarrier

ESNR
D. Halperin, W. Hu, A. Sheth and D. Wetherall, Predictable 802.11 Packet
Delivery from Wireless Channel Measurements, ACM SIGCOMM, 2010
Consider frequency selective fading

Frequency-Aware Rate
Adaptation (FARA)
H. Rahul, F. Edalat, D. Katabi, C. Sodini
MOBICOM 2009

logies that
es, such as
wider fremade 100unlicensed
ed discardo access the
. This shift
frequency
other in the
requencies

RKF Engineering Solutions, LLC

Frequency Diversity
30
25
SNR (dB)

Technology

20
15
10
5
0
-40

-20

0
Freq (Mhz)

20

40

Frequency
diverse
across
of
802.11a
Figure 1: Frequency
diversity
across 100100MHz
MHz of 802.11a
specaware rate
trum as observed by two receivers for transmissions from the
rtures fromspectrum
same sender. The figure shows that the SNRs of different frequens a scheme
ciesSNRs
can differof
by asdifferent
much as 20 dB onfrequencies
a single link. Further, different

The
can be as
data transreceivers prefer different frequencies.
ink, it en-much as 20dB on a single link
requencies
Different receivers could prefer different
traditional
C protocolfrequencies
use of digital TV whitespaces that occupy 100-250 MHz of spectrum

FARA

te Adaptation
and MACthe
Protocols
Instead of assigning
same rate to the

entire
frequency
band, it allows
each

z Edalat , Dina Katabi , and Charles Sodini

sub-carrier to pick a modulation
OFDM
f Technology
RKF Engineering Solutions, LLC
and a code rate that match its SNR
54Mb/s

25
SNR (dB)

ologies that
es, such as
d wider fremade 100unlicensed
ed discardo access the
g. This shift
t frequency
other in the
frequencies

30

20
15
10
5

6Mb/s

0
-40

-20

0
Freq (Mhz)

20

40

a new
ntains
which

es not
d, the
bband
ceived
urthernoise
d sent
ce alThus,
om evsince

Receiver
driver protocol
$"%&'#()*+,-../('0-1/

%/571/':'%/-*./

2%3

""#

!""#

FARA

%23

%/,7189).-

ot bits

&7189)./

!"#$$#$#"!

371/':'!*./(9/

(5)

!"#$$#$#"!

$"%&'4/5/-6/('0-1/

(b) Schematic of FARA-enabled 802.11 PHY

Initially, the sender transmit few symbols using

the lowest
bit-rate
for allwith
sub-carriers
Figure
3: OFDM
PHY semantics
and without FARA. In

FARA-enabled devices, the choice of modulation and FEC code rate

The receiver selects the bit-rate based on an
is done independently for each OFDM subband.

SNR-Rate mapping table

Minimum Required SNR
<3.5 dB
3.5 dB
5.0 dB
5.5 dB
8.5 dB
12.0 dB
15.5 dB
20.0 dB
21.0 dB

Modulation
Coding
Suppress subband
BPSK
1/2
BPSK
3/4
4-QAM
1/2
4-QAM
3/4
16-QAM
1/2
16-QAM
3/4
64-QAM
2/3
64-QAM
3/4

Predictable 802.11 Packet Delivery

from Wireless Channel Measurements
D. Halperin, W. Hu, A. Sheth and D. Wetherall
ACM SIGCOMM, 2010

SNR-based Rate Adaptation

SNR-based rate adaptation is usually
inaccurate because we

Assume frequency-flat fading

Select the bit-rate based on average SNR
across bins

However, this will over-estimate the

channel quality because

A packet will fail to pass the CRC check even if

only few bits are erroneous due to frequencyselective fading

MIMO Stream
Separation

OFDM
Demodulator

Effective SNR

(0)
Received
signal

Deinterle

(2)
Scrambled,
interleaved,
coded bits

Bias toward the weaker sub-carrier SNRs

1
=
BERk (SNR
Figure 3: TheBER
802.11n
decoding
process. MIMO r
eff ,k MIMO-OFDM
s)
52
Demodulation converts the separated
signals into bits (2). Bits from th
!1correct errors. Finally, scramblin
by convolutional
decoding
(4)
to
SNReff ,k = BERk (BEReff ,k )
processed (5).

Modulation

Bits/Symbol (k)

BPSK
QPSK
QAM-16

1
2
4

QAM-64

BERk ()

Q 2

3
Q
/5
4

7
Q
/21
12

Table 2: Bit error rate as a function of the symbol SNR for

Effective SNR

Figure 4: Our indoor 802.11n testbeds, T1 and T2. T1 consists of

spread
overup
20 000
feet. The
nodesusing
are placed
to ensure a
Look
thesquare
SNR-MAP
table
ESNR
nodes, and diverse scattering characteristics.
24

SNR (dB)

Subcarrier SNRs

20

Packet SNR
QAM-64

16

QAM-16

12
8
-28

QPSK
BPSK

-14

14

28

Subcarrier index
Figure 5: Sample faded link showing the packet SNR and ef-