Lecture 8

Spread Spectrum and OFDM

+ 2

Time Domain View (Sieve)

Channel

Direct Sequence Spread Spectrum

+ 3

Spread Spectrum

n Usually the spectrum of a signal is related to the data

(symbol) rate
n The null-to-null bandwidth @ 1/T
n T is the symbol duration

n Spread-spectrum
n The spectrum is much wider than 1/T
n The spreading is achieved using a “spreading signal” also called a
“code signal” or “spreading code”
n The receiver uses correlation or matched filtering to recover the
original data
+ 4

Types of Spread Spectrum

n Direct-sequence spread spectrum (DSSS)

n Each information symbol is “chipped” into a pattern of smaller
symbols
n The pattern is called the spread-spectrum “code” or “sequence”
n It is used in IS-95, W-CDMA, cdma2000 and IEEE 802.11

n Frequency hopping spread spectrum (FHSS)

n Symbols or packets are transmitted on different frequency carriers
each time
n Slow frequency hopping – the same frequency carrier is used over
several symbols or a packet (common)
n Fast frequency hopping – the frequency carrier is changed within a
symbol period
n Used in GSM, IEEE 802.11 (legacy) and Bluetooth
+ 5

Systems using Spread

Spectrum
nDSSS is employed in 2G CDMA systems
n IS-95, cdma2000

nDSSS is employed in all 3G cellular systems

n UMTS and HSPA

nDSSS was used in legacy IEEE 802.11 (WiFi)

+ 6

DSSS Modulation
n The original data stream Data Bit Data In
is “chipped” up into a
pattern of pulses of 1 2 3 4 5 6 7 8 9 10 11

smaller duration

n Good autocorrelation
properties “Spread” Bits

n Good cross-correlation
Spreading
properties with other
Code In
patterns

n Each pattern is called a

spread spectrum code or
spread spectrum chip
sequence
Periodic Spreading Code
+ 7

DSSS details

n Instead of transmitting a rectangular pulse for a zero or a one, we

transmit a sequence of narrower rectangular pulses

n The narrow pulses are called “chips”

n You often see references to “chips/sec” instead of bits/sec

n The easiest way of creating a DSSS signal is to multiply one period of the
spreading sequence with each data symbol
n Example: IEEE 802.11
n Barker sequence: [1 1 1 -1 -1 -1 1 -1 -1 1 -1]
n To transmit a “0”, you send [1 1 1 -1 -1 -1 1 -1 -1 1 -1]
n To transmit a “1” you send [-1 -1 -1 1 1 1 -1 1 1 -1 1]
n Sometimes parts of the spreading sequence are multiplied with
the data symbol
+ 8

Processing gain

n Definition of processing gain

n The duration of a chip is usually represented by Tc
n The duration of the bit is T
n The ratio T/Tc = N is called the “processing gain” of the
DSSS system
n The processing gain is also the ratio between the bandwidth of
the spread signal to the bandwidth of the data signal

n In many cases, this is also the ratio of the height of the

autocorrelation peak to the maximum sidelobe
n This ratio depends on the spreading code properties
+ 9

Operation of a DSSS
Transceiver

Demodulation involves a
process called “correlation”
+ 10

Spectrum and Autocorrelation

Original signal
E Autocorrelation
PSD
of Rectangle

⌧ 1 1
fc fc + f
T T

Spread Signal
PSD
E Autocorrelation
of Barker-11
1
fc
Tc
1
fc +
⌧ Tc
f
+ 11

Autocorrelation properties of the Barker

sequence

n The width of the mainlobe is 2T/11

n About one-tenth the width of the autocorrelation of the
rectangular pulse
n The height of the mainlobe is 11 times the height of
the sidelobes
n The ratio of mainlobe peak to sidelobe is an important
measure of how “good” a spreading code is
+ 12

7- Chip M-sequence
Ts
Data Bit
time time
chip
Ts
(a) (b)
[1 -1 -1 1 -1 1 1 ]
Spreading
Code
7

chip
Periodic
time time
Autocorrelation
-1
(c)
Ts
+ 13

Autocorrelation

n Consider the spreading sequence

n [1 -1 -1 1 -1 1 1]
Aperiodic
autocorrelation
1 -1 -1 1 -1 1 1

1 -1 -1 1 -1 1 1 Result: 1 x -1 + -1 x 1 + -1 x 1 = -3

1 -1 -1 1 -1 1 1 Periodic
autocorrelation
1 -1 -1 1 -1 1 1 1 -1 -1 1 -1 1 1

Result: 1 x -1 + -1 x 1 + -1 x 1 + 1 x 1 + -1 x -1 + 1 x -1 + 1 x 1= -1
+ 14

Example in a two-path channel

n Random data sequence of ten data bits

n Spreading by 11 chips using a Barker pulse

n Two path channel with inter-path delay of 17 chips > bit duration
n Multipath amplitudes
n Main path: 1
n Second path: 1.1

n Just for illustration!

n Reality:
n Many multipath components
n Rayleigh fading amplitudes
n Noise!
+ 15

Data and Channel

2
0 0 0 0 1 1 0 0 1 0
1.5

0.5

-0.5

-1

-1.5

-2
10 20 30 40 50 60 70 80 90 100 110
+ 16

Output without spreading

Errors
introduced by
Signal after correlation the channel
is sampled at green lines
Output of a Matched Filter
15 25

0 0 0 0 1 1 0 0 1 0 20
10
0 0 0 0 1 1 0 0 1 0
15

5
10

0 5

0
-5

-5
-10
-10

-15 -15
0 20 40 60 80 100 120 0 20 40 60 80 100 120

Without Multipath With Multipath

+ 17

Output with spreading

Errors introduced
by the channel are
removed
Output of a Matched Filter
15 15

10 0 0 0 0 1 1 0 0 1 0 10 0 0 0 0 1 1 0 0 1 0

5
5

0
0

-5
-5

-10
-10

-15
0 20 40 60 80 100 120 -15
0 20 40 60 80 100 120

Without Multipath With Multipath

+ 18

Summary of DSSS and Combatting

Multipath

Traditional Transmission
Symbol 1 Symbol 2

Data Bit
time
Intersymbol Interference
Ts

DSSS Transmission
Ts

time Channel
chip Reduced Intersymbol Interference
& In-band Diversity
+ 19

The RAKE receiver

n Observe the peaks in the channel output in the previous slides that are
NOT sampled (Peaks that are not at the green vertical line)
n They contain the “same” information as the sampled peaks – but these
peaks are delayed!

n A RAKE receiver consists of a tapped delay-line that samples these peaks

n Each peak usually suffers independent fading

n This is a form of diversity inherently available in DSSS systems

n In IS-95 systems the RAKE receiver has three “fingers”

n It can sample three such peaks simultaneously
n A 4th finger is used to listen to adjacent cells for RSS measurements and
to support soft hand-off
n The mobile station is temporarily connected to more than one base station
+ 20

Principle of RAKE Receiver

n Steps
n Multiple versions of a signal arrive more than one chip interval apart
n Receiver attempts to recover signals from multiple paths and combine them

n This method achieves better performance than simply recovering

dominant signal and treating remaining signals as noise
+
CDMA/DSSS Summary
+ 22

CDMA Properties: Near-Far

Problem
n A CDMA receiver cannot successfully
de-spread the desired signal in a n Power control and channel problems!
high multiple-access-interference
environment

n Unless a transmitter close to the

receiver transmits at power lower
than a transmitter farther away, the
far transmitter cannot be heard

n Power control must be used to

mitigate the near-far problem

n Mobiles transmit at such power

levels to ensure that received power
levels are equal at base station Base station
+ 23

CDMA Deployment Issues

n Radio planning in CDMA systems is different from

standard TDMA/FDMA systems
n Reuse is defined differently

n Capacity calculations are different

+ 24

Network planning for CDMA

n There is no concept of co-channel or adjacent

channel interference
n Interference arises from users in the same cell and from
neighboring cells
n Coding and spread spectrum play a very important role in
the mitigation of interference

n Instead of defining an Srbased on signal strength, it

is more common to use a value of Eb/It that provides
a given “quality of signal”
n Usually this is the value that provides a frame error rate of
1% – this provides a good MOS for voice
n The quantity It is the total interference
+ 25

More on Eb/It

n The value of Eb/It depends greatly on

n Propagation conditions
n Transmit powers of the interfering users
n Speed of the MS
n Number of multipath signals that can be used for diversity

n Cell breathing
n The boundary of a CDMA cell is not fixed and depends on
where the Eb/It is reached
n Capacity must be offloaded to other carriers to overcome
this effect
+ 26

Coverage holes in CDMA

High
Soft interference
Single hole
handoff
regions CDMA Cell

Multiple CDMA
Cells
n Power control, soft handoff and RSS thresholds play
a very important role in the design
n If too many BSs (or sectors) cover an area, this may create
a “coverage hole”
n Usually, not more than three BSs or sectors should cover
an area
+ 27

Approach

n Somewhat simplified, but works in general for M users in a cell

n Let us consider the reverse link (uplink)
n There are two components of the interference
n Own cell interference - Io
n Other cell interference – Ioc

n Assuming perfect power control, the own cell interference is

given by:
Io = (M-1) S vf
n S is the average power received from each of the M mobile stations
n The reverse link “activity factor” is vf

n The activity factor is a measure of what fraction of time a

transmission occurs
+ 28

Other Cell Interference

n Interference from other cells fluctuates as a function

of the load
n The average value Ioc can be expressed as follows
Ioc = f M S vf
n Assumption is that all other cells are similar to the current
cell
n The factor f indicates fraction of other cell received power
compared to the own-cell received power
n In some ways, f is a measure of the reuse factor

n The factor f depends on the size of the given cell, the

path loss exponent, shadow fading distribution, soft
handoff parameters, etc.
+ 29

Approach (II)

n Total interference is given by:

Itotal = Io + Ioc = [(1+f)M-1] vf S = [M/h - 1] vf S
n Here the term h refers to the “reuse efficiency”
n Suppose there is imperfect power control, we can represent this by a
factor hc
Itotal = [M/h - 1] vf (S/hc)

n In general, the required SIR must be smaller than the observed

SIR
(Eb/It)req < (SIR)system
n Ignore thermal noise

n The desired signal has a power S multiplied by the “processing

gain” Gp
+ 30

Approach (III)
n Proceeding further, we get:
Eb S Gp S Gp
= =
It [M/h - 1] vf (S/hc) [(1+f)M-1] vf S/hc

1 Gp hc
M= +
1+f (Eb/It) vf (1+f)

n Solving for M we get:

Gp hc
Mmax = 1 +
(Eb/It) vf (1+f)
n Mmax is called the “pole point” or asymptotic cell capacity
+ 31

Cell Loading and Pole Point in IS-95

n Cell loading
n A measure of the total interference in the system compared to
thermal noise
n Represented by the quantity r = M/Mmax
n You can show that it is also approximately equal to the ratio of
the total interference to the thermal noise

n Sample calculation
n Let (Eb/It)reqd = 6 dB = 4, R = 9.6 kbps, Rc = 1.2288 Mcps, hc = 0.8,
vf = 0.5, f = 0.67
n Then, the pole point or Mmax will be:
n Mmax = 1 + (1.2288 ´ 106/9.6 ´ 103)(0.8/(4 ´ 0.5 ´ [1+0.67]) = 1 +
30.65 = 32
n If a 3 sector antenna is used, typically, the gain in capacity is by a
factor of 2.55 so that the pole point is: 31.65´ 2.55 = 81
+ 32

Comparison with AMPS/TDMA

n In AMPS, each service provider has 12.5 MHz BW

n With a 3 sector antenna, we can have a frequency reuse of 7
n There are 30 kHz channels per voice call
n Number of channels/cell =
(12.5 ´ 106 / 30 ´ 103) ´ (1/7) = 57

n In the case of IS-136, with a 3 sector antenna, we can have a frequency

reuse of 4
n Each 30 kHz channel can carry 3 voice calls
n Number of channels/cell =
(12.5 ´ 106 / 30 ´ 103) ´ (1/4) ´ 3 = 312.5
n What was the pole point of IS-95?
n 81 per carrier per cell sector
n With 8 cdma carriers in a 12.5 MHz bandwidth, we can have up to 648
channels per cell sector
n With 10 cdma carriers in a 12.5 MHz bandwidth, we can have up to 810
channels per cell sector
+ 33

Remarks

n Ranges of values
n Power control inefficiency hc varies between 0.7 and 0.85
n Voice activity factor vf varies between 0.4 and 0.6
n The other cell interference f varies between 0.56 and 1.28
for a path loss exponent of 4 and a standard deviation of
shadow fading of 6 to 10 dB
+ 34

Other issues

n Forward Link
n We have to be worried about the pilot, sync, paging and traffic channels in IS-95 and
many more in cdma2000 and UMTS
n The strength of the pilot channel effectively determines the size of the cell
n Interference is from clusters of high power transmitters rather than many distributed low
power transmitters
n Design should try to make the forward and reverse link capacities as close to one another
as possible
n This will reduce the amount of unnecessary interference and enable smooth handoffs
between cells

n PN Sequence Reuse
n How closely should the same pilot offsets be used? (later when we do IS-95)

n How does the link budget affect the capacity?

n How does soft handoff affect the capacity?

+ 35

Frequency Domain View (Gate)

Channel

OR

Orthogonal Frequency Division Multiplexing

+ 36

Diversity (continued) – Frequency

Hopping
n Traditional
n Transmitter/receiver pair communicate on a fixed frequency
channel.

n Frequency Hopping Idea

n Noise, fading and interference change with frequency band
in time
n Move from band to band
n Time spent on a single frequency is termed the “dwell time”

n Originally developed for military communications

n Spend a short amount of time in one frequency band
n Prevent interception or jamming
+ 37

Frequency Hopping Spread

Spectrum

Developed during WWII by actress Hedy

Lammar and classical composer
George Antheil

Patent given to government

+ 38

Frequency Hopping Spread

Spectrum
n Two types of systems
n Slow Hopping
n Dwell time long enough to transmit several bits in a row (timeslot)
n Fast Hopping
n Dwell time on the order of a bit or fraction of a bit (primarily for military systems)

n Transmitter and receiver must know hopping pattern or algorithm

that determines the pattern before communications.
n Cyclic pattern – best for low number of frequencies and combating small-scale
fading :
n Example with four frequencies: f4, f2, f1, f3, f4, f2, f1, f3, ….
n Random pattern – best for large number of frequencies, combating co-channel
interference, and interference averaging
n Example with six frequencies: f1, f3, f2, f1, f6, f5, f4, f2, f6, …
n Use random number generator with same seed at both ends
+ 39

Frequency Hopping concept

CLK C B A fc

t0 1 0 0 f4

t1 0 1 0 f2

t2 1 0 1 f5

C B A t3 f6
1 1 0

t4 1 1 1 f7
CLK
t5 0 1 1 f3
f7
One Period of Sequence = 1 0 0 1 0 1 1
t6 f1
f6
0 0 1

frequency channels
t7 1 0 0 f4
f5
f4
f3

f2
f1
time
+
Combatting Time Dispersion
Hop
Frequencies
Received SNR

frequency
Transmission Retransmission
Lost Here Here Successful
 + 41

Example Systems
Collision Different Users

n GSM (2G Cellular)

…… n Very slow hopping

n Original IEEE 802.11

Received SNR

n Slow hopping

1
MHz
n Bluetooth
n Also slow hopping
2.402 2.480 frequencyover 79 frequencies
each 1 MHz wide
GHz GHz

n Per packet hopping

+
How do you utilize the entire
bandwidth?

Idea in IEEE 802.11g/a

= OFDM!
+ 43

Orthogonal Frequency Division

Multiplexing
n Idea in frequency domain:
n Coherence bandwidth limits the maximum data rate of the channel
n Send data in several parallel sub-channels each at a lower data rate and
different carrier frequency

n Idea in time domain:

n By using several sub-channels and reducing the data rate on each
channel, the symbol duration in each channel is increased
n If the symbol duration in each channel is larger than the multipath delay
spread, we have few errors

n OFDM enables
n Spacing carriers (sub-channels) as closely as possible
n Implementing the system completely in digital eliminating analog VCOs
+ 44

What is OFDM?

n Modulation/Multiplexing technique

n Usual transmission
n Transmits single high-rate data stream over a single
carrier
n With OFDM
n Multiple parallel low-rate data streams
n Low-rate data streams transmitted on orthogonal
subcarriers
n Allows spectral overlap of sub-channels
+ 45

OFDM Remarks

n It is NOT a new technology but has found new importance because

of applications
n DSL modems where the channel is not uniform
n Digital audio and video broadcast
n Wireless LAN applications
n IEEE 802.11a and HIPERLAN-2

n Fast implementation using FFT’s is now possible

n Can be adaptive (used in 802.11a)
n Problems
n Synchronization between carriers
n Peak-to-average power (PAP) ratios
n Requires linear amplifiers
+ 46

OFDM Advantages

n Bandwidth efficiency

n Reduction of ISI
n Needs simpler equalizers

n Robust to narrowband interference and frequency

selective fading
n Possibility of improving channel capacity using
adaptive bit loading over multiple channels
+ 47

OFDM in frequency and time domains

n Note orthogonality in both domains

n What is one “OFDM symbol”?
Power Spectrum Fourier transform of symbol

Channel

Bc

single carrier

frequency

sub-carrier Sub-Carriers
Amplitude

time
Frequency
4 Carriers Spanning the Bandwidth
of One Carrier
+ 48

OFDM Signal/Symbol

Df
+ 49

OFDM Symbol

n One OFDM “symbol” lasts for say Ts seconds

n The symbol consists of the sum of the individual symbols from the many sub-carriers
n Example: Consider QPSK on each carrier

n In general
n For N subchannels, the N samples of the i-th transmitted OFDM symbol can be written
as

Complex
IFFT
Number
+ 50

Guard Time and Cyclic Prefix

n Guard time eliminates ISI if larger than expected delay spread

occurs
n If the guard time has no signal, intercarrier interference (ICI)
may occur
n ICI is like a cross talk between subcarriers

n A cyclic prefix eliminates ICI

n Ensures that delayed replicas of OFDM symbols always have
integer number of cycles within the FFT interval
n Maintains orthogonality between subcarriers
n Cyclic prefix is removed at the receiver
+ 51

OFDM Transmission – basic system

n N consecutive complex symbols are converted into a group of N

parallel data streams, which then are modulated over
orthogonal subcarriers

Channel Symbol Serial to N-Point Parallel Guard/CP

Encoding Mapping Parallel IFFT To Serial Insertion

Radio
Channel
+ AWGN

Channel Parallel to N-Point Serial to Guard/CP

Detector
Decoding Serial FFT Parallel Removal
+ 52

Adaptive OFDM
|H(f)|

frequency
+ 53

Channel Partitioning for Multicarrier

Modulation

n As the channel is frequency Noise PSD/|H(f)|

selective, it makes sense to
split the channel into several
smaller parts
n Each smaller chunk is now
an AWGN channel
n Each AWGN channel
provides a different SNR
Allocation of power
n Question: How do we allocate
transmit powers/modulation
Water-filling algorithm
schemes to each chunk? What
Allocate more energy where
is the most optimal?
the SNR is better!
+ 54

Adaptive OFDM

n Improve channel capacity further

n Change modulation scheme
n Allocating bits/power per subcarrier according to the quality of each
subchannel

AOFDM Components
Adaptive Loading/Allocation Algorithm

+
Set of Modulation Schemes
+
Channel Quality Estimator*
+ 55

Adaptive Modulation

No transmission (0 bits)
BPSK (1 bit/symbol)
Set of Modulation
Schemes QPSK (2 bits/symbol)
8-QAM (3 bits/symbol)
16-QAM (4 bits/symbol)
+ 56

Adaptive Modulation on Parallel

Channels

SNR (dB) BW Efficiency

4 bits

3 bits

2 bits

1 bit

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Number of Subcarriers
+ 57

Operation of Adaptive Algorithms

Channel
Quality
Estimator
Based on optimal Channel Quality time
“Water-Filling” Information, e.g. SNR
Power Distribution
Adaptive Modulation Scheme Selection
Algorithm

Subcarrier 1

Bits Subcarrier 2
And Subcarrier 3
Power Allocation
+

Subcarrier N
+ 58

OFDM Based Wireless LANs – IEEE

802.11a
n Operates in the U-NII Band
n 5.15–5.25, 5.25–5.35, and 5.725–5.825 GHz
n Provides multiple transmission modes/rates depending on
channel conditions.
n 6, 9, 12, 18, 24, 36, 48, and 54 Mbps
n 4 digital modulations: BPSK, QPSK, 16-QAM, 64 QAM
n Radio spectrum is divided into 8 separate segments/channels,
20 MHz each
n 52 carriers (subchannels) per channel
n Each subcarrier has bandwidth of ~300 kHz
n 48 for data modulation, and 4 for pilot signal
+
Recent Trends

n MIMO with OFDM

n IEEE 802.11n, 802.11ac
n Data rates greater than 100 Mbps

n OFDM for wide area data services

n LTE and WiMax

n Other PHY technologies

n UWB with OFDM
n MC-CDMA
+ 60

Revisiting “Data Rates” in Wireless

n Home A/V networks are expected to need 1-

10 Gbps
n Assuming a spectral efficiency of 1 bps/Hz, we need
at least 1 GHz of spectrum
n Have ignored the effects of multipath fading
n Brute force approach
n May not meet the technology, regulatory and
cost requirements
n Can we increase the bps/Hz in wireless
systems?
+ 61

Edholm’s Law
n Phil Edholm

Eventual convergence
n Nortel’s CTO

n Three Telecom
Categories
n Wireline
n Nomadic (Portable)

n Wireless (Mobile)

n Data rates increase

exponentially
n There is a
predictable time lag
between wireless
and wireline
systems Source: IEEE Spectrum - July 2004
+ 62

How can we increase data rates?

n Traditional ways
n Reduce the symbol duration
n Needs larger bandwidth
n Leads to a wideband channel and frequency selectivity -
irreducible error rates
n Increase the number of bits/symbol
n Error rates increase with M for the same Eb/N0

n MIMO systems
n There is no need to increase the bandwidth or power
n But what are the limitations?
n Use multiple transmit (Tx) and receive (Rx) antennas
n Increases spectral efficiency to several tens of bps/Hz
+ 63

What is MIMO?

n So far we have considered Single Input

Single Output or SISO systems
n Both transmitter and receiver have
one antenna each
n Simplest form of transceiver
architecture

n Single input multiple-output (SIMO)

systems
n Receiver has multiple antennas

n Multiple input multiple output (MIMO)

systems
n Both transmitter and receiver have
multiple antennas
n Strictly: Each antenna has its own RF
chain (modulator, encoder and so on)
+ 64

Performance enhancements due to

MIMO
n Diversity gain
n Ability to receive multiple copies of the signal with
independent fading
n Spatial multiplexing gain
n Send different information bits over different
antennas and recover the information
n Interference reduction
n Reduce the region of interference thereby
increasing capacity