EHB 453, Introduction to Mobile

Communications
Lecture6: OFDM
Prof. Mustafa Ergen
Outline
• What is OFDM?
• Coding
• Synchronization
• Equalization
• Peak to Average Power Ratio
• Ex: IEEE 802.11a
What is OFDM?
• Multi-carrier modulation/multiplexing technique
• Available bandwidth is divided into several subchannel
• Data is serial-to-parallel converted
• Symbols are transmitted on different subcarriers
OFDM

• Spectra overlap, but signals are orthogonal.

• Example: Rectangular waveform -> Sinc spectrum
Wireless Multipath Channel

•Delay spread •Doppler spread

Time

FT FT

Frequency Time

Frequency

Frequency
Effects of Multipath Delay and
Doppler
Narrowband Wideband OFDM
QAM

Time
Time

Time

Frequency Frequency Frequency

OFDM
• Advantages
– Spectral efficiency
– Simple implementation
– Tolerant to ISI
• Disadvantages
– BW loss due guard time
– Prone to frequency and phase offset errors
– Peakto average power-problem
Applications
• Fixed / Wireline:
• ADSL Asymmetric Digital Subscriber Line
• Mobile / Radio:
• Digital Audio Broadcasting (DAB)
• Digital Video Broadcasting - Terrestrial (DVB-T)
• Hiperlan II
• Wireless 1394
• Wireless LAN: IEEE 802.11a/n
• 4G: OFDMA based WiMAX, LTE
What is subchannel and subcarrier?
• Frequency-selective channel is divided into flat
fading subchannels
• Fast serial data stream is transformed into slow
parallel data streams – Longer symbol durations
Subchannels
• OFDM
spectrum
subchannel
spacings are
selected so,
that they are
mathematically
orthogonal to
each other

• Subchannels
overlap on
each other –
Sinc-shaped
spectra
OFDM Subcarrier Spectra

•before channel

•after channel

Frequency
Pilots
• Transmitted instead of data – known by the receiver
– Pilots are transmitted first in each burst
• 802.11a/g uses 4 subchannels as pilots
• Some timeslots can be used as pilots
• Data can be normalized by pilot components
• Pilots are designed for easy detection
• Pilots are used for channel estimation
• Frequency and phase offsets
• Can be used for synchronization
Guard time: Cyclic Extension
Guard Time vs Delay spread

Guard Time increases the transmit energy and reduces the bit rate
however is required to combat with ISI
A simplified OFDM system
Mechanics
• The incoming data is converted from serial to
parallel and grouped into bits each to form a
complex number x after modulation in order to
be transmitted over
N low-rate data streams.
• Each low-rate data stream is associated with a
subcarrier of the form
j 2 πf k t
€
€
φ k (t) = e
N −1
1
s(t) =
N
∑x φ k k (t), 0<t <T
k= 0
I-FFT: OFDM Transmission
• Transmission of QAM symbols on parallel subcarriers
• Overlapping, yet orthogonal subcarriers

User φ 0 (t)
symbols
φ1 (t)

=
s(t)

Parallel-to-
Serial-to-

I-FFT
Serial-to-
parallel

Parallel

Serial
φ k (t)
€
φ N −1 (t)
€

€
€
OFDM Symbol Different Modulation

Same Modulation
Received Signal Number of
Received Signal Subcarriers

N −1
1
y(t) = s(t) * h(t) =
N
∑H k x k φ k (t), 0<t <T
Delay Spread k= 0
τh
Hk = ∫ h(t)e j 2 πf k t dt Channel Impulse Response
0

y k = H k x k, k = 0,...,N −1 Received signal after sampling

^ yk
xk = ^
Hk Channel estimation

Estimated Input
Example of OFDM
• Lets we have following information bits
– 1, 1, -1, -1, 1, 1, 1, -1, 1, -1, -1, -1, -1, 1, -1, -
1, …
• Just converts the serials bits to parallel bits
C1 C2 C3 C4

1 1 -1 -1
1 1 1 -1
1 -1 -1 -1
-1 1 -1 -1
-1 1 1 -1
-1 -1 1 1
Example of OFDM cont..
Modulate each column with corresponding sub-carrier using BPSK

Modulated signal for C1 Modulated signal for C2

Modulated signal for C3 Modulated signal for C4

Example of OFDM cont..
• Final OFDM Signal = Sum of all signal
V(t)
N -1

V (t ) = å I (t ) sin( 2pnt )
n =0
n

Generated OFDM signal, V(t)

OFDM Architecture
Coding
• In a multipath environment, all subcarriers will experience
different fading environment and all will arrive with
different amplitudes.
– Block coding
– Convolutional coding
– Concatenated coding
– Turbo coding
– Trellis coding
Interleaving
• Interleaving
– Scatters error bursts
– Can be done in time or in
frequency domain
– Block Interleaving
• Write row-by-row
• Read column-by-column
• Additional matrix
permutation is possible
Block coding
• A way of mapping k symbols to n symbols with n>k
• We call k symbols a messageword,
• and the block of n symbols a codeword.
• Mapping k message symbols to n code symbols is
called encoding,
• and the reverse process is called decoding.
Systematic if messageword is
directly recovered by removing the
parity bit.

Linear if the sum of two codewords

is always a valid codeword, and a
scalar multiple of any codeword is
also a valid codeword.
Cyclic Redundancy Check
• Error detection
– CRC-n can detect errors less than n with a
probability of 1 – 2-n
Error Correction
• Hamming weight
– Number of nonzero symbols in codeword
• Codeword is 122200001 then Hamming weight is
5.
• Hamming distance
– Comparison metric between two codewords
by the number of places where the codewords
differ
• Codewords 100111 and 110000, the Hamming
distance is 4.
• Minimum distance of a code
For a minimum distance d,
– Lowest Hamming distance in linear codewords
we can correct up to
distance floor ((d-1)/2)
Reed-Solomon Coding
• Popular
coding
scheme of
today
• Cyclic code
and linear
code.

• Cyclic
codewords
are still a
codeword
when
shifted.
Reed-Solomon Coding
• The best minimum distance obtainable codes.
• They are linear, cyclic and their generator polynomial
This is a very popular value
has well-defined roots (2t). because of the prevalence
of byte-oriented computer
• Easy to encode and decode systems.

• A popular Reed-Solomon code is RS(255,223) with 8-

bit symbols where n=255 and k=223 and s=8 and
2t=32. The decoder can correct any 16 symbol errors
in the codeword.
Reed-Solomon code
• The Reed–Solomon code, like the convolutional code, is a transparent
code.

• This means that if the channel symbols have been inverted somewhere
along the line, the decoders will still operate.

• The result will be the inversion of the original data. However, the Reed–
Solomon code loses its transparency when the code is shortened.

• The "missing" bits in a shortened code need to be filled by either zeros or

ones, depending on whether the data is complemented or not.

• (To put it another way, if the symbols are inverted, then the zero-fill needs
to be inverted to a one-fill.)

• For this reason it is mandatory that the sense of the data (i.e., true or
complemented) be resolved before Reed–Solomon decoding.
Convolutional Coding
• Operates on serial streams rather than blocks.

• Described by two parameters: the code rate (k/n) and

constraint length (K) – length of the convolutional
encoder which denotes the number of stages and the
cycles an input bit retains in the convolutional encoder.

• Viterbi decoding or sequential decoding are used.

• Viterbi decoding has fixed decoding time – it is preferable
in hw implementation, allows soft decision decoding.
Convolutional coding
• Encoder
– K=3 (7,5), code rate is ½ and m=2.
– (7,5) is code generator polynomials
Convolutional coding
Viterbi Decoder
• Viterbi decoder decides about the original bits
according to accumulated error metric. At time
instant t, the smallest accumulated error metric is
selected according to the history of what states
preceded the states. This method is called
traceback method.
Viterbi decoding
• For each branch, there is accumulated error
metric and minimum is selected. If accumulated
error metric is equal, then the decoder decides by
looking forward. For instance, at time t =3 and
t=12, there are errors, and the decoder can end up
with the same accumulated error.
Concatenated coding
Concatenated coding
• The inner convolutional code performs superior error
correction with soft decision decoding, and if
convolutional code makes an error, it causes a large
burst, since Viterbi algorithm may pick a wrong
sequence. In this case, we know that block coding,
especially an interleaved Reed-Solomon coding, is
superior correcting the bursty errors.
Concatenated coding
provides means to
constructing long
codes, and it also con-
firms Shannon’s
channel coding
theorem by stating
that if the code is long
enough, any error can
be corrected.
Trellis Coding
• Combination of coding and modulation.
• Trellis coding adds redundant constellation points rather than
redundant bits or symbols. Consequently, bit rate increases but
the symbol rate stays the same and it conserves bandwidth.
Increasing the constellation size reduces Euclidean distance
between the constellation points, but sequence coding offers a
coding gain that over- comes the power disadvantage of going to
higher constellation.
Trellis coding
Adds one extra bit and expands the constellation without
increasing the signal energy. The signal energy is kept the
same, since the distance between the symbols decreases.
Although it sounds like a disadvantage, advantage comes
from the restriction on what transitions are allowed in the
constellation.
Turbo coding RSC is (1,G2/G1) but
Convolutional codes
• A typical turbo are (G1,G2)
coding11
includes parallel
concatenated
convolutional
codes where the
information bits
are coded by
two or more
recursive
systematic
convolutional
(RSC) codes,
which are
typically
interleaved and
optionally
punctured
Turbo coding
• Turbo codes increase data rate
without increasing the power of
a transmission, or they can be
used to decrease the amount of
power used to transmit at a
certain data rate.
• Turbo coding shows high error
correction performance
because of its structure based
on interleaving in conjunction
with concatenated coding and
iterative decoding using
(almost) uncorrelated extrinsic
information.
• Turbo coding such as block
turbo coding and convolutional
turbo coding are included in
IEEE 802.16 as sup- ported
coding schemes.
LPDC Coding (Low-density parity
check)
• A (n,k) LDPC encoder would have an H matrix, which
is mxn in size where m=n-k.
LDPC decoding
• Hard decision decoding
Synchronization
• OFDM needs to employ time and frequency synchronization.
– Time synchronization is to decide for the symbol boundaries. Commonly, a
sequence of known symbols preamble are used to detect the symbol
boundaries.
– It has less sensitivity to timing offset as compared with singlecarrier
systems since timing offset does not violate the orthogonality of
subcarriers in OFDM system, but causes ISI in single-carrier systems.
– Unlike time synchronization, frequency synchronization, which is to
estimate the frequency offset in the oscillators in order to align the
oscillators in the transmitter and receiver, is essential otherwise ICI occurs,
since subcarriers could be shifted from its original position and the
receiver may experience non-orthogonal signals.
Synchronization
• Timing Offset
– OFDM is insensitive to timing offset as long
as offset is within the guard time.
• No ISI and ICI is guaranteed.

• Frequency Offset
– OFDM is sensitive to the frequency offset
since it causes ICI.
Pilot-assisted Time/Frequency
Synchronization
• Pilots are transmitted – known at the receiver.
– Symbol timing and carrier frequency offset
can be estimated at the receiver.
– Frame Synchronization by preamble- known
pilot symbols at the beginning of the frame.

There is also blind time-frequency synchronization: It is

pilotless and based on maximum likelihood estimation.
The parameters that need to be estimated requires longer
observation.
Detection and Channel Estimation
• To estimate the transmitted bits at the receiver,
channel knowledge is required in addition to the
estimates of random phase shift and amplitude
change, caused by carrier frequency offset and
timing offset.
Coherent Detection
• Channel
variation in time
and frequency
are used to
determine the
minimum pilot
spacing in time
and frequency.
Pilot spacing in time

Pilot spacing in frequency

 The least-squares (LS) channel
Channel Estimation estimation minimizes ||y-Xh||

^
−1
y = Xh + n h LS = X y
^ ^
h MMSE = A h LS
−1
A=R ^ R ^
h h LS
€ Linear minimum mean squared
h h LS

error estimate (LMMSE)

Equalization
• Used to combat for intersymbol interference (ISI)
Equalization in Single Carrier
Complexity of equalizer in single-carrier system increases with data rates, since as
data rates increase, the receiver needs to get more frequent samples to
compensate for the delay spread and consequently sample clock (t) decreases.
These in- crease the number of delay taps in the equalizer and makes it almost
impossible to meet rates above 100 Mbps.
MMSE Equalizer
Peak-to-Average Power Ratio and
Clipping
• One of the significant drawback of OFDM system is
the possibility to experience large peaks since the
signal shows a random variable characteristic since it
is sum of N independent complex random variables.
• These different carriers may all line up in phase at
some instant and consequently produce a high peak,
which is quantified by peak-to-average-power ratio
(PAPR).
Clipping
Clipping
Clipping
Application: IEEE 802.11a
Application: IEEE 802.11a
References
• Mobile Broadband by M. Ergen
• OFDM by K. M. Shazad