Multicarrier transmission systems

The basic idea of multi-carrier modulation is quite simple and follows nat-
urally from competing desire for high data rates and inter symbol interference
(ISI) free transmission. In other words a high bit rate data stream is demulti-
plexed into N parallel substreams and thus the bitrate (bit duration) of individ-
ual substream increases substantially. In order to eliminate the effects of ISI,
symbol duration Ts needs to be longer than the channel delay spread τmax .

Xn (f )

X0 (f ) f

Figure 1: orthogonal and non-orthogonal frequency division multiplexing.

The received baseband equivalent of the OFDM modulated multi-carrier signal

is defined as
X NX
−1
x(t) = ak [n]gk (t − nTs ) (1)
n k=0

where,
n is the time index
k is the subcarrier index
N is the numer of carriers
Ts = N · T is the strected symbol duration
ak [n] is the n-th data symbol on k-th subcarrier
gk (t) is the k-th baseband pulse modulating k-th subcarrier

1
and here gk (t) is defined as
fk t
gk (t) = ej2π Ts · Π(t/Ts ) (2)
k = 0, · · · , N − 1

where the first term is the complex valued carrier with frequency fk = k/Ts .
The subcarrier separation is 1/Ts , the second term is a rectangular pulse. The
modulated signal for k-1, k, k+1 -th subcarrier is

t − nTs j2πfk−1 t/Ts

xk−1 = ak−1 Π( )e
Ts
t − nTs j2πfk t/Ts
xk = ak Π( )e
Ts
t − nTs j2πfk+1 t/Ts
xk+1 = ak+1 Π( )e
Ts
The subcarrier pulses are orthogonal
Ts
1
Z
gn (t) · gm
∗
(t)dt = δn−m , 0 ≤ n, m ≤ N − 1. (3)
Ts t=0

Due to orthogonality, there is no inter carrier interference. The data can be re-

x1 (t) x1 (t)

ej2πf1t e−j2πf1t
Parallel to Serial

x2 (t) x2 (t)
Serial to Parallel

QAM Symbols
ej2πf2t e−j2πf2t
Sk x(t) x(t) sk
0 ≤ t ≤ Ts

xN (t) xN (t)

ej2πfN t e−j2πfN t

Figure 2: Graphical illustration of multicarrier transmission systems.

constructed by means of matched filter without any performance loss compared

to single carrier Nyquist system. The received signal can be defined as
(n+1)Ts
1
Z
kt
yk [n] = r(t)e−j2π Ts dt (4)
Ts nTs

2
and the transmitted signal can be found from the modulated signal as
Z (n+1)Ts
1 kt
ak [n] = x(t)e−j2π Ts dt (5)
Ts nTs
The practical implementation of orthogonal frequency division scheme can be
conveniently through discrete Fourier transform and inverse discrete Fourier
transforms. The transform pair is defined as
N −1
1 X nk
x[n] = s[k]ej2π N (6)
N
k=0
N −1
nk
X
s[k] = x[n]e−j2π N (7)
n=0

where x[n] is the OFDM modulated and s[k] is the actual transmitted data
symbol on k-th subcarrier. The spectral representation of the OFDM signal is
illustrated below,
The time domain waveform of the transmitted signal is illustrated in the figure

∆F

0.6

0.4
Amplitude

0.2

−0.2
-8 -6 -4 -2 0 2 4 6 8
subcarrier index [k]

Figure 3: The spectral representation of the OFDM signal.

below. The actual transmitted signal x[k] is the sum of the modulated wave-
forms obtained from each subcarrier as in (??). Fig. ?? below illustrates an

3
 ℜ{x(t)} 2

−2

0 0.5TS Ts

Figure 4: Time domain waveform of OFDM modulated signal.

OFDM symbol spanning symbol duration Ts = N · T . From the fig.?? it is also

evident that modulated output looks like a random (complex) waveform which
depends on the input data sequence modulated on the subcarriers. It must also
be mentioned that there exists one-to-one mapping between input and output
signals. The peak amplitude of the waveform is significantly higher than the
average values and this is a serious short-comming of the OFDM systems.
The overall (end-to-end) transmission system looks like

FEC QAM S/P IFFT P/S

CP DAC
Encoder Modulator Convertor Convertor

uk sk xk x(t) x(t)

S/P IFFT P/S QAM FEC

ADC CP Convertor Convertor Demodulator Decoder

x(t) xk ŝk ûk

Figure 5: Block diagram for OFDM transmission and reception systems

Cyclic Prefix
A guard interval are included in the serial data stream of OFDM transmitter
so as to overcome the effect of ISI produced by the signal transmission over a

4
frequency selective channel. To take care of this problem, the OFDM symbol is
cyclically extended to create a gaurd interval. Specifically the cyclic extension
of an OFDM symbol is the periodic extension of DFT output as shown by

x(−k) = x(N − k) for k = 1, · · · , Ncp (8)

The graphical illustration of (??) is illustrated in the fig. ??.

111
000
000
111 xi 11
00
000
111
00
11
000
111 xi+1 11
00
00
11
xi [N−Ncp −1] xi+1 [N−Ncp −1]
xi [0] xi [N − 1]xi+1 [0] xi+1 [N−1]

Figure 6: Graphical illustration of guard band insertion process.

The minimal length of CP should be Ncp ≥ Lh , this condition ensures that

linear convolution is transformed into a circulant convolution. It must be said
that transmission of redundant CP causes a certain drop in throughput rate as
Ncp +N samples are transmitted instead of N , but this loss is a small price to
pay for simplicity of implementation and equalization process.
The matrix representation of the transmitted vector x in the presence of a
frequency selective channel and additive noise is defined as
     
y[N −Ncp −1] x[N −Ncp −1] n[N −Ncp −1]
 
h[0]
.. .. ..
     
      

 .  h[1] h[0]
  
 .  
  . 

 y[N −1]  h[2] h[1] h[0]   x[N −1]   n[N −1] 
      
      
y[0]  h[3] h[2] h[1] h[0] . . . x[0] n[0]
      
    
      
y[1] . x[1] n[1]
h[3] h[2] h[1] . .
      

=
  h[0] 
+
  
y[2] x[2] n[2]
  
.
      
.. h[3] h[2] . . h[1] h[0]  .. ..
      
     
 .  
.
 .   . 
h[3] . .
      
 ..   h[2] h[1] 
  ..   .. 

 . 
 

..  .  
  . 


 y[N −2]  
  . h[3] h[2]  x[N −2]   n[N −2] 
   

     
y[N −1] h[3] x[N −1] n[N −1]
(9)

5
The process of adding cyclic prefix basically transform the diagonal channel
matrix H into a circulant matrix Hc as

     
y[0]   x[0] n[0]
  h[0] h[3] h[2] h[1]    
 y[1]     x[1]   n[1] 
  h[1] h[0] h[3] h[2]    
       
 y[2]     x[2]   n[2] 
  h[2] h[1] h[0] h[3]    
 ..     ..   .. 
 . = .. · . +
  . 
  h[3] h[2] h[1] h[0] .   
 ..     ..   .. 

.
  ..  
.
 
.



 
  h[3] h[2] h[1] . h[0]
 
 
 
 


y[N − 2]   x[N − 2] n[N − 2]
..
h[3] h[2] . h[1] h[0]
     
y[N − 1] x[N − 1] n[N − 1]
(10)

Plainly speaking, insertion of cylic prefix at the start of every OFDM symbol
converts the linear convolution between transmitted data and channel impulse
response into circulant convolution.
From the study of linear algebra we know that any circulant matrix can be
transformed into diagonal matrix by left and right multiplication with (I)DFT
matrices. These (I)DFT operations are performed at transmitter and receiver
and the channel between them is converted into a circulant channel due to in-
sertion of cyclic prefix.

 
  h[0] h[1] h[2] h[1]  
1 1 ··· 1
  1 1 ··· 1
 h[1] h[0] h[3] h[2] 
 
 
1
 WN
−2
··· WN−N  
 h[2] h[1]
 1
h[3] 
2
WN ··· N
WN 

   
1 WN
−4
··· WN
−2N  
..  1 4
WN ··· WN2N 

. .. .. 
 h[3] h[2] . 
 . .. .. 

 ..   ..

. ··· .   .. . ··· . 

2

 h[3] . h[0]



2
1 WN
−N
··· WN
−N 
..
 1 N
WN ··· N
WN
. h[1] h[0]
 
H[0] ···
 

 H[1] ··· 

 .. 
=
 . 
 (11)
 . .. .. .. 
 . .
 . . .


H[N −1]

6
 nk
where WNnk = ej2π N . Writing the above expression in compact matrix notation

y = F Hc FH s + n


(12)
y = Λh s + ñ (13)

where Λh is the diagonalized channel frequency response due to right and left
multiplication with (I)DFT matrices. The zero forcing equalization of the diag-
onalized matrix can be calculated easily since calculation of the inverse matrix
for a diagonal is very low complexity.

Λh ΛH ΛH H
ΛH

−1
−1
h h y = Λh Λh h Λh s + ñ

ŝ = Λh ΛH ΛH

−1
h h y (14)

The consequence of diagonalization process are illustrated in the fig. ??.

H (f ) H[N−1]
H[0]

0 N-1

Figure 7: Diagonalization of circulant channel matrix translates frequency se-

lective channel into frequency flat channel frequency response.

Limitations/Problems with OFDM systems

OFDM systems despite of its salient features such as ease of implementation,
high bandwidth efficiency, it has found application in several industrial stan-
dards such as ADSL, DAB/DVB , WiFi, WiMax and LTE. But OFDM does
suffer from the following problems

• Peak to average power ratio.

• Spectral leakage.

• Carrier frequency offset.

7
 • Direct current offset.

• Phase noise.

• IQ imbalance.

It is beyond the scope of this lecture to study the effects of the above mentioned
artifacts.
Applications of Orthogonal Frequency Division Multiplexing
OFDM has become and integral part of several transmission system owing to
its simple implementation and low complexity equalization. Just some of the
techniques are listed here

• Multi-input Multi-output OFDM.

• Multi-carrier code division multiple access.

• Alamouti scheme based on OFDM.

Orthogonal Frequency Division Multiple Access

Orthogonal frequency division multiplexing can be used as means of multiple
access. Different subcarriers can be assigned to different users who can com-
municate with base station on orthogonal subcarriers as long as synchroniza-
tion/timing constraints are satisfied. The graphical illustration of this process is
illustrated in the fig.??. This scheme is widely adopted in industrial standards
such as WiMax and LTE systems.

8
 user 1

Base Station
user 2

user 3

Figure 8: Orthogonal frequency division multiple access scheme scenario.