Application of Orthogonal Frequency Division Multiplexing

with Concatenated Codes for Wireless Broadband

Communications

What is presented:

• theoretical and simulated bit error rate performances of concatenated coding schemes with
orthogonal frequency division multiplexing (OFDM) in multipath fading channels.

• Outer Reed Solomon code is concatenated with different trellis based inner codes in order to achieve a
high coding gain. Trellis coded modulation (TCM), turbo code with multilevel mapping (TC), turbo
trellis coded modulation (T-TCM) and convolutional codes (CC) are used as inner codes.

• A 512-carrier OFDM system is selected to facilitate the transmission of bit rates up to 155 Mb/s,
while the high gain concatenated scheme is used to obtain a remarkable bit error rate performance.

• An interleaver is introduced between RS outer coder and trellis based inner coder to randomize the
burst errors from the inner decoder. A channel interleaver is also introduced to combat the correlated
fading of the channel.

• Theoretical upper bound of the bit error rate is obtained for comparison. Windowing is used to reduce
the high peak-to-average power ratio of the OFDM signal. Bit error rate performance with the
windowing is also presented.

I. INTRODUCTION
• In the next few years, increasing attention to wireless broadband services is expected due to
the benefits, primarily mobility and installation flexibility. More generally the demand of
wide-band multimedia services arising from wireless users has stimulated particular interest
in the design of high quality, high speed (10 Mbit/s and up to 155 Mbit/s) wireless networks,
which are able to provide access services and capabilities of broad-band fixed networks.

1/33
• It is expected that most future broadband networks will be based on asynchronous transfer
mode (ATM). ATM transport is an efficient transmission and multiplexing platform for
integrating a variety of fixed and wireless services.

• Third and perhaps second generation wireless access will be integrated with ATM based
backbone transport and intelligent networks capable of providing many wireless services.
This is because ATM can provide flexible bandwidth assignments. A realistic way to convert
conventional synchronous transfer mode (STM) based networks in to ATM networks with
wireless multimedia handling capabilities is shown in [1].

• Wireless ATM can satisfy the need of delivering bandwidth on demand to portable terminals.
A reliable radio link capable of handling data rates of tens of Mb/s is the critical component
to meet this goal. Coded orthogonal frequency division multiplexing is promising solution to
this.

• The most popular application of OFDM is for digital audio broadcasting (DAB) [3, 4].
Outside broadcast link, asynchronous digital subscriber lines, digital video broadcasting are
potential applications [5, 6].

• In 1998, the IEEE 802.11 standardization group decided to select OFDM as the basis for their
new 5GHz standard, targeting a range of data rates from 6 up to 54 Mb/s. This new standard
is the first to use OFDM in packet-based communications, while all above-mentioned
applications are continuous transmission systems.

• Following the IEEE 802.11 decision, high performance LNA (HIPERLAN) type 2 and
multimedia mobile access communication (MMAC) also adopted OFDM for their physical
layer standards [8].

• There are several studies done in the area of OFDM for wireless broad band applications. A
modem, which can transmit high rate ATM cells over an indoor radio channel, was discussed
in [9]. A similar study was done in [10] to transmit variable rate ATM cells over high-speed
indoor wireless channels (100 to 155 Mb/s). Radio access to an ATM network is considered
in [11], in order to make all broadband services available to mobile stations.

2/33
• Especially, a time division duplex frame structure is analyzed and described considering
different multiple access techniques. The system operates at 2.5GHz and a channel
bandwidth of 8 MHz is assumed. The study concentrates on medium access control (MAC),
automatic repeat request (ARQ) and a hand-over protocol, but it does not consider the
transmission aspects of OFDM.

• Magic wireless ATM network demonstrator (WAND) and the MEDIAN projects demonstrate
prototypes wireless ATM networks operating at a raw data rate from 20 Mb/s up to 155 Mb/s.
Both projects will expand the services of the fixed B-ISDN using local wireless broadband
transmission. OFDM is for the transmission of B-ISDN originated ATM cells with
maximum efficiency. It envisages a transparent data flow from the fixed ATM network to
the mobile ATM capable multimedia terminal.

• Coded OFDM is essential for all the applications discussed above. Transmission over a
frequency-selective channel implies that some of the subcarriers are strongly attenuated and
cause errors even at high average signal power. In this flat fading situation an efficient
channel coding leads to a very high coding gain, especially if soft decision decoding is
applied.

• Here we are investigating the bit error rate performance of an OFDM system with some of
the powerful forward error correcting codes available today. Trellis coded modulation
(TCM), turbo coding (TC), turbo trellis coded modulation (T-TCM) and convolutional code
(CC) concatenated with Reed Solomon (RS) code are investigated.

• Turbo codes, the newest code in this family of codes has a very high coding gain [12]. Turbo
codes result from concatenation of two recursive systematic convolutional codes.

• Combination of turbo codes with TCM gives a bandwidth efficient coding scheme termed as
turbo trellis coded modulation (T-TCM) [13, 14]. T-TCM is an extension of binary turbo
codes, where recursive systematic convolutional codes are replaced by recursive systematic
TCM codes. Application of T-TCM is preferred where high spectral efficiency and good bit
error rate performance is expected [15].

3/33
II. SYSTEM DESCRIPTION

• In OFDM, a block of N symbols { X n , n = 0,1,..., N − 1} is formed with each symbol

modulating one of a set of N sub-carriers, { fn , n = 0,1,..., N − 1} . The N sub-carriers are

chosen to be orthogonal, that is fn = n∆f , where ∆f = 1/ NT and T the is original symbol

period. The resulting signal in the discrete time domain can be expressed as

N− 1
xk = ∑ X n e j 2πnk / N (1)
n= 0

where k =0,1,… N-1, is the time index. A cyclic prefix (called guard interval) is added to the
resulting signal in order to avoid the intersymbol interference (ISI), which occurs in multipath
channels. At the receiver the guard interval is removed and only the time interval [0, T] is
evaluated. The guard interval is usually a periodic extension of the symbol over the interval [-
TCP, 0], resulting a symbol of length [-TCP, T].

A. Peak to average power ratio (PAPR)

The main disadvantage of OFDM is the large PAPR, which is defined as
2
max x(t )
PAPR = 2
(2)
E[ x (t ) ]

where x(t) is the digital to analog converted signal of xk given in (1) and E[x] is the expected
value of x. When N signals are added with the same phase, they produce a peak power which is N
times the average power. Therefore maximum theoretical PAPR of an OFDM signal is N.
Several schemes have been proposed to reduce this undesirable high PAPR of the OFDM signal
[16-19].

The block diagram of the system is shown in Figure 1. Here µ is the modulation index (number
of bits per symbol) of the inner encoder. Therefore input to the inner encoder is µ parallel bits.

4/33
The ATM cell-mapping scheme described above is valid for the selected encoder structures,
which depends on overall code rate and the signal-mapping scheme used. For different encoder
structures a suitable mapping scheme can easily be found in a similar manner.

• We have selected the ATM as the transport technique. The outer RS coder receives the
unprotected ATM cells at the transmitter. Four ATM cells are mapped to a single block of
data for the RS coder input. The total length of the data word is then 212 bytes. The output
code word length is then 244 bytes after being coded by RS (244,212) encoder.

• The output from the RS coder is divided into two data blocks. Four trailing bits (zeros) are
appended to each block for trellis termination. The inner encoder maps these two blocks of
data into 8PSK signal points when TCM and T-TCM inner codes are used and into 16QAM
signal points when turbo code and convolutional code are used. Specifications of the inner
codes used are given below.

1. Ungerboeck rate 2/3, 8 states TCM code with 8PSK mapping – Viterbi decoding (TCM1).
The encoder is shown in Figure 2.

2. Ungerboeck rate 2/3, 16 states TCM code with 8PSK mapping - Viterbi decoding (TCM2).
The encoder is shown in Figure 3

3. T-TCM with Ungerboeck rate 2/3 8 states component code with 50% puncturing and 8PSK
mapping, - Iterative decoding with max log MAP algorithm (T-TCM). The encoder
polynomial of the component TCM codes is given by
1 0 0 1 
G3×4 = 0 0 1 1  (3)
 

0 1 0 0 

In this recursive systematic code a parity check bit is appended to every two information bits
and the resulting triple bits are used to select one of 8PSK symbols according to
Ungerboeck's rules. There is no uncoded bit in these RSC codes and hence no parallel
transitions in T-TCM. T-TCM encode is shown in Figure 4

5/33
4. Turbo code with rate ½, 8 states convolutional component codes with 50% puncturing of
parity bits and 16QAM mapping - Iterative decoding with max log MAP algorithm (TC). The
generator polynomial of the component recursive systematic convolutional encoder is given
by
1 1 0 1
G2×4 =   (4)
1 1 1 1 
5. Convolutional code with 16QAM mapping – Viterbi decoding (CC). The industry standard
constraint length 7, rate ½ convolutional code is used. The convolutional encoder having
generator polynomials g1=1338 and g2=1718.

• This complex baseband signal is then multiplexed in to 512 orthogonal carriers using the
inverse fast Fourier transform (IFFT) processor. The block interleaver used between RS
coder and inner encoder converts long error bursts of the inner decoder output to short error
bursts with length of a few bytes. These short error bursts are then randomly distributed over
RS-code words, which can effectively be corrected by the RS decoder. For a channel with
correlated fading, a channel interleaver is introduced at the output of the IFFT processor. The
mapping of data (ATM cells) into OFDM symbols is shown in Figure 5.

• The choice of a carrier frequency for a radio system usually depends on its bandwidth and the
wave propagation characteristics of the selected frequency in the expected environment of
operation. For broadband systems, we have to consider both path loss and impulse response
of the channel for the selected frequency. 60GHz band is currently under investigation for
broadband wireless applications in both indoor and outdoor environments as it is having
desirable properties anticipated [20]. Therefore, we have selected 60GHz frequency for the
simulations.

• The selection of suitable number of carriers for the OFDM system depends on several
parameters. Assuming enough bandwidth (94MHz) can be allocated in the 60GHz frequency
band, the suitable number of carriers are selected using the following relationship.

6/33
 RbTm
( µ N )min = (5)
δRc
where Rb is the data rate, Tm is the maximum delay spread of the channel, δis the proportion of
the guard interval to the OFDM symbol duration, Rs is the code rate and N is number of sub
carriers. System specifications are presented in Table 1.

Table 1 System specifications

Bit rate 155.52Mb/s
Number of carriers 512
Delay spread of the channel 250ns
Number of paths 6
Length of the guard interval 50 (10% of the total OFDM symbol
duration)
Speed (for correlated fading) 120km/s
Spectral efficiency 2bits/s/Hz for all schemes
Reed Solomon code RS(244,212)
Frequency 60GHz
Size of the interleaver between 10.9µs (4*244 byte interleaver)
RS code and inner code
Size of the channel interleaver 21.8µs (1024*8 symbol interleaver)
Total interleaver delay 65.4µs

III. FREQUENCY SELECTIVE FADING CHANNEL MODEL

• The discrete channel impulse response of the multipath channel [21] can be expressed as

M− 1
h(t ) = ∑ hme jφm δ(t − τm ) (6)
m =0

where, M is the number of paths, τm is the delay corresponding to the mth path, hm is Rayleigh

distributed fading amplitude and φm is uniformly distributed phase. It is assumed that the

7/33
channel is static during an OFDM symbol. Then the frequency response of the channel [22] can
be found as
M−1

m =0
(
H n = ∑ hm e jϕ m )e − j 2πnm / N
, n = 0,1,...,N-1 (7)

where N is the number of sub-carriers of the OFDM scheme. The average signal power and the
total multipath power can be normalized to one.

• The frequency response of the channel is thus obtained by taking the FFT of the fading
coefficients (as hm=0 for m>L). Statistical properties of AWGN is same in both time and
frequency domains. Therefore, received signal at the output of the FFT processor [22] is
obtained by

Rn = H n X n + Z n (8)

• Xn is the modulated data symbol Zn is the additive white Gaussian noise. In this case the noise
variance depends on the signal to noise ratio (Eb/N0), code rate (Rc) and the spectral efficiency
(η) of the modulation scheme.
• This is given by
1
σn2 = (9)
2η Rc ( Eb / N o )
Figure 6 shows the equivalent channel model.

8/33
 IV. CODING SCHEMES AND DECODING ALGORITHMS

A. Trellis coded modulation (TCM)

• TCM schemes employ redundant non-binary modulation in combination with finite-state
encoder, which governs the selection of modulation signals to generate coded signals [23]. In
the receiver, a soft-decision maximum likelihood sequence decoder decodes the noisy signals.
Non systematic, eight-state encoder depicted in Figure 2 is used in the simulations. Zero bits
are appended to both data streams d0 and d1 for the trellis termination purpose. The set
partition of 8PSK constellation, which achieves higher euclidean distances, is shown in
Figure 7 [24]. Corresponding trellis diagram and the signal constellation is depicted in Figure
8.

• Viterbi decoding is done at the receiver. The branch metric of the Viterbi decoder for the
OFDM in multipath fading channel is the squared euclidean distance between the received
signal and the branch symbol multiplied by the fading coefficient. We have assumed that the
perfect channel state information is available at the receiver. The branch matrix ( Bnj )

corresponds to the received symbol at the branch can be expressed as

2
Bnj = Rn − H n snj (10)

Here Hn is given in (6) and snj is the branch symbol (8PSK) corresponds to the jth branch. Both

TCM schemes use the above decoding algorithm.

B. Binary turbo code (TC)

• Turbo code is constructed by concatenating two recursive systematic convolutional (RSC)

component codes through an interleaver. A random interleaver is used for the simulation. We
used a ½ rate 8-state RSCs as component codes, having a generator polynomials given by (3).
The code words of the above turbo code consist of the input bits (systematic) followed by the
parity check bits from the two encoders. Parity check bits of each RSC output are punctured

9/33
 alternatively increasing the code rate to ½. The states of trellises are forced to the zero state at
the end of each block of information. This process is termed as trellis termination.

C. Turbo trellis coded modulation (T-TCM)

• T-TCM encoder is similar to the turbo encoder structure. But in T-TCM encoder RSCs are
replaced by recursive systematic TCM encoders. In this case, the output from the RSC TCM
encoders is 8PSK symbols. Random interleaver is again used but this time pair wise
interleaving is done. Puncturing the every other symbol from the two encoders alternatively
forms output symbol sequence.

• SISO decoder using max log MAP algorithm is used for T-TCM decoding [14]. The
logarithm of the transition probability corresponding to the transition from state σn to state σn-
1 of the trellis can be found as

2
Rn − H n snj
log[γσ
( n , σn − 1 )] = − log(2πσ ) −2
+ log( Pa ) (11)
2σn2
n

Pa = p (σn / σn − 1 ) (12)

where Rn is the received symbol, Pa is the a priori information, snj is the branch symbol

corresponds to the branch connecting states σn to σn-1 and Hn is given by (6). The rest of the
decoding process is done as described in [14].

D. Extracting soft values for convolutional code and turbo code decoding

• The 16QAM constellation with Gray mapping is used in both turbo code and convolutional
code. The a posteriori probability of transmitted symbol at interval n can be expressed as

p( Rn | snj ) p ( snj )
p( snj | Rn ) = (13)
p (rn )
10/3
3
where snj is the symbols of 16QAM constellation corresponding to the jth branch to the trellis.

p( Rn / snj ) is the conditional probability observed when signal snj is given. p( snj ) is the a priori

probability of the jth signal being transmitted at interval n. This is usually a constant for each
symbol being transmitted. The probability of receiving symbol Rn can be expressed as

M−1
p( Rn ) = ∑ p( Rn | snj ) p( snj ) (14)
j =0

If un,i is the ith bit of the nth received symbol then:

p(un,i = 0 ) = ∑
un,i = 0
p ( snj | Rn )

∑
un,i =0
p( Rn | snj ) p( snj ) (15)
= M− 1

∑
j =0
p ( Rn | snj ) p (snj )

If the a priori probability p( snj ) is the same for all symbols, (14) can be simplified as follows:

p(un,i = 0 ) = ∑
un,i = 0
p ( Rn | snj ) (16)

If σN ,is the variance of the noise, p( Rn / snj ) in (15) can be expressed as

1  | R − s j |2 
p( Rn | snj ) = exp − n 2 n  (17)
2πσN  2σ N 

In the logarithmic domain the probability in (15) can be expressed approximately as

L(un ,i = 0) = max[ln( p ( Rn | snj ))]

un ,i = 0

 R sj 2
 (18)
= max − 
n− n

un ,i = 0  2σ N2 
 
11/3
3
where ‘ln’ is the natural logarithm. Then L(un,I=0) can be found simply by taking the minimum
value of the Euclidean distances calculated between the received signal and the signals of the
16QAM constellation where ith bit equal to zero. Similarly L(un,I=1) is also found. For the
convolutional decoder (Soft input Viterbi) branch matrices are calculated as shown in the Table
2.

Table 2. Calculation of the branch metric for soft input Viterbi decoding

Bits at the branch (n) Branch metric (n)

n(1) n(2)
0 0 L(un,I=0) + L(un,I=0)
0 1 L(un,I=0) + L(un,I=1)
1 0 L(un,I=1) + L(un,I=0)
1 1 L(un,I=1) + L(un,I=1)

For turbo decoder (max log MAP algorithm) the logarithm of likelihood ratio (LLR) associated
with each bit has to be calculated. The LLR of the ith bit in the nth interval [25] can be expressed
as
 p(un ,i = 0 | Rn ) 
LLR(un ,i ) = K log 
 p (u = 1| R )   (19)
 n ,i n 

where K is a constant. Based on this, the result obtained in (17) and for K=1, we can find the LLR
of the ith bit as

LLR(un ,i ) = L(un ,i = 0) − L(un ,i = 1) (20)

The transition probability corresponding to the transition from state σn to state σn-1 can then be
found using the LLR values obtained in (19).

( n , σn− 1 ) = LLR(systematic) + LLR(parity)

γσ (21)

12/3
3
• The investigated coding schemes possess characteristics such as high spectral efficient
modulation and high coding gain required by broadband wireless communication systems.

• The concatenated schemes RS/TC and RS/T-TCM are capable of achieving good bit error
rate performances in all the channels considered. Both concatenated schemes have about 2dB
gain compared to other coding schemes. But the receiver complexities of these two codes are
higher in terms of number of operations performed at the decoder. Therefore complexity of
decoding has to be considered when these coding schemes are compared with each other.

E. Synchronization and sensitivity to frequency offset errors

There are two undesirable effects caused by frequency offset. First one is the reduction of signal
amplitude in the output of the filters matched to each of the carriers. Second one is the
introduction of inter channel interference (ICI) from the other carriers which are now no longer
orthogonal. This is due to the fact, in OFDM the carriers are closely spaced in frequency
compared with the channel bandwidth. Maintaining sufficient accuracy in frequency is difficult
in mobile radio links, which also introduces significant Doppler shift [26].

The OFDM symbol representation with a frequency-offset yk is given by (22). It consists of N

complex sinusoids, which have been modulated with N complex modulation values Xn [26].

N− 1
1
yk =
N
∑
n=0
X n H n e j 2πn ( n + ε )/ N + wn (22)

where Hn is given in (6), ε is the relative frequency offset of the channel (the ratio of the actual
frequency offset to the intercarrier spacing), and wn is the complex envelope of the AWGN. It is
assumed that the impulse response of the channel does not change during the symbol interval.
The demodulation process, which is implemented with a FFT, consists of three components. The
nth element [26] at the demodulator output Yn is given by

Yn = ( X n H n ){(sin πε ) / N sin(πε / N )}.e jπε ( N − 1) / N + I n + Wn (23)

13/3
3
The first component is the modulation value Xn modified by the channel transfer function. This
component experiences an amplitude reduction and phase shift due to the frequency offset. Since
N is always much greater than πε , Nsin(πε/N) can be replaced by . The second term In is the
interchannel interference (ICI) caused by the frequency offset and is given by

N− 1
I n = ∑ ( X l H l ){(sin πε ) /( N sin(π(l − nε ) / N ))}.e jπε ( N − 1)/ N .e − jπ( l − n )/ N
. (24)
l =0
l≠n

The relationships for Yn and In can be used to simulate the effect of frequency offset of the carrier
on the average bit error rate.

V. SIMULATION RESULTS
• Performances of the concatenated schemes were investigated in three different multipath
channels. Power delay profiles of these theoretical channel models [9] can be expressed as

 m −3
(1 + 3.64 M ) 0 ≤m ≤M power 3 decline

hm = exp(− 4.6 )
m
0 ≤m ≤ M exp onential
 M

 1 0 ≤m ≤ M uniform


where M is the total number of delayed paths. Corresponding channel models are depicted in
Figure 9.

• The bit error rate performances of the coded OFDM systems in these multipath channels are
depicted in Figure 10-Figure 12. For these results fading of the channels is considered to be
independent. Figure 10 depicts the bit error rate performance of five concatenated coding
schemes in a channel with a uniform power delay profile, having six delayed paths. It is clear
that the lowest possible bit error rate is obtained using the T-TCM and TC concatenated

14/3
3
 scheme. The performances of T-TCM and TC are obtained after six iterations, which increase
the complexity of these schemes compared to the others (Table 3).

• T-TCM and TC schemes are capable of achieving a bit error rate less than 10-5 when the
Eb/No is 6dB. They further achieves coding gain of about 1.5dB at bit error rate 10-5
compared to other coding schemes. Similar bit error rate performances were observed in
other two channels as shown in Figure 11 and Figure 12.

• T-TCM and TC schemes have lower bit error rates compared to TCM1, TCM2 and CC
schemes. We note that the bit error rate performance of TCM2 is better than that of TCM1. It
is well known that the bit error rate performance becomes better when the number of states of
a TCM scheme is increased [23]. This is true for TCM schemes used in OFDM systems also.
Bit error rate performance of CC scheme is close to the performance of TCM2 scheme.

• CC scheme with 64 states is a popular coding scheme in wireless communications. The

forward error correction codes used in IEEE 802.11 and HIPERLAN are based on this. As is
seen, the bit error rate performance of both T-TCM and TC schemes are better than the
performance of the industry standard CC code.

• 16QAM modulation is used with both TC and CC schemes so that the spectral efficiency is
2bits/s/Hz. This spectral efficiency is same as those used in other three coding schemes
allowing a fair comparison of bit error rate performances. Next section presents the bit error
rate performances observed in channels with correlated fading.

A. Bit error rate performance in a channel with correlated fading

• The simulation results given in Figure 13-Figure 15 were obtained under the assumption that
the fading on the signal is correlated. In general mobile environment is frequency selective.
In this case the correlation in phase and amplitude between two signals separated in
frequency, is high for small frequency separation and falls essentially to zero as the
separation substantially exceeds the correlation bandwidth.
• A channel interleaver is to be used in order to break the correlation of the signal. The

15/3
3
 proposed channel interleaver is a 1024 8, symbol-block interleaver. The span and the depth of
the interleaver is selected according to the following rules.

KL
SP >
REN
LI ×SP > t (r )

Where SP and LI are the span and the depth of the interleaver, KL and REN are the constraint
length of the trellis and the code rate of the encoder. t(r) is the average fade duration.

• It is assumed that a fade occurs when the normalized signal power falls below -3dB. Then the
average fade duration is 82µs, which is equal to 15 OFDM symbols. This resembles a slowly
varying channel compared to the OFDM symbol duration. Therefore the channel can
effectively be considered as static during an OFDM signal. For the TCM1 scheme the code
rate of the encoder is 2/3 and the constraint length of the system is four. Therefore the span
must be greater than six.

• Figure 13 depicts the bit error rate performances of coded OFDM schemes in a channel with
uniform power delay profile and correlated fading. In this case also T-TCM and TC schemes
perform better than other three schemes.

• It can be seen that both TC and T-TCM schemes have similar performances even in a fading
channel. TC scheme has a coding gain greater than 2dB at bit error rate of 10-5 compared to
TCM1, TCM2 and CC schemes. Both T-TCM and TC schemes are capable of achieving a bit
error rate of 10-5 with an average Eb/N0 of 8dB.

• This is a degradation of about 2dB compared to the bit error rate performance in a fully
interleaved channel shown in Figure 10. Correlation of the fading in the channel degrades the
bit error rate performance of the system.

• The use of interleaving reduces the degradation significantly. Although the channel
interleaver used in our simulations is capable of achieving significant improvement, further
improvement of the bit error rate performance is expected using a proper interleaving.
16/3
3
• Figure 14 and Figure 15 depict the performance of coding schemes in a correlated fading
channels with exponential power delay profile and power three decline power delay profile
respectively. There again both T-TCM and TC schemes show better bit error rate
performances.

• Maximum frequency offset expected in the system is equal to the maximum possible Doppler
shift (fc=6.67 kHz). The relative frequency offset is then calculated with respect to the carrier
separation. Carrier separation of the system considered is 183.5kHz. Therefore corresponding
relative frequency offset is 0.036933.

• Figure 16 shows the bit error rate performance of the coding schemes in a channel with
frequency offset. A channel with uniform power delay profile and a frequency offset is used
for the simulation. As is seen, the degradation due to this frequency offset is negligible for all
coding schemes considered.

• There is a 0.75dB degradation of the performance of T-TCM and 0.5dB degradation for TC
scheme, while degradation in other schemes are much less. But with the increase in the
frequency offset the performance degrades rapidly.

• For a frequency offset of 2fc there is a degradation of 3 dB at a bit error rate of 10-4 for both
T-TCM and TC schemes, while TCM scheme degrades by 2dB under that same conditions. It
is interesting to note that the CC scheme does not degrade much compared to the other
schemes, when the frequency offset increases. Degradation of CC scheme is only about
0.3dB at bit error rate of 10-4 for frequency offset of fc and it is about 1dB when the frequency
offset is 2fc.

• Figure 17 depicts the complementary cumulative density function (CCDF) of the PAPR
(Pr(PAPR<PAPRo)) of a 512 carrier OFDM system. CCDF is the probability that the PAPR
is less than a given threshold. It is clear that not all code-words of an OFDM signal result in a
high PAPR. High peaks of the transmitted signal occur with very low probability [19]. From
the figure it can be seen that the high peaks causing PAPR greater than 11.5dB occur with a
probability less than 10-3.
17/3
3
• Therefore clipping or windowing the high peak would not have considerable effect on bit
error rate performance. But, If we deliberately clip the high peaks it will cause another
undesirable effect, spectrum splatter. For this reason peak windowing is proposed in [19]. As
it is already shown there, a window with good spectral properties has to be used. The window
should be as narrowband as possible to reduce the effect of windowing on the other samples.
We used a Hamming window of length 10 in our simulations. The corresponding windowing
function is given by
 A
1 − 1 − W (n) 0 ≤n ≤10 (25)
 X
where W(n), A and X are Hamming window, clip level and the amplitude of the corresponding
peak respectively.

• Bit error rate performances of TCM1 scheme, obtained with different clip levels are shown in
Figure 18. Curve 1 presents the bit error rate without peak windowing, while curves 2-5
present peak widowing at 6dB, 5dB, 3dB and 2dB respectively. As is seen, PAPR can be
restricted to 6dB without noticeable degradation in bit error rate.

• When the peak windowing is used at 3dB degradation of more than 1dB is observed.
Therefore it is clear that the peak windowing can be used to get the considerable peak power
reduction without much degradation of the bit error rate performances. The complexity of
using windowing to reduce PAPR is much less compared to the complexities of other
schemes [16, 17]. Therefore, we suggest peak windowing for the proposed system. By using
peak windowing complex PAPR reduction methods can be avoided.

VI. CONCLUSIONS
• The performances of orthogonal frequency division multiplexing with concatenated coding
schemes are investigated in fading channels. ATM is used as the transport mechanism and a
suitable mapping of ATM cells into OFDM symbols is proposed. A 512-carrier OFDM
18/3
3
 system with a 10% guard interval, which is longer than the maximum delay spread of the
channel is used.

• RS/T-TCM and RS/TC concatenated schemes have a high coding gain compared to other
coding schemes and both schemes outperform TCM/RS scheme and RS/CC scheme in a
frequency selective fading channels considered.

• When the complexity of the schemes are considered, TCM scheme has the lowest decoding
complexity, while CC has a moderate complexity. But both T-TCM and TC schemes have
higher complexities.

• If our objective is to find a suitable coding scheme with highest coding gain T-TCM would
be our choice as it has lower complexity than TC although both have similar bit error rate
performances. But, if we want to go for a scheme with a lowest complexity and having a
reasonable bit-error-rate performance, TCM1 scheme would be our choice.

• Industry standard CC code with soft decision decoding falls in between above two categories,
having a moderate decoding complexity and slightly better bit error rate performance
compared to TCM. The performance degradation due to possible frequency offset due to
Doppler effect is negligible in this system.

• But, it is observed that CC scheme has less sensitivity to frequency offset than other coding
schemes. A relatively simple scheme of windowing the high peaks is used to limit the PAPR
of the OFDM signal. PAPR can be reduced by about 5.5dB at 0.1% PAPR without noticeable
bit error rate degradation. It can be concluded that investigated concatenated schemes will be
potential candidates for the future broadband wireless networks. The performances of the
proposed schemes are expected to improve with the use of a suitable ARQ scheme.

19/3
3
 REFERENCES
[1] H. Nakamura, H. Tsuboya, M. Nakano, and M. Nakajima, "Applying ATM to mobile
infrastructure networks," IEEE Commun. Mag., vol. 36, no. 1, pp. 66-73, Jan. 1998.
[2] L. J. Cimini, "Analysis and Simulation of a Digital Mobile Channel Using Orthogonal Frequency
Division Multiplexing," IEEE Trans. Commun., vol. COM-33, no. 7, pp. 665-675, July 1985.
[3] M. Alard and R. Lassalle, "Principles of modulation and channel coding for digital broadcasting
for mobile receivers,", EBU Technical review, no 224, 1987.
[4] B. L. Floch, R. Halbert-Lassalle, and D. Castelain, "Digital sound broadcasting to mobile
receivers," IEEE Trans. Consum. Elect., vol. 35, no. 3, pp. 493-503, Aug. 1989.
[5] J. S. Chow, J. C. Tu, and J. M. Coiffi, "A discrete multi-tone transceiver system for HDSL
applications," IEEE J. Select. Areas Commun., vol. 9, no. 6, pp. 895-908, Aug. 1991.
[6] S. Moriyama, K. Tsuchida, M. Sasaki, and S. Yamazaki, "Digital Outside-Broadcasting-Link
Using OFDM Modulation Scheme," IEEE Trans. Broad., vol. 42, no. 3, pp. 266-277, Sept. 1996.
[7] W. A. C. Fernando and R. M. A. P. Rajatheva, "Performance of turbo and trellis coded OFDM for
LEO satellite channels in global mobile communications," in Proc. IEEE ICC, Atlanta, USA, pp.
56-60,1998.
[8] R. van-Nee and G. Awater, "New High-Rate Wireless LAN Standard," IEEE Trans. Commun.,
vol. 37, no. 12, pp. 82-88, Dec. 1999.
[9] A. Chini, M. S. El-Tanany, and S. A. Mahmoud, "Transmission of High Rate ATM Packets over
Indoor Radio channels," IEEE J. Select. Areas. in Comm., vol. 14, no. 3, pp. 469-476, Apr. 1996.
[10] D. Dardari and V. Tralli, "High-Speed Indoor Wireless COFDM Systems at 60GHz: Performance
and Design Criteria," in Proc. IEEE GLOBECOM, New York, NY, USA, pp. 1306-1311, 1997.
[11] H. Rohling and R. Gruneid, "Performance comparison of different multiple access schemes for
the downlink of an OFDM communication system," in Proc. VTC, New York, NY, USA., pp.
1365-1369,1997.
[12] C. Berrou, A. Glavieux, and P. Thitimajshima, "Near Shannon limit error-correcting coding and
decoding: Turbo-codes. 1," in Proc. ICC, Geneva., pp. 1064-1070, 1993.
[13] P. Robertson and T. Worz, "A novel bandwidth efficient coding scheme employing turbo codes,"
in Proc. ICC, pp. 962-967,1996.
[14] P. Robertson and T. Worz, "Bandwidth-Efficient Turbo Trellis-Coded Modulation Using
Punctured Component Codes," IEEE J. Select. Areas in Commun., vol. 16, no. 2, pp. 206-218,
1998.
[15] N. H. Ha and R. M. A. P. Rajatheva., "Performance of turbo trellis-coded modulation (T-TCM) on
frequency-selective Rayleigh fading channels," IEE Proc. Commun., vol. 146, no. 4, pp. 251-257,
Aug. 1999.
20/3
3
[16] S. H. Muller and J. B. Huber, "A Comparision of Peak Power Reduction Schemes for OFDM," in
Proc. IEEE GLOBECOM '97, New York, NY USA, pp. 1-5, 1997.
[17] J. A. Davis and J. Jedwab, "Peak-to-Mean Power Control in OFDM, Golay Complementary
Sequences and Reed-Muller codes," IEEE Trans. Commun., vol. 45, no. 7, pp. 2397-2417, Nov.
1999.
[18] D. Wulich and L. Goldfeld, "Reduction of peak factor in orthogonal multicarrier modulation by
amplitude limitting and coding," IEEE Trans. Commun., vol. 47, no. 1, pp. 18-21, Jan. 1999.
[19] R. van-Nee and A. D. Wild, "Reducing the Peak-to-Average Power Ratio of OFDM," in Proc.
IEEE VTC, pp. 2072-2076, 1998.
[20] L. M. Correia and R. Prasad, "An Overview of Wireless Broadband Communications," IEEE
Commun. Mag., vol. 35, no. 1, pp. 28-33, Jan. 1997.
[21] J. G. Proakis, Digital communications, 3rd ed. New York: McGraw-Hill, 1995.
[22] A. Chini, Y. Wu, M. El-Tanany, and S. Mahmoud, "Filtered Decision Feedback Channel
Estimation for OFDM Based DTV Terrestrial Broadcasting System," IEEE Trans. Broad., vol.
44, no. 1, pp. 2-11, Mar. 1998.
[23] G. Ungerboeck, "Channel coding with multilevel/phase signals," IEEE Trans.Info. Theory, vol.
IT-28, no. 1, pp. 55-67, Jan. 1982.
[24] S. B. Wicker, Error control systems for digital communications and storage: Prentice Hall, 1995.
[25] S. L. Goff, A. Glavieux, and C. Berrou, "Turbo-codes and high spectral efficiency modulation," in
Proc. ICC, New Orleans, pp. 645-649,1996.
[26] P. Moose, "A Technique for Orthogonal Frequency Division Multiplexing Frequency Offset
Correction," IEEE Trans. Commun., vol. 42, no. 10, pp. 2908-2914, Oct. 1994.
[27] E. Biglieri, Introduction to trellis-coded modulation, with applications. New York: Macmillan,
1991.
[28] B. Vucetic, "Bandwidth efficient concatenated coding schemes for fading channels," IEEE Trans.
Commun., vol. 41, no. 1, pp. 50-61, Jan. 1993.

21/3
3
 Channel
Channel Channel
Interleaver Deinterleaver

Parellel to Serial to
Serial Parellel
Converter Converter
xk xˆk
IFFT IFFT

Xn Xˆn

Serial to Parellel to
Parellel Serial
Converter Converter

Inner Inner
Encoder Decoder

µ bits

Deinterlea
Interleaver
ver
Decoder
Encoder

RS (n0 , k 0 ) RS (n 0 , k 0 )
BER
Input
Data

Figure 1 System model.

22/3
3
 S0 [n]
d1[n]
input
data
S1[ n] S2 [n]
d0 [n]

+ y2 [n] 00001111
+ 00110011 Output X[n]
y1[n]
01010101 (complex signal)
S-Encoder states y0 [n] 01234567
d- input bits
y-convlolutional encoder output
X- 8PSK signals

Figure 2. 8-state Ungerboeck encoder.

S0 [n] S2 [n]
d0[n]
input
data
S1[ n] S3 [n]
d1[n]

+
y2 [n] 00001111
+ 00110011 Output
y1[n]
01010101 (complex signal)
S-Encoder states y0 [n] 01234567
d- input bits
y-convlolutional encoder output
X- 8PSK signals

Figure 3. 16-state Ungerboeck encoder.

23/3
3
 Information bit pairs
Recursive 8-PSK Symbols Output symbol
Systematic sequence
TCM Encoder 1

Pairwise Symbol
Interleaver De-interleaver

Recursive
Systematic
TCM Encoder 2 8-PSK Symbols

Figure 4 T-TCM encoder.

4 ATM Cells
48 5 48 5 48 5 48 5
212 bytes (10 .91 µ s )

data 32 check bytes

RS ( 244 , 212 ) coder output − 1952 bits

976 bits +4 zeros appended 976 bits +4 zeros appended

490 8 − PSK complex signals 490 8 − PSK complex signals

512 point IFFT output 512 point IFFT output

512 50 512 50

OFDM − 1 5 .45 µs OFDM − 2 5.45 µs

Figure 5 Mapping of ATM cells into OFDM symbols.

24/3
3
White Gaussian White Gaussian White Gaussian White Gaussian White Gaussian White Gaussian
Noise Source Noise Source Noise Source Noise Source ... Noise Source Noise Source

Doppler Doppler Doppler Doppler Doppler Doppler

Filter j Filter Filter j Filter Filter j Filter
...

FFT

Transmitted Gaussian
Signal Noise
X
Received
Signal
+

Figure 6 Frequency selective fading channel model for OFDM transmission.

0 1

0 1
0 1

0 1 0 1 0 1
0 1

000 100 010 110 001 101 011 111

Figure 7 Set partition of the 8PSK constellation.

25/3
3
 0 0 0
0 4 2 6 S0

1 5 3 7 2 4 S1
2
4 0 6 2 S2 3 1

5 1 7 3 S3
4 0
2 6 0 4 S4

3 7 1 5 S5 5 7

6 2 4 0 S6 6

7 3 5 1 S7
(b)
(a)

Figure 8 Rate 2/3 , 8 states Ungerboeck encoder using 8PSK mapping (a). Trellis diagram
(b).signal constellation.

26/3
3
 0.7 0.7

0.6 0.6

0.5 0.5
normalized power

normalized power
0.4 0.4

0.3 0.3

0.2 0.2

0.1 0.1

0 0
0 50 100 150 200 250 0 50 100 150 200 250
delay (ns) delay (ns)

Power 3 decline power delay profile Exponential Power delay profile

0.35

0.3

0.25
normalized power

0.2

0.15

0.1

0.05

0
0 50 100 150 200 250
delay (ns)

Uniform power delay profile

Figure 9 Channel Models.

27/3
3
 0
10
Channel - Uniform
power delay

-1 1. TTCM
10
2. TC
3. TCM2
4. CC
-2 5. TCM1
10

-3

BER
10

-4
10

3 4
2
-5
10 5

-6
10
2 4 6 8 10 12
Eb/No (dB)

Figure 10 Comparison of performance of different coding schemes in a channel with

uniform power delay profile and independent fading.
0
10
Channel - Eponential
power delay
1. TTCM
-1 2. TC
10
3. TCM2
4. CC
5. TCM1
-2
10

-3
BER

10

-4
10
5
2 1
3
-5
10 4

-6
10
2 4 6 8 10 12
Eb/No (dB)

Figure 11 Comparison of performance of different coding schemes in a channel with

exponential power delay profile and independent fading.

28/3
3
 0
10
Channel -Power three
decline
1. TTCM
-1 2. TC
10
3. TCM2
4. CC
5. TCM1
-2
10

-3
10

-4
10
4
2 3

-5 1 5
10

-6
10
2 4 6 8 10 12 14

Figure 12 Comparison of performance of different coding schemes in a channel with power-

three decline power delay profile and independent fading.
0
10
Channel -Uniform
power delay
1. TTCM
-1 2. TC
10
3. TCM2
4. CC
5. TCM1
-2
10

-3
BER

10

-4 1
10

3
-5 2
10 4
5

-6
10
4 5 6 7 8 9 10 11 12
Eb/No (dB)

Figure 13 performance comparison of different concatenated coding schemes in a

correlated fading channel with uniform power delay profile.

29/3
3
 0
10
Channel -Exponential
power delay
1. TTCM
-1 2. TC
10
3. TCM2
4. CC
5. TCM1
-2
10

-3

BER
10

-4
10
2
4
5
1
-5 3
10

-6
10
4 6 8 10 12 14
Eb/No (dB)

Figure 14 performance comparison of different concatenated coding schemes in a

correlated fading channel with exponential power delay profile.
0
10
Channel -Power three
decline
1. TTCM
-1 2. TC
10
3. TCM2
4. CC
5. TCM1
-2
10

-3
BER

10

-4
10 2 4

3 5
1
-5
10

-6
10
4 6 8 10 12 14
Eb/No (dB)

Figure 15 performance comparison of different concatenated coding schemes in a

correlated fading channel with power three decline delay profile.

30/3
3
 -0
10
Channel - Uniform Power delay
with a frequency offset
1. TTCM
2. TC
3. CC
-1
10 4. TCM1

-2 4
BER

10

3
2 4 3
1 1 2
-3 3 2 4
10

4
No Frequency offset
Frequency offset = fC
Frequency offset = 2 × fC 1
-4
10
0 2 4 6 8 10
Eb/No (dB)

Figure 16 RS/TCM performance with frequency offset.

0
10

-1
10
Pr(PAPR <PAPRo)

-2
10

-3
10
5 6 7 8 9 10 11 12 13 14

PAPR (dB)

Figure 17 CCDF of an OFDM signal with 512 carriers.

31/3
3
 -0
10
1. No windowing
2. Windowing at 6dB
3. Windowing at 5dB
4. Windowing at 3dB
5. Windowing at 2dB
-1
10

-2
BER
10

-3
10
5
1 3
4
2
-4
10
0 2 4 6 8 10
Eb/No (dB)

Figure 18 Bit error rate performance with peak windowing (TCM1).

-1
10
1. RS+ TCM (8 states) - theoretical
2. RS+ TCM (8 states) - simulation
3. RS+ TCM (16 states) - simulation
4. RS+ TCM (16 states) - theoretical

-2
10

-3
BER

10

-4
10 2

4
1

-5
10
0 2 4 6 8 10 12 14 16
Eb/No (dB)

Figure 19 Theoretical upper bound and the simulation results for Ungerboeck’s 8 state
8PSK code (TCM) in concatenation with RS(244,212) code.

32/3
3
 -1
10 1. RS+TCM (8 states) - theoretical
2. RS+TCM (8 states) - simulation
3. RS+TCM (16 states) - simulation
4. RS+TCM (16 states) - theoretical
5. RS+CC (64 states) - simulatedl
4. RS+CC (64 states) - theoretical
-2
10

-3
BER

10

-4 6
10 2

5
4
1

-5
10
0 2 4 6 8 10 12 14 16
Eb/No (dB)

Figure 20 Comparison of theoretical and simulated results

33/3
3