International Journal of Engineering Trends and Technology (IJETT) – Volume 4 Issue 9- September 2013
The adaptive bit-interleaved coded modulation based
on OFDM for mobile radio system
M NEELIMA1, B SHOBAN BABU2
1
Student, 2Associate professor, svcet chittoor, 1,2ECE department, 1,2Sri Venkateshwara College of Engineering and technology,
Chittoor, Andhrapradesh, India-517127
Abstract — Now a day’s industries require some special
techniques due to increase in wireless communications. These
techniques should solve the communication related problems
and challenges and improves the quality of service. When the
signal propagates from transmitter to receiver it undergoes to
some random fluctuations called Noise. This will happened in
both time and frequency domain. We need to predict such
noise and eliminate it. The need of predicting noise and
measure of it is referred as Channel state information simply
CSI. This is the measurement of noise prediction. But, here
the measurement will be observed after some time units. The
transmitter selects appropriate modulation technique
depending on the Channel state information. This process is
called as Adoption. In this process transmitter and receiver
both need to send acknowledgement to each other to confirm
whether it received or not. This method is referred as
Channel state information feedback.
These techniques combined together is referred as the
adaptive bit interleaved coded modulation simply ABICM.
There are so many techniques are there in this method. They
are UAM, ATCM and BICM. Previous method called BICM
was based on Bhattacharya bound. This method was works
based on minimum distance of constellation and a nominal
non-adaptive BICM. It works by determine the constellation
size and transmission power. ATCM method or adaptive
trellis coded modulation gives better performance but
outdated. So, Proposing ABICM was based on expurgation
bounded aided by fading prediction. This ABICM method
improves the accuracy of bit error rate (BER). This method
also gives better spectral efficiency and performance.
Keywords— CSI- Channel state information, ATCMAdaptive trellis coded modulation, OFDM- orthogonal
frequency division multiplexing, ABICM- adaptive bit
interleaved coded modulation.
I. INTRODUCTION
The adaptive modulation methods have been
studied in the past to combat the severe channel
conditions encountered in wired and wireless
channels. Among them, the adaptive MQAM
that adapts the constellation sizes according to
the channel condition significantly improves the
SE of the system when the knowledge of CSI is
perfect at the transmitter. To further improve the
SE, the ATCM are employed to achieve coding
ISSN: 2231-5381
gain over the original uncoded MQAM when
the CSI is reliable. However, the delay caused
by processing and feeding CSI from the receiver
back to the transmitter would significantly
degrade the quality of CSI. The performance of
both ATCM and uncoded MQAM decreases
dramatically as the reliability of CSI degrades
while the coding gain of ATCM over uncoded
MQAM diminishes at the same time. For
extremely unreliable CSI, the uncoded MQAM
even outperforms ATCM.
To improve the reliability of CSI, the LRP
techniques are proposed. Compared with
outdated CSI caused by delay the LRP
technique provides more accurate CSI for
practical delay values. However, even when
advanced LRP techniques are employed, the
SEs of ATCM and UAM are still reduced
noticeable by the prediction errors at medium
prediction range (0.2-0.5 carrier wavelength).
Therefore, for mobile radio systems where the
delay falls in that prediction range, it is
necessary to find adaptive modulation schemes
are less sensitive to the prediction errors.
As one of the diversity methods, bit-interleaved
coded modulation (BICM) has shown good
performance over the unreliable wireless
communication channel. Com-pared with the
bandwidth efficient trellis coded modulation
(TCM) scheme, it achieves higher code
diversity, but at the price of reduced Euclidean
distance. Moreover, the adaptive bit-interleaved
coded modulation (ABICM) was proposed in to
combat the unreliable CSI by utilizing its high
code diversity. The simulation results show that
the ABICM maintains the SE even for highly
unreliable CSI.
http://www.ijettjournal.org
Page 3909
International Journal of Engineering Trends and Technology (IJETT) – Volume 4 Issue 9- September 2013
In the original ABICM method the
Bhattacharyya bound of single-bit-error
probability and only the minimum distance
between coded bit sequences are considered to
determine the constellation size while
maintaining the target BER. This method only
provides coarse estimates to the single-bit-error
probability. The result is compared with the
single-bit-error probability of a non-adaptive
BICM scheme, whose BER performance is
known from simulation, to infer the BER of
ABICM. Then the constellation size is selected
to maintain the target BER. Due to inaccurate
estimates of BER, the simulated BER of this
ABICM method significantly deviates from the
specified target BER. Hence, additional
experimental energy adaptations are required to
bring the BER to the desired level.
In this dissertation, we advocate the usage of
expurgated bound for ABICM. The expurgated
bound, proposed in provides accurate BER
estimates to the BICM in AWGN channel as
well as Rician and Rayleigh fading channels.
New ABICM methods based on expurgated
bound are developed. The resulting BER of
these methods are closer to the target BER then
the original ABICM method.
II. ADAPTIVE OFDM AIDEDBY LONGRANGE PREDICTION
OFDM:
It is a method of encoding digital data on
multiple carrier frequencies. OFDM has developed
in to wideband digital communication. There are
two types of multicarrier systems are there.
They are
1. Single carrier system
2. Multi carrier system.
At the transmitter, the information bits are
coded through a convolution encoder, whose
outputs are interleaved. The adaptive bit and power
loading algorithm maps the interleaved bits into
MQAM symbols for all subcarriers, and determines
their constellation sizes and energies using the CSI
fed back from the transmitter. To facilitate the LRP,
pilot symbols are inserted. Then inverse fast Fourier
transform (IFFT) is employed to get the OFDM
symbol in the time domain. To prevent inter-
ISSN: 2231-5381
symbol interference (ISI), a cyclic prefix (CP) is
added to the OFDM symbol before transmission.
The transmitted signal experiences frequencyselective Rayleigh fading channel. At the receiver,
the CP is removed and the FFT restores the
transmitted symbols in the frequency domain. The
data symbols are demodulated, de-interleaved and
decoded to obtain the decisions of information bits.
Based on the observed pilot symbols, the future CSI
is predicted and fed back to the transmitter.
Figure: LRP-enabled Adaptive orthogonal multicarrier
system with ABICM.
III. ABICM WITH IMPERFECT CSI
ABICM METHOD IN
ABICM will achieve the coding again and
Rayleigh fading channel. It will minimize the error
probability by using interleaving technique. This
interleaving is referred as bit interleaver. It will
correct the forward error codes. Many
communication channels are not memory less. Here
we need to refer the following terms. They are
Bit Error Rate(BER): Is the number of bit
errors divided by the total number of transferred
bits during a studied time interval.
Channel State Information(CSI): It refers to
known channel properties of a communication link.
This information describes how a signal propagates
from the transmitter to the receiver.
Expurgated Bound: It is the probability of
error associated with transmission over discrete
memory less channels.
In ABICM, the candidate Gray-labelled
MQAM constellations are X m  of sizes
http://www.ijettjournal.org
Page 3910
International Journal of Engineering Trends and Technology (IJETT) – Volume 4 Issue 9- September 2013
X m  2 m , m  M, where M is the set of possible guidelines for selecting the nominal scheme are not
provided. As a result, the simulated BER can
bits per symbol employed by the adaptive
deviate significantly from the target BER, and the
modulator. Let d min, m and E m denote the minimum
actual transmit power has to be adjusted by
Euclidean distance and average symbol energy of simulations to meet the desired BER.
constellation
x m , respectively. The error ABICM method based on the Expurgated Bound
The expurgated bound proposed in [5]
probability of choosing the symbol Xˆ  xm at the
provides an accurate BER estimate for non-adaptive
receiver when X  xm is transmitted (the single BICM, which does not require the CSI knowledge.
symbol error probability) is considered in [1]. We develop a method based on the expurgated
Given the predicted channel coefficient Hˆ ( n, l ) , bound for ABICM systems with predicted CSI.
Suppose the transmitted coded bit sequence
the Bhattacharyya bound of this error probability is
and
its
estimate are c and ĉ. , respectively. These
[1], [2]
sequences originate and terminate at the same state
1 K
KC  ,

PX  Xˆ Hˆ (n, l ) 
exp 

and differ by d bits. The channel predictions and
1 K  C
1

K

C


2
2
Where
and constellations associated with these d error bits are
K | Hˆ ( n, l ) | / 
ˆ
ˆ1 ˆ 2 ˆ d
2
2
and
  [  1  2 ... d ]
,
C  d min,m ( K  1) /(4 N 0 ). To relate this bound to H  [ H H ...H ]
th
For the i error, the constellation size
the target BER, a non-adaptive BICM scheme respectively.
i
i
m
called the nominal scheme, which uses fixed |  | 2 , i  [1, d ]. . Due to the assumption of ideal
MQAM modulation, is employed in [1]. The interleave, the corresponding channel coefficients
nominal scheme and the ABICM scheme under H  [ Hˆ 1 Hˆ 2 ...Hˆ d ] are
independent
random
investigation use the same convolution encoder. For
variables and the conditional probability density
the nominal scheme, the SNR required to achieve
function (PDF) is p( H | Hˆ )  p ( H 1 | Hˆ 1 )... p ( H d | Hˆ d ) .
the target BER is determined by simulation, and the
corresponding Bhattacharyya bound on the single With the knowledge of Ĥ and  , the pair-wise
symbol error probability is denotedD0(This bound error probability (PEP) is bounded by [5, eq. (48)]
  j d
1
ds
is computed from (3) by setting Hˆ ( n, l )  0 since
P(c  cˆ , Hˆ )  f ex (d , ,  , Hˆ )
exi (s) s ,
2j  j i 1
the nominal scheme is not adaptive). For example, a
nominal scheme that uses QPSK constellation and where  is the levelling rule[5,eq. (49)],  is a
rate2/3, 4-state convolutional encoder achieves the small positive number [16], and
1 m 1
target BER=10−5 at 17.2 dB, and the resulting D0
ˆ
ˆ
exi (s)  E[e s( X ,Z ) ]  i m    H( X ,Zˆ ) (s).
=0.0367[1]. It is argued in [1] that the BER of
m 2 p1 c0 X
ABICM can be maintained approximately at the
In (6), m i is the size of constellation  i ,  ci , p
target BER if the Bhattacharyya bound (3) for
i
th
ABICM is equal to or smaller than D0. Hence, to is the subset of  where the p bit takes on the
maintain the target BER, the energy required to use value c,  ( X , Zˆ ) is the metric difference between
the constellation x ml for the l th subcarrier is
two symbols X and Ẑ , where Ẑ is the unique
i
i
i
i
c, p
KC
 1 K

E Hˆ ( n ,l ) ( m1 )  arg min 
exp( 
)  D0 
E ml
1 K  C
1  K  C

nearest neighbour of
X in  i that satisfies
Zˆ   ci , p .We use c to denote the complement of bit
There are potential problems with this
approach, which we refer to as the original ABICM
ˆi
c. Finally,  H( X , Zˆ ) ( s ) is the Laplace transform of the
method. First, single symbol error probability is
considered in (3), while the performance is usually PDF of  ( X , Zˆ )
measured by the BER. Second, the Bhattacharyya
ˆ
ˆ
H( X ,Zˆ ) E[e s( X ,Z ) | Hˆ i , X , Zˆ ] .
bound in (3) is inaccurate. Finally, the utilization of
the nominal scheme is not justified, and the
i
ISSN: 2231-5381
http://www.ijettjournal.org
Page 3911
International Journal of Engineering Trends and Technology (IJETT) – Volume 4 Issue 9- September 2013
To calculate P( X  Zˆ | Hˆ i ) , the maximum
likelihood (ML) detection with the knowledge of
perfect CSI H i at the time of detection is assumed.
Denote the received symbol as Y. The conditional
probability of Y is
.
1
pH
i
(Y | X ) 
N
channel with K factor | Hˆ | 2 /  2 . It can be solved
numerically using the method proposed in [16].
For a rate k c / nc convolution encoder with
the free Hamming distance d free , the union bound
on the BER is
exp(  | Y  H i X | 2 / N 0 )
Pb 
0
Then the metric difference between X and Ẑ at the
decoder is
( X , Zˆ )  log pH (Y | X )  pH i (Y | Zˆ ) .
i
Using (8) and (9) in (7), we obtain [5,eq.(50)]:
 s(1  sNˆ 0 )K | X  Zˆ |2 /(K  1) 

exp 
 1  s(1  sNˆ )K | X  Zˆ |2 /(K  1) 
ˆi

0

H( X , Zˆ ) 
1  s(1  sNˆ 0 )K | X  Zˆ | 2 /(K  1)
1
kc

W (d ) fˆ
I
ex
( d ,  ,  m , Hˆ ),
d  d free
Where WI (d ) are the weights of the error events at
the Hamming distance d . Using (9), the average
symbol energy that satisfies the BER constraint is
derived:
 1
E Hˆ (m)  arg min 
Em
 k c

W (d ) fˆ
I
d d free
ex

(d ,  ,  m , Hˆ  BERtg 

Where BERtg is the target BER. The result of (13)
can be evaluated numerically given the predicted
channel coefficient Ĥ , the MMSE of prediction  2 ,
the constellation, the noise power, and the free
Hamming distance of convolution encoder. We
refer to this method as an ABICM scheme based on
the Expurgated bound
Note that it is too complex to be computed
in real-time due to the rapidly-varying channel
conditions. However, it is possible to compute the
thresholds offline and to employ a look-up table in
mobile communication equipment in practice. The
computational speed of (13) is much higher than for
simulations-based methods, which are often used to
0
determine the thresholds in complex adaptive
 i0   m . This assumption is justified by the modulation systems and need to run millions of bits
observation that the bit-loading algorithm selects to obtain reliable results.
the symbol energy and the constellation to maintain
IV. NUMERICAL RESULTS
the same performance for all symbols, and SIMULATION SETUP
 exi (s ) determines the error probability of each
We consider an orthogonal multicarrier
symbol. Thus, a simplified expurgated bound on system that has 120 sub carriers and the sub carrier
PEP, which employs Ĥ and  m instead of vectors spacing of 10.94 KHz. The total bandwidth of this
system is about 1.3MHz. We assume that the carrier
Ĥ and  , is defined as
frequency is2.5GHz, which is typical for the Mobile
fˆex (d ,  ,  m , Hˆ )  fˆex (d ,  ,  i , Hˆ i )
WiMAX deployment. The vehicular speed is set
  j
1
to86.4km/hour, and the corresponding maximum
i
d ds

[ ex ( s )]
2j  j
s
Doppler frequency is fdm=200HZ. The pilot spacing
The above equation is exactly the is 8 in both frequency and time domains. The
expurgated bound on PEP for non-adaptive BICM fading predictor has filter orders 3 and 20 in the
that uses constellation  m min the Rician fading frequency and time domains, respectively. The
fading channel is generated by the ETSI Vehicular
B model, which has six paths and RMS delay
Where Nˆ 0  ( 2  | Hˆ i | 2 ) N 0 and K | Hˆ i | 2 /  2
As stated earlier, for each bit rate m, our
goal is to determine the symbol energy that satisfies
the target BER given predicted channel
coefficient Ĥ . Although the analytical BER needed
to achieve this goal can be derived from PEP, the
calculation of PEP in (5) requires the knowledge of
channel predictions, transmission energies, and
constellation sizes for all d errors. Due to random
interleaving, this knowledge is not available in
practical systems. Therefore, we simplify (5) by
assuming  exi ( s )   exi0 ( s ) for all i  [1, d ], where
i is one of the d indices with given Hˆ i0  Hˆ and
0
0
0
ISSN: 2231-5381
http://www.ijettjournal.org
Page 3912
International Journal of Engineering Trends and Technology (IJETT) – Volume 4 Issue 9- September 2013
spread of 4 s . In the simulations, SNR is defined
as Er/(LxNoxR) . Transmitted pilot energy is
assumed to be equal to the average symbol energy
i.e.,E p=ET/L. In all simulations, the target BER is
BERtg=10 -5. If not specified otherwise, the rate 2/3,
4-state optimal convolution encoder (constraint
length K=2) is employed for both ABICM and
BICM. The set of constellation sizes is {0, 4, 16, 64}
for all adaptive modulation methods, i.e. ABICM,
uncoded adaptive modulation, and ATCM. For
ATCM,1/2rate, 4-state Unger boeck encoder and
set-partition method in are utilized.
COMPARISON WITH THE ORIGINAL ABICM
METHOD
First, we investigate the BER achieved by
the original method and our ABICM method
discussed in Section III. For the original ABICM
method, two nominal schemes that employ
quadrature phase shift keying (QPSK) and 64QAM
in Rayleigh fading channel are used, and the
resulting D0 values are 0.0367 and 0.0624,
respectively. Figure 6.3 shows the simulated BER
vs. prediction range for the two ABICM methods at
SNR= 15 dB. In practical mobile communication
systems, the prediction range is usually 0.1 − 0.5λ.
Within this prediction range, the BER of the
original ABICM method is either significantly
greater or lower than the target BER depending on
the choice of the nominal scheme. Moreover, the
resulting BER of the original ABICM method
varies dramatically as the prediction range changes.
On the other hand, the method based on the
Expurgated Bound maintains the target BER.
Results in confirm this conclusion for other SNR
values. In wireless communication system design, it
is desirable to obtain as accurate estimate of the
BER as possible. While the original ABICM
method sometimes achieves a lower BER than the
target BER, a system designer would prefer to save
the transmission power by maintaining the BER at a
higher but acceptable rate. Therefore, we employ
our ABICM method in the remaining simulations.
spectral efficiency for medium to high SNR since
its minimum Euclidean distance is larger than for
BICM. On the other hand, the spectral efficiency of
ATCM is significantly degraded by imperfect CSI,
while ABICM still maintains high spectral
efficiency for σ2 = 0.1, a typical MMSE for realistic
mobile radio conditions. Both ABICM and ATCM
outperform un coded adaptive modulation, although
the coding gain of ATCM is small when the CSI is
not reliable. We also provide the spectral
efficiencies of non-adaptive BICM schemes that
use BPSK and QPSK. Note that ABICM
significantly outperforms these non-adaptive
schemes while the spectral efficiency of BICM with
QPSK is similar to those of un coded adaptive
modulation and ATCM when σ2 = 0.1 due to the
coding gain and diversity provided by BICM. This
comparison demonstrates that in the slow fading
channel, adaptive modulation methods that do not
require interleaving, e.g., ATCM, achieve the best
spectral efficiency. On the other hand, ABICM is
the best choice for practical mobile wireless
channels. Fig. 6.5 illustrates the dependency of the
spectral efficiency
on the normalized spatial prediction range FDMΤ.
Figure BER of ABICM vs Prediction Range
SPECTRAL EFFICIENCY COMPARISON
Two prediction MMSE values are
considered, σ2 = 0.1 and 0.001, which for FDM =
200 Hz correspond to 2ms (0.4λ) and 0.1ms (0.2λ)
prediction ranges at 30dB SNR. When CSI is Figure Comparison of Spectral Efficiencies vs SNR.
reliable (σ2 = 0.001), ATCM achieves the highest
ISSN: 2231-5381
http://www.ijettjournal.org
Page 3913
International Journal of Engineering Trends and Technology (IJETT) – Volume 4 Issue 9- September 2013
. Figure Comparison of Spectral Efficiencies vs
SNR
CONCLUSION
Improved ABICM Based on Expurgated Bound
for Cellular Radio Orthogonal Multicarrier System
was successfully implemented using MATLAB.
Adaptive modulation techniques are much less
sensitive to prediction error. The Exactness of Bit
Error Ratio is maintained by the ABICM. Future
extension of this work is to present operation of the
variable rate Turbo Bit interleaved coded
modulation in the fast desertion environment.
REFERENCES
1.
Figure Spectral Efficiency vs Normalized Spatial
Prediction Range.
L. Goeckel, “Adaptive coding for time-varying channels using
outdated fading estimates,” IEEE Trans. Commun., vol. 47, no. 6,
pp. 844–854, July 1999.
2.
Duel-Hallen, S. Hu, and H. Hallen, “Long-range prediction of
fading signals,” IEEE Signal Process. Mag., vol. 17, no. 3, pp. 62–75,
May 2000.
3.
J. Proakis, Digital Communications, 4th edition. McGraw-Hill,
2001.
4.
K. N. Lau, “Performance analysis of variable rate: symbol-bysymbol adaptive bit interleaved coded modulation for Rayleigh fading
channels,” IEEE Trans. Veh. Technol., vol. 51, no. 3, pp. 537–550, May
2002.
5.
K. Song, A. Ekbal, S. T. Chung, and J. M. Cioffi, “Adaptive
modulation and coding (AMC) for bit- interleaved coded OFDM (BICOFDM),” IEEE Trans. Wireless Commun., vol. 5, pp. 16851694, July
2006.
6.
R. Liu, J. Luo, and P. Spasojevic, “Adaptive transmission with
variable rate turbo bit-interleaved coded modulation,” IEEE Trans.
Wireless Commun., vol. 6, pp. 3926–3936, Nov. 2007.
7.
Duel-Hallen, “Fading channel prediction for mobile radio
adaptive transmission systems,” Proc. IEEE, vol. 95, pp. 2299–2313,
Dec. 2007.
BIOGRAPHIES
Figure Spectral Efficiency vs Normalized Spatial
Prediction Range.
Figure Spectral Efficiency Comparison of ABICM.
ISSN: 2231-5381
http://www.ijettjournal.org
M NEELIMA,
M.Tech, (VLSI) student,
Sri Venkateswara College of
Engineering and Technology,
Chittoor, Andrapradesh, India517127
Mr. B.ShobanBabu received his
graduation degree from JNTUCE,
Anantapur in 1998, post graduation
degree from VTU, Belgaum in the
year 2003, and pursuing Ph.D. from
SVUCE, Tirupati in the Image
Processing Domain.
He worked as Assistant and Associate
professor at M.I.T.S., Madanapalle, as
Associate Professor at PRIT,Medak.
He is working as Associate Professor
in SVCET,Chittoor,since 2012.
Page 3914