Wireless Sensor Network, 2009, 1, 268-275
doi:10.4236/wsn.2009.14033 Published Online November 2009 (http://www.scirp.org/journal/wsn).
Performance Improvement of the DSRC System
Using a Novel S and Π-Decision Demapper
Jeich MAR, Chi-Cheng KUO
Department of Communications Engineering, Yuan Ze University, Taiwan, China
Received March 31, 2009; revised June 20, 2009; accepted June 23, 2009
Abstract
Based on the constellation diagram of the different modulations, a novel S and Π-decision rule is designed
for the analog demapper of the orthogonal frequency-division multiplexing (OFDM) systems. The dedicated
short-range communications (DSRC) system is chosen as an OFDM platform to compare the performance
among the proposed S and Π-decision decoder, hard-decision and soft-decision decoders. Simulation results
demonstrate that both the complexity and performance of S and Π-decision demapper used for M-ary quad-
rature amplitude modulation (QAM) OFDM system can be greatly improved. The number of decisions be-
tween the received symbol and constellation points can be simplified to look up table2
log
M
times for M-ary
QAM OFDM system.
Keywords: S and Π-decision Rule, Analog Demapper, DSRC System, OFDM
1. Introduction
In the traditional digital communication system where
the digital Viterbi decoder is used, the maximum likeli-
hood decision rule is applied to both the demapper and
digital hard-decision or soft-decision Viterbi decoder.
The soft-decision decoder is the recommended scheme to
be used in the digital Viterbi decoder because it provides
a coding gain over the hard-decision decoder [1,2]. A
simplified algorithm of the soft-decision Viterbi decoder
for the 16-quadrature amplitude modulation (QAM) and
64-QAM constellations was presented in [3], which al-
lows the complexity of the demapper to be maintained at
almost the same level for all the possible modes of
HIPERLAN/2. In [4], it presented that for M-ary QAM
systems the complexity of the demapper in a soft-deci-
sion Viterbi decoder used for bit-interleaved coded
modulation can be significantly lowered without com-
promising the performance. Four types of analog-input
Viterbi decoders are described and compared in [5],
where the analog-to-digital converter (A/D) converter is
included as part of a digital Viterbi decoder. The analog
circuit flaws of the previously used add-compare-select
(ACS) chips are included in the comparison. It concludes
the analog Viterbi decoder is able to outperform the
digital Viterbi decoder, as well as achieve 3-bit or higher
decoding resolution. In this paper, we propose a new S
and Π-decision decoder, where the S and Π-decision
rules are designed for the analog demapper of the or-
thogonal frequency-division multiplexing (OFDM) sys-
tems. The ACS and path memory (PM) modules, which
are parts of the analog Viterbi decoder introduced in [6],
are used to perform the Viterbi decoding process. The
proposed S and Π-decision demapper combined with the
digital Viterbi decoder is another alternative to using the
S and Π-decision decoder. Because analog Viterbi de-
coder outperforms digital Viterbi decoder and analog
very-large-scale integration (VLSI) implementation is in
general more area and power efficient than digital im-
plementation, the performance comparison of the S and
Π-decision decoder using the digital Viterbi decoder is
not included in the paper.
The dedicated short-range communications (DSRC)
system [7], which employs OFDM technique, provides
wireless communications over a short distance between
the roadside and high-speed mobile radio units or be-
tween high-speed vehicles. The DSRC system will work
in a mobile environment with time-varying characteris-
tics. The time-varying fading effect of the DSRC system
may not be effectively compensated by using the long
symbol training and pilot-based frequency synchronizer
[8]. The combined interleaving and convolution coding,
which provides the time diversity function, may further
improve the performance of the DSRC system. We
choose DSRC system as a basis for the performance

J. MAR ET AL. 269

comparison among different decoders. Based on the
analog-input Viterbi decoder, the coding gain of the
DSRC system achieved by replacing the hard-decision
and soft-decision decoders with the proposed S and
Π-decision decoder will be confirmed with the simula-
tions.
The rest of this paper is organized as follows. In Sec-
tion II, several channel decoding schemes are described.
The proposed S and Π-decision decoder is depicted in
detail for the binary phase-shift keying (BPSK), quadra-
ture phase-shift keying (QPSK), 16-QAM and 64-QAM
OFDM signals of the DSRC systems. Section III briefly
describes the base band model of the DSRC system,
which is used to compare the bit error rate (BER) per-
formance of the DSRC system for the different decoders.
Simulation results are given in Section IV, which show
the coding gain achieved by the proposed S and
Π-decision decoder compared to the hard-decision and
soft-decision decoders. Finally, conclusions are drawn in
Section V.
2. S and Decision Decoder for the OFDM
Systems
In the traditional digital communication system, the
digital Viterbi decoding uses a maximum likelihood rule
which is ideal for an additive white Gaussian noise
(AWGN) channel. For a hard-decision Viterbi decoder,
the samples matching to a single bit of a code word are
quantized to the two levels zero and one, a decision is
made as whether each transmitted bit in a code word is
zero or one. The coding gain of the soft-decision Viterbi
decoder for the hard-decision Viterbi decoder in
Rayleigh fading channel increases to about 2dB [1]. A
four-level discrete symmetric channel model [2] is used
for the soft-decision decoder. The demapper assigns one
of four values to each received signal. The path metrics
in the Viterbi algorithm are calculated by weighting the
square of the Hamming distance between the soft-deci-
sion and the reference value. The four-level soft-decision
Viterbi decoder is almost exactly as shown for the hard-
decision case with the only difference being the in-
creased number of path metrics.
The block diagram of the proposed S and Π-decision
decoder for the OFDM system is shown in Figure 1,
where it consists of a S and Π-demapper and an analog
Viterbi decoder. The S and Π-decision decoder is a non-
uniform infinite-level quantization decoder. The S and
Π-decision demapper assigns an analog complex value to
the analog Viterbi decoder for each received signal z(y)
according to a combination of S and Π functions, as
shown in Figure 2. The Π function [9], as shown in Fig-
ure 2(a), is defined as follows:
(();,/2,) ()
((); ,)
1(();,/2,) ()
SzyM RM RMforzyM
zy RMSzyMMR MRforzy M
 




(1)
where z(y) represents either or and y is
the received signal after channel compensation. The con-
stellation decoder estimate the mth symbol gained
through the received signal after channel compensation,
, can be found in the signal
space diagrams [10] for BPSK, QPSK, 16-QAM and
64-QAM, respectively, where the values of
ˆ()
m
Iy ˆ()
m
Qy
ˆm
ˆ ˆˆ
() ()()
mm m
Yy Iy jQy
()
I
y and
are serially decoded according to the modulation
type. The Π-function goes to zero at the points
ˆ()
m
Qy
()
z
yMR

 (2)
while the Π-function goes to 0.5 at the crossover points
() 2
R
zyM

 (3)
Notice the parameter R is now equal to one, which is
the total width at the crossover points; parameter M is
now equal to zero, which is the middle point of the
Π-function. The S-function, as shown in Figure 2(b), is
defined as follows:
2
2
0 ()
()
2() ()
(();,, )()
12() ()
1 ()
for zy
zy forz y
Szy zy forz y
for zy





 
















(4)
For BPSK modulation, the value of ˆ()
m
I
y is in the
interval of (-1,1) of constellation diagram. From BPSK
constellation diagram, the original one bit binary data (b0)
is decided as using the S-decision rule. Using (4) for
α=-1, β=0 and γ=1, the values of are produced as
follows: =-1 for z(y)≦-1; =1 for z(y)1; ≧
=for -1<z(y)<1.
0
ˆ
b
0
ˆ
b
0
ˆ
b
0
ˆ
b
)) y
0
ˆ
b((Sz
For QPSK, the values of ˆ()
m
I
y
(2)
b

and in the
constellation diagram are found in the intervals (-1, 1).
The original two-bit vector = (b0, b1) is also esti-
mated using the S-function as a decision rule. The
S-decision rule in (4) for α=-1, β=0 and γ=1 is used to
determine from the received I-channel signal part
ˆ()
m
Qy
0
ˆ
b
ˆ()
m
I
y
part ˆ
and determine from the Q-channel signal
1
ˆ
b
()
.
m
Qy

270 J. MAR ET AL.
ˆ ˆ() ()Ix Qx
ˆ()Qx
ˆ()Ix
()
ˆ
i
b

Figure 1. Block diagram of the S and Π-decision decoder.
The constellation diagram of
Fi
16-QAM is shown in
1
gure 3, where the values of ˆ()
m
I
y and ˆ()
m
Qy for
16-QAM modulation are found intervals of (-3, -1,
1, 3), respectively. The message points in each quadrant
are assigned with Gray-encoded four-bit vector (3) b
the in

=
(b0 b1 b2 b3). The first two bits (b0 b1) and last twots
(b2 b
3) in (3)
b

are transmitted in I and Q-channel, re-
spectively. h first two bits (boldface) of the I-channel
from left to right message points and last two bits (Nor-
mal) of the Q-channel from bottom to top message points
have the same 8-bit pattern 00 01 11 10 in Figure 3. The
first four odd bits are 0 0 1 1 and the second four even
bits are 0 1 1 0 can be estimated by using the S and
Π'-decision rule, respectively, as shown in Figure 4. The
Π'-decision rule as shown in Figure 4(a) is defined as
0()-3zy

bi
Bot
'
(();-3,-2,-1)-3() -1
(()) 1-1()
1-(( );1,2,3)1( )3
0()
Szy zy
zy zy
Szy zy
zy



 





1
3
(5)
where the S-function is defined in (4). The first two bits
(0
ˆ
b 1
ˆ
b) and the last two bits (2
ˆ
b 3
ˆ
b) of the demapper
ouufor each message point arestimated from the
values of ˆ()
m
tp t e
I
y and ˆ()
m
Qy, respectively. The S-de-
cision rule ed to ine 0
ˆ
b and 2
ˆ
b and the
Π'-decision rule is used to determin 1
ˆ
b an3
ˆ
b. The
number of decision needed to obtain Snd Π-decision
demapper output in 16-QAM OFDM system is four.
The same design principle of the S and Π-decision
de
is usdeterm
ed
a
mapper used for 16-QAM is applied for 64-QAM.
The values of ˆ()
m
I
y and ˆ()
m
Qyin the constellation
diagram for 64-QAM modulare found in the in-
tervals of (-7, -5,-3, -1, 1, 3, 5, 7), respectively. Simi-
larly, the six-bit vector (4)
ˆ
b

= (0
ˆ
b 1
ˆ
b 2
ˆ
b 3
ˆ
b 4
ˆ
b 5
ˆ
b)
ation
for each message point of 64-QAM modulation is de
Figure 2. (a) Π-decisionule; (b) S-decision rule. r
Figure 3. The constellation diagram of 16-QAM.
-3 -2 -10 1 23
0
0.5
1
-3 -2 -10 1 23
0
0.5
1
z(y)
(a)
'(())
z
y

(())Szy
z(y)
(b)
0101
Figure 4. (a) Π'-decision rule; (b) S-decision rule for 16-QAM.
0011
-1-0.8 -0.6 -0.4 -0.2 00.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
1
-1-0.8 -0.6 -0.4 -0.2 00.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
Π(x)
S(x)
x
x
)
)
(a
(b
-1-0.8 -0.6 -0.4 -0.2 00.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
1
1
0.8
-1-0.8 -0.6 -0.4 -0.2 00.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
Π(x)
S(x)
x
x
)
)
(a
(b

J. MAR ET AL. 271
termined using the , and S-decision rules,
which are designed witt pattern 000 001 011
010 110 111 is generated from first
three bits from left essage points and last
three bits from bottotssage points, for both I
and Q-channels. Tht pattern 0 0 0 0 1 1 1 1
can be determined frohsion rules as shown
in Figure 5(a). The le 8-bit pattern 0 0 1 1
1 1 0 0 can be estimated b using the -decision
function which0 is defined as
3
1
The first three bits () and the last three bits
) of each t are estimated from
es of , respectively. The
rulrmine and
"
1

h
101 100, whi
to ri
m to
e first
m t
second m
"
2

24-bi
ch
ght m
op me
8-bi
e S-deci
idd
y"
1

''
1
(();-5,-4,-3)-5()-3
(()) 1-3()3
1-( ();3,4,5)3()5
0()5
Szy zy
zy zy
Szy zy
zy



 





(6)
The third 8-bit pattern 0 1 1 0 0 1 1 0 can be estimated
from the "
2
-decision rule as shown in Figure 5(b),
which is defined as
0

0()-5zy

''
2
()-7
((); -7, -6, -5)-7()-5
1-5()
1-((); -3, -2, -1)3()1
(()) 01()
(();1,2,3)1() 3
13()5
zy
Szy zy
zy
Szy zy
zy zy
Szy zy
zy




 


  




1-(();5,6,7)5( )7
07()
Szy zy
zy




(7)
0
ˆ
b
me
()
an
use
1
ˆ
b
ssa
d
d t
2
ˆ
b
ge poin
ˆm
Qy
o det
(3
ˆ
b
th
S-de
4
ˆ
b
e valu
cisi
5
ˆ
b
on
ˆm
Iy
e is
()
e0
ˆ
b 3
ˆ
b,
-8 -6 -4 -2 0 2 4 6 8
0
0.5
1
-8 -6 -4 -2 0 2 4 6 8
0
0.5
1
-8 -6 -4 -2 0 2 4 6 8
0
0.5
1
(())Szy
(a)
(c)
z(y)
0000 1111
Figure 5. (a) S-decision rule; (b) -decision rule (c)
''
1(())zy
''
2(())zy
(b)
0101 1 010
0011 0011
''
1
 ''
2

-
"
1

decision rule for 64-QAM.
the -decision rule is used to determine and
and t
1
ˆ
b4
ˆ
b
he "
2

-decision rule is used to determ
mber of decisions n
n 64-QAM OFDM system
that for Gray encoded M-ary QA
systeme number of decisions for
apper can be reduced to
ine
eeded for S and
is si
anal
2
ˆ
b
x. I
M OFD
og S and
and
t is
M
5
ˆ
b
Π
con
Π
. Th
–dema
clud
s,
-dem
e nu
pper i
ed
th
2
log
M
. If th
ion rules are ca
lapper decisisim
e analog Vi
64-state decoderce bn
n
rb
cod
trics.
e S, Π
in
ied as
gned
le
, Π'
the
to
in
and
ed
and
tab
Fi
Π''
gure 1
decis
e, each dem
look up table. Th
is a
lculated and s
on can be
terbi decoder
with a tra
alog com
tored
plif
desi
ack gth
of 24, 48, 96 and 144 for BPSK, QPSK, 16-QAM
64-QAM, respectively. The aplex values of
the constellation decoding vectors (()
ˆi
b) gain
through the combinational S and Π-decision rules are
input to analog deinterleaver and analog Vitei de-
er in turn. The analog deinterleaver [11] permutes
the analog demapper output according to the switching
order performed in the interleaver. The analog Viterbi
decoder consists of the analog ACS module and a
digital PM module [6]. ACS module performs the cal-
culation of the analog path meThe transmitted
message bits are decoded by PM module using the
trace back through the trellis architecture. The decod-
ing algorithm and the sequence control for analog
Viterbi decoder remain identical with the digital
Viterbi decoder.
3. Base Band Model of the DSRC System
The block diagram of the DSRC system is shown in Fig-
ure 6. The protocol data unit (PDU) trains are applied to
the physical layer for transmission. A 127 pseudorandom
sequence is used to scramble the data out of the binary
sequence before the convolutional encoding. The pur-
pose of the scrambler is to prevent a long sequence of 1s
or 0s to aid the timing recovery at the receiver. The gen-
erator polynomial [10] of the pseudorandom sequence is
47
()

g
DDDD  (8)
where D is the unit-delay. The different initialization
value is decided by the first 7 bits of each PDU train. The
crambled data sesquence is encoded with a rate 1/2 con-
tion
volutional code with the generator polynomial g(1)(D) for
he upper connection and g(2)(D) for the lower connect
as follows:
(1)2 35 6
()1
g
DDDDD  (9)
(2)2 3 6
()1
g
DDDDD  (10)
where D is the unit-delay for convolutional codes and the
lowest-order term in the polynomial matches the input
stage of the shift register. The puncturing pattern [10] is

272 J. MAR ET AL.
Mapper S/P Pilot
Insertion IFFT Cyclic Prefixing P/S
P/S S/PFFT Remove
CP
Channel
Estimation
S and Π
Demapper
Time-Varing
Channel
+
Binary
Data
O
utpu
t
Data
AWGN
m
X

m
x

c
m
x

m
Y

m
y

c
m
y

m
Convolutional
h

Scrambler
Descrambl er
Encoder
Analog
Viterbi
Decoder
Inteleaver
Analog
DeInteleaver D/A
48 64 6480 48 64 6480
m
w

48 64 64 80 48 64 6480
Figure 6. The block diagram of the DSRC base band model.
used to make a rate 3/4 convolutional code from the rate
1/2 convolutional code. A convolutional code may cor-
rect many well-spaced errors, while being unable to han-
dle an error burst introduced by the fading channel. The
block interleaver or deinterleaver pair [10] applied to the
DSRC system can spread the burst error across onto
nonadjacent subcarriers and mapped alternately onto less
and more significant bits of
rocessing the scrambler, convolution encoder and inter-
the constellation. After
p
leaver, followed by mapping to BPSK or QAM constel-
lation points, the transmitting data stream is divided into
several parallel bit streams. An OFDM signal is built
using an 64-points inverse fast Fourier transform (IFFT).
The input vector to the IFFT is given as
,0,1, 1
[ ,...,]T
mmmmN
XXX, X



(11)
where Xm,k represents the kth subcarrier of the mth OFDM
symbol and N is 64 in the DSRC system. The IFFT out-
put signal vector is
,0,1,1
[ , ,...]T
mmmmN
xxx, x


 (12)
mth OFDM sym-where xm,n is the nth sample point of the
bol.

,, ,
0
exp(2)IFFT
mn mkmk
k
xXjnkX
N



 (13)
The cyclic prefixes (CP), which are produced with the
copies of the
1
1N
last parts of the OFDM symbol, are pre-
pended to the front of each vector m
x
. The cyclic pre-
fixing ou put signal vector is represented as
c
m
x

t
(14)
w
,0 ,1,1
,,1,1,0,1,
[ ,...,]
[,,...,,,...,
ccc T
mmmNq
mN qmN qmNmmmN
xx, x
xx xxx,x

 

1 ]T

here ,
c
mn
x
is the nth sample point of the mth OFDM
symbol and q is the length of the CP. Therefore, the re-
ceived signal vector is given by
,0 ,1,1
[ ,...,
cccc c
mm m mmmmNq
yxhwyy, y

 

 ]
T
(15)
Where stands for linear convolution,
m
h

and
m
w
are the channel impulse response vector anaddi-
tive white Gaussian noise (AWGN) vector for the mth
OFDM symbol, respectively.
d the
,
c
mn
y
is the nth sample
point of the mth OFDM symmth received signal
vector
bol in the
c
m
y

. The channel im response vector pulse
,0 ,,1, 1
[,... ]T
mmm mN
hhh,h



can besented by [12]: repre
2 1
,
0
N
mn i
i
hhe

(),
i
 0
Di
jfTn
mnN
 


1


se i
T
and τi
 (16)
ntercarrier interfer-
ence (IChe received signals; is the sam
λ spread index; is the ith-
where s the complex impulse response of the mth
OFDM symbol in the ith path; fDi is the ith-path Doppler
frequency shift, which may cau
m
i
hi
I) for t
is the delay
mal
ple period;
path delay
time norized by sampling time.
After removing the CP, the received signal vector m y

is
,0,1,1,,1,1
[,...,] [,...,]
Tccc T
mmmmNmqmqmNq
yyy,yyy,y



(17)
where ym, is the nth sample point of the mth OFDM sym-
bol. Thodulated received signal vector is
..., ]T ,Y

n
e dem
,0 ,1
[ ,
mm
m
YYY

, 1
mN
(18)
where

,, ,
0
FFT
N
mk mn mn
n
ye y


 (19)
Suppose the guard interval is longer than the length of
the channel impulse response, that is, there is no inter-
symbol interference between the OFDM symbols, the
demodulated sample vector m
Y

can
2
1
Njnk
Y


then be represented
as
mm
W
[13]
mmm
YXHI


,
,,H
(2) 0
,0 ,11
[ ,...]T
mmmmN
HHH 

(21)
,0,1, 1
[ ,...,]T
mmm mN
III, I



(22)
2
)N
1


,
0
sin( i
Di
i
i
jl
jfT D
m
mki
D
fT
Hhe e
fT



, 01kN
i



 
(23)

J. MAR ET AL.273
2() 2
11
,,
2()
00
11 , 01kN
1
Di
i
Di
jfkK
NjK
mN
mkimkjfkK
iK N
Kk
e
IhXe
Ne












n
ic
ted signals Xm,k. Im,k is the ICI part of the received
signal at the kth subcarrier of the m
pending on the signal values Xm,k m
carriers.
On the highway, the maximum vehicle speed is 200
km/hr. The DSRC system needs mo
and phase synchronization technology. Four uniform pilot
subcarriers, which are inserted in the positions of the ,
20th, 34th, and 48th subcarriers for each of the transmitted
DSRC data symbols, are applied for the DSRC receiver to
estimate the frequency and track the phase of the rece
signals. A pilot-based frequency synchronizer mechanism
including least squares estimation (LSE) and interpolation
is used for equalizing the pilot signal-aided frequency and
phase synchronization [14].
M OFDM (24Mbps) and
4-QAM OFDM (27Mbps) modulations are 9, 12, 17, 24,
ctively, which are used as a basis for
ver performance. The 3Mbps, 6Mbps,
(24)
where {}
mm
WFFTw
. Hm,k is recognized as the accu-
rate channel frequency respose at the kth subcarrier of
the mth OFDM symbol, whh is independent of the
transmit
th OFDM symbol, de-
odulated on all sub-
re robust frequency
6th
ived
4. Simulations
Packet detection, timing synchronization and coarse fre-
quency offset estimation of the DSRC receiver are per-
formed according to the algorithms provided in [15]. The
simulations focus on comparing the DSRC system per-
formance among the proposed S and Π-decision decoder,
hard-decision and soft-decision decoders. The DSRC
system is specified in the 5.85-5.925GHz ITS radio ser-
vices band. In a DSRC system, one frame has 100
OFDM symbols [7]. The total number of subcarriers is
64 including four uniformly distributed pilot subcarriers
and 12 guard subcarriers. According to the IEEE 802.11p
standard, the minimum input signal to noise ratio values
of the DSRC receiver for BPSK OFDM (3Mbps), QPSK
OFDM (6Mbps), 16-QAM OFDM (12Mbps), 16-QAM
OFDM (18Mbps), 64-QA
6
25 and 27dB respe
valuating the receie
and 12Mbps data transmission rates are made by using
1/2-rate convolutional code. The 18Mbps and 27Mbps
data transmission rates are produced by using 3/4-rate
convolutional code. The 24Mbps data transmission rate
is produced by 2/3-rate convolutional code. The quanti-
zation loss for the digital decoding is also considered in
the decoding BERs of Figures 9 and 12.
Based on analysis results in [16], the quantization
loss for convolutional decoding relative to the con-
tinuous case for two-level, four-level and eight-level
digital Viterbi decoders are evaluated as 7.73, 1.78,
and 0.52 dB with the 1/2-rate convolutional code;
11.43, 2.19 and 0.55 dB with the 3/4-rate convolutional
code. These losses remain roughly constant across the
range of BER plotted. Referring to the SPICE simula-
tion results in [6], the loss of real analog Viterbi de-
coder with 0.002 ACS noise and 10% comparator off-
set is less than 0.2 dB roughly compared with the ideal
analog Viterbi decoders. Therefore, the ACS noise and
the comparator offset will not be considered in the fol-
lowing simulations.
In the DSRC system, the coherence time Tc is calcu-
lated by
0 4230.4230.423645.3 sec
c
DDDc
. c
Tfvvf

 (25)
where fD=fDi for i=1 and 2. The maximum Doppler shift
is given by D
D
v
f

, where vD is the maximum vehicle
speed, fc is the carrier frequency, and c is the velocity of
light. Jakes’ channel model [17] is used to produce a
time- varying Rayleigh fading channel simulator. The
effects of AWGN and carrier frequency shift are also
considered in the DSRC channel. Simulations are carried
out for the vehicle speed vD = 200 km/hr, delay spread τ
= 200 nsec, 100 data symbols and different decision
Viterbi decoder.
Figure 7 shows when the delay spread exceeds 150
nsec, the severer frequency selective channel fading will
be
nce of the DSRC receiver with BPSK OFDM
(3Mbps), QPSK OFDM (6Mbps), 16-QAM OFDM
(12Mbps), 16-QAM OFDM (18Mbps), 64-QAM
(24Mbps) and 64-QAM OFDM (27Mbps) mo
are shown in Figure 8–12 respectively. Figure 8 shows
g three diffe
caused by reducing coherent bandwidth. The BER
performa
OFDM
dulations
the BER of QPSK OFDM modulations for Viterbi de-
coders usinrent decision rules are reduced to
0246810 12 14 16 18 20
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
BER
SNR(dB)
Delay spread = 100ns
Delay spread = 110ns
Delay spread = 120ns
Delay spread = 130ns
Delay spread = 140ns
Delay spread = 150ns
Delay spread = 200ns
Figure 7. BERs of the DSRC system using pilot subcarrier-
aided equalizer in different delay spread for 16-QAM
OFDM modulations.

274 J. MAR ET AL.
02 4 6 810 12
10
-6
10
-4
10
-2
10
0
BER
SNR(dB)
hard-decision
four-level soft-decision
S and -decision
Figure 8. Comparisons of BERs of the QPSK DSRC system
less than 10-5 at the minimum signal-to-noise ratio (SNR),
which meets the requirements specified in the IEEE
802.11p standard.
When the quantization loss for the convolutional
decoding is considered, the BER of 16-QAM OFDM
DSRC system is shown in Figure 9, where using a
hard- decision Viterbi decoder is higher than 10-5 at the
minimum SNR (17dB) for the case of the 12Mbps data
transmission rate. The 16-QAM OFDM DSRC system
using the four-level and eight-level soft-decision and
analog Viterbi decoders will be reduced to less than
10-5 at the minimum SNR (17dB), which meet the re-
quirements specified in the IEEE802.11p standard. It is
noted the S and Π-decision decoders results in a coding
gain of 1.5 dB and 5.2dB compared to the eight-level
using the hard-decision de-
oder cannot be lower than 10-5, when the data trans-
-ray Rayleigh fading channel envi-
ro
shown in Figure 12, where the quantization loss for the
(6 Mbits/sec; 200km/h; code-rate is 1/2) in terms of the dif-
ferent decision decoders.
and four-level soft-decision decoders, respectively,
when the quantization loss for the convolutional de-
coding is considered. Figure 10 shows the BER of
16-QAM DSRC system
c
mission rate increases to 18Mbps. The hard-decision
curve has a floor that is generated by the hard-decision
loss under the two
nment. The 16-QAM DSRC system using both the
soft-decision decoder and the S and Π-decision decod-
ers will reduce the BER to less than 10-5 at the mini-
mum SNR (24dB), which meets the requirements
specified in the IEEE802.11p standard. The S and
Π-decision decoder has a 1.5dB coding gain compared
with the four-level soft-decision decoder.
Figure 11 shows the BER of 64-QAM DSRC system
under the conditions of 18 Mbits/sec data rate and
200km/h vehicle speed cannot be lower than 10-5 for
three different Viterbi decoders. All three BER curves
have the floors that are caused by the high-order 64-
QAM modulation DSRC system operated under the fast
fading channel. The BER of 64-QAM DSRC system is
0510 15 20 25
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
BER
SNR(dB)
hard decision
four- level so ft-decision
eight-level soft-decision
S and  -decision
Figure 9. Comparisons of BERs of the 16-QAM DSRC sys-
tem (12 Mbits/sec; 200km/h; code-rate is 1/2) in terms of
the different decision decoders with quantization loss.
05 1015 20 2530
10
-6
10
-4
10
-2
10
0
BER
SNR(dB)
hard-decision
four-level soft-decision
S a nd

-decision
Figure 10. Comparisons of BERs of the 16-QAM DSRC
system (18 Mbits/sec; 200km/h; code-rate is 3/4) in terms of
the different decision decoders.
05 10 152025 30
10-6
10-4
10-2
100
BER
SNR(dB)
hard-decision
fou r-level soft -decisi on
S and  -decision
Figure 11. Comparisons of BERs of the 64-QAM DSRC
system (24 Mbits/sec; 200km/h; code-rate is 2/3) in terms of
the different decision decoders.

J. MAR ET AL. 275
NSC 96-2219-E-155-005.
7. References
[1] K. M. Lee, D. S. Han, and K. B. Kim, “Performance of
the Viterbi decoder for DVB-T in Rayleigh fading chan-
nels,” IEEE Trans. Consumer Electron., Vol. 44, pp.
994–1000, August 1998.
[2] S. B. Wicker, “Error control systems for digital communi-
cation and storage,” New Jersey, Prentice-Hall, Inc., 1995.
[3] F. Tosato and P. Bisaglia, “Simplified soft-output de-
mapper for binary interleaved COFDM with application
to HIPERLAN/2,” in Proc. IEEE Int. Conf. Communica-
tions, Vol. 2, pp. 664–668, April 2002.
[4] E. Akay and E. Ayanoglu, “Low complexity decoding of
bit-interleaved coded Modulation for M-ary QAM,” in
Proc. IEEE Int. Conf. Communications, Vol. 2, pp. 901–
905, June 2004.
01020 3040
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
BER
SNR(dB)
har d decision
four-l evel soft -deci si on
eight-l evel soft-deci sion
S and

-decision
Figure 12. Comparisons of BERs of the 64-QAM DSRC
system (27 Mbits/sec; 120km/h; code-rate is 3/4) in terms of
the different decision decoders with quantization loss.
owed the 64-QAM DSRC
system using three different decoders can be reduced to
less than 10-5 at the minimum SNR (27dB), which meets
the requirements specified in IEEE802.11p standard. The S
and Π-decision decoder results in a coding gain of 2.5 dB,
5dB and 15 dB compared to the eight-level soft-decision,
four-level soft-decision and hard-decision Viterbi decoders,
respectively, when the quantization loss for the digital
Viterbi decoding is considered.
5. Conclusions
A new S and Π-decision rule is proposed for the analog
demapper of the OFDM system. The DSRC system is
chosen as an OFDM platform to compare the perform
ance of the S and Π-decision decoder, which consists
ain of the S and Π-decision decoder relative t
convolutional decoding is considered and the vehicle speed
reduces to vm = 120 km/hr. It sh
-
of
an S and Π-decision demapper, analog deinterleaver and
an analog Viterbi decoder, with hard-decision and soft-
decision decoders. Simulation results show the coding
og
four-level and eight-level soft-decision decoders are
evaluated as 5.4dB and 1.5 dB, respectively, with the
16-QAM DSRC system; and 5dB and 2.5 dB with the
64-QAM DSRC system when the quantization loss for
the convolution decoding is considered. Each analog
demapper output can be determined by looking-up S and
Π table 2
log
M
times for M-ary QAM OFDM systems.
Many other applications related to OFDM with the pro-
posed S and Π-decision decoder are possible.
6. Acknowledgement
he authors would like to acknowledge gratefully the
search grants from National Science Council, Taiwan,
] K. He and G. Cauwenberghs, “Performance of analog
idwest Symposium on Circuits
ust 1999.
een Roadside and Vehicle Sys-
tem–5Ghz Band Dedicated Short Range Communications
ccess Control (Mac) and Physical Layer
ations, IEEE802.11p, December 2005.
decoders,” IEEE
main processing,” in
peci-
ett., Vol. 7, No. 9, pp. 446–448, Sep-
Communications, New
York: IEEE Press, 1974.
[5
Viterbi decoding,” 42nd M
and Systems, Vol. 1, Aug
[6] K. He and G. Cauwenberghs, “Itergrated 64-state parallel
analog Viterbi decoder,” ISCAS 2000, Vol. 4, May 2000.
[7] Standard Specification for Telecommunications and Infor-
mation Exchange Betw
(DSRC) Medium A
(PHY) Specific
[8] S. Coleri, M. Ergen, and A. Bahai, “Channel estimation
techniques based on pilot arrangement in OFDM sys-
tems,” IEEE Trans. Broadcast., Vol. 48, pp. 223–229,
September 2002.
[9] J. Giarratano and G. Riley, Expert Systems, 2nd. Edition,
PWS Publishing Company, 1994.
[10] J. Heiskala and J. Terry, OFDM Wireless LANs: A Theo-
retical and Practical Guide, Sams, 2001.
[11] V. C. Gaudet, R. J. Gaudet, and P. G Gulak, “Programma-
ble interleaver design for analog iterative
Trans. Circuits Syst. II, Vol. 49, pp. 457–464, July 2002.
[12] D. G. Manolakis, V. K. Ingle, and S. M. Kogon, “Statis-
tical and adaptive signal processing,” McGraw-Hill, Edu-
cation, 2000.
[13] Y. Zhao and A. Huang, “A novel channel estimation
method for OFDM mobile communication systems based
on pilot signals and transform-do
Proc. IEEE VTC, Vol. 3, pp. 2089–2093, May 1997.
[14] J. Rinnie and M. Renfors, “Pilot spacing in orthogonal
frequency division multiplexing systems on practical
channels,” IEEE Trans. Consumer Electron., Vol. 42, No.
4, pp. 959–962, November 1996.
[15] IEEE 802.11a, IEEE standard for Wireless LAN Medium
Access Control (MAC) and Physical Layer (PHY) s
fications, 1999.
[16] M. R. G. Butler and A. R. Nix, “Quantization loss for
convolutional decoding in Rayleigh-fading channels,”
IEEE Commun. L
tember 2003.
[17] W. C. Jakes, Microwave Mobile
T
re