Вы находитесь на странице: 1из 2

# : and the bit error probability can be upper bounded assuming that is

Signal set for ring-BCM half of the symbol error probability, so that:
J. Castifieira Moreira, B. Honary and P.G. Farrell
A new signal set is constructed using 4QAM and Walsh-Hadamard
functions and is applied in ring-block coded modulation
(ring-BCM) schemes. These 4QAM (M/2)-dimensional signal set As an example, the upper bound on bit error probability for the case
ring-BCM schemes show better performance than equivalent k= m =4, which corresponds to a transmission using the 4QAM
MPSK ring-BCM schemes. Asymptotic coding gains of 7 to 8 dB 8-dimensional signal set, predicts a coding gain of around 5 dB at
are obtained, in comparison to nncoded BPSK. Pbe= which is confirmed by simulation.

Introduction: Block-coded modulation (BCM) is a combined coding Ring-block-coded modulation using 4QAM (M/Z)-dimensional signal
and modulation technique that uses coding and modulation as one set: In the following Sections, results for circulant and pseudocyclic
entity, as suggested by Massey [l]. Forney 121 generalises this concept multilevel codes , designed over MPSK constellations, are
by defining the so-called signal space coding schemes. In a signal extended to the proposed 4QAM (M/2)-dimensional signal set. The
space code, we can identify a multilevel encoder, a signal space, and a block diagram of these schemes is shown in Fig. 1. The multilevel
mapping procedure between these two entities. Multilevel (non- source provides a sequence a = [ala* . . . ad of k elements of Z,, that
binary) encoders and decoders are found to be easily designed and are input to a multilevel encoder, which generates a sequence
effective if they operate over the ring of integers modulo-Q, Z,. This c=[c1c2.. .e,,] of n elements of the same ring, encoded such that
concept, introduced separately and simultaneously by Massey and n > k. Each element of a ring of integers modulo-Q is mapped into a
Mittelholzer  and Baldini et al. , is selected as the coding symbol s(c,) that belongs to the 4QAM (M/2)-dimensional signal set.
technique for our BCM schemes, which will be referred to as ring-
block-coded modulation (ring-BCM) schemes.

--
In this Letter we propose an efficient signal set that splits the s@,)
encoding between a multilevel encoder and a baseband encoder aez; C€Z;; constellation
, I signal

## rature modulator. The new signal set is called a 4QAM (M/2)-dimen-

sional signal set and is used in combined block coding and modulation
schemes over rings. The systematic linear circulant and pseudocyclic
1 - 1 I
codes proposed in [ 5 ] are used in these ring-BCM schemes, redesigned
for the new signal set, to provide an improvement in performance by 2 Fig. 1 Block diagram of modulator for ring-BCM scheme
to 3 dB over equivalent MPSK ring-BCM schemes (presented in ) in
comparison with uncoded BPSK. Thus, these signal space coding
schemes are improved by optimising the Euclidean distance properties
of the corresponding signal set. These schemes offer asymptotic coding NRI and RI systematic circulant and pseudocyclic linear ring-
gains of 7 to 8 dB for the AWGN channel, in comparison to uncoded block-coded modulation schemes using 4QAM 4-dimensional signal
BPSK. set: Table 1 lists the parameters of RI systematic circulant linear 213-
rate ring-BCM schemes over Z, for the 4QAM 4-dimensional signal
set. The sequence of integers from Z, for each code is the first column
4QAM (M/2)-dimensional signal sei: The proposed set is designed as of the circulant matrix forming the generator matrix of the code. N is
a combined coding and modulation signal set. The baseband encoder the number of nearest neighbours of each codeword in the signal
takes m (m = 2‘, i = 1, 2, 3, . . .) basic Walsh-Hadamard waveforms space. The asymptotic coding gain (ACG) is evaluated as in . Table
corresponding to an uncoded word of m bits, to form an orthogonal 2 lists the parameters of pseudocyclic linear 2/3-rate ring-BCM
code. By taking the inverses of these signals, a biorthogonal code with schemes over Z,. Here the sequence of integers represents the top
2m signals is generated. The in-phase input of a quadrature modulator row of the pseudocyclic matrix that when cyclically shifted, forms the
is then fed with one of the 2/71 biorthogonal waveforms and the remaining rows of the generator matrix of the code.
quadrature input with either the same waveform or its inverse to
generate one of the M=4m possible modulated signals. This is
equivalent to the use of a repetition code, which is a maximum
separable distance code, and it maximises the minimum Euclidean
distance of the signal set. The signal set is characterised by a code rate
(in bits per signal element) of: n 2/3 rate ring-BCM schemes, diC ACG, N
40AM 4-dimensional signal set over BPSK (dB)
r = - log, M -
2 + log, m 6 5677
I

12.0 4.77
j I

96
m m 9 165777 12.0 4.77 20
For example, if m = 2, then there are two basic signals in the orthogonal 12 3 17742 10 16.0 6.02 167
code, four signals in the biorthogonal code, and eight modulated signals
generated by the quadrature modulator. The rate for this case is r = 312.
Other examples are for m = 4 and m = 8, which have M = 16 and Table 2: RI pseudocyclic 2/3 rate ring-BCM schemes using
M= 32 modulated signals, with rates of unity and 5/8, respectively. It is 4QAM 4-dimensional signal set
noted that these signal sets are geometrically uniform (GU) .
A mapping between the Q elements of the ring of integers modulo-Q
and the A4 signals of the signal set is defined, where M= Q. When 4QAM 4-dimensional signal set over BPSK (dB)
m=2, the first two Walsh-Hadamard functions are combined with 677700
4QAM to generate a set of eight signals that can be mapped into the 05 1670000 12.0 4.77 20
ring of integers modulo-8. This signal set is useful to overcome encoder 12 366773000000 16.0 6.02 190
rate penalties of up to 213. When m = 4, the first four Walsh-Hadamard
waveforms are combined with 4QAM to generate a set of 16 signals
that can be mapped into the ring of integers modulo-16. This set of 16
signals is not able to cancel an encoder rate penalty. If m > 4 a rate NRI and RI systematic circulant and pseudocyclic linear ring-block-
penalty, as a consequence of (l), has to be accepted for these 4QAM coded modulation schemes using 4QAM &dimensional signal
(M/2)-dimensional signal sets. set: Tables 3 and 4 list the parameters of RI systematic circulant
For schemes using these signal sets, the symbol error probability can and pseudocyclic linear 1/2-rate ring-BCM schemes over Z I 6for the
be upper bounded (union bound) using the same procedure presented in 4QAM 8-dimensional signal set, respectively.

## ELECTRONICS LETTERS 4th July 2002 Vol. 38 No. 14 72 1

Table 3: RI systematic circulant linear 1/2 rate ring-BCM schemes J. Castiiieira Moreira (Electronic Department, School of Engineering,
using 4QAM 8-dimensional signal set Mar del Plata University, .I B. Justo 4302, Mar del Plata, 7600,
Argentina)

## over BPSK (dB) E-mail: casti@fi.mdp.edu.ar

4QAM 8-dimensional signal set
12 8 14 15 B. Honary and P.G. Farrell (Department of Conzmunication Systems
10 8 13 14 15 15 40.0 6.98 60 (DCS), Lancaster University, Lancaster, LA I 4 YR. United Kingdom)
12 10 12 9 15 3 0 48.0 7.78 396

References
1 MASSEY, J.L.: ‘Coding and modulation in digital communications’. Int.
Zurich Seminar on Digital Communication, Zurich, Switzerland, March
1974
2 FORNEY, G.D : ‘Geometrically uniform codes’, IEEE Trans. Inf Theory,
n ring-BCM schemes,
1 / 2 rate d:, ACG, N
1991,37, (5), pp. 1241-1260
4QAM &dimensional signal set over BPSK (dB) 3 MASSEY, J.L., and MITTELHOLZER, T.: ‘Codes over rings - practical
10 I
13 15 12 15 15 7 0 0 0 0 40.0 6.98 1IO necessity’. AAECC Symp., Toulouse, France, June 1989
12 I 11 8 10 7 1 14 15 0 0 0 0 0 I 48.0 I 7.78 I 456 4 BALDINI FILHO, R., PESSOA. A.C., and ARANTES, D.S.: ‘Systematic linear
codes over a ring for encoded phase modulation’. Int. Symp. on
Information and Coding Theory (ISICT’87), Campinas, S.P., Brazil,
July-August 1987
It is interesting to note that the circulant schemes offer the same ACG 5 BALDMI FILHO, R., and FARRELL, P.G.: ‘Coded modulation based on rings
as the pseudocyclic schemes, but with significantly fewer nearest of integers modulo-q. Part 1: Block codes’, IEE Proc., Commun., 1994,
neighbours. 141, (3), pp. 129-136
For all the cases, equivalent NRI schemes with similar distance 6 LINDSEY, W.C., and SIMON, M.K.: ‘Telecommunication systems

properties were also found, with the same performance as the corre- engineering’ (Prentice-Hall, NJ, USA, 1973)
7 CASTIEIRA MOREIRA, J.: ‘Signal space coding over rings’. PhD Thesis,
sponding RI schemes, so the latter only are listed, as they are more
Lancaster University, May 2000
useful in practice. Fig. 2 shows the performance of the bit error rate for
the (12, 8) (3 1 7 7 4 2 1 0) RI systematic circulant linear ring-BCM
scheme (Table 1) for the 4QAM 4-dimensional signal set, in compar-
ison with BPSK. Decoding is performed using a versatile soft decision
decoder, based on the Chase algorithm . TTCM assisted genetic-algorithm aided
reduced-complexity multiuser detection
S.X. Ng, K. Yen and L. Hanzo
10-1 r
Turbo trellis coded modulation assisted genetic-algorithm aided
reduced-complexity multiuser detection (TTCM-GA-MUD)is capable
10-2 I
of providing a considerable coding gain of 2.9 dB without any
bandwidth expansion at a bit error ratio of while maintaining
10-3 a complexity reduction factor 8.59 x IO6 compared to the optimum
a f i multiuser detector in the context of 16QAM transmissions, while
10-4:
supporting K = 10 users.

## 10-5 ; Introduction: The optimal code division multiple access (CDMA)

multiuser detector (MUD) [ l ] based on the maximum-likelihood
(ML) detection rule perfoms an exhaustive search of all the possible
104
combinations of the K numtier of users’ M-ary symbol sequences and
I
then selects the most likely combination from the set of legitimate
10-2 combinations as the K-component M-ary detected symbol sequence.
-2 0 2 4 6 8 1 0 1 2
EbINo, dB
Since an exhaustive search I S conducted, the computational complex-
ity of the detector increases exponentially with the number of users K
Fig. 2 Bit error rate for (12, 8) (3 I 7 7 4 2 1 0) RI systematic circulant as well as with the number of phasors in the modulation constellation
ring-BCM scheme (Table I ) for 4QAM 4-dimensional signal set, in employed. Genetic algorithms (GAS) have been used for efficiently
coniparison with BPSK (AWGN channel) solving combinatorial optirnisation problems in numerous applica-
BPSK tions . Recently, GA assisted MUD (GA-MUD) has been studied
-0-0- (12, 8) (3 1 7 7 4 2 1 0) RI systematic circulant ring-BCM scheme using binary phase shift keying (BPSK) modulation in the context of a
CDMA system [3, 41, which exhibits a substantially reduced
complexity compared to the optimum MUD, while maintaining a
Conclusions: The proposed signal set for combined coding and
similar performance. Turbo trellis coded modulation (TTCM) [ 5 ] is a
modulation schemes is designed by expanding the biorthogonality
channel coding scheme, which has a structure similar to that of the
family of power efficient binary turbo codes, but avoids extending
two-dimensional modulation and the inherent use of coding in the
the required bandwidth by absorbing the parity bits upon extending
Walsh-Hadamard biorthogonal signal set are combined to provide the
the phasor constellation.
improved distance properties and performance of the new scheme.
4QAM 4-dimensional signal set ring-BCM schemes designed for this
constellation show asymptotic coding gains of up to 6.02 dB. 4QAM
TTCM-GA-MUD: We consider a synchronous CDMA uplink system
8-dimensional signal set ring-BCM schemes offer up to 7.78 dB.
as shown in Fig. 1, where K users simultaneously transmit data
Though this performance will be reduced somewhat due to the effect
packets of 1000 TTCM encoded symbols using M-ary modulation
of the relatively large number of nearest neighbours (since the code
to a MUD over AWGN channels. The TTCM scheme employed a
distance profile is very uniform), it is better than that of equivalent
code memory of three and four turbo decoding iterations. The
MPSK ring-BCM schemes by about 2 to 3 dB.
received signal can be written as: r(t)=s(t)+n(t), where s(t) is the
sum of the transmitted signals of all users and n(t) is the zero-mean
0IEE 2002 16 November 2001 complex AWGN with independent real and imaginary components,
Electronics Letters Online No: 20020488 each having a double-sided power spectral density of a2=N0/2. It can
DOI: 10.1049/e1:20020488 be shown that the log-likelihood function (LLF) for a vector 6 is given