Simplified Soft-Output Demapper for Binary Interleaved

COFDM with Application to HIPERLAN/2

Filippo Tosato1, Paola Bisaglia
Imaging Systems Laboratory
HP Laboratories Bristol
HPL-2001-246
October 10th , 2001*
In this paper a comparison in performance between hard
and soft decision Viterbi decoding, with application to the
HIPERLAN/2 standard, is presented. The results show
that when channel state information (CSI) is included in
the generation of the soft decision information, the soft
decision method greatly outperforms the hard method.
Moreover, a simplified algorithm for the soft-output
demapper for the 16-QAM and 64-QAM constellations is
developed, which allows the complexity of the demapper to
be maintained at almost the same level for all the possible
modes of HIPERLAN/2.

* Internal Accession Date Only

Approved for External Publication

Dipartimento di Elettronica e Informatica, University of Padova, Via Gradenigo 6/A, 35131 Padova,
Italy
Copyright Hewlett-Packard Company 2001

Simplied Soft-Output Demapper for Binary Interleaved COFDM

with Application to HIPERLAN/2
Filippo Tosato 1 and Paola Bisaglia 2
1 Dipartimento di Elettronica
2 Hewlett{Packard

e Informatica, University of Padova, Via Gradenigo 6/A, 35131 Padova, Italy

Laboratories, Filton Road, Stoke Giord, Bristol BS34 8QZ, U.K.

Abstract | In this paper a comparison in performance between hard and soft decision Viterbi decoding, with application to the HIPERLAN/2 standard, is presented. The
results show that when channel state information (CSI)
is included in the generation of the soft decision information, the soft decision method greatly outperforms the
hard method. Moreover, a simplied algorithm for the
soft-output demapper for the 16-QAM and 64-QAM constellations is developed, which allows the complexity of the
demapper to be maintained at almost the same level for all
the possible modes of HIPERLAN/2.

I. Introduction

HIPERLAN/2, the European standard recently specied by ETSI BRAN, has been designed to provide several
data rates, selected according to the channel conditions,
for short-range communications in indoor and outdoor environments. OFDM has been chosen as the modulation
format because of its good performance in highly dispersive channels. The data rate ranging from 6 Mbit/s to
54 Mbit/s, can be varied by using various signal alphabets
for modulating the OFDM sub{carriers and by applying
dierent puncturing patterns to a mother convolutional
code. BPSK, QPSK and 16-QAM are used as mandatory
modulation formats, whereas 64-QAM is optional [1].
Previous results on HIPERLAN/2 have suggested the
use of a hard decision Viterbi decoder for decoding the
convolutional code, while the soft decision Viterbi decoder
has not been considered because of its greater computational complexity, especially when higher modulation formats are employed. In fact, in the case of the 16-QAM and
64-QAM constellations, each axis carries more than one
bit and the metric functions, in the soft-output demapper, to determine the soft information for each bit, are in
general quite complicated. Instead, in the case of BPSK
and QPSK, the soft information is simply proportional to
the distance from the decision boundary.
In this paper, after reviewing two dierent approaches
of implementing the soft-output demapper and soft-input
Viterbi decoder for multi-level modulations, we adapt the
Pyndiah's algorithm [2] to COFDM systems and propose
a simplied version of it which allows us to reduce the
complexity in the case of higher order constellations.
Simulation results show that in a typical HIPERLAN/2
channel a signicant improvement in performance is obtained using soft decision decoding compared to harddecision decoding.

The structure of this paper is as follows. In Section II

the HIPERLAN/2 system model is introduced, in Section
III two soft decoding schemes with bit metric calculation
and soft bit information are described and compared, in
Section IV a simplied soft bit computation is proposed.
Finally, in Section V some numerical results are presented
on the performance of HIPERLAN/2 with the proposed
low complexity soft Viterbi decoding.
II. System model

In Figure 1 the HIPERLAN/2 system model on which

we base our analysis is shown. At the transmitter, the
input bits, after scrambling, are convolutionally encoded
with a rate 1/2, constraint length 7 convolutional code1 ,
bit-by-bit interleaved and then converted into QAM symbols, according to Gray-coded constellation mappings.
This scheme is also called Bit-Interleaved Coded Modulation (BICM) [3]. The complex symbols are then fed to an
OFDM modulator, arranged into a physical frame called
physical (PHY) burst with an appropriate preamble, and
nally transmitted.
Let M = 22m be the number of symbols of the generic
square QAM constellation, so that m interleaved bits are
mapped into the in-phase and quadrature components of
the complex symbol2 . Let a[i] = aI [i] + jaQ [i] denote
the QAM symbol transmitted in the i-th sub-carrier and
fbI;1; : : : ; bI;k ; : : : ; bI;m; bQ;1 ; : : : ; bQ;k ; : : : ; bQ;m g the corresponding bit sequence. Assuming that the cyclic prex
completely eliminates ISI (Inter OFDM Symbol Interference) and ICI (Inter Channel Interference), then the received signal in the generic sub-carrier can be written as

r[i] = Gch (i)
a[i] + w[i] ;

(1)

where Gch (i) is the Channel Frequency Response (CFR)

complex coecient in the i-th sub-carrier and w[i] is the
complex AWGN noise with variance  2 = N0 .
At the receiver, the OFDM demodulator performs synchronization and channel estimation, the latter being used
by a zero-forcing equalizer which compensates for attenuation and phase shift in each data sub-carrier. If the
channel estimate is error free, the output of the one-tap
equalizer is given by
1 Two dierent puncturing patterns can be applied to obtain code
rates 3/4 and 9/16 [1].
2 For BPSK only the in-phase component is present.

PDU
train
from
DLC

PDU
train
to
DLC

bit
scrambling

convolutional
coding

bit
interleaving

QAM
mapping

OFDM
modulator

PHY
bursts

RF
transmitter

preamble
dispersive
channel

preamble
de
scrambling

de
coding

de
interleaving

QAM
demapping

OFDM
demodulator

useful
data

RF
receiver

Fig. 1. Block diagram of HIPERLAN/2 PHY layer.

y[i] = a[i] + w[i]=Gch (i) = a[i] + w0 [i] ;

(2)
where w0 [i] is still complex AWGN noise with variance
0 2 (i) = 2 =jGch (i)j2 .
III. Soft Maximum Likelihood Decoding for
Binary Interleaved Coded Modulation

Since the conditional pdf of r[i] is complex Gaussian

1
p (r[i] j a[i] = ) = p
exp
2

1 jr[i]
2

Gch (i) j2
2

(5)
and as r[i] = Gch (i)
y [i], the metrics (4) are equivalent to

mc (bI;k ) = jGch (i)j2
min( ) jy[i] j2 ; c = 0; 1 : (6)

2S

Soft Viterbi decoding for BICM employing BPSK or

QPSK modulations is straightforward as the soft bit information, before being weighted by the CSI coecients, is
simply given by the received signals for BPSK and by their
in-phase and quadrature components for QPSK. Therefore in the following discussion we will focus on the higher
modulation levels, namely 16- and 64-QAM.

Finally, these metrics are de-interleaved, i.e. each couple

(m0 ; m1 ) is assigned to the bit position in the decoded
sequence according to the de-interleaver map, and fed to
the Viterbi decoder which selects the binary sequence with
the smallest cumulative sum of metrics.

A. ML bit metrics

B. LLR soft bit values

Because interleaving is applied to the encoded bits before the QAM modulator, Maximum Likelihood Decoding
(MLD) of multi-level BICM signals would require joint demodulation and convolutional decoding and is therefore
quite complex to implement in practice [4]. In fact, according to the MAPSE (Maximum A Posteriori Sequence
Estimation) criterion the following maximization should
be performed to estimate the encoded bit sequence b:
bb = arg max P [bjr] ;
(3)
b
where r is the received sequence of QAM signals. We
also assume that perfect CSI is available, i.e. fGch (i)g
are known to the receiver. Thus, all possible coded and
interleaved bit sequences would have to be considered for
(3) to be evaluated. In [5], Zehavi proposed a decoding
scheme which consists in calculating sub-optimal simplied bit metrics to be used inside a Viterbi decoder for
path metric computation. For each symbol r[i], 4m metrics need to be derived, two for each in-phase and quadrature bit bI;k , bQ;k , corresponding to possible values 0; 1.
For bit bI;k (the same applies to bit bQ;k ) rst the QAM
constellation is split into two partitions of complex sym(0)
bols, namely SI;k
comprising the symbols with a `0' in
(1)
position (I; k ) and SI;k
which is complementary. Then
the two metrics are obtained by
m0c (bI;k ) = max( ) log p (r[i] j a[i] = ) ; c = 0; 1 : (4)
c
 2 SI;k

c
I;k

BICM allows a dierent MAPSE-based decoding

scheme in which the received QAM signals are rst demodulated by a soft-output demapper and de-interleaved,
and then passed to a standard binary soft-input Viterbi
decoder [6]. The idea is to demap the received signal into
soft bits which have the same sign as provided by a hard
detector and whose absolute value indicates the reliability
of the decision.
The optimum hard decision on bit bI;k (the same applies
for bit bQ;k ) is given by the rule
bbI;k =  if P [bI;k =  j r[i]] > P [bI;k = (1  ) j r[i]] ;
 = 0; 1 : (7)
Set  = 1, then (7) can be rewritten as

bbI;k = 1

if

log

P [bI;k = 1 j r[i]]
>0:
P [bI;k = 0 j r[i]]

(8)

Thus, the Log-Likelihood Ratio (LLR) of decision bbI;k is

dened as3
P [bI;k = 1 j r[i]]
LLR(bI;k ) = log
P [bI;k = 0 j r[i]]
P
 2 S (1) P [a[i] =  j r[i]]
= log P
; (9)
 2 S (0) P [a[i] =  j r[i]]
I;k

I;k

3 If

we set  = 0 in (7), then the LLR function shall be dened

with the opposite sign.

which is the soft bit information assigned to bit bI;k .

By applying Bayes rule and assuming that the transmitted symbols are equally distributed, relation (9) yields

P
 2 S p (r[i] j a[i] = )
LLR(bI;k ) = log P
:
 2 S p (r[i] j a[i] = )

c)
where subset S 0 (I;k
contains the real parts of the complex
(c)
symbols of subset SI;k
, for c = 0; 1.
Im
00 10

(1)

(10)

I;k

(0)

01 10

LLR(bI;k )  log

max 2 S (0) p (r[i] j a[i] = )

00 11

01 11

-3

-1

00 01

01 01

min jy [i]

(0)
 2 SI;k

= [m0 (bI;k )

00 00

01 00

(0)
S
I;1

min jy [i]

(1)
 2 SI;k

m1 (bI;k )]=4 :

(12)

(0)
(1)
Figure 2 shows the partitions (SI;k
; SI;k
) for the generic
(0)
(1)
bit bI;k , and (SQ;k ; SQ;k ) for the bit bQ;k , in the case of
the 16-QAM constellation. As can be seen they are delimited by either horizontal or vertical boundaries. Therefore, the two symbols within the two subsets, nearest to
the received equalized signal, always lie in the same row
if the partition boundaries are vertical (bits bI;1 and bI;2
in Figure 2) or in the same column if the boundaries are
horizontal (bits bQ;1 and bQ;2 in Figure 2). The same observation holds true for the 64-QAM constellation. As a
consequence, equation (12) can be rewritten as:
LLR(bI;k ) =

jGch (i)j
4

min(0) (yI [i]

I 2 S I;k
0

, jGch (i)j
DI;k ;

2

I )2

min(1) (yI [i]

I 2 S I;k
0

11 01

10 01

11 00

10 00

Re

-1

-3
(0)
S
I;2

(1)
S
I;1

1
-1
(1)
S
I;2

Re

(0)
S
I;2

-3

bI;1

bI;2
Im

(0)

(1)

S
Q;2

(1)
Q;1

IV. Simplified LLR computation

-3

(11)

In the Appendix we demonstrate that using the approximate bit metrics (6) for path metric calculation inside the
Viterbi Algorithm (VA) is equivalent to demodulating the
received signals into soft bit values according to (12) and
then employing a soft binary VA for decoding.
Yet, expression (12) allows further simplication with
signicant reduction of computational complexity and
negligible soft decoding performance loss. Moreover, with
the second scheme the same standard implementation of
soft-input Viterbi decoder for BPSK signals can be used
also for multi-level modulations.

(
2

10 11

Im

Using (5) in (11) and normalizing by 2= 2 the nal soft

bit values can be calculated as:
LLR(bI;k ) =
2

11 11

-1

I;k

(
j
Gch (i)j2
=

3
2dI;2

Sub-optimal simplied LLR

P can be obtained by the logsum approximation: log j zj  maxj log zj , which is
good as long as the sum in the left-hand side is dominated by the largest term, as typically occurs in channels
with high Signal-to-Noise Ratio (SNR). Thus
I;k

10 10

I;k

max 2 S (1) p (r[i] j a[i] = )

Im

11 10

I )2

(13)

1
-3

S
Q;2

-1

1
-1
S

Re

-1

-3

2dQ;2
1

-1

Re

(0)

Q;1

-3

(0)

S
Q;2

-3

bQ;1
bQ;2
Fig. 2. Partitions of the 16-QAM constellation used in HIPERLAN/2 PHY layer.

Evaluation of the terms DI;k for the in-phase bits of a

16-QAM symbol yields

8
>
jyI [i]j  2
<yI [i] ;
DI;1 = 2(yI [i] 1) ; yI [i] > 2
>
:2(yI [i] + 1) ; yI [i] < 2
DI;2 = jyI [i]j + 2 :

(14)
(15)

It can be easily veried that the DQ;k functions for the

two quadrature bits are the same as (14) and (15) with
yI [i] replaced by yQ [i].
Similarly, formulas can be derived from (13) for the inphase bits of the 64-QAM constellation:

8
yI [i] ;
jyI [i]j  2
>
>
>
>
>
2(
y
[
i
]
1)
;
2 < yI [i]  4
I
>
>
>
>
<3(yI [i] 2) ; 4 < yI [i]  6
DI;1 = 4(yI [i] 3) ; yI [i] > 6
>
>
>
2(yI [i] + 1) ;
4  yI [i] < 2
>
>
>
>
3(
y
[
i
]
+
2)
;
6  y [i] < 4
>
>
:4(yI [i] + 3) ; y [i] < I 6
I
8 I
>
<2( jyI [i]j + 3) ; jyI [i]j  2
DI;2 = 4 jyI [i]j ;
2 < jyI [i]j  6
>
:2( jyI [i]j + 5) ; jyI [i]j > 6
(
jy [i]j 2 ; jyI [i]j  4 :
DI;3 = I
jyI [i]j + 6 ; jyI [i]j > 4

(16)

(17)
(18)

SNRi / jGch (i)j2 :

(19)

Expressions (14) and (16)-(17) are still cumbersome to

evaluate, so we introduce a further simplication. Expression (14) can be approximated as follows:
DI;1 ' yI [i] ;
(20)
and the same applies to the rst quadrature bit.
The approximate expressions for (16)-(18) are given by

DI;1 ' yI [i]

DI;2 ' jyI [i]j + 4
DI;3 = jjyI [i]j 4j + 2 :

(21)
(22)
(23)

These approximations correspond to calculating jDI;k j

(or jDQ;k j) as the distance of the received equalized signal
y[i] from the nearest partition boundary and assigning to
DI;k (or DQ;k ) the sign + or - according to which partition
y[i] falls in.
In Figures 3 and 4, the approximate functions are plotted versus the theoretical ones for both 16- and 64-QAM.
6

DI;2

DI;1

Theoretical
Approximation

6
4

0
yI [i]

6
4

yI [i]

Fig. 3. Approximate versus exact LLR functions for the in-phase

bits of the 16-QAM constellation of Figure 2.

Formulas (20) and (15), and (21)-(23) can be generalized for any square QAM constellation with similar Gray

10

10
Theoretical
Approximation

DI;3

DI;2

Expressions (14)-(15) are similar to the LLR functions

for bit reliability derived by Pyndiah in [2], in the case of a
single carrier system over a Gaussian channel and applied
to near optimum decoding of product codes. However, unlike a single carrier system in which all data symbols are
aected by the same SNR on average, in a multi-carrier
OFDM system the various carriers suer from dierent
channel attenuation levels and so data conveyed by subcarriers having a high SNR are a priori more reliable than
those transmitted in sub-carriers with low SNR. This additional information is supplied to the decoder by weighting
the LLR functions by the square modulus of the CFR,
which represents the Channel State Information (CSI), as
can be seen in (12). Note also that the CSI coecient is
proportional to the SNR in the i-th sub-channel:

DI;1

10

10

10

yI [i]

10

yI [i]

yI [i]

Fig. 4. Approximate versus exact LLR functions for the in-phase

bits of the 64-QAM constellation.

labelling4 . Let dI;k and dQ;k denote half the distance between the partition boundaries relative to bit bI;k and
bQ;k , with k > 1 (see Figure 2). Then, the LLR functions can be approximated as:

yI [i] ;
k=1
jDI;k 1j + dI;k ; k > 1
LLR(bI;k ) = jGch (i)j2
DI;k ; k  1 ;
DI;k '

and

yQ [i] ;
k=1
jDQ;k 1 j + dQ;k ; k > 1
LLR(bQ;k ) = jGch (i)j2
DQ;k ; k  1 :
DQ;k '

(24)

(25)

Note that dI;k = dQ;k for all k , for the Gray labelling we
have considered.
V. Numerical results

The results, obtained by computer simulations, are

given in terms of Bit Error Rate (BER) versus Eb =N0 ,
where Eb is the energy per information bit. In order to
verify the performance of the hard and soft decoding for
the four possible modulation formats, modes 1, 4, 5 and 7
of HIPERLAN/2, using BPSK, QPSK, 16-QAM and 64QAM, respectively, have been employed. We note however
that similar results should be expected with the other
modes since they only dier in the puncturing schemes;
moreover from [7], mode 2 and mode 3 should perform
as mode 4 and mode 1, respectively. The curves have
been generated averaging over 2000 dierent realizations
of channel model A [7], typical for large oce environment with non-line-of sight propagation. In all cases, the
decision depth in the Viterbi decoder has been xed to 60.
Results for mode 1 (continuous line) and mode 4 (dashed
line) in Figure 5 show that at a BER = 10 4 the softdecoding gain compared to hard-decoding is equal to
4.5 dB and 6.5 dB, respectively. In Figure 6, where mode 5
(continuous line) and mode 7 (dashed line) are used, the
4 For

other Gray labelling patterns only minor changes are needed.

gain of the soft-decoding compared to hard-decoding is

equal to 6.5 dB and 8.5 dB, respectively, at a BER of 10 4 .
As it can be observed from Figure 6, in both cases of the
16-QAM and 64-QAM constellations, there is no performance degradation from the theoretical formulas given in
(14)-(15) and (16)-(18), using our proposed approximated
formulas in (24)-(25); therefore the latter can be used with
a reduction in computational complexity.
0

10

Mode 1: Uncoded
Mode 1: Hard Decoding
Mode 1: Soft Decoding
Mode 4: Uncoded
Mode 4: Hard Decoding
Mode 4: Soft Decoding

10

Appendix
Equivalence between soft decoding schemes
with bit metrics (6) and soft bits (12)

We indicate the soft decoding scheme with bit metric

generation (6) and Viterbi decoding as method 1 whilst
that with soft demapping (12) will be referred to as method
2. Table I shows the bit metrics for the two methods,
which are used by the VA to calculate the cumulative path
metrics along the trellis.
Method 1
Bit metric (decoded `0')

BER

Bit metric (decoded `1')

m0
m1
TABLE I

soft decoding methods.

3

10

10

15

20

25

30

35

E /N (dB)
b

Fig. 5. Comparison between hard and soft decoding for mode 1 and
mode 4 of HIPERLAN/2, with channel model A.
0

10

Mode 5: Uncoded
Mode 5: Hard Decoding
Mode 5: Soft Decoding, Theor.
Mode 5: Soft Decoding, Approx.
Mode 7: Uncoded
Mode 7: Hard Decoding
Mode 7: Soft Decoding, Theor.
Mode 7: Soft Decoding, Approx.

BER

10

10

10

mc

for

c = 0; 1

is given by (6).

Note that the dierence between the bit metrics for

the decoded `0' and `1' is the same for the two methods,
namely (m0 m1 ).
b be the decoded sequence chosen by the VA with
Let b
method 1, then its path metric is the smallest with ref(1)
b ) = Mmin
erence to method 1, i.e. M (1) (b
. We want to
b is also the decoded sequence obtained
demonstrate that b
(2)
b ) = Mmin
with method 2, that is M (2) (b
.
b0 6=
Assume that the decoded sequence by method 2 is b
bb and then M (2)(bb 0 ) < M (2)(bb ). Because the dierence
between the bit metrics relative to decoded `0' and `1'
is the same with both methods, it follows that also the
dierence between the overall metrics of two generic paths
in the trellis is the same, and then

M (1) (bb 0 ) M (1) (bb ) = M (2) (bb 0 ) M (2) (bb ) < 0 : (26)
b 0 ) < M (1) (bb ), which is in contrast with
Therefore M (1) (b
b is the path with the smallest cumuthe hypothesis that b

10

10

m1 ) + 1 2

m1 ) 1 2

Comparison between bit metric increments used by the two

10

10

Method 2

1
(m0
4
1 (m
0
4

15

20

25

30

35

E /N (dB)
b

Fig. 6. Comparison between hard and soft decoding with exact

and approximated computation of soft bit values for mode 5 and
mode 7 of HIPERLAN/2, with channel model A.
VI. Conclusions

In this paper we have presented some new results on

the performance of HIPERLAN/2 transmission modes
with soft Viterbi decoding. We have adopted a decoding scheme in which the received complex symbols are
demapped into soft bit information which is weighted by
the CSI coecients and then fed to a conventional soft
binary Viterbi decoder. This scheme is equivalent to that
proposed by Zehavi for BICM, requiring soft metric calculation. Simplied formulas for the LLR computation have
been proposed which show no performance loss compared
to the theoretical ones. Gains from 4.5 dB (for mode 1) up
to 8.5 dB (for mode 7), at BER = 10 4 have been obtained
over hard decoding.

lative sum of bit metrics for method 1. We can conclude

that the VA always selects the same decoded bit sequence
with either method.
References

[1] ETSI TS 101 475, \Broadband Radio Access Networks

(BRAN); HIPERLAN Type 2; Physical (PHY) layer, v1.2.2",
2001.
[2] R. Pyndiah, A. Picard and A. Glavieux, \Performance of Block
Turbo Coded 16-QAM and 64-QAM Modulations," in Proc. of
IEEE GLOBECOM '95, pp. 1039-43, Singapore, Nov. 1995.
[3] G. Caire, G. Taricco and E. Biglieri, \Bit-Interleaved Coded
Modulation," IEEE Trans. on Inf. Theory, vol. 44, pp. 927946, May 1998.
[4] X. Li and J.A. Ritcey, \Bit-Interleaved Coded Modulation with
Iterative Decoding," in Proc. of IEEE ICC '99, pp. 858-63,
June 1999.
[5] E. Zehavi, \8-PSK Trellis Codes for a Rayleigh Channel," IEEE
Trans. on Comm., vol. 40, pp. 873-884, May 1992.
[6] M. Speth, A. Senst and H. Meyr, \Low Complexity SpaceFrequency MLSE for Multi-User COFDM," in Proc. of IEEE
GLOBECOM '99, pp. 2395-99, Dec. 1999.
[7] A. Doufexi, S. Armour, A. Nix and D. Bull, \A Comparison
of HIPERLAN/2 and IEEE.11a Phisical and MAC Layers,"
VTC'2000.