Multicarrier Modulation Techniques for TETRA

II
Xianhui Zhang 1, Dongping Yao 2
State Key Laboratory of Rail Traffic Control and Safety
Beijing Jiaotong University
Beijing 100044, P.R. China
E-mail: [email protected] 1, [email protected] 2

TETRA standard, a multi sub-carrier approach with 8

sub-carriers per 25 kHz is used, i.e. 8, 16, 32 and 48 subcarriers in 25 kHz, 50 kHz, 100 kHz and 150 kHz carriers
respectively. The modulation types used on each subcarrier shall be 4-QAM, 16-QAM or 64 QAM. The
modulation symbol rate on each sub-carrier shall be 2400
symbols/s. The modulation gross bit rates are given in
table 1[3].

AbstractTETRA is a digital trunked communication

system. TETRA adopt Filtered Multitone Modulation
(FMT) multicarrier scheme to provide high speed data
service. This paper introduce modulation parameters which
are defined in TETRA standard and efficient implement,
then simulate and analyze the BER performance in AWGN
and multipath channel. 
Keywords-TETRA;multicarrier; FMT; DFE

I.

TABLE I.

INTRODUCTION

Terrestrial Trunked Radio (TETRA) is a digital

trunked communication system. TETRA supports voice
and dispatch services, also short data service and packet
data service. TETRA is used by government agencies,
emergency services, public safety networks, rail
transportation staff, transport services and the military in
more than one hundred countries in the world now[1][2].
In China, TETRA played a important role in 2008 Beijing
Olympic Games security. Beijing. Guangzhou and
Shenzhen subway are using TETRA for dispatching. With
increasing requirement for high speed data service,
TETRA has developed TETRA Enhanced Data Service
(TEDS) in TETRA Release 2. TEDS is a new TETRA
High Speed Data (HSD) service using different RF
channel bandwidths and data rates in multicarrier
approach. At present, TETRA equipments in market are
generations of TETRA I. Multicarrier modulation in
TETRA II is on the research and design stage. The
efficient implement and performance are concerned. In
this paper, we simulate the BER performance in AWGN
channel and multipath channel after equalized.
II.

GROSS BIT RATES FOR QAM CARRIERS (KBIT/S)

Table Column Head

Modulation
Type

25 kHz

50 kHz

100 kHz

150 kHz

4-QAM
16-QAM
64-QAM

38.4
76.8
115.2

76.8
153.6
230.4

153.6
307.2
460.8

230.4
460,8
691.2

Figure 1. Modulation Processing

fk

The standard gives the sub-carrier centre frequency

in Hz by the following formula:

THE MULTICARRIER MODULATION IN

TETRA II

fk

(0.5625  (M / 2  k ) u1.125) / T 

0,1,..., M  1 , where M is the number of subcarriers and T is the symbol duration which is 2400
for k

Multicarrier system divides data into several parallel

streams, so each subcarrier has a low symbol rate. That
means the channel time delay is smaller than symbol
period. Therefore multicarrier system is insensitive to time
dispersion and dont need complicated equalizer. In

symbols/s. This leads to a sub-carrier spacing of 2.7 kHz.

A block diagram of the modulation process is shown on
Fig. 1. g (t ) is the square root raised cosine pulse shape

___________________________________
978-1-4244-8268-9/11/$26.00 2011 IEEE





as

B. FMT efficient implementation

FMT is actually a application of uniform DFT filter
bank technology. In fact, FMT direct implementation is
complicated. The filters are required working on a high
rate T which is K times of symbol rate T0 . To achieve

According to the standard, the multi sub-carrier

modulation is Filtered Multitone Modulation (FMT). FMT
is proposed in [4]. FMT is a discrete time implementation
of a multicarrier

high level of subchannel spectral containment the length

of prototype filter always large that makes calculation
complicated. Therefore, efficient implementations are
necessary to make FMT practical.
This implementation has been proposed in [7].We
assume the sub-carrier to be f k k /( MT ) . If we

filter and
follows:

its

G( f ) 1
0.5(1  sin(S (2 f T  1) / 2D ))

G( f )
G( f )

G( f )

spectrum

f d (1  D ) / 2T

for

(1-D )/2T d f d (1  D ) / 2T

a1 (lT0 )

aM1(lT0)

nK

g(nT)

nK

g(nT)

nK

f t (1  D ) / 2T

ej2Sf0iT

ej2S f0iT
a0 (lT0 )

j2S f iT
e

g(nT)

noise

x(iT )

defined

for

for

is

channel

z0 (lT0 )

j2Sf iT h(nT)

pK

h(nT)

pK

h(nT)

pK

e
y(iT)

ej2SfM1iT

ej2SfM1iT

a 0 ( lT0 )

equalizer

z1 (lT0 )

a 1 ( lT0 )

equalizer

zM1(lT0)

aM 1(lT0 )

equalizer

compute the polyphase decomposition of (3) with period

,
assuming
T1 M1T

M1

l.c.m.(M , K )
x (mT1 )

Figure 2.

PM

LK [4],

x(iT  mT1 )

Baseband block of the FMT system.

i=0,...,M1  1, m =

M-1

system where sub-carriers are uniformly spaced and the

sub-channel pulses are identical [5].

= ak (lT0 ) g (iT  mT1  lT0 )e

FMT is different with OFDM scheme in the channel

shaping filers. FMT can deploys rectangular, Gaussian or
root raised cosine pulse shaping, but OFDM must use
rectangular pulse [6].

= ak (lT0 )e

M-1

M1 elements.

Therefore, the step to get the FMT signal is: an M-point

IDFT, cyclic extension of the outputs, filtering with the
i

pulses g (lT0 )

g (iT  lT0 ) , sampling with period

LT0 , and P/S conversion.

Similarly, the kth output signal of the FMT
demodulator is given by
M1 1

zk (lT0 )

y (iT  mT1 )h(lT0  iT  mT1 )e

 j 2S k

iT  m
M

m= i 0
M1 1

y(iT  mT1 )h(lT0  iT  mT1 )e

j

2S
ik
MT

i 0 m=

M1 1

= Z (lT0 )e

i = (3)

j

2S
k iP
M1T

i 0

(6)

The signal x (iT ) is transmitted over channel. The

where

Zi (lT0 )

received signal y (iT ) then passes through each subchannel. The outputs are

y(iT )e

ak (lT0 ) via an M-

point IDFT with a cyclic extension to

k 0 l=

zk (lT0 )

(5)

where A (lT0 ) are obtained from

At time iT the signal x (iT ) input to the channel is

given by
k

g (iT  mT1  lT0 )

l=

spacing 'f
f k  f k 1 is larger than 1/ T0 the scheme
is said to be non-critically sampled FMT, otherwise if
'f 1/ T0 it is said to be critically sampled FMT [7].

a (lT ) g (iT  lT )e

2S
ik
MT

= A i (lT0 ) g i (mLT0  lT0 )

K, each sub-channel symbol period is T T0 / K , then

each symbol stream is filtered by a baseband filter
g ( nT ) , referred to as a prototype filter. If the sub-carrier

x(iT )

l= k=0

complex-valued modulation symbols ak (lT0 ) is M-QAM

symbol which is transmitted on the sub-
  
M-1 at symbol rate 1/ T0 . After upsampling by a factor of

j 2S f k iT

1
( iT  mT1 )
MT

k=0 l=

A. FMT direct implementation

Fig. 2 shows a FMT communication system. M is the
number of sub-channels; f k is the kth sub-carrier. The

M 1

j 2S k

 j 2S f k iT

y(iT  mT )h(lT
1

 iT  mT1 )

m=

h(lT0  iT ) k=0,...,M-1

i=

= yi (mT1 )h  i (lT0  mT1 )

(7)

m=

Therefore, the efficient implementation demodulation step

is S/P conversion of the received signal, interpolating the

(4)

This work is supported by the Program for Changjiang Scholars and

Innovative Research Team in University (No. IRT0949).


M 1 polyphase components of the input signal by a factor

i

L, then filtering with pulses h (lT0 )

followed by a

hoverall

h(lT0  iT ) ,

M 1 -point DFT and sampling the outputs

nk

h o vera ll

Hard
decision

[ f 0 , f1 ,..., f Nf 1 ]T

(11)
(12)

the feedforward and feedback filter coefficients are

calculated base on MMSE criterion [9] [10] . They can be
designed as follows

EQULIZATION IN FMT

FMT system has a high level of spectral containment,

such that the ICI is negligible compared to other noise
signal. The
ak

(10)

d [d 0 , d1 ,..., d Nd 1 ]T

with index iP.

III.

[h0 , h1 ,..., hNc ]

(CH PC  O ) 1 CH tG

(13)

d = MCf

ak

(14)

O is SNR, t G is a column vector whose elements

are all zeros except G th element is 1.

where

d
prototype filter design in fact cant meet the Nyquist
criterion. So, ISI is brought in the system inevitably. Here
we equalize the transmit filters and transmission channel
[8].

P (I Nd uG 0Nd uNd I Nd uNh- Nd -G )

(15)

M (0Nd uG I Nd uNd 0Nd uNh- Nd -G )

(16)

There are three common equalization method, i.e.,

maximum likelihood sequence estimation, linear and
decision feedback equalization. MLSE is the optimal
method for BER performance, but it isnt practical for
most actual system because of large computational
complexity. The linear equalization  effectively
overcome the effect on the transmit signal brought by
nonlinear channel. So, we use DFE equalizer. Shown as
Fig. 2, the Decision Feedback Equalizer (DFE) is inserted
at the output of each subchannel.

Figure 3.

DFE equalizer structure.

Because of almost no ICIeach subchannel can be

viewed independently and equalized separately. In the kth
subcarrier,
z k (iT ) ((a 'k (nT ) g (nT ) u e j 2S fk iT ) c(nT ) u e  j 2S fk iT ) h(nT )
=a 'k (nT ) g (nT ) (c(nT ) u e  j 2S fk (i 1)T ) h(nT )

IV.

(8)

overall impulse response of the kth subcarrier is given by

g (iT ) (c(iT ) u e j 2S fk (i 1)T ) h(iT )

'f

K / MT0 ; the

Fig. 4 shows the BER curve in the AWGN channel

when the filter length is 36 with DFE equalizer and
without it. It is clear that the DFE equalizer improves the
BER performance as a result of eliminating ISI brought by
filter. Fig. 5 compares symbol constellations between
before and after equalized when the eb/n0 is 18 db. Fig. 6
shows the BER in the channel c [1, 0.5, 0.1] with and
without DFE equalizer. We can figure out that the DFE
equalizer efficiently compensates the transmit signal and
has a good performance on BER.

The DFE structure is shown in Fig. 3. In the figure,

hoverall , f , d represent overall subchannel, feedforward
equalizer output, i.e. a k

The simulation parameters are set according to the

TETRA standard. The 4-QAM symbol rate T0 is 2400
symbols/second each subchannel; the subcarrier spacing
'f is 2700 Hz; the number of subcarrier M is 8; the

transmit period T T0 / K ; g(iT)=h(iT) is root raised

cosine pulse shape filter, roll factor D is 0.125. The
equalizer parameter Nf , Nd , G is 25, 25, 24
p Krespectively.

(9)

filter, feedback filter respectively.

(17)

SIMULATION AND RESULTS

sampling factor K=9 is given by

where

hk ,overall (iT )

h0

h1

hNc
N N
Nhu Nf

pK

pK

a 'k (nT ) ak (nT0 ) n K , c ( nT ) is the

transmission channel, represents convolution. the

h0 0
h h
0
1

h1

hNc
0 hNc

G is delay of the

ak G .

Let us define three vectors, the overall subchannel

feedforward and feedback filters coefficients



V.

CONCLUTION

BER in AWGN and multipath channel

10

This paper described the application of FMT

technology in TETRA II. We adopted DFE to equalize
filter channel and transmission channel. The performance
results shows the DFE is an effective scheme to eliminate
the ISI in FMT system.

-1

10

-2

BER

10

This paper only uses one equalization method. We will

compare performance of different equalization methods
for FMT in the future.

-3

10

-4

10

in AWGN with DFE

in channel c with DFE
in channel c without DFE

BER in AWGN channel

10

-5

10

without DFE
with DFE

10

12

14

16

18

eb/n0

-1

10

Figure 7. BER in AWGN and channel c.

-2

BER

10

-3

10

REFERENCES
-4

10

[1]
[2]
[3]

http://en.wikipedia.org/wiki/Terrestrial_Trunked_Radio
http://www.tetramou.com/
ETSI EN 300 392-2 v3.2.1 Terrestrial Trunked Radio (TETRA);
Voice plus Data (V+D); Part 2: Air Interface (AI)
[4] Giovanni Cherubini, Eleftheriou. E, Olcer. S, Filetered Multitone
Modulation for Very High-speed Digital Subscriber Lines, IEEE
JSAC, pp. 1016-1028, June 2002.
[5] Giovanni Cherubini, Eleftheriou. E, Olcer. S, Cioffi. J, Filter
Bank Modulation Techniques for Very High-Speed Digital
Subscriber Lines, IEEE Comm. Mag, vol. 38, pp. 98-104, May
2000.
[6] Andrea M. Tonello, Perfomance Limits for Filtered Multitone
Modulation in Fading Channels, IEEE Tr. Wireless Comm, pp.
2121, Sep. 2002.
[7] Andrea M. Tonello,     
Implementations of FMT Modulation Architectures, in Proc.
IEEE Inernational Conference on Acoustics, Speech and Signal
Processing (ICASSP 2006), Toulouse, vol. 4, pp. 625-628, Mar.
2006.
[8] Benvenuto. N, Tomasin. S, and Tomba. L, Equalizaton Methods
in OFDM and FMT Systems for Broadband Wireless
Communications, IEEE TCOMM, pp. 1413-1418, Sep. 2002.
[9] C. Richard Johnson, Philip Schniter, Thomas J. Endres, James D.
Behm, Donald R. Brown, Raul A. Casas, Blind Equalization
Using the Constant Modulus Criterion: A Review, in Proceedings
of the IEEE, Vol. 86, pp. 1927-1950, Oct. 1998.
[10] J. M. Cioffi, G. P. Dudevoir, M. V. Eyugoblu, G. D. Forney,
MMSE decision-feedback equalizers and coding Part I:
Equalization results, IEEE Trans. Commun., vol. 43, pp. 25822594,
Oct.
1995

-5

10

10

12

14

16

18

eb/n0

Figure 4. BER in AWGN channel

Figure 5. Constellation before DFE

Figure 6. Constellation after DFE