= C zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA which tail - - PDF document

SLIDE 1

Software Radio structure for second generation mobile communication systems

Anne Wiesler, zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA student member IEEE and Friedrich Jondral, senior member IEEE Institut fur Nachrichtentechnik, Universitat Karlsruhe

Abstract-Third generation mobile communication systems like the European UMTS will enable a flexible communication, free from standard specific regulation

• f modulation, channel coding, baud rate and multiple

access schemes. This flexibility can only be reached by a radio structure which performs all baseband functions in software and is therefore totally software programmable [l]. To ensure that different software configurations can be understood and supported by all mobile terminal architectures a general programming language is required to describe the used air interface components. By the example of the second generation mobile communication systems like the European GSM, the Japanese PDC, the American IS-54 (respectively IS-136) and the European wireless communications system DECT, a common de- scription and implementation of the transceiver functions like channel coding, modulation and equalisation have been developed.

• I. INTRODUCTION

Third generation systems like the European UMTS will realize two goals with a more flexible communication technology: Firstly a global roaming will be possible. Secondly a dynamical adaption of the air interface to the time variant radio channel, the service and the cell type can be executed during each communication session. This flexibility can only be achieved by a so called software radio, this means with a totally software programmable radio structure. Changing to another mobile communi- cation system can be done by downloading the needed software and reconfiguring the mobile terminal. To ensure that the configurations can be understood and supported by different mobile terminal architectures a general programming language is required to describe the used air interface components. As a first approach a common description and implemen- tation of the second generation mobile communication systems GSM, PDC, IS-54 and the wireless communi- cations system DECT have been developed. As they all are TDMA/FDMA systems a lot of functions are used in a similar way and zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

so a parametrized software

implementation is proposed. This has the advantage that not the whole software of a system has to be downloaded. The reconfiguration can be done by exchanging a set of parameters, which enables a seamless and fast change of zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

0-7803-4320-4/98/$5.00 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

0 1998 IEEE

the air interface. In the next section a common transmitter structure is described. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA A common receiver structure is proposed in Section I11 with main emphasis on a general equalizer.

• 11. COMMON TRANSMITTER STRUCTURE

The generation of the transmission signal is exactly specified in the standards. Functions can be defined with a few parameters, for example the convolutional channel coding can be exactly described by the generator polynomials (in binary representation), the code rate (which is not always equal to the number of generator polynomials), the constraint length, a parameter that sets a termination on or off. In the case of termination the number of bits in one block and a parameter that specifies which tail bits are used must be set additionally. In most cases the tail bits are zeros, but in the IS-54 System for FACCH-bits there are the first information bits used. In the same way the block channel coding, the puncterer, the interleaver, the scrambler (for DECT) and a general burst builder can be described and implemented. It is more difficult to find a common structure for the modulation function.

• A. rl4-DQPSK

The IS-54 and PDC use the .n/4-DQPSK, a special case

• f QPSK [3]. The complex envelope of a n/4-DQPSK mod-

ulated signal with the symbol duration T is

03

s D Q p s K ( t ) zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

= C zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

2 , .g(t -

.T) (1)

n=O

00

= C

exp [@(.)I

.

g ( t -

n ~ ) . (2)

n = O

The phases of the complex symbols z, = exp [@(.)I are At a time two bits are assigned to one of four possible differential phases like it is shown in Tab. 1. Thus the information is placed into the differ- ential phase of two successive symbols. The sym- bols z, are alternately out of {-1,1, - j , j } and zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

{ z (

1 + j ) , &(-l + j ) ,

*(-I

• j ) ,
• $1. The

2363 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

V C

‘98

SLIDE 2

Table 1: Bit to differential phase assignment pulse former zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

g ( t ) is a square root raised cosine filter with

roll off factor zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

Q zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

= 0.35 for IS-54 and Q = 0.5 for PDC.

This means that the two nyquist criteria are fulfilled and no intersymbol interference (ISI) is produced by the

• filtering. The modulation is not linear because of the

differential assignmen: zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA (3), but it can be realized with a I/Q-Modulator. A disadvantage of this theoretically very bandwidth efficient modulation is the high amplitude fluc-

• tuation. However the fluctuations are smaller compared

to the one of the original DPSK [5], the requirements at the linearity of the end power amplifier are still very

• high. Economical C-power amplifiers can not be used in

Japanese and American mobile terminals because of the high nonlinear distortions causing a strong increase of the side lobes.

• B. GMSII’

The nonlinear GMSK modulation is used in the Euro- pean GSM and DECT systems. GMSK is a special kind

• f a 2-level FSK with modulation index h = 0.5 [ll].

The complex envelope of an GMSK modulated signal is with the NRZ-stream dn zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

E {-1,l) and the frequency im-

pulse zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

g ( t ) .

In the GSM-system the bits b; are first encoded differentially as follows

bi = bi ebi-1, b-1 = 1

(5)

with @ means modulo-2 addition. After that the NRZ- bitstream is build by transforming bi = 1 to di = -1 and zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

bi = 0 to di = 1. The DECT-system builds the NRZ-

bitstream directly from the incoming bits by transforming

b; = 1 to d; = 1 and b; = 0 to di = -1. Instead of a

rectangular-impulse which is used at MSK, here the fre- quency impulse

• is used. h

~ ~ ~ ~ ~ ( t ) ist the known Gaussian impulse with the time bandwidth product BT. This causes a reduction of bandwidth, but with the trade of a controlled intersymbol interference (ISI). In the GSM-system the factor BT = 0.3 was chosen, that results in a IS1 over about 2 symbols but a small bandwidth. This IS1 is equalized together with the IS1 caused by the mobile channel at the receiver (Section 111). Indoor or micro cell channels do not cause long path delays and therefore at the DECT-system BT = 0.5 was chosen which results in minor IS1 combined with a smaller bandwidth efficiency. For the realisation of the theoretically infinite long Gaus- sian impulse zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

as FIR-filter it is cut to the length LT, with

L 2 3. Thus at time t = n T the phase reply is zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

t zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

q(t)

= J s ( r ) d r .

(7) zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

• ca

q ( t ) is shown in Fig. 1 .

q(t)

• Fig. 1: Phase reply q ( t ) for GMSK with L =

4, BT = 0.3

At t > LT q ( t ) is equal 0.5. With q(t) the GMSK signal

s ~ ~ s ~ ( t )

can be described as follows

00

]

(8)

S G M S K ( ~ )

=

exp j2nh

d,q(t -

nT) The nth NRZ-bit dn causes a change of phase about rq(t)

• r -nq(t), which is added to the changes of the previ-
• us symbols. Here the information lies in the direction of

the rotating complex signal vector. In [9] it is shown that

S G M S K ( ~ )

can be builded by superposition of N, = 2L-1 impulses CK:

[

n = O

with

n L-1

i=O 1=1

This representation can be used for all CPM-signals, for example the complex envelope of a MSK-signal (here is L = 1) can be written

03

s ~ s ~ ( t )

=

exp [ j ~ h

2

di] Co(t -

nT) (12)

n=O

L i=o

J

0-7803-4320-4/981$5.00

0 1998 LEEE

2364 VTC ‘98

SLIDE 3

with sin(nt/2T) zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

0 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

5 t 5 2T

• therwise zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

CO=

{

This is the well known representation of MSK modulation. (13) as 0-QPSK zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

0.5 - 0.0-

• 0 5 -

For GMSK with L = 4 the exact superposition is made by eight impulses. This makes the realisation of the mod- ulation according to zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA (9) complicated [7]. In Fig. 2 the two first impulses CO and C

1 are displayed. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

CO

has 99% of the signal energy.

• Fig. 2: Impulse CO

and C

1 for L =

4, BT = 0.3 Thus the following approximation of zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

S G M S K (t)

is obvi-

• us:

M

r zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

n i

,=o

L

i=O

J

00

n=O

with z,

E {-1,1, - j , j } . Herewith the approximated

GMSK-signal can be built similar to the r/4-DQPSK- signal (2). In the same way the symbols z, are obtained by accumulation and filtering by g ( t ) respectively Co(t). The common modulator structure (Fig. 3) can be used and ex- tended to other modulations (QAM, PSK, appr. CPM).

FIR PRECODER NRZ MBIT’ISYMBOL 01 = zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA 0.35

BT =

. 3

BT

=

0.5

• Fig. 4: Complex envelope of the approximated GMSK

power amplifier in C-mode without a loss of efficiency. This makes the approximation not attractive for single mode GSM mobile terminals, however the spectrum of the ap- proximated GMSK-signal fulfils the requirements of the GSM-standard. On the other hand the envelope of the r/4- DQPSK shows much higher fluctuations than the approx- imated GMSK and has higher demands to the linearity of the power amplifier. That means that in a multimode mo- bile terminal the use of the approximated GMSK means no loss of efficiency.

• 111. COMMON RECEIVER STRUCTURE

To develop a common receiver structure it is necessary

to differentiate between the methods for narrowband and broadband systems. In the case of broadband systems the channel is frequency selective, this means that IS1 occurs during the transmission. The number of symbols which are overlapped depends on the symbol duration T and the maximum path delay rmax

• f the channel.

For the GSM system several channel models have been developed [2]. Here r

, , ,

= 20ps is assumed, with T = 3 . 6 9 ~ ~

this results in an IS1 of about 6 symbols. Due to a more efficient speech encoder and the four level modulation scheme the symbol duration of the IS-54 is 41.14 ps and 4 7 . 6 ~ ~ for PDC. In both cases r,,, is assumed to be equal to T [4], [8], [6]. This means an IS1

• f two successive symbols. However a path delay of 4 7 . 6 ~ s

is rather rare and most PDC mobile terminals do not use an equalizer [lo]. At the DECT system (T = 0 . 8 7 ~ s ) the indoor channel has r

, , ,

= 0.5ps and the outdoor

channel shows r

, , ,

= 1.Ops [12]. The DECT system is

in most cases a narrowband system, but in some special environments an equalization may be necessary.

• Fig. 3: Common I/&-modulator
• A. Equalization

It should be mentioned that the approximation of the GMSK has the effect that the complex envelope is not ex- actly constant any more. Fig. 4 shows the complex repre- sentation of the approximated GMSK-signal for BT = 0.5 and BT = 0.3. Though the fluctuations of the ampli- tude are not very high, it is not possible to use an end In all systems MLSE equalization is applied with the viterbi algorithm because this is the best solution in the sense of the ML-criterion. In a software radio a general, flexible viterbi-equalizer is required. The discrete channel model in Fig. 5 can be described as follows: The received

0-7803-4320-4/98/$5.00

0 1998

IEEE

2365 VTC ‘

9 8

SLIDE 4 1

• 1
• 1
• zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

3 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA 57 I 26 I zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA 1 T DATA zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA SF TRAINING S DATA

I I I zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA Pn-N

3 825 T C T zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

• Fig. 5: discrete channel model

4 1 b T T

signal y(t) is

' 111 10 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA 8 I I 5 111 6 DATA s w i m mvcc SF SAKH DATA GT zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA 00

y(t) =

z, . h(t -

nT)

+

T ( t )

(16)

n = O

21 I2 IM DATA SYNCH SACCI1

with the impulse response

I1 I 1 11 CDVCT DATA KSVD

h(t)

=

S T ( t )

* k ( r , t )

* g R ( q

(17) including the transmitter filter zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

gT

(t),

the time variant mo- bile channel k ( r , t ) and the receiver filter gR(t). The com- plete impulse response has the length N +

• 1. Additionally

there is the filtered AWGN ~ ( t )

= n(t) * gR(t).

The sam- pling of (16) with symbol clock leads to zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA N

p=O

The discrete channel model with constraint length N can be described by a trellis diagram. The number of states

( z , - ~ ,

~ ~ - 2 , .

. . ,

z , - ~ ) is M N ,

with the number M of pos- sible symbols at the same time. The number of possible transitions from one state to another is MN-'. This leads to different trellis for the different mobile systems. The trellis for PDC and IS-54 is shown in Fig. 6, here is N = 1 and A

4 =

• 4. The number of states is only four, that means

low calculation effort for the viterbi algorithm.

• Fig. 6: Trellis of x/4-DQPSK with N = 1

For GSM the IS1 of the modulation leads to a channel constraint of about 7, but 128 states would cause too much calculation effort. So in most applications 16 or 32 states are used. The trellis is shown in Fig. 7 with N = 3. For MLSE the channel impulse response hp,, must be

• estimated. With a 0.577 ms long burst (Fig. 8) the channel
• Fig. 7: Trellis of GMSK with N = 3
• Fig. 8: Burst of GSM, PDC and IS-54

response can be assumed to be stationary during one GSM

• burst. The channel estimation can be reduced to one per
• burst. The IS-54 and the PDC bursts (Fig. 8) are both

6.67 ms long and therefore channel tracking is necessary. A known training or synchronisation sequence is used for the channel estimation. At GSM this sequence is placed in the middle of the burst. The estimation is performed by correlation of the received training sequence with the known training sequence. The equalisation is performed in forward and backward mode by starting in the middle of the burst. The three tail bits at the end and at the beginning of the burst make the decision for the correct trellis better. The general viterbi equalizer therefore has the following parameters:

0 equalization turned ON/OFF

the constraint length of the channel N

0 the symbols of the modulation and the possible tran-

sitions

0 the number of symbols of one block

the mode (forward/backward)

0 the number of tail bits

• B. Demodulation

For all systems a differential demodulation can be used. The (equalized) symbols 2(n) are multiplied with i" (n- 1).

0-7803-4320-4198B5.00

0 1998 IEEE

2366 VTC '98

SLIDE 5 exp zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA ( - j ( n r f o zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

t

Aw)

t ) zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

[6] K.A. Hamied and G. L. Stueber. An adaptive trun- cated MLSE receiver for japanese personal digital cel-

• lular. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

IEEE Transactions on Vehicular Technology, VTC-45(1):41-50, February 1996. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

p p zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

! Channel Estim.!

• - - - - - - - - _ - - - I

[7] P. Jung. Laurent’s representation of binary digital

• Fig. 9: Receiver structure

In the case of approximated GMSK with (14) this leads to:

= exp[jnhd,]

(22) The sign of the imaginary part of A Z G M s K ( n ) is the wanted dn. The transmission from NRZ to the binary rep- resentation yields &

, . At GSM the precoding (5) has to be

canceled with zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

A

• .

.

b, = b, @ b,-1.

(23) With the same multiplication the demodulation can be done for 7r/4-DQPSK (2): The phase of A.ZDQ~SK

(n)

is the wanted differential phase, with Tab. 1 the bits b, are obtained. The common receiver structure Fig. 9 can be used in soft- ware radios for second generation mobile systems. The equalizer can be turned off for narrowband systems.

References

Software radios. IEEE Communications Magazine, 1995. GSM 05.05 (ETS 300 577). Radio transmission and

• reception. ETSI.
• P. A. Baker. Phase-modulation data sets for serial

transmission at 2000 and 2400 bit per second. AIEE

• trans. Commun. Electron., pages 166-171, Juli 1961.

ETI/TIA. Interim Standard - 627. 1992.

• K. Feher. Modems for emerging digital cellular-mobile

radio systems. IEEE Trans. Vehicular Technology, 40:355-365 , 199

1.

continuous phase modulated signals with modulation index 112 revisited. IEEE Trans. Communcations, 42~221-224, 1994.

[8] Gustav Larsson, Bjorn Gudmunson, and Krister

• Raith. Receiver performance for the north american

digital cellular system. Proc. IEEE Vehicular Tech- nology Conf., Vol. 41:l-6, May 1991. [9] P.A. Laurent. Exact and approximate construction

• f digital phase modulations by superposition of am-

plitude modulated pulses (amp). IEEE Tmns. Com- mun., COM-34:150-160, February 1986. [lo] R.W. Lorenz. Vergleich der digitalen Mobilfunksys- teme in Europa (GSM) und in Japan (JDC) unter besonderer Berucksichtigung der Wirtschaftlichkeit-

• saspekte. Der Fernmelde Ingenieur, 1/’93,2/’93,1993.

GMSK modulation for digital mobile radio telephony. IEEE Trans. Commu- nications, 29(7) : 1044-1050, July 1981.

[ll]

• K. Murota and K. Hirade.

[12] T.A. Wilkinson. Radio propagation channel modelling for the DECT test bed, volume TD(91). COST 231.

0-7803-4320-4/98/$5.00 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

0 1998 IEEE

2367 VTC ‘98