On the Downlink Capacity of WCDMA Systems

with Transmit Diversity

Vaibhav Singh, Oya Yilmaz, Jialing Wang, Karthigeyan Reddy, and S. Ben Slimane
Radio Communication Systems
Department of Signals, Sensors, and Systems
Royal Institute of Technology (KTH)
Electrum 418, 164 40 KISTA, Sweden

Abstract— Transmitter diversity at the base station has been mobile and each base station is modeled explicitly. We do not
included in WCDMA standard for third generation mobile model the pilot channel transmitted by each base-station.
systems as a means of increasing the downlink system capacity.
In previous works, it has been shown that open loop transmitter
diversity can significantly increase the downlink capacity. How- b1 , b 2
ever, the performance in multipath environment is not known. In
r1 , r2
our study, we present the simulation results for downlink system b1 , b 2 b̂1 , b̂2
STTD
a(t) RAKE
capacity taking into account the effect of Doppler frequency and receiver Dec.
multipath fading. The results show that even in severe multipath
wireless channel, diversity scheme can greatly improve the system
capacity when no soft handover is used. −b∗2 , b∗1

I. I NTRODUCTION Fig. 1. WCDMA with Alamouti STTD encoding and decoding.

The Space Time Transmit Diversity (STTD) scheme

adopted by 3GPP in the current WCDMA standard is based B. Channel Model
on a modification of the Alamouti space time coding scheme In cellular systems, the signals are exposed to both long term
[1]. However this scheme is based on orthogonal design. In a fading and short term fading effects. In general, the long term
WCDMA system, due to multipath propagation, the signals effect is modeled by a deterministic path loss and a stochastic
received will no longer be orthogonal. Previous works on part modeled as a lognormal process. In this paper, we model
this topic [2], [3], [4] have mainly focused on the impact the long term effect between antenna 1 of the mth sector and
of multipath fading on link level performance. In [3], [4], mobile user i as [11]:
system level results have been presented for STTD but they
do not consider the increase in interference caused by mul- gmi = d−µ
mi Gmi (1)
tipath propagation. In this paper, we present results for the where d−µ
mi is the distance dependent pathloss, dmi is the dis-
downlink capacity of a WCDMA system both with STTD and tance from the mobile to the mth cell site, and µ is the pathloss
without STTD. We focus on the STTD performance in typical exponent. Gmi represents the log-normal shadow fading, i.e.,
urban macro cellular environment, assuming either the ITU 10 log10 (Gmi ) is a zero mean Gaussian random variable with
Vehicular-A or the Pedestrian-A multi-path delay profile [5]. standard deviation σs dB. As the transmit antennas of each
We consider only speech traffic over dedicated channel. In sector are co-located, they are assumed to have the same long
Section II, the concept of WCDMA system is introduced and term effect.
the transmitter diversity scheme STTD is discussed. In Section The short term effect is modeled as frequency selective fad-
III, the system model used in the simulation is described. In ing channel with equivalent lowpass channel impulse response
Section IV, the simulation results are presented and analyzed. given by
In the last section, a conclusion is drawn. N


h(t, τ ) = hi δ(τ − τi ) (2)
II. WCDMA C ELLULAR S YSTEM i=1

A. Cellular System Model where N is the total number of paths, hi is zero-mean complex
We consider a multi-cellular WCDMA system with hexag- Gaussian random variable with power spectral density 2σi2
onal cells and three sectors per cell. Each sector is using an representing the gain of path i, and τi is the time delay of path
STTD encoding scheme according to Figure 1. In this work we i. We further assume a normalized channel fading coefficients,
only consider R = 12.2 kbps speech service over dedicated i.e.,

N
N
channels in the downlink. In order to model the intra-cell |hi |2 = 2σi2 = 1. (3)
and inter-cell interference adequately, the link between each i=1 i=1
 Two multipath delay profiles are considered in this pa- and
per: The ITU Pedestrian-A (Ped-A) and Vehicular-A (Veh-A)  1
power delay profiles.The power of the multi-paths in the power R, τ =0
var[ρii (τ )] = ζ
(8)
delay profile has a chi-square distribution with two degrees
W , τ = 0
of freedom while the amplitude is Rayleigh distributed with
a classical Doppler spectrum. The channels of the different where ζ = 1 for synchronous waveforms and ζ = 1/3 for
transmit antennas are assumed independent. asynchronous waveforms.
Assuming perfect channel state information and that the
C. Receiver Model delays {τi }N
i=1 are known, the RAKE receiver output samples
The receiver (mobile unit side) consists of one omni- corresponding to the transmitted symbols, b1 , b2 , are obtained
directional antenna and a RAKE receiver having a total num- as follows:
ber of fingers equals to the number of paths N . Considering a N 

lT
certain mobile unit within the coverage area of sector m, say rl = au (t)h∗i1 r(t + τi1 )dt, l = 1, 2 (9)
mobile u, the equivalent lowpass of the received signal can be i=1 (l−1)T
written as
Km
where au (t) is the spreading code of mobile user u. Assuming
1


N
BPSK modulation, the two symbols, b1 and b2 , after STTD
r(t) = hil skl (t − τil ) + ie (t) + z(t) (4)
combining are estimated as
l=0 i=1 k=1

where Km is the total number of active mobile users within b̂1 = sgn{y1 } = sgn {r1 + r2∗ } (10)
sector m, ie (t) is the external interference coming from other
sectors, z(t) is complex Gaussian random process with zero- b̂2 = sgn{y2 } = sgn {r2 − r1∗ } (11)
mean and power spectral density N0 , hil is a complex process
representing the fading coefficient of path i, skl (t) is the Considering the receiver of mobile station 1 in sector m, the
transmitted signal of user k from transmit antenna l, combined sample at the mobile station, y1 , can be represented
 as:
skl (t) = φk gm pm ak (t)bk (t),
N

 
where each data symbol is assumed to have unit amplitude, y1 = |hn,1 |2 + |hn,2 |2 φ1 gm pm ρ11 (0)
i.e., |bk (t)|2 = 1 and the spreading waveform is denoted by n=1

N
N
ak (t). Here pm is the total transmitted power of base station  ∗ 
+ hn,1 (hp,1 − hp,2 ) + hn,2 (h∗p,1 + h∗p,2 )
m, gm is the link gain between the mobile receiver and base
n=1 p=1
station m, and φk is the portion of power allocated to user p=n
k. The spreading bandwidth is W ≈ 1/Tc where Tc is the 
× φ1 gm pm ρ11 (τn,1 − τp,1 )
chip duration. When the period of the spreading waveforms is Km
N

N

larger than the information symbol duration T , the correlation
 ∗ 
+ hn,1 (hp,1 − hp,2 ) + hn,2 (h∗p,1 + h∗p,2 )
properties follow n=1 p=1 k=2

 × φk gm pm ρk1 (τn,1 − τp,1 )
T
1 N


ρik (τ ) = √ ai (t − τ )ak (t)bi (t − τ ) dt  ∗ 
T 0 + hn,1 i1,n + hn,2 i2,n
 T n=1
1
≈ √ ai (t − τ )ak (t) dt. (5)
N
 
T 0 + h∗n,1 z1,n + hn,2 z2,n (12)
For a spreading waveform defined as n=1

+∞

where
ai (t) = ai,n Tc (t − nTc ) (6)  lT
n=−∞ 1
zl,n = √ z(t + τil )a1 (t) dt (13)
T (l−1)T
where ai,n = ±1 with equal probability and T (t) is a
rectangular pulse of length T and amplitude equals to one, is zero-mean complex Gaussian with variance N0 ,
the cross correlations (5) can be determined from the results
 lT
of [6]. Using our notation and considering users within the 1
same cell, it can be shown that when i = k, il,n =√ ie (t + τil )a1 (t) dt (14)
T (l−1)T

0, τ =0
var[ρik (τ )] = (7) is the external interference sample, and ρkm (τ ) is as defined
ζ
W , τ = 0
earlier.

2
 The communication quality of interest for mobile user 1, Assuming that each user requires a maximum tolerable bit
related to the bit error probability, is given by error probability that can be mapped into equivalent minimum

Eb /I0 value, denoted γt . Hence, the signal quality of mobile
Eb E[yk |bk (t)]E∗ [yk |bk (t)]
= user 1 can be measured using the outage probability defined
I0 1 var[yk |bk (t)]
N 2 as 


  φ1 gm pm Eb
2 2 O1 = P r < γt (20)
= |hn,1 | + |hn,2 | (15) I0 1
n=1
2Rσ12
where γt is the minimum required SIR for good signal quality.
where E is the expectation operator, R is the information data
rate, and the factor 2 comes from the fact that the user power is III. S IMULATION M ODEL
split between the two transmit antennas. With the assumption The simulation is performed over a system with 19 sites and
of the terms in the combined sample being Gaussian and mu- each site has 3 sectors. Wrap-around is used to eliminate the
tually independent, the variance of the interference becomes edge effects. The available system bandwidth is W = 5 MHz.
N 

  The system offers speech service at a data rate of R = 12.2
2 2 2
σ1 = |hn,1 | + |hn,2 | kbps. The maximum base station power is limited to 20 W [4].
n=1
  The maximum transmission power per link is 2 W [4] and the
ζ ζ

B initial power is set to 0.1 W. The environment we assume is
× θN g0 p0 + gi pi + N0 (16) an urban environment. So we set the propagation constant µ
W W i=2
to 4. The gain constant is chosen to be −28 dB. The standard
where B is the total number of interfering base stations, θN deviation for lognormal shadow fading is chosen as 10 dB.
is the total orthogonality factor caused by fading multipath We use soft handover with a handover margin of 3 dB. The
channels and is given by SIR balancing algorithm is used for power control and the

N ∗
2 power control is done in steps of 1 dB. Hence in transmitter

N

hn,1 (hp,1 − hp,2 ) + hn,2 (h∗p,1 + h∗p,2 ) diversity case, the power of each antenna is changed in steps
θN = N
n=1 p=1 2 m=1 (|hm,1 |2 + |hm,2 |2 ) of 1 dB. In our model, the active mobile users are assumed
p=n to be uniformly distributed over the cells. The cell load is
Replacing (16) in (15) and simplifying, the bit-energy-to- Poisson distributed. We consider two types of traffic models.
interference-spectral-density ratio of mobile user 1 can then • Mobiles with low mobility (3 km/hr)
be rewritten as • Mobiles with high speed (60 km/hr)
  

N
|h |2
+ |h |2
φ1 gm pm The SINR threshold after despreading, γt , is set to 7 dB corre-
Eb W n=1 n,1 n,2
= B (17) sponding to a BER of 10−3 (standard for voice communication
I0 1 2R θN g0 p0 ζ + ζ i=2 gi pi + N0 W systems) [2]. All the simulation parameters considered in the
For the case of one transmit antenna, the bit-energy-to- paper are summarized in the following table:
interference-spectral-density ratio of mobile user 1 takes the TABLE I
following form: S IMULATION PARAMETERS
 

N
|h |2
φ1 gm pm
Eb W n=1 n,1 Item Symbol Value
= B (18) System Bandwidth W 5 MHz
I0 1 R θ̃N g0 p0 ζ + ζ i=2 gi pi + N0 W Chip Rate Rc 3.84 Mcps
Information Bit Rate R 12.2 Kbps
where now the total orthogonality factor, denoted θ̃N , is given Gain Constant c −28 dB
by Cell Radius r 1000 m
Propagation Constant µ 4

N
N pmax 33 dBm [4]
|hn,1 |2 |hp,1 |2 Maximum Transmission Power
θ̃N = N (19) per link
2
n=1 p=1 m=1 |hm,1 | Initial Transmission Power pinit 20 dBm
p=n Total Base Station Power Ptot 20 W
Noise Floor N0 −107 dBm [4]
Comparing the two schemes, it is observed that the STTD SIR Threshold γt 7 dB
scheme has a better diversity order and has the capability to Lognormal Standard Deviation σs 10 dB
resolve the fading multipath components from the two transmit
antennas. However, it is not clear which of the two methods
perform better since they have different total orthogonality IV. S IMULATION R ESULTS
factors. The strength of this orthogonality factor will affect The simulation is done for with and without diversity cases.
the system performance and it will be interesting to see the The results obtained without diversity are used as reference to
interaction between the increase in diversity order and the evaluate the gain obtained with STTD. This is done for the
increase in experienced interference when STTD is used in two channel profiles and the two different mobile speeds. The
WCDMA systems. SIR of the user after despreading is averaged over the last five

3
frames after the power control algorithm has converged. If this 0
Veh A Model, Low Mobility (3km/h)
10
averaged SIR is less than the target SIR of 7 dB, the user is No STTD
considered to be in outage. The outage probability is the ratio STTD
of the number of users in outage to the total number of users
in the system. To reduce the statistical variance, we run five 10
-1
independent simulations and average the results obtained for

Outage Probability
these simulations.
The outage probability as a function of the cell load for -2
10
the case of no TD and the case of STTD for low mobility and
Ped-A channel profile is shown in Figure 2. It is observed that
STTD provides significant gain in WCDMA system capacity.
For example, for a 2% outage probability, a 30% capacity gain 10
-3
is obtained with STTD.

Ped A Model, Low Mobility (3km/h) -4

0 10
10 50 55 60 65 70 75
data1 Cell Load (Erlang/Cell)
data2
Fig. 3. Outage Probability as a function of the cell load for Vehicular-A
-1
10 model and low mobility (3 km/h).
Outage Probability

-2
A. System Capacity for High Mobility
10
In order to find the effect of mobility on capacity, we
also evaluate the system capacity for high mobility. Results
obtained for Ped-A channel are shown in Figure 4. Figure
-3 5 shows the capacity for Veh-A environment. The gain in
10

Ped A Model, High Mobility (60km/h)

-1
10
-4
10
40 50 60 70 80 90
Cell Load (Erlang/Cell)

Fig. 2. Outage Probability as a function of the cell load for Pedestrian-A -2

model (3 km/h) and low mobility with data1 corresponding to the case of No 10
Outage Probability

STTD and data2 to the case with STTD.

The results obtained with Vehicular-A model are shown in

Figure 3. For this case, the capacity gain for 2% outage is 13%. -3
10
It is seen that the gain obtained for Veh-A channel is less than
that for the Ped-A channel. Since the number of paths in Veh-
A model is higher, there is already a higher degree of multi- No STTD
path diversity which the RAKE receiver can take advantage STTD
-4
of. This is the reason why the gain with STTD is less for 10
50 55 60 65 70 75 80 85 90
Veh-A environment than for Ped-A environment. Using STTD Cell Load (Erlang/Cell)
helps to ”smooth out” the fading effect, which results in a
lower transmitted power. If a user is in a bad position, the Fig. 4. Outage Probability as a function of the cell load for Pedestrian-A
transmitted power for that user has to be increased in order model and high mobility (60 km/h)
to achieve the target SIR. This increases the interference for
the other users. For low mobility case, the fading will last system capacity for 2% outage is 8% for Ped-A channel. For
for a longer time. By using transmitter diversity, we are able Veh-A channel the gain obtained is very small. Compared to
to negate the effect of fading and achieve improved system low mobility case, we have much lower gain with STTD in
capacity. Thus for low mobility there is more diversity gain high mobility case. When we have high mobility, a user does
and less interference averaging. not stay in a bad position for long. Thus the interference is
averaged out. Hence the interference in the system decreases

4
 Veh A Model, High Mobility (60km/h) CDF of the orthogonality factor (θ)
0
10 1
No TD
0.9
STTD
0.8
-1
10
0.7
Outage Probability

0.6
-2

F(θ)
10 0.5

0.4

-3 0.3
10
Ped−A No TD
0.2
Ped−A STTD
Veh−A No TD
0.1 Veh−A STTD
-4
10
50 55 60 65 70 75 80 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Cell Load (Erlang/Cell)
θ

Fig. 5. Outage Probability as a function of the cell load for Vehicular-A Fig. 6. The cumulative distribution function of the orthogonality factor for
model and high mobility (60 km/h) the 2 channel models with and without STTD

leading to an increase in capacity. Since in high mobility [3] Engstrom, B.,Ericson, M., ”WCDMA System Level Performance With
we already have an interference averaging effect, the gain Fast Fading And Non-Ideal Power Control”, Proceedings IEEE Vehicu-
obtained by introducing diversity is lower as compared to the lar Technology ConferenceSept. 1999.
[4] Technical Specification Group Terminals; Terminal Conformance Spec-
low mobility case. ification; Radio Transmission And Reception (FDD) (Release 5); Tech-
The cumulative distribution function of the orthogonality nical Specification: 3GPP TS 34.121 V5.3.1 (2004-04)
factor for the two channel models is illustrated in Figure 6 [5] 3rd Generation Partnership Project. Technical Specification Radio Ac-
cess Network Physical Layer Procedures. 3GPP TS 25.214 V3.0.0.
for the case of WCDMA with single transmit antenna and the October 1999
case of WCDMA with STTD. [6] Pursley MB., “Performance evaluation for phase-coded spread spectrum
multiple access communication – Part I,” IEEE Trans. on Commun., Vol.
It is observed that the orthogonality factor increases when 25, No. 8, pp. 795-799, 1977.
the number of paths from the base-station to the mobile in- [7] 3GPP, ”Technical Specification 25.211, Physical Channels and Mapping
crease. It is also seen that introducing transmitter diversity into of Transport Channels onto Physical Channels” July, 1999
[8] Parkvall, S., Karlsson, M., Samuelsson, M., Hedlund, L., Goransson,
the system increases the orthogonality factor and hence the B., ”Transmit Diversity In WCDMA: Link And System Level Results”,
intra-cell interference increases. In low mobility, the diversity Vehicular Technology Conference Proceedings, 2000
gain is sufficient to compensate for the increased interference [9] Canales, M., Valdovinos, A., Gallego, J.R., Gutierrez, F., ”Performance
Analysis Of Diversity Transmission Modes In UTRA FDD Under
and additionally provide a capacity gain. Time-Varying Multipath Channels”, Personal, Indoor and Mobile Radio
Communications, 2002. The 13th IEEE International Symposium on
V. C ONCLUSIONS Volume: 3, 15-18 Sept. 2002
[10] Hamalainen, S., Holma, H., Toskala, A., Laukkanenz, M., ”Analysis
The results have shown that transmitter diversity provides of CDMA Downlink Capacity Enhancements”,Personal, Indoor and
downlink capacity improvement. The gain obtained is much Mobile Radio Communications, 1997. The 8th IEEE International
Symposium on Volume: 1 , 1-4 Sept. 1997
higher for low mobility case as compared to high mobility [11] Jalloul, L., Rohani, K., ”CDMA Forward Link Capacity And Coverage
case. The gain obtained decreases as the number of paths In A Multipath Fading Channel”,Vehicular Technology Conference, 1997
increases. The effect of using STTD in combination with IEEE 47th Volume: 3 , 4-7 May 1997
[12] Selection Procedures For The Choice Of Radio Transmission Technolo-
various receiver diversity combining schemes has been studied gies Of The UMTS. UMTS 30.03 version 3.2.0; Technical Report; April,
on the link level in [13]. Future work can look into the system 1998
level gain obtained when using both transmit and receiver [13] Bjerke, B. A., Proakis, J. G., Zvonar, Z.,”Antenna Diversity Combining
Schemes in Fading Multipath Channels”, IEEE Wireless Communica-
diversity in WCDMA systems. tions pp 97- 106, January 2004

R EFERENCES
[1] Alamouti, S.M., ”A Simple Transmit Diversity Technique For Wireless
Communications”, IEEE J. Select. Areas, Vol 16, No. 8, pp. 1451-1458,
October 1998.
[2] Cai, M., Zhang, X., Zhou, N., Slimane, B.,”On The Capacity Of
CDMA With Successive Interference Cancellation”,EPMCC 2003,Glas-
gow, Scotland, April 2003.