2006 IEEE Ninth International Symposium on Spread Spectrum Techniques and Applications
Uplink Throughput Enhancements and Delay
Reductions in 3G WCDMA Systems Using 3.5G
Enhanced Uplink Techniques
E.J.B. Rodrigues' 2
'Federal University of Rio Grande do Norte - UFRN - GEMA, 59072-970, Natal - RN, Brazil
2Federal Center of Technological Education of Paraiba - CEFET-PB - CETRON, 58015-430, Joao Pessoa - PB, Brazil
eduardo_telecomgyahoo.com.br
Abstract-This work aims to analyze the performance of three
Enhanced Uplink (EUL) schemes for uplink expansion in third
generation (3G) wideband CDMA systems. By means of
computational simulations the throughput increase was
investigated for Node-B controlled Hybrid Automatic Repeat
Request (HARQ) mechanisms and Multi-User Detection (MUD)
schemes, besides the delay reduction combining HARQ schemes
and shorter TTI (Transmission Time Interval) size. The results
show that throughput can be increased by 21% to 74%,
depending on the applied technique and the mobility scenario.
Keywords - 3.5G EUL, delay reductions, HARQ, MUD,
throughput enhancements, WCDMIL4.
shorter TTI,
I. INTRODUCTION
To support services as real time gaming, P2P applications
and video telephony, throughput enhancements as well as
retransmission delays reductions in the uplink (UL) for 3G
WCDMA systems have been reached with the implementation
of the new 3.5G schemes, proposed in 3rd Generation
Partnership Project (3GPP) Release-6. These expansions are
known as High Speed Uplink Packet Access (HSUPA) or
Enhanced Uplink (EUL) [1]. This technique is the UL version
for High Speed Downlink Packet Access (HSDPA) proposed
for the DL [2]. The practical implementation of EUL schemes
in a mobile network is made through the inclusion of the
E-DCH (Enhanced Dedicated Channel), besides software
updates and hardware implementations. The coexistence
between current 3G WCDMA DCH (Dedicated Channel) and
E-DCH must be guaranteed. Diverse candidate approaches for
enhance 3G DCH capabilities were proposed in recent works.
Some of them are the Adaptive Modulation and Codification
(AMC) [3]-[4], fast Node-B scheduling [5], fast E-DCH
initialization [4], smart antennas systems [6], [7], users
admission control [8], shorter TTI (Transmission Time
Interval) size [1], [3], [9], Hybrid Automatic Repeat Request
(HARQ) [5], [10] and Multi-User Detection (MUD) schemes
with Interference Cancellation (IC) [5], [11]-[12].
In this work I focus on the three last schemes. The
performance is evaluated by computational analysis in Matlab.
Three different ITU Power Delay Profiles (PDP) were adopted
for simulations, two macro-cell and one micro-cell, according
to simulations parameters in [13]. The paper is organized as
0-7803-9780-0/06/$20.00 ©2006 IEEE
follows. First, in Section II, a brief description of the analyzed
EUL techniques is done. In section III theoretical
considerations for simulations are described. In section IV the
simulation scenarios are presented. In section V simulations
results are shown and conclusions are done in section VI.
II. ANALYZED EUL TECHNIQUES
The 3.5G EUL techniques studied in this work for 3G
WCDMA UL are described bellow.
A. Shorter TTI Size
Third generation WCDMA systems support frame sizes
with durations equal to 80, 40, 20 and 10 ms. The frame sizes
supported by E-DCH are 2 ms and 10 ms, showing that E-DCH
allows reduction of five times in the 3G WCDMA TTI size.
Reduce TTI fromlO ms to lowers values can mean reductions
in data transference delays, in the air interface, and the delays
of frame alignment (the data to be transmitted have that to wait
until the start of the following TTI) [6]. Moreover, this
reduction allows fast data processing, reduction in the amount
of buffer information in both User Equipment (UE) and
Node-B. As a consequence, the end user throughput also can be
expanded with a better use of the resources in terms of delays
[1]. Fig. 1 shows the TTI used in the simulations.
B. Hybrid Automatic Repeat Request
Due to fading limitations in 3G propagation channels, it is
necessary the use of error-correction codes and retransmission
techniques. The ARQ protocols apply error-detection codes in
the data packages. When an error occurs, the receiver of a
message discards the received package and sends information
of negative acknowledgment (NACK) to the transmitter,
through a feedback channel [14]. If the transmitter received
acknowledgment (ACK) message, which indicates that the
message was correctly received, it will send the next data
sequence. The main problem of these techniques is the delay
introduced to the system. The Forward Error Correction (FEC)
protocols make possible the reduction in the number of
necessary retransmissions. The junction of FEC and ARQ is
called Hybrid-ARQ (HARQ), which may be done with or
without combining techniques.
223
 L: LL2 HACK
TTI= lOsu TTI= iOnu TTI=2msu iondea
data transMis

DCH Frame E-DCH Frame B-DCH reduced

fraro (a)
L2 Ll
LI NACK
Fig. 1. TTI sizes used for simulations. i rdata d at =i
Ira"sIn
In HARQ schemes, two types of combining techniques are L2 NACK --

widely used Chase Combining (CC) and Incremental (b

Redundancy (IR) [10], [13]. If a retransmission is necessary in
3G WCDMA systems, the flow of signalization messages is
done between UE and RNC (Radio Network Controller), in the Fig. 2. Messages flow in a retransmission process:
data link layer (Layer 2) [6]. This process, called L2 ARQ, (a) ARQ in layer L2. (b) HARQ in layer LI
shown in Fig. 2(a), produces a longer delay in the transmission,
which can represent problems in real-time services as gaming, -
for example. With EUL schemes, Node-B controlled HARQ
mechanisms (Layer 1 HARQ), shown in Fig. 2(b), are N
implemented in UL, as well as had been in the DL with NR= 1- PG (i+1) (1)
HSDPA [2], for reduce longer delays and fading. The E +I
retransmission delays reduction allows the operation of the No
physical channels with somewhat higher error probability,
which turns into increased system capacity and better spectral
efficiency [5].
The UL load factor, qUL, is a parameter used in the
determination of throughput in WCDMA systems. 17UL is
C. Multi-User Detection function of the WCDMA chip rate (W), own user bit rate (Rj),
In a conventional CDMA system, simultaneous users in a voice activity factor (vj ) and is given by:
same cell and users of neighbouring cells interfere ones on the
others. Theoretically, significant increases in capacity and 77 = (1 ± i) 1 =1- 1
improvements in the near-far effect can be obtained if the = W 1 NR (2)
interference that each user has on the others could be cancelled (Eb/ V
R. -v.
(or attenuated). The MUD can be applied for this purpose. In R *v
contrast to the conventional CDMA detection, with banks of
matched filters, in the MUD the signals of different users are Equation (2) is called load equation and it predicts the
jointly used to detect the signal of each user individually. One amount of NR over thermal noise due to interference. The
way to execute the MUD is through the Interference relation between NR and 71UL is given as:
Cancellation (IC). The basic principle of the IC is the creation,
in the BS receiver, of separate estimates components of the
Multiple Access Interference (MAI), from each user to NR(dB) = -1 0 log10 (1 - 77L) (3)
eliminate, overall or partially, MAI caused by simultaneous
users [6], [11]. The MUD efficiency, EfMUD , is the used
parameter to determine how much of the MAI can be The WCDMA UL cell throughput (I) is function of bit rate
eliminated. EfMUD is directly related to the complexity of (R) and BLER for a specific EbIN0 given as:
practical implementation of this MUD receiver.

III. THEORETICAL CONSIDERATIONS FOR SIMULATIONS

To verify the application of the EUL techniques, shown in T L1 - BLER Eb _+_ R I
77_ u (4)
section II, in a 3G WCDMA system, a link level simulator was L K No Eb_E i±+
developed and the system performance was analyzed in terms N
of reduction of the retransmission delays and throughput in °
function of noise rise (NR). More detailed analytical
formulation for simulations can be found in [6], [11], [13], For MUD schemes, (1) can be rewrite as [15]:
[15]. In HARQ schemes the NR is function of the number of
simultaneous users (N), the processing gain (PG), the other-to-
own cell interference ratio (i) and the ratio of signal energy per N&AD=1+ (1
N*(i
N +1)
±)
bit divided by noise spectral density (FI/No), required to meet a +(______- i+-Eg9 5
predefined block error rate, (BLER). Noise includes both (b
thermal noise and interference [6]. NR is given in (1). t/O

224
 The MUD throughput gain, TGMuD, is given by [15]: determine some signal parameters like EINo and BLER.
Results derived from AVI tables were applied in this work like
PG in [5], [16]-[17]. For all the simulated scenarios the value of the
(E XA+ (1 EfMUD) WCDMA chip rate W=3.84 Mcps and the bit rate for packet
No i± +1 data services R=64 kbps were considered. A summary of the
TGAMl.D= PDG |XiI f.U ( 6) main simulation parameters is shown in TableI1[13].

No V. SIMULATIONS RESULTS
As previously defined, the simulations were divided into
If several traffic classes are considered, equations (I)-(6) three different scenarios.
need to be more complete, considering PG, vj, EbINO of each A. Delay reductions with HARQ and Shorter TTI
service. See more details in [6], [13], [15]. The most relevant simulations results about the estimated
time needed to transmit a complete data frame, with
IV. SIMULATIONS SCENARIOS TTI=10 ms, in UL of a WCDMA network are shown in
Table II. These results show that for BLERtarget=20% and
TheL2 retransmission delay reductions
LI and
analyzed through the combination of Shorter TTI and HARQ,
were
~~~~~~~~~~~maximum
reducinund
number of
30 wa Li1
retransmissions M=5. A delay
obtainedswhen apyn EUL el
lae'I
in~~~~~~~~~~~~~~~~~
in the layer Li, techniques.
ehius th The h UUL throughput
hogptgi gain inna3
a 3G reduction- around
HARQ 34% was
B/C schemes. obtained
Table III presents applying
when the results for
WCDMA system was verified through the implementation of TTI= s.hInthi
C
scenIo the reducti for
Node-B controlled HARQ (here called EUL HARQ LI) TTI= 2 ms. In thas scenard o the delays reductron for
schemes and MUD receivers with IC type PIC (Parallel BLERarget=20%
potential
was around 41.5% that demonstrates the
Interference Cancellation) [12]. For simulations ITU Power sytems eerse link.two techniques for enhance 3G WCDMA
of these
Delay Profiles (PDP) for macro-cell and micro-cell systemsreverse link.
environments were used: ITU Vehicular Type A (here called
VA) and the ITU Pedestrian Type A (here called PA). The UE B. Throughput enhancements with HARQ
speeds are 50 km/h and 3 km/h (here called VA 3, VA_50 and In these simulations, the larger UL throughput expansion
PA 3). For the throughput and delay analysis, with EUL with Node-B controlled HARQ schemes were obtained for
HARQ LI, three different scenarios were considered. The first VA_50 PDP. Fig. 3 plots the throughput in function of NR
scenario, called Rel-99 ARQ L2, ARQ retransmissions occurs obtained when HARQ techniques were applied, i.e., the
between the RNC and the UEs. In the second one, called EUL BLERtarget were increased from 1% to 10%. The highest
HARQ - A, it was adopted the HARQ - I, without throughput enhancements were obtained in high mobility
combination techniques. In last, called EUL HARQ - B/C, the macro-cellular channels, with UEs speeds of 50 km/h, due to
HARQ type II and III were adopted, with CC and IR highest spectral efficiency gain when BLERtarget were increased
combination techniques [14]. The retransmission delays of the from 1% to 10%. For NR=6 dB, that means qUL = 75%, the
three scenarios were analyzed considering the UL of a cell with 3.5G HARQ schemes had provided throughput gains around
omnidirectional antennas. Simulations were done for TTI = 2 21.5%. A comparison of the results for all the PDPs simulated,
and 10 ms, and a maximum number of retransmissions, Ml for NR=6 dB is reported in Table IV.
were five. The presented analyses considered that the
implementation of EUL HARQ LI allows the system operation C. Throughput enhancements with MUD
with higher block error rate, BLERtarget, than allowed in the 3G
WCDMA [5]. The BLERtarget reduction allows a systematic In order to determine the cell throughput enhancements
reduction in the EbINO ratio for the simultaneous UEs connected with MUD schemes, the simulator estimates the capacity gains
in the network and proportional throughput increases, forthethree consideredvalues of EfMUD inthiswork.
dependent on the analyzed PDP [6].
In this paper, MUD receivers were classified according to 700 - Re-99ARQL2
HARO-AVA 50
-------V-----------A------------
the complexity of practical implementation. Thus, high + EUL
EUL HARQ B/C VOA' '
complexity of practical implementation means EfuD=80%,
average
complexity EftDu=40%, and low complexity 500 ------------------------------ ---------------- l
ForEfmuD=20%. MUD
the analyses we considered a network 400 ---------- ---------- --------------------I--------
composed by 4 Node-B, with 3 sectors each, resulting in 12
cells. The MUD mechanism was implemented in one cell. In
theothers,
ARQ L2Rel-9 mechanisms were implemented, 200 ---------- -----------------------------------------------
that is a typical 3G WCDMA retransmission processes [6]. 10
Although the other-to-own cell interference ratio, i, be defined ti--l
°°
by a random variable, only its average values were considered, i-X __ I __ __ I __ _
according to the PDP under analysis. The AVI (Actual Value Nose Rise (dB 6
Interface) tables were used in recent works in order to
Fig. 3. Throughput in function of NR.

225
 Fig. 4 plots the WCDMA analytical qUL in function of the the 3.5G EUL studied techniques. Node-B controlled HARQ
cell throughput gain with MUD. These results were obtained results shown that these schemes are able to enhance
for Ef,D = 4000 and diverse values of i. The cell throughput throughput capacity and reduce delays, for both TTI= 2 and
10 ms frame sizes. The most significant enhancements were
enhancements simulations were done for the three ITU PDPs found for ITU macro-cellular PDPs, for UEs speed about
(macro-cell and micro-cell). The results for Ef,,,glD and 50 km/h. The MUD schemes also shown large potential for
BLER=20%, 10% and 1% are given in Tables V, VI and VII. In throughput expansion, specially in low mobility scenarios (ITU
these simulations, the larger UL throughput expansions with micro-cellular PDPs). The capability of expansion is directly
MUD receivers were found for PA_3 PDP. For EfMD = 40% proportional to the efficiency of detection and to the
the gain was around 27.02%, for 'uL =75%, that turns this complexity of practical implementation of MUD receivers in
average complexity MUD receiver an option for practical BSs.
implementations for enhances in UL cell throughput in current
3G WCDMA networks. Fig. 5 plots the throughput in function REFERENCES
of NR results obtained for NR=3 and 6 dB, for all simulated [1] 3GPP TR25.896 V6.0.0 Feasibility study for enhanced uplink for UTRA
EfmUD values, in a PA_3 PDP with BLER=20% and i=40%, FDD. March 2004.
compared with 3G WCDMA performance. [2] 3GPP TR25.858 V5.0.0. HSDPA Physical Layer Aspects. March 2002.
[3] G. Sharma, G. Kumar, "Moving towards HSUPA (high speed uplink
packet access): a complete 3.5G wireless system", Personal Wireless
Communications, 2005. ICPWC 2005. 2005 IEEE International
Conference, pp. 174- 177, January 2005.
---------- -------- -~r--2-2---n------ ---'---------- ---------- [4] R. Ratasuk, A.Ghosh, et all. "Predicting Link Level Performance for
Enhanced Uplink", Vehicular Technology Conference, 2004. VTC2004-
----------n-tA--.@o--------------n---------'---------r---------Fall. 2004 IEEE 60th Vol. 6, pp. 4185 - 4189, September. 2004.
/f . . . [5] C. Rosa, J. Outes, K. Dimou, T. S0rensen, J. Wigard, F. Frederiksen,
.a . and P. Mogensen. "Performance of Fast Node B Scheduling and LI
HARQ Schemes in WCDMA Uplink Packet Access", IEEE Proceedings
---------I-------- --------- ------ of the 59th Vehicular Tech. Conf., VTC-2004, pp.1635-1639 May 2004.
// [6] H. Holma, Z. Honkasalo, S. Hamalainen, J. Laiho, K. Sipila, A. Wacker.
-M -- i= O/. WCDMA for UMTS: Radio Access for Third Generation Mobile
i=25%
I. , , . , i-40% ,Communications. John Wiley & Sons. Third Edition, 2004.
E [_j-5E% [7] H. Holma And A. Tolli. "Simulated and Measured Performance of 4-
i=85 Branch Uplink Reception in WCDMA", IEEE Proceedings of the 53rd
i=100% Veh. Tec. Conference, VTC-2001, Vol. 4, pp. 2640- 2644, May 2001.
40 -0
10 0 L ____ | ____
O 10 20 30 0 70 [8] E. AXELL. "Coexistence Real Time and Best Effort Services
0~~~~~Preta
20oghu 30
Gain0[]E.AEL
4
Cell Throughput Gain
Percentual Cel Ceitne of fRa ieadBstEfr evcsiin
Enhanced Uplink" WCDMA. Master thesis. Department of E.E.,
Linkopings University, Sweden. January 2005.
D. Malladi, J. Damnijanovic, X. Zhang and S. Willenegger.
[9]
Fig. 4. WCDMA analytical 'JUL in function of the percentual "Introduction to WCDMA Enhanced Uplink", Qualcomm Inc. IEE, 3G
Mobile Comm. Technology 2004, pp. 240-244, October 2004
cell throughput gain for EfmuD =40%. [10] M.Bertinelli, E. Malkamaki. "HARQ for WCDMA Enhanced Uplink:
Link Level Performance in SHO ", 15th IEEE Intemnational Symposium
1800 _- Lion Personal, Indoor and Mobile Radio Communications, 2004, PIMRC-
---------
~~~~~~~~2004. Vol. 4, pp. 2856 -2860, September 2004.
1600 [11] N. Souto, J. Silva, A. Correia, F. Cercas, A. Rodrigues, A. Santos.
1400 "Multi-user detector schemes for the UMTS uplink transmission" IEEE
Eighth Intemnational Symposium Spread Spectrum Techniques and
1200 Applications, ISSSTA-2004, pp. 364-368, August/September 2004.
1000*N=E;
dB E_Ul[UD=8[12] A. ToskalaHommloinen S. and H. Holma. "Link and System Level
Performance of Multi-user Detection CDfM Uplink", Wireless Personal
Communications 8, No. 3, pp. 301-320, 1998.
Fig.5.Throughputinfunction ofNRforPA 3. [13] Rodrigues, E. J. B. "Um estudo do aumento da capacidade de sistemas
400 ~~~~~~~~~~~~~~~3G
WCDMA usando t6cnicas 3 .5G de expansdo do enlace reverso,"
200 aM.S. thesis, Dept. Elec.. Eng., UFRN., Natal, Brazil, December 2005
[14] F. Bauer, F.Berens, D. Huy, J. Rodriguez, and X. Yang. "D4.6 System
Level Simulation Results - multi-cell environment", MATRICE iST-
enhancementsforthe EG
MUD MA reverse linkwhen g2001-32620, September 2004.
NR = dB
3
[15] N. Gikonomopoulos, P. Cosimini. "System Level Performance of Multi-
user Detection for W-CD MA", 3G Mobile Communication
Technologies, 2002. pp. 276-280. May 2002.
Fig. 5. Throughput in function of NR for PA_3. [16] J. Outes, Uplink Capacity Enhancement in WCDAM. Ph.D. Thesis,
VI. CONCLUSIONS
i [17 Wacer F. Fredeikse, and T. KdIng. "Performanc an Mdeino
 TABLE I. SIMULATIONS PARAMETERS
Parameters PDP
ITU Macro-cell ITU Micro-6ell
Path loss (dB) for distance in (km), BS antenna height equals 15 m and WCDMA FDD L[dB] = 128.1 + 37.6* log10 d L[dB] = 147.7 + 40 . log10 d
carrier frequency 2 GHz.
Other-to-own cell interference ratio seen by the BS receiver. Macro-cell with 3 sectors: i = 65 % Micro-cell with 3 sectors: i = 40 %
AVI values o6f EI as function ofBLER for considered ITU PDP (dB) VA_3 VA_50 PA_3
BLERtarget= 20% 2.80 3.40 2.62
BLERtarget= 10% 3.20 3.90 2.80
BLERtarget = 1% 3.90 5.30 3.00
Number ofusers for simulations (V) in function of BLER
BLERtarget= 20% 85 77 70
BLERtarget= 10% 95 80 75
BLERtarget= 1% 130 90 80

TABLE II. RETRANSMISSION DELAYS FOR AND TTI= 10 ms

Delay (ms)
Simulated BLERPre
Scenarios 20% 10% 1%
Rel-99 ARQ L2 109.27 86.07 71.31
EUL HARQ - A 82.56 75.57 70.51
EUL HARQ - B/C 81.41 75.55 70.50

TABLE III RETRANSMISSION DELAYS FOR AND TTI=2 ms

Delay (ms)
Simulated BLER",
scenarios 20% 10% 1%
Rel-99 ARQ L2 102.5 84.44 71.31
EUL HARQ - A 72.53 66.67 62.42
EUL HARQ - B/C 72.43 66.66 62.41

TABLE IV. THROUGHPUT ENHANCEMENTS WITH HARQ FOR ALL SIMULATED PDPS AND NR= 6 dB

PDP Throuhput (ls)

Rel-99 ARQ L2 EUL HARQ - A EUL HARQ - B/C
VA_3 692.98 749.12 756.73
PA_3 956.55 969.40 979.02
VA_50 508.41 645.83 651.68

TABLE V. THROUGHPUT ENHANCEMENTS WITH 3.5G MUD RECEIVER. 7?UL = 75% AND BLER=20%
Throuht (kbs)
PDP
Rel-99 EfMUD=20% EfMUD =40% EfMUD =80%
VA_3 752.49 832.05 931.51 1220.63
PA_3 998.85 1115.21 1268.74 1739.89
VA_50 660.14 729.94 817.19 1070.82

TABLE VI. THROUGHPUT ENHANCEMENTS WITH 3.5G MUD RECEIVER. 7?UL = 75% AND BLER=10%
ThouWgut (kbp)
PDP
Rel-99 EJPJD =20° EfpgD =40% EfJgD =80%
VA_3 774.06 855.90 958.21 1255.61
PA_3 924.65 1032.38 1174.50 1610.66
VA_50 671.20 742.17 830.88 1088.76

TABLE VII. THROUGHPUT ENHANCEMENTS WITH 3.5G MUD RECEIVER. 7?UL = 75% AND BLER=1%
Throughput (Wks)
PDP Rel-99 EJMUD =20% EfMUD =40% EfMUD 80%
VA_3 722.20 798.56 894.01 1171.49
PA_3 956.62 1068.07 1215.10 1666.34
VA_50 531.15 593.95 664.95 871.33

227