Vous êtes sur la page 1sur 6

3G HSPA for Broadband Communications with High

Speed Vehicles
Santiago Tenorio#1, Paul Spence*2, Beatriz Garriga#3, Javier Lpez#4, Aitor Garca#5, Miguel Arranz #6
#

Vodafone Technology Networks, Vodafone Spain


Isabel Colbrand, 22, Madrid, Spain
1
Santiago.Tenorio@vodafone.com
3
beatriz.garriga@vodafone.com
4
javier.lopez3@vodafone.com
5
aitor.garcia@vodafone.com
6
miguel.arranz@vodafone.com
*

McLaren Electronic Systems


Woking, Surrey, GU21 4YH, United Kingdom
2
Paul.Spence@mclarenelectronics.com
Keywords: HSPA, Doppler, Mobility, High Speed, Telemetry, Formula 1
Other known standard mobile communication systems
Abstract This paper presents a proof of concept for a
encounter equal or worse technical challenges as for instance
continuous
superior
quality
Broadband
Vehicular
for any Orthogonal Frequency Division Multiplex (OFDM)
communication system enabled through 3G HSPA in very high
based system, frequency offsets, phase noise, and Doppler in a
speed mobility scenarios (beyond 300 km/h), suitable for
time varying channel quickly result in a significant degraded
telemetry applications in trains, emergency vehicles and motor
performance. In addition, this is coupled with impact from
sport events. The system is quite unique as radio transmission for
inter-carrier interference (ICI) between the OFDM sub-carriers
telemetry services under extreme speed conditions requires not
only superior Quality of Service guarantees but must also be able
and increasing complexity in the carrier estimation.
to satisfy these performance requirements under extreme and
This paper addresses the challenge of utilising and optimising
arbitrarily demanding environments as are typical during any
an existing commercial 3G HSPA Broadband system for
e.g. Formula 1 racing event.
communication to very high speed vehicles. As the main
immediate objective, this activity sought to establish a
Issues related to the Doppler Effect and abrupt changes of the
working baseline for a new generation of telemetry systems
serving HSPA channel are analyzed and addressed here.
suitable for high speed applications and delivers a Proof of
Conclusions show how a special 3G network design can help to
Concept in a Formula 1 environment.
mitigate Doppler Effect impacts. The processes carried out by
In this paper, we introduce the setup utilised and the results
both the UE and the network to cope with this high speed
environment has proven essential to sustain the service in these
obtained from several tests performed in different high speed
conditions, as it has the use of suitable receiver Types in the UE.
scenarios covering controlled environments.
Using derived guidelines and conclusions, a unique system has
been developed, built and tested in a Formula 1 environment
with very promising results.

I. BACKGROUND
rd

3 Generation Partnership Project (3GPP) has standardized


WCDMA-based packet-switched air interfaces for both
downlink and uplink called High-Speed Downlink Packet
Access (HSDPA) and High-Speed Uplink Packet Access
(HSUPA) respectively [1]. Under conditions where the signal
strength on the source cell is rapidly deteriorating (as in high
speed scenario) it can occur that the UE may not be able to
reliably decode the necessary mobility information, the
Service Cell Change (SCC) and Radio Resource Control
(RRC) message(s) from the source cell leading to a call drop.
An attempt to re-establish the call as defined in the standard
[2] under these extreme speed conditions is also a challenge as
verified in live testing environment.

II. EQUIPMENT DESCRIPTION


The UTRAN network was provided by a 3G equipment
manufacturer using products and features commercially
available.
Five types of devices were tested [3]

Type 1 Receiver, cat 8 HSDPA and cat 5 HSUPA


Type 2 Receiver, cat 8 HSDPA and cat 5 HSUPA
Type 3 Receiver, cat 8 HSDPA and cat 5 HSUPA
Type 3 Receiver, cat 10 HSDPA and cat 6 HSUPA
Type 3 Receiver, cat 14 HSDPA and cat 6 HSUPA
III. SCENARIO DESCRIPTION

All testing took place in a high speed circuit - IDIADA - in the


north east of Spain. A dedicated 3G network was installed in
the 2.1GHz band using four Nodes B covering the entire
circuit, and in addition a compact RNC and CORE network

978-1-4244-2519-8/10/$26.00 2010 IEEE

was also installed on-site. Two of the sites covered the curves
in the circuit and were located in a strategic position in order
to avoid Doppler Effect impact. The remaining two sites were
installed close to the straight sections of the track - straights
- to fully analyse the Doppler Effect.
For coverage purposes, 20m masts were used. Sites in the
curves and straights had tri-sector and bi-sector design with
65 and 33 cross-polar antenna beam width respectively.

The main outcomes regarding Doppler Effect on HSPA


performance are:

In downlink, in neither RSCP nor Ec/No no significant


degradation was detected. The HSDPA throughput loss
was 16% in the worst case
In uplink, the BLER and retransmission rate critically
increases at speeds beyond 180km/h where throughput
drops to almost 0kbps if no Doppler compensation
functionality is activated in the node B

Figure 3 shows the effect of Doppler Speed on HSUPA UL


data traffic. As speed increases the number of retransmissions
increases impacting on throughput.
180 Km/h

Car Speed

-180 Km/h

Doppler Speed

5 Mbps

Test scenario, IDIADA circuit

HSUPA Throughput
0,5 Mbps
7

Several tests were carried out to measure and quantify the


impact of both Doppler spread and mobility management
limitations in extreme conditions i.e. a high speed vehicle.
Particular attention was given to the evaluation of data service
performance itself, in particular measuring the throughput, the
RTT (Round Trip Time), and the effect of cell change when
the vehicle was traveling across two or more cells at speeds of
around 250km/hr and beyond.
A) DOPPLER SHIFT IMPACT
The main affection of high speed scenario is Doppler shift,
also known as Doppler Effect. This effect is the change in
frequency in of a wave for an observer moving relative to the
source of the waves.
For waves that propagate in a medium, such as radio waves,
the velocities of the observer and of the source are relative to
the medium in which the waves are transmitted. Doppler shift
follows the next formula, also represented in the figure 2:

Retransmission Number

0
65 dB

UE Power Headroom

20 dB

Fig. 3

CAT6 HSUPA UL throughput degradation above 180km/h

Figure 4 shows the effect of Doppler Speed on HSDPA DL


data throughput depending on the UE receiver Type. Type 3
receiver achieved the best performance under different speeds ,
and Type 2 showed the biggest throughput degradation of
around 16% respect to 30km/h. Regarding HSDPA coverage
loss measured by CQI, all devices showed similar degradation
trend at high speeds i.e. approximately 1.2dB less compared to
low speeds.
UE Receiver Type: --- Type 3

CQI

IV. DOPPLER SHIFT AND MOVILITY ANALYSIS

---Type 2

---Type 1

27

6000

26

5500

25

5000
4500

24

4000

23

3500

22

3000

f
f d = v cos
C

21

2500

: angle between UE mobility and signal propagation


directions
v: vehicle rate
C: radio spread rate
f: carrier frequency

Fig. 4

20

kbps

Fig. 1

2000
30 Km/h

50 Km/h

80 Km/h

120 Km/h 150 Km/h 180 Km/h 220 Km/h 250 Km/h

dot lines CQI / continious lines MAC throughput

CAT8 HSDPA DL throughput and CQI vs. receiver type and


Doppler Speed

To mitigate the effect of Doppler shift in user performance, the


following algorithms were implemented in the network side:
1) Frequency offset estimation
2) Frequency offset compensation

Fig. 2

Doppler shift components

There are different methods to perform frequency offset


estimation and compensation - although most of them are
vendor proprietary algorithms.
Figure 5 shows the HSUPA throughput gain with a Doppler
shift compensation algorithm on. At 250Km/h, there is no

throughput degradation comparing to 80 and 150km/h cases.


Regarding UE power, 17dB less is required to get 3.7Mbps
more if Doppler shift compensation is put in place.

DISTANCE REQUIRED FOR SHO AND HSDPA CELL CHANGE

Distance (m)

Physical UL Throughput (Mbps)


4.5

4.0

3.7

4.1

4.0

Cell addition

300

Cell deletion

HSDPA Cell Change

250
200
150
100
50
0

30

2
1
0
80Km/h

150Km/h

33.0

30.9

31.2

36.0

32.1

30
14.2

20
10
0
80Km/h

150Km/h

DOPPLER Feature Off

Fig. 5

80

120

Fig. 6

250Km/h

Average Power Headroom (dB)


40

50

150

180

220

250

300

350

Speed (Km/h)
Note: Time to trigger 1a: 100ms; Time to trigger 1b: 640ms; Time

0.4

250Km/h

DOPPLER Feature On

CAT6 HSUPA Throughput and Remaining available power in the


UE vs. Doppler speed

Distance per event

Inter frequency and Inter system handover: The


process takes approximately 1.4s ~ 2s in case of inter
frequency handover and 1.4s in case of intersystem
handover under normal conditions.
Cell reselection: When camped on a cell, the mobile
shall regularly search for a better cell according to the
cell reselection criteria. Cell reselection failure was
frequently detected in high speed scenarios because UE
has changed cells before the cell reselection timer
expires.
Performance improvement in high speed mobility scenarios
requires:

B) MOBILITY IMPACT
1.
Current 3G network mobility processes are optimized to
operate at medium-low speed mobility (<150km/h) so under
high speed scenarios a different parameterization and design is
needed.
Soft handover: New target cell addition time is about
400~800ms (from new cell detection by UE till active
set cell update complete message). Additionally,
HSDPA DL service requires cell change when a new
target cell becomes x-dBs better than the actual cell
level in terms of EcNo. This process takes usually a
longer period of time since physical channel
reconfigurations must be performed in the UE. The
average time required to perform such action is about
1.2~3.3s with 1.75s the average and 0.71s the standard
deviation (from new cell change condition detection by
UE till physical channel reconfiguration complete
message). This broad range is due to the fact that many
3G vendors implement specific timers to avoid ping
pong effect during HSDPA cell change. A full HSDPA
cell change procedure requires soft handover plus cell
change actions. Figure 6 shows the average distance
traverse during the cell addition, cell deletion and
HSDPA Reference Cell Change event depending on the
speed. The minimum total overlap distance required
between 2 cells would be around 260m at 300Km/h,
345m if being more conservative.

Avoid inter frequency and inter system handover to


reduce call drop rate and zero throughput periods.
Make usage of single cell configurations to avoid
handover and ping pong effect due to pilot pollution as
much
as
possible.
This
impact
increases
proportionately at higher speeds. Figure 7 represents
HSDPA performance when changing cell on a polluted
area. The CQI, HS-SCCH success rate and HS-DSCH
BLER degradation are bigger at high speed impacting
on throughput

2.

Reference Cell

Reference Cell

SC 25
SC 1

CQI

30

CQI

15

HS-SCCH Success rate

HS-SCCH Success rate

100%
50%

BLER HS-DSCH
30%
0%
6.9 Mbps

MAC HSDPA Throughput

BLER HS-DSCH

MAC HSDPA Throughput

0.5 Mbps

Pilot Pollution at 250 Km/h

Fig. 7

3.
4.

Pilot Pollution at 50 Km/h

Pilot Pollution effect on HSDPA performance vs. speed

Optimize network design maximizing handover


overlapping distance taking into account high speeds
Optimize parameters to accelerate UE decisions and
network reaction (e.g. accelerate 1A trigger, to make

difficult to trigger 1B configuration)


To optimize cell reselection timers (Treselection,
Qoffset and Qhyst) and reduce the system
information
update
time.
Additionally,
call
reestablishment timers should be optimized.

5.

Figure 8 shows MAC HSDPA throughput variation during cell


change by modifying 1d event reporting parameters (target
cell signal level strength over serving cell level Hys1d and
time to satisfy the threshold Ttrig) when triggering fast,
medium or slow cell change procedures. As seen on the figure,
being reactive and trying to be on the best cell always or
conservative delaying cell change procedure are not the best
optimization implementations to improve performance.
THROUGHPUT (Mbps)

Fast HSDPA Cell Change

Medium HSDPA Cell Change

Slow HSDPA Cell Change

3.7

3.7
3.6
3.5
3.4
3.3
3.2
3.1

3.5
3.3

Fast: Hys1d = 2 dB; Ttrig = 320 msec


Medium: Hys1d = 3 dB; Ttrig = 200 msec
Slow: Hys1d = 4 dB; Ttrig = 320 msec

Fig. 8

HSDPA Performance comparison during cell change with


different mobility optimization strategies

To avoid mobility issues and increase UL capacity, a feature


offered by several infra vendors called Multi-RRU has been
tested. This feature permits several physical cells to work as a
single cell for down link transmission and as independent cells
for up link reception
The following benefits are obtained from the use of this
feature in high speed scenarios:
Reduction of number of handovers controlled by the RNC
Flexibility on the coverage area of one cell to adapt the
network design to specific high speed scenarios needs
Enhancement of the uplink capacity (Throughput increase)
due to different sector RTWP management
Downlink diversity gain in the overlap coverage area of
different RRU's.

Application throughput (Mpbs)

Boths user in
same sector

Boths user in
different sectors

Boths user in
same sector

10
9
8
7

Figure 9 shows the effect of the feature on uplink cell capacity,


when both users are located in the overlap area of both sectors
the traffic in UL is shared between both (4.5Mbps) but when
each user is located under the coverage area of a different
physical cell the throughput increase as they dont have to
share common resources (around 4.5Mbps each)
V. HIGH SPEED PERFORMANCE IN AN OPTIMIZED
NETWORK
A) HSDPA PERFORMANCE (250 KM/H)
Figure 10 shows a HSPA/HSPA+ performance comparison
among CAT8, CAT10 and CAT14 devices at 250Km/h. Both
CAT10 and CAT14 devices benefit from 15 codes usage with
QPSK modulation and the Enhanced Layer 2 3GPP feature,
thus improving the HS-DSCH BLER. The benefit of using 15
codes with a new advance 3GPP Release 7 receiver type is
significant with up to 88% more throughput achieved with
only 5 more codes available. The reason for this is not only the
number of codes available but also the improved HSDPA
coverage measured by the CQI which is 3dBs better due to
improved receiver type. CAT14 HSPA+ 64QAM device only
gets 6% more throughput than CAT10. This is due to the fact
that there was an Iub limitation and the maximum achievable
throughput was 16Mbps (although Iub limitations aside, a
peak of 21.8Mbps would have been possible).
24.4

13

15.0

10 10.5
5.0

Average
MAC
Throughput
(Mbps)

Fig. 10

28 26.5

CAT8

14 14.6
9.7

8 8.2

6.2

Peak MAC
Throughput
(Mbps)

CAT10
CAT14

4.6

Average
code

Average CQI

Average
BLER (%)

CAT8, CAT10 and CAT14 HSDPA/HSPA+ performance


comparison at 250km/h

However, the main conclusion is that utilizing this Rx Type in


the UE permitted 64QAM modulation to be utilized in up to
27% of samples with the BLER measured as below the BLER
target of 10%. 64QAM modulation is not affected negatively
with high speed

6
5

B) HSUPA PERFORMANCE (310 KM/H)

4
3
2
1
0

Static UE

Fig. 9

Mobile UE

Total UL Cell Bandwidth

MRRU feature performance

Figure 11 shows a HSUPA performance comparison among


CAT5 and CAT6 devices at 310Km/h. The CAT5 device
achieved high and stable throughput providing 1.8Mbps on
average as measured at the physical layer (10% below the

CAT5

CAT6

7.9

Avg RTT Cat 5

Avg RTT Cat 6

Stand Deviation RTT Cat 5

Stand Deviation RTT Cat 6

300

264

250
RTT (ms)

maximum throughput). The CAT6 device provided more


unstable throughput but did achieve 4.1Mbps on average (29%
below the maximum throughput). Using the CAT6 device
required around 8.6dBs more power compared to the CAT5
device to achieve 127% more throughput

200

164

150
100
50

158

102
49

47 39
11 11

32

44
9

4.1

1.8

Fig. 13

HSUPA Physical Throughput


[Mbps]

Fig. 11

HSUPA performance

UL Cell capacity with multi-user was assessed at 250km/h.


For this purpose, 6 CAT5 HSUPA devices were placed in the
same car performing UDP uploads. For each of the users a
100kbps UL Guaranteed Bit Rate (GBR) was setup in the
Node B and 2 different maximum UL RTWP increase levels
relative to background noise were analysed: 30dB and 10dB.
Figure 12 shows the results obtained. The UL cell throughput
is slightly degraded by around 15% with a low RTWP increase
but the GBR is achieved 92% of the time whilst only 45.8%
was reached in the other case. Achieving 100% GBR would
require an even lower RTWP increase allowance.
% Time Samples

Total UL Application Throughput


100%

92.0%

2.00

80%

1.50

60%
45.8%

1.00

40%
30.2%

Time samples (%)

HSUPA Throughput CAT5 (Mbps)

2.50

21.2%
0.50

20%
2.8%

0.3%

0.00

Round Trip Time variation with soft handover for CAT8


HSDPA/CAT5 and CAT6 HSUPA devices

6.7%
1.0%

VI. CONCLUSIONS
The trial in IDIADA demonstrated that the 3GPP HSPA
technology works with only a slight degradation at speeds
beyond 300km/h.
A Doppler shift compensation feature is required in the node B
when the UE moves at speeds over 180km/h to avoid
throughput degradation. No significant Doppler effect has
been seen on the UE side.
HSPA performance enhances in single cell configuration with
a multi RRU feature improves uplink cell capacity.
Ad-hoc optimization and design solution(s) are mandatory to
avoid cell change ping pong effect due to pilot pollution.
Inter-cell overlapping distances over 300m are recommended
to avoid HSPA performance loss due to lack of soft handover
or cell change time availability.
Network performance is improved using HSPA+
CAT14/HSUPA CAT6 devices in comparison to legacy
devices i.e. better downlink and uplink throughput, round trip
times were achieved and adapted better to the coverage and
radio environment.

0%

ROT 10dB ROT 30dB ROT 10dB ROT 30dB ROT 10dB ROT 30dB ROT 10dB ROT 30dB
3

VII. ACKNOWLEDGEMENTS

Simultaneous Users
Fig. 12

The Round Trip Time of a HSUPA CAT5 device was also


measured via simulating load in addition to another CAT5
device performing FTP uploads at 310Km/h. On average, the
RTT value in loaded conditions increased from 59ms to 80ms
(35% increase).

-0.7
Tx Power (dBm)

2
Number of cells in active set

Average HSUPA application user throughput vs. number of


satisfied users with GBR

C) ROUND TRIP TIME RESULTS (250 KM/H)


Figure 13 shows the Round Trip Time performance
comparison for CAT5 and CAT6 HSUPA devices when
sending 64bytes ping packets at 250Km/h. As seen in the
figure, CAT6 HSUPA device achieves 8ms lower RTT values
on average (-17%) comparing to CAT5 device. Besides, both
average and standard deviation RTT values increase critically
when the CAT5 device was in soft handover.

The authors would like to acknowledge Qualcomm, Huawei


and ZTE Corporation for facilitating the necessary UE,
network infrastructure and related support, and in particular to
the Vodafone McLaren Mercedes Racing Team for their
support and access to their high-end engineering facilities.
VIII. REFERENCES
[1]
[2]
[3]

3GPP Rel-7 and Rel-8 White Paper (3G Americas).


www.3gamericas.org
3GPP TS 25.331 V8.9.0 Radio Resource Control (RRC); Protocol
Specification.
3GPP TR 25.101 V9.0.0 (2009-05) User Equipment (UE) radio
transmission and reception (FDD).

Vous aimerez peut-être aussi