Beruflich Dokumente
Kultur Dokumente
(3)
For purposes in this paper a CPICH RSCP = 30dB shall
be considered as it is the average value in the area of testing.
To obtain the quality, Ec/No, at the UE receiver antenna
[6] suggests that the following relationship is used as given in
(4).
E
c
N
o
=
Pilot Tx Power
Lp
(kTBNF) +
Total Cell Power
Lp
+
100
i=0
M
i
1000
(5)
In (5) L
1
is the output of the layer 1 lter and M
i
is the
sampled CPICH RSCP from the measurement report.
2) Layer 3 Filter: To accomplish the layer 3 ltering an
exponential averaging formula as described in [4]. Mathemat-
ically the layer 3 lter is expressed as in (6).
F
n
= (1 a)F
n1
+aM
n
(6)
The coefcient a in (6) is known as the layer 3 coefcient and
can also be adjusted/tuned, F
n
indicates the layer 3 ltered
measurement output for the current output and F
n1
the layer
3 lter output from the previous time step. M
n
is the layer 1
lter output L
1
.
3) Soft Handover Inequalities: The standards prescribed in
[3] describe the decision algorithm used to determine when
radio links should be added to the active set or removed from
it. To consider a radio link to be added to the active set the
following inequality is considered:
10Log(M
new
) W 10Log(
N
A
i=1
M
i
) + (1 W)
10Log(M
best
) (R
1a
H
1a
2
) (7)
For purposes of this paper W or the weight term shall be
considered as 0 for both event1a and event1b. Thus for event1a
the inequality in (7) becomes:
10Log(M
new
) W 10Log(M
best
) R
1a
+
H
1a
2
(8)
In (8), M
new
is the newly measured radio link and M
best
is
the best radio link currently in the active set. Also in (8) are
the terms R
1a
and H
1a
representing the reportingrange1a and
hysteresis1a parameters.
The inequality to determine if a radio link should be
removed from the active set according to event1b is described
as follows according to [3]:
10Log(M
old
) 10Log(M
best
) R
1b
+
H
1b
2
(9)
191
Where M
old
is a old radio link in the active set and R
1b
and
H
1b
are the reportingrange1b and hysteresis1b parameters.
IV. MONTE CARLO METHOD
To compute a problem subdued to a random input several
executions of the problem is required. Executing a problem for
every possible random input to the system is computationally
expensive. The Monte Carlo method assumes the output from a
problem is representative of the whole population or possible
outcomes. Accordingly only a certain amount of executions
are necessary for outputs within a certain level of condence.
V. MONTE CARLO SIMULATION
According to [8], [9], [10] and [11] the Monte Carlo method
can be viewed as a general method of numerical integration.
Function f(x) is to be evaluated by a denite integral to
determine the area otherwise stated as approximating the
function or integral.
=
1
0
f(x)dx (10)
The function f(x) in (10) might be multidimensional, which is
usually the case with functions being analysed with the Monte
Carlo simulations, and can be rewritten as:
=
1
0
1
0
1
0
f(x
1
, x
2
, , x
n
)dx
1
dx
2
dx
n
=
(0,1)
n
f(x)dx (11)
Equation (11) is a denite integral providing a deterministic
answer. If x contains random numbers
1
uniformly distributed
throughout the area of the integral, the function f(x) becomes
a random variable.
The random variable f(x) has an expected value:
2
E[f(x)] =
(0,1)
1
f(x)dx (12)
By comparing (11) with (12) a probalistic expression for the
integral is found to be:
= E[f(x)] (13)
A statistical estimator for the expression in (13) is given in
[10] to be:
H = f(x) (14)
The estimator in (14) is unbiased and indicates the random
variable f(x) has a mean ,
k=1
f(x
k
) (15)
1
Numbers chosen randomly within a certain statistical distribution
2
The sum of the values of a random variable divided by the number of
values for discrete random variables
In [8], (15) is referred to as the crude Monte Carlo estimator
with a standard deviation
n
. In practise the variance is
obtained through:
s
2
=
1
n 1
n
k=1
(x
k
x)
2
(16)
Where x is the average of all values contained in vector x. The
standard deviation, , is then found to by taking the square
root of s in (16).
=
s
2
=
1
n 1
n
k=1
(x
k
x)
2
(17)
According to [11] the crude Monte Carlo estimator is similar
to the theory of large numbers. The theory of large numbers
implies that a quantity
Q can be approximated if N samples
of independent values Q
1
, Q
2
, , Q
N
are used to calculate
the mean.
Q =
Q
1
, Q
2
, , Q
N
N
(18)
For (18) to approximate Q to a satisfactory level of accuracy
N has to be very large, leading to computationally expensive
solution. For purposes in this project N was chosen to be 100
as Q lead to a converging solution with 100 iterations.
VI. RESULTS
Using the outputs of the models discussed in the previous
sections one obtains counts of the amount of events 1a and 1b
that has been reported by the UE. The result of the amount
of reported events due to the reporting range and hysteresis
parameters are also included. Another outcome is the insight
into the distance away from the serving and measured base
stations.
A. Using only reportingrange1a and reportingrange1b
The rst set of results looks at the amount of event1a and
event1b reported according to (8) and (9). In gure 4 the
amount of event1a and event1b is indicated also with a trend
line indicating that for reporting range between 2 and 4, the
least amount of event reports occur.
0 2 4 6 8 10 12 14
500
0
500
1000
1500
2000
2500
3000
Reportingrange1a and Reportingrange1b
A
m
o
u
n
t
o
f
e
1
a
o
r
e
1
b
r
e
p
o
r
t
e
d
Event1A
cubic
Event1B
data 1
data 2
Fig. 4. The amount of events 1a and 1b reported with only reportingrange1a
and reportingrange1b set.
192
B. Taking into consideration the reportingrange and hysteresis
parameters
When considering (8) and (9) and values for hysteresis are
included. We now obtain results as given in gures 5 and 6. For
0 2 4 6 8 10 12 14
0
500
1000
1500
2000
2500
Reportingrange1a + Hyst1a
A
m
o
u
n
t
o
f
r
e
p
o
r
t
e
d
e
v
e
n
t
s
data 2
cubic
Fig. 5. The amount of event1as reported with reportingrange1a + hyst1a.
the reportingrange1a and hyst1a parameters, the least amount
of reports occur when the sum of the two parameters are 4.
The least amount of event1bs are reported when the sum of
reportingrange1b and hyst1b are between 10 and 12 as shown
in gure 6.
2 4 6 8 10 12 14 16
0
5
10
15
20
25
30
35
40
45
50
Reportingrange1b + Hyst1b
A
m
o
u
n
t
o
f
r
e
p
o
r
t
e
d
e
v
e
n
t
s
data 2
cubic
Fig. 6. The amount of event1b being reported when taking into consideration
reportingrange1b + hyst1b.
C. Distance away from the NodeB
The aim is to have the event1a and event1b reported from
the UE as close as possible to the middle of the distance
between the two base stations. Thus at a distance of 2.5 km
away from the transmitting base station, which is assumed
to be on the right hand side of the graph. In gure 7 the
distance from the base station is optimal for event1a when the
sum of the reportingrange and hyteresis values are between 2
and 4. At the aforementioned point in gure 7 the distance
from the serving base station is roughly 2.3 km. For event1b
the distance from the base station remained around 2.5 km
for various values of the reportingrange1b and hysteresis1b
parameters.
0 2 4 6 8 10 12
0
500
1000
1500
2000
2500
3000
3500
4000
4500
Reportingrange1a+Hyst1a and Reportingrange1b+Hyst1b
D
i
s
t
a
n
c
e
f
r
o
m
B
S
1
Event1a
Event1b
Fig. 7. Distance from the transmitting Nodeb of the rst reported event1a and
event1b. Negative distance indicates the distance from the measured nodeB.
D. Effects of implementing simulated results in the network
Once the results from the simulation was processed a deci-
sion was made that the optimal setting would be as follows:
reportingrange1a = 2.5 dB
reportingrange1b = 3 dB
hyst1a = 5 dB
hyst1b = 5 dB
The aim was to have a small amount of events reported as
close as possible to the midpoint between the two NodeBs. It
should be however noted that this simulation does not include
event1a and event1b reports being acknowledged explaining
that more events appear to be reported with the simulation
than in the actual network condition. From the gure shown
below (gure 8), drops due to soft handover failure and loss
of synchronisation in the uplink has decreased over time.
The time that the parameter settings were implemented is
indicated in the area between the two vertical red lines. Figure
0
10
20
30
40
50
60
70
1
0
/
2
2
1
6
:
0
0
1
0
/
2
3
1
1
:
0
0
1
0
/
2
4
0
6
:
0
0
1
0
/
2
5
0
1
:
0
0
1
0
/
2
5
2
0
:
0
0
1
0
/
2
6
1
5
:
0
0
1
0
/
2
7
1
0
:
0
0
1
0
/
2
8
0
5
:
0
0
1
0
/
2
9
0
0
:
0
0
1
0
/
2
9
2
0
:
0
0
1
0
/
3
0
1
5
:
0
0
1
0
/
3
1
1
0
:
0
0
1
1
/
0
4
0
1
:
0
0
1
1
/
0
4
2
0
:
0
0
1
1
/
0
5
1
5
:
0
0
1
1
/
0
6
1
0
:
0
0
1
1
/
0
7
0
5
:
0
0
1
1
/
0
8
0
0
:
0
0
1
1
/
0
8
1
9
:
0
0
1
1
/
0
9
1
5
:
0
0
1
1
/
1
0
1
0
:
0
0
1
1
/
1
1
0
5
:
0
0
1
1
/
1
2
0
5
:
0
0
1
1
/
1
3
0
0
:
0
0
1
1
/
1
3
1
9
:
0
0
1
1
/
1
4
1
4
:
0
0
1
1
/
1
5
0
9
:
0
0
1
1
/
1
6
0
4
:
0
0
1
1
/
1
6
2
3
:
0
0
1
1
/
1
7
1
8
:
0
0
1
1
/
1
8
1
3
:
0
0
1
1
/
1
9
0
8
:
0
0
1
1
/
2
0
0
3
:
0
0
1
1
/
2
0
2
2
:
0
0
1
1
/
2
1
1
7
:
0
0
1
1
/
2
2
1
2
:
0
0
Sumof PMNOSYSRELSPEECHULSYNCH
Sumof PMNOSYSRELSPEECHSOHO
Sumof PMNOSYSRELSPEECHNEIGHBR
Linear (Sum of PMNOSYSRELSPEECHULSYNCH)
Linear (Sum of PMNOSYSRELSPEECHSOHO)
Linear (Sum of PMNOSYSRELSPEECHNEIGHBR)
Fig. 8. A decrease is noticed since the implementation of the soft handover
parameters as indicated to be optimal by the simulation.
9 shows a signicant decrease in the dropped call rate on the
RNC which translates to improved subscriber experience. In
[1] it is mentioned that changing handover parameters can
affect resource utilisation of both the NodeB and RNC. It was
decided to obtain results in terms of NodeB resource utilisation
and RNC processor load during the trial period. The data was
193
Fig. 9. A denite decrease in the dropped call rate on the RNC after the
implementation of the soft handover parameters.
normalised and plotted as in gure 10. As shown in gure
Fig. 10. NodeB and RNC resource utilisation during the trial period.
10, no perceived increase in resource utilisation on either the
RNC or NodeB was observed indicating that with improved
retainability resource utilisation remained the same.
VII. CONCLUSION
The Monte Carlo simulation results provided an optimal
relation between hysteresis and reporting range parameters.
From the simulation results it is clear that small values are
required for reporting range parameters for both event1a and
event1b. Larger parameter values cause delayed soft handovers
with the possibility of dropped calls. A slightly larger value
for hysteresis parameters can be considered than for reporting
range parameters. Results from the simulation were used to
implement values with as few as possible reporting events
roughly halfway between the serving cell and measured cell.
Signalling load is impacted by unnecessary active set updates
caused by too many reporting events, however infrequent
reporting could delay the handover. Signalling load also in-
creases the amount of hardware and other resources being
utilised in the UTRAN. It was proved from results that with
the optimal settings no noticeable increase in resource usage
or signalling is observed.
Soft handover parameters also affect the distance from the
serving base station at which the measurement reports are sent
from the UE. Small values for event1a parameters indicated
measurement reports to be sent from the UE at just before the
half way point between two base stations and proved to be a
optimal position for the handover process to be reported and
timely to the point where chances of a dropped call is at a
minimum. Parameter values obtained from the simulation that
were implemented in a live network did show a signicant
improvement in handovers and a reduction in dropped calls.
Further work may include a propagation model which can
now be used to simulate different morphologies and to opti-
mise the soft handover parameters for various morphologies.
REFERENCES
[1] J. Kim, D. Kim, and P. Song, Design of optimum parameters for
handover initiation in wcdma, IEEE - ISBN 0-7803-7005-8/01, 2001.
[2] M. Schinnenburg, I. Forkel, and B. Haverkamp, Realization and opti-
misation of soft and softer handover in umts networks, IEEE, 2007.
[3] Radio resource control protocol for the ue-utran radio interface, 3GPP
TS 25.331 v8.1.0.
[4] Services provided by physical layer, 3GPP TS 25.302 Ver 6.2.0,
December 2004.
[5] P. Zanier, R.Guerzoni, and D. Soldani, Detection of interference, dom-
inance and coverage problems in wcdma networks, PIMRC Helsinki,
2006.
[6] J. Laiho, A. Wacker, and T. Novosad, Radio Network Planning and
Optimisation, 2nd ed.
[7] E. R. Systems, Wcdma radio network design, Training Document,
2003.
[8] J. Hammersley and D. Handscomb, Monte Carlo Methods. London:
Methuen & Co Ltd, 1964.
[9] C. Robert and G. Casella, Monte Carlo Statistical Methods. USA:
Spring-Verlag, 2004.
[10] Unknown, Monte carlo method, [Online] available at
http://www.riskglossary.com/link/monte carlo method.htm, Date
acessed: 17 March 2009.
[11] Y. Shreider, Method of Statistical Testing. London: Elsevier Publishing
Company, 1964.
194