Beruflich Dokumente
Kultur Dokumente
responsible editor:
hermann.buddendick@awe-communications.com
Hermann Buddendick AWE Communications GmbH Otto-Lilienthal-Str. 36 D-71034 Bblingen Phone: +49 70 31 71 49 7 - 16 Fax: +49 70 31 71 49 7 - 12
Issue V1.0
1 Motivation
System simulations are required for the network planning process on order to come to an cost effective investment in air interface network infrastructure. This is achieved by reducing the number of Node B (and sites) to a minimum, still fulfilling capacity and service quality constraints. System simulations help to evaluate mobile network performance in different environments with different configurations or network layouts. Performance indicators of interest are network capacity and service availability. There are mainly three approaches for the simulation of UMTS networks: dynamic simulations, semi-dynamic simulations (Monte Carlo simulator) and static simulations.
Dynamic System Simulations One way to model the network behavior is to use dynamic system simulations with quite realistic models for most aspects and effects (power control, soft handover, mobility, ...), and to simulate the time-variant behavior of the network. Detailed results can be obtained by using this method, but it is very time consuming. It is mainly used to validate or optimize small parts of the network or to perform parameter studies to tune the network settings. Static Simulation / Analytic Approach To speed up the required simulation time and to enable the simulation of larger areas the consideration of individual mobiles and discrete mobile distributions has to be given up, and the network capacity is to be predicted with analytical methods. Therefore, in a second approach (static or analytical) the mobile terminals are considered to be distributed continuously over the simulation area, i.e. for each pixel of the simulation area a kind of fractional mobile is assumed, and the complete simulation area has to be scanned only once. With this static/analytic approach the capacity of a given UMTS FDD network layout can be estimated based on the propagation conditions in a very fast way. This is very useful for a first rough network planning with a few iterations necessary to find a proper network layout that fulfills your needs. Monte-Carlo Simulations Another option for network simulations is to use Monte-Carlo (MC) simulations with less detailed models (compared to the dynamic approach explained above): time variant effects are not considered but many (independent) snapshots with random mobile distributions are evaluated. These simulations are faster than dynamic simulations but to get statistically reliable results many snapshots have to be carried out and the simulation time depends on the number of snapshots and on the number of mobile stations. Practical examples show that the size of the area that can be considered is limited to a few hundred cells.
The Monte Carlo (MC) method consists in repeating an experience many times with different randomly determined data in order to draw statistical conclusions. It can be applied for mobile networks simulation. In this case the repeated experience is called a snapshot and represents a set of Mobile Stations in the network with random position, state and parameters. The results given by a high number of snapshots are considered to be representative of all the possible states of the network. The aim of a Monte Carlo simulator is to provide an analysis tool that allows the fast and accurate evaluation of the performance of a UMTS network during the planning and optimization phases.
February 2006
2 Simulation Approach
The results given by the Monte-Carlo simulator are essentially related to two concepts: The coverage (proportion of the area in which a communication link can be successfully established) The capacity (maximum traffic supported by the network) The main features of the Monte Carlo system simulator are the following: Fast coverage and capacity prediction Mean noise rise and mean power per cell evaluation Power distribution histograms for DL and UL Determination of the maximum capacity of the network
The last two options require ray-optical propagation models to determine whether a possible receiver location, i.e. a pixel, is in Line of Sight (LOS) condition, or to compute the Channel Impulse Response (CIR). As the determination of the CIR with the propagation model is quite time consuming, this feature eliminates a part of the Monte-Carlo simulators benefits compared to full dynamic simulations (although it has to be performed only once for each fixed antenna/cell configuration and it is available via look-up tables afterwards). So in most cases the OF will be set to a constant (average) value.
February 2006
OF (value) 1 0..1 0
Typical values for the orthogonality factor are OF = 0.5 for the ITU Vehicular A profile and 0.9 for the ITU Pedestrian A channel profile. For the constant OF model an appropriate (mean) value for the expected OF has to be used.
The required Eb/No for each Mobile Station is known by link level simulation and is provided by the system supplier. It is defined as:
February 2006
Eb No
Where:
=
DL , MSi
(1)
Eb/No is the signal to noise ratio (on a net bit basis) PTXi is the required transmit power for the link (Serving sector of a Base Station BS(i) of Mobile Station i to MS i aBS(i),i is the linear attenuation for the path BS(i) to MSi G is the processing gain is the downlink orthogonality factor for the considered link (UL: = 0) Iown is the interference level from the own cell Iother is the interference level from the other cells The previously mentioned interference levels are given by the following: downlink
I own = PTX j aBS (i ),i + PCCH BS ( i ) aBS (i ),i j _ served _ by _ BS ( i ) j i I other = _ by _ u PTX u , j au ,i + PCCHu au ,i u BS ( i ) j _ served
P UL , j a j , BS (i ) TX
(3)
I other =
jother _ cell
PTXUL , j a j , BS ( i )
(5)
A PTX = B
Bi = Eb No
DL , MSi
C PTX = D
Di = Eb No PN
UL , BS ( i )
(7) (9)
Eb aBS (i ),i (1 );(1) N o MSi E aBS ( j ),i ; (2) Ai , j = b N o MS i aBS (i ),i G;(3)
(10)
(11)
with: (1) is valid if the MS j is different from the MS i and belongs to the same cell as the MS i; i.e. they have the same serving sector BS(i)=BS(j). (2) is used when the MS j belongs to another cell, i.e. BS(j)BS(i) (3) is used when i=j
February 2006
The condition checked by the simulator is the ability for a MS to communicate with its serving BS with a sufficient uplink Eb/N0 ratio assuming that the MS is transmitting with full power and the interference in the network is fixed (predefined). The point is that the interference contributions are unknown at this step of the calculations and therefore a fixed load defined by the user defined maximum load factor and a scaling factor is used to assumes a typical interference level.
February 2006
3 Examples simulation
To demonstrate the performance of the Monte Carlo simulator, some examples and case studies will now be presented. The first example illustrates the case of a bad scenario that has to be improved with the help of the Monte Carlo simulator.
February 2006
Figure 3-1: Example network layout in a dense urban area (Paris France). Extract of the result file: Parameters: Thu Dec 08 14:33:30 2005 Number of snapshots: 100 Traffic scale factor: 18.900 Average Number of MS: 25.0 Repartition of the generated MS per service: Service 384 -> 100.00 % Coverage for Service 384: Coverage -> 99.84 % Nr MS (according to traffic parameters) -> 25.00 <=> 1.19 users per cell Mean Nr MS in the coverage zone -> 24.96 <=> 1.19 users per cell Mean CPICH C/I -> -7.79 dB Mean throughput per cell -> DL: 456.41 kb/s UL: 76.22 kb/s Capacity: Mean total Nr of MS per cell: 1.19 Mean total throughput per cell: DL: 456.41 kb/s UL: 76.22 kb/s Result of the simulations: Downlink -> 18 % of the Snapshots are valid Uplink -> 100 % of the Snapshots are valid Mean power per Cell: 1999.229 mW <=> 33.01 dBm (mean over all BS of all valid snapshots) Mean Noise Rise per Cell: 1.107860 <=> 0.445 dB (average over all BS of all valid snapshots)
On the previous example, the Monte-Carlo simulator indicates that the layout and network parameters should be improved because the capacity of the network does not satisfy the requirements. You can try to decrease the CPICH level in order to minimize the interferences. Of
by AWE Communications GmbH February 2006
course the impact on coverage has to be observed. The effect of the CPICH power at the BS can be checked by making an additional simulation with a CPICH level divided by two compared to the previous simulation (now: 30dBm). A part of the result file is listed below Coverage Service 384 : Coverage -> 99.00 % Nr MS (according to traffic parameters) -> 25.00 <=> 1.19 users per cell Mean Nr MS in the coverage zone -> 24.75 <=> 1.18 users per cell Mean CPICH C/I -> -10.79dB Mean throughput per cell-> DL: 452.57 kb/s UL: 75.58 kb/s Capacity Service 384 : Mean total Nr of MS per cell: 1.19 Mean total throughput per cell: DL: 452.57 kb/s UL: 75.58 kb/s Result of the simulations: Downlink -> 65 % of the Snapshots are valid Uplink -> 100 % of the Snapshots are valid Mean power per Cell: 1219.734 [mW] <=> 30.86 dBm (mean over all BS of all valid snapshots) Mean Noise Rise per Cell: 1.103706 <=> 0.429 dB (average over all BS of all valid snapshots) From the results the influence of the CPICH power level can be clearly seen: 65% of valid snapshots for DL instead of 18% with the first simulation) impact on coverage is negligible (99% instead of 99.84%)
February 2006
You can notice that the previous change has the following consequences on the simulation results: 78% of downlink valid snapshots instead of 65% whereas the traffic has been increased similar coverage, mean power per cell, and noise rise values If the orthogonality factor was set to 1.0 (perfect orthogonality) 97% of the snapshots would be valid for DL under the same conditions. This shows the great impact of the downlink orthogonality factor. The low capacity of this network is due to the simple design of the layout (orientation of the antennas, location of the base stations).
Figure 3-2: Percentage of successful snapshots for DL (left) and average DL power per cell (right).
Figure 3-3: Example for the noise rise output (with increasing traffic load)
February 2006
10
Figure 3-4: DL link power distribution (left) and UE power distribution (right)
4 Propagation Prediction
The 3G Monte-Carlo simulation approach is based on the predicted path loss matrices for all base stations. So after having defined a network scenario, first the propagation prediction has to be performed, and in a second step the network prediction (i.e. 3g Monte-Carlo simulation) can be run. In small scenarios the propagation may be predicted for each transmitter for the complete simulation area. Considering large areas a prediction radius can be defined for each transmitter. Note that the definition of the prediction radius has to ensure sufficient overlap of neighboring cells. This overlap is not only required for best server determination, but also for interference considerations, i.e. the overlap should be designed to account for the interference of approx. two rings of interfering cells at each pixel. That means that the prediction radius should be approx. 2-3 times the inter-site distance. Many different propagation models are available within ProMan. Depending on the scenario, the environment, and the available database the best suitable propagation model must be chosen. The Monte Carlo simulator is able to handle the propagation results generated by all available models. Further information about the propagation models can be found on the following web site: http://www.awe-communications.com/Propagation/index.htm Furthermore it should be mentioned that additional information about the propagation models is also available in form of application notes. See the following web site: http://www.awe-communications.com/ApplicationNotes/index.html
5 Further Information
For further information you are invited to visit AWE Communications website
http://www.awe-communications.com http://www.awe-communications.com/Network/3G/SimulatorsOverview/MonteCarlo.html
or to send an e-mail to the responsible editor of this document
net@awe-communications.com
February 2006