Internal
UMTS Capacity Planning ISSUE 4.0
www.huawei.com
RNP Staff Training Dept.
HUAWEI TECHNOLOGIES CO., LTD.
Huawei Confidential
Foreword
WCDMA is a self-interference system
WCDMA system system capacity is closely related to coverage
WCDMA network network capacity has the “soft capacity” feature
The capacity planning of the WCDMA network network is performed under certain traffic models
Page 3
Foreword
WCDMA is a self-interference system
WCDMA system system capacity is closely related to coverage
WCDMA network network capacity has the “soft capacity” feature
The capacity planning of the WCDMA network network is performed under certain traffic models
Page 3
Objectives After this course, you will：
Understand the factors that restrict the WCDMA network capacity
Understand the methods and procedures of estimating multi-service capacity
Understand the key technologies for
enhancing network capacity
Page 4
Contents 1 Traffic Model 2 Uplink capacity analysis 3 Downlink capacity analysis 4 Multi-service capacity estimation 5 Network estimation procedure 6 Capacity enhancement technologies
Page 5
Contents 1 Traffic Model 1.1 Overview of traffic model 1.2 CS traffic model 1.3 PS traffic model
Page 6
Service Overview
The WCDMA system supports multiple services
Variable-rate services (e.g. AMR voice)
Combined services (e.g. CS & PS)
High-speed data packet services (384k service)
Asymmetrical
services (e.g. stream service )
Large-capacity and flexible service bearing
Page 7
QoS Type Service c R a e t e a g l o t i m r y e c N a o t e n g r o e r a y l t i m e
characters
examples
It is necessary to maintain the time relationship Conversationa between the information entities in the stream. l Small time delay tolerance, requiring data rate symmetry
Voice service, videophone
Streaming
Typically unidirectional services, high requirements Streaming on error tolerance, high requirements on data rate multimedia
Interactive
Request-response mode, data integrity must be maintained. High requirements on error tolerance, low requirements on time delay tolerance
Background
Data integrity should be maintained. Small delay restriction, requiring correct transmission
Page 8
Web page browse, network game
Background download of Email
Objectives of Setting Up Traffic Model
In order to determine the system configuration, we need to determine the capacity of the air interface first
In the data service, different transmission model will generate different system capacities
We need to set up an expected data transmission model of the customer so that we can plan the network properly
In order to set up a right model, the operator should provide some statistic data as reference
Page 9
Traffic Model
Traffic model is a means of researching the capacity features of each service type and the QoS expected by the users who are using the service from perspective of data transmission
In the data application, the user behaviour research mainly forecasts the service types available from the 3G, the number
of users of each service type, frequency of using the service, and the distribution of users in different regions
Page 10
The Contents of Traffic Model
Service pattern Traffic Model Results User behaviour
Capacity planning Network configuration Coverage planning
Page 11
Typical Service Features Description
Typical service features include the following feature parameters:
User type (indoor ,outdoor, vehicle)
User’s average moving speed
Service Type
Uplink and downlink service rates
Spreading factor
Time delay requirements of the service
QoS requirements of the service
Page 12
Contents 1 Traffic Model 1.1 Overview of traffic model 1.2 CS traffic model 1.3 PS traffic model
Page 13
CS Traffic Model
Voice service is a typical CS services.
Key parameters of the model
Penetration rate
BHCA Mean busy-hour call attempts
Mean call duration (s)
Activation
factor
Mean rate of service (kbps)
Page 14
CS Traffic Model Parameters
Mean busy-hour traffic (Erlang) per user = BHCA * mean call duration /3600
Mean busy hour throughput per user (kbit) (G) = BHCA * mean call duration * activation factor * mean rate
Mean busy hour throughput per user (bps) (H) = mean busy hour throughput per user * 1000/3600
Page 15
Contents 1 Traffic Model 1.1 Overview of traffic model 1.2 CS traffic model 1.3 PS traffic model
Page 16
PS Traffic Model
The most frequently used model is the packet service session process model described in ETSI UMTS30.03
Page 17
PS Traffic Model Session
Packet Call
Packet Call
Downloading
Active
Downloading
Dormant
Dormant
Active
Packet Call
Data Burst
Data Burst
Active
Data Burst
Dormant
Page 18
PS Traffic Model Parameters
Packet Call Num/Session Packet Num/Packet Call Packet Size (bytes) Reading Time (sec) Typical Bear Rate (kbps) BLER
Page 19
Parameter Determining
The basic parameters in the traffic model are determined in the following ways:
Obtain numerous basic parameter sample data from the existing network
Obtain the probability distribution of the parameters through processing of the sample data
Take the distribution most proximate to the standard probability as the corresponding parameter distribution through comparison with the standard distribution function
Page 20
PS Traffic Model Parameters
Typical Bearer Rate (kbps)
Bearer rate is variable in the actual transmission process
BLER
In the PS service, when calculating the data transmission time, the retransmission caused by erroneous blocks should be considered. Suppose the data volume of service source is N, the air interface block error rate is BLER, the total required data volume to be transmitted via the air interface is:
N N * BLER N * BLER 2 N * BLER 3 N * BLER n
1 1 BLER
Page 21
* N
PS User Behaviour Parameters
Penetrating Rate
User behavior
BHSA User Distribution (High, Medium, Low end)
Page 22
PS User Behaviour Parameters
Penetration Rate
BHSA
The percentage of the users that activates this service to all the users registered in the network. The times of single-user busy hour sessions of this service
User Distribution (High, Medium, Low end)
The users are divided into high-end, mid-end and low-end users. Different operators and different application situations will have different user distributions
Page 23
PS Traffic Model Parameters
Session traffic volume （Byte）： Average traffic of single session of the service
SessionTra fficVolume ( PacketSize) * ( PacketNum/ PacketCall ) * ( PacketCall Num / Session)
Data transmission time (s) ： The time in a single session of service for purpose of transmitting data.
DataTransmissionTime ( s )
1 SessionTra fficVolume* 8 / 1000 * 1 BLER TypicalRat e
Holding Time（s）： Average duration of a single session
of service HoldingTime ( PackketCal lNum / Session 1) * Re adingTime DataTransmissionTime( s) Page 24
PS Traffic Model Parameters
Active factor:
The weight of the time of service full-rate transmission among the duration of a single session.
ActiveFact or
DataTransmissionTime
HoldingTime
Busy hour throughput per user (Kb):
BusyHourThroughput / user
BHSA * SessionTrafficVolume* 8 / 1000
PS throughput equivalent Erlang formula (Erlang)
Data _ Erlang
( Percentage OfDiffrent User Penetratin gRate
BusyHourThroughputUnderTypical Applicatio nEviroment ) TypicalBea redRate 3600 ActiveFact or
Page 25
Contents 1 Traffic Model 2 Uplink capacity analysis 3 Downlink capacity analysis 4 Multi-service capacity estimation 5 Network estimation procedure 6 Capacity enhancement technologies
Page 26
Basic Principles
In the WCDMA system, all the cells share the same frequency, which is beneficial to improve the system capacity. However, co-frequency multiplexing causes interference between users. This multi-access interference restricts the capacity
The radio system capacity is decided by uplink and downlink. When planning the capacity, we must analyze from both uplink and downlink perspectives
Page 27
Uplink Interference Composition
I TOT I own I other P N
I own
：Interference from the users of
this cell
I other ：Interference from users of
adjacent cell P N ：Noise floor of the receiver
Page 28
Uplink Interference Composition
Receiver noise floor PN P N 10 log( K * T *W ) NF – K：Boltzmann constant, 1.38×10
23
J / K
– T：Kelvin temperature, normal temperature: 290 K – W：Signal bandwidth, WCDMA signal bandwidth 3.84MHz
– 10lg(KTW) = -108dBm/3.84MHz NF P
N
= 3dB (typical value of macro cell BTS)
10log( K * T *W ) NF 105dBm / 3.84 MHz
Page 29
Uplink Interference Composition
I own：Interference from users of this cell
Interference that every user must overcome: I total
P j
P j is the receiving power of the user j , V j is active factor P j W 1 Under the ideal power control : Eb / No j I TOT P j R j v j I TOT 1 W 1 j 1 Eb / No j R j v j The interference from users of this cell is the sum of power of all the users arriving at the receiver
Hence, P : P j
N
I o wn
P j 1
Page 30
Uplink Interference Composition
I other : Interference from users of adjacent cell
The interference from users of adjacent cell is difficult to analyze theoretically, because it is related to user distribution, cell layout, and antenna direction diagram.
Adjacent When
cell interference factor ：i
I other I own
the users are distributed evenly
- For omni cell, the typical value of adjacent cell interference factor is 0.55
- For the 3-sector directional cell, the typical value of adjacent cell interference factor is 0.65
Page 31
Uplink Interference Analysis I TOT I own I other P N N
I TOT 1
1 i 1
Define
1
Eb / No j
W
1
P N
R j v j
1
L j 1
1
Eb / No j
W 1
R j v j N
Then
I TOT I TOT 1 i L j P N 1
Page 32
Uplink Interference Analysis I TOT P N
Obtain
1 1 1 i
N
L
j
1
Suppose that: All
the users are 12.2 kbps voice
users, the demodulation threshold Eb/No = 5dB
Voice activation factor vj = 0.67
Adjacent
cell
– interference factor – i = 0.55
Page 33
Uplink Load Factor
Define the uplink load factor UL
N
N
1
1
1
1 i L j 1 i
1
1
EbvsNo j
W 1
R j v j
When the load factor is 1, I TOT is infinite, and the corresponding capacity is called “threshold capacity”. Under the above assumption, the threshold capacity is approx 96 users.
Page 34
Load Factor and Interference
According to the above mentioned relationship, the noise will rise: NoiseRise
I TOT P N
1 1 1 i
N
L j
1
50% Load — 3dB 60% Load — 4dB 75% Load — 6dB
Page 35
1 1 UL
Limitation of the Current Method
The above mentioned theoretic analysis uses the following simplifying explicitly or implicitly:
No consideration of the influence of soft handover – The users in the soft handover state generates the interference which is slightly less than that generated by ordinary users.
No consideration of the influence of AMRC and hybrid service – AMRC reduces the voice service rate of some users, and makes them generate less interference, and make the system support more users. (But call quality of such users will be deteriorated) – Different services have different data rates and demodulation thresholds. So, we should use the previous methods for analysis, but it will complicate the calculation process. – Since the time-variable feature of the mobile transmission environment, the demodulation threshold even for the same service is time-variable.
Page 36
Limitation of the Current Method
Ideal power control assumption – The power control commands of the actual system have certain error codes so that the power control process is not ideal, and reduces the system capacity
Assume
that the users are distributed evenly, and the adjacent cell interference is constant
Considering the above factors, the system simulation is a more accurate method: – Static simulation: Monte_Carlo method – Dynamic simulation
Page 37
Contents 1 Traffic Model 2 Uplink capacity analysis 3 Downlink capacity analysis 4 Multi-service capacity estimation 5 Network estimation procedure 6 Capacity enhancement technologies
Page 38
Downlink Interference Composition
I TOT I own I other P N
I own : Interference from the users of this cell
I other : Interference from the users of adjacent cell
P N : Noise floor of the receiver
Page 39
Downlink Interference Composition
Receiver noise floor PN
P N 10 log( K *T *W ) NF 23 – K Boltzmann constant, = 1.38* 10 J / K
– T Kelvin temperature, normal temperature 290 K – W Signal bandwidth, WCDMA signal bandwidth 3.84MHz – NF: Receiver noise figure 10lg(KTW) NF
= -108dBm/3.84MHz
= 7dB （ UE typical value ）
P N 10log( K * T *W ) NF 101dBm / 3.84 MHz
Page 40
Downlink Interference Composition
I own:Interference from downlink DCH of this cell
The downlink users are identified with the mutually orthogonal OVSF codes. In the static propagation conditions without multipath, no mutual interference exists.
In case of multi-path propagation, certain energy will be detected by the RAKE receiver, and become interference signals. We define the orthogonal factor α to describe this phenomenon. P I own j 1 j T PL j
– In the formula, PT is a total transmitting power of BTS, which includes the dedicated channel transmitting power and the common channel transmitting power
P T P CCH
N
P
P
UE1
T
j
1
UE2
UEj Page 41
PL j
Downlink Interference Composition
I other : Interference from the downlink DCH of adjacent cell
The transmitting signal of the adjacent cell BTS will cause interference to the users in the current cell. Since the scrambling codes of users are different, such interference is non-orthogonal
Assume
the service is distributed evenly, the transmitting power of all BTSs will be equal. k,j In the system, there are K adjacent cell BTSs, where path loss from the number k BTS to the user j is PLk,j. Hence we obtain:
PT Cell 1
K
I other j P T
PL1, j
PL2, j
UE1
PT
1
1 PLk , j
Cell 2
UE2
UEj
PT PLk, j
Cell k Page 42
Downlink Interference Composition I TOT I own I other P N
1 j
P T PL j
K
P T 1
1 PLk , j
P N
Suppose the power control is desired, we obtain
P j
EbvsNo j
PL j W 1
I TOT j
R j v j
Then
P j EbvsNo j
R j W
v j I TOT j PL j
Page 43
Downlink Interference Composition P T P CCH
Because
N
P j
1
Then R j P T P CCH EbvsNo j v j I TOT j PL j W 1 N K R j P T 1 P CCH EbvsNo j v j PL j 1 j P T P N W PL j 1 1 PLk , j N
K R j PL j P CCH EbvsNo j v j 1 j P T P T P N PL j W 1 1 PLk , j N
Page 44
Downlink Interference Composition Resolve PT to obtain
R j v j PL j P CCH P N EbvsNo j W 1 P T N R j 1 1 j i j EbvsNo j v j W 1 N
where i j is the adjacent cell interference factor of the user, defined as:
i j
K
PL j
PL 1
k , j
Page 45
Downlink Interference Analysis
According to the above analysis, we can define the downlink load factor:
R j 1 j i j EbvsNo j v j W 1 N
DL
When the downlink load factor is 100%, the transmitting power of the BTS is infinite, and the corresponding capacity is called “threshold capacity”.
As different from the theoretic calculation of uplink capacity, a j and i j in the downlink capacity formula are variable related to user position.
Namely, the downlink capacity is related to the spatial distribution of the users, and can only be determined through system simulation.
Page 46
Simulation Result
Page 47
Simulation Result Analysis
When the transmitting power of the BTS is 43dBm (20W), the supported maximum number of users is approx 114.
In order to ensure system stability, we do not allow the mean transmitting power of the BTS to be more than 80% of the maximum transmitting power, namely, 42dBm. This way, the supported number of users is 111.
Page 48
Contents 1 Traffic Model 2 Uplink capacity analysis 3 Downlink capacity analysis 4 Multi-service capacity estimation 5 Network estimation procedure 6 Capacity enhancement technologies
Page 49
Contents 4 Multi-service capacity estimation 4.1 Network capacity restriction factors 4.2 Typical capacity design methods
Page 50
Capacity Restriction Factors
The WCDMA network capacity restriction factors in the radio network part include the following:
Uplink interference
Downlink power
Downlink channel code resources (OVSF)
Channel element (CE)
Iub interface transmission resources
Page 51
Downlink Transmit Power
The downlink transmit power has two parts: one part is used for common channel, and the
other part for dedicated (traffic) channel.
P T P CC H
N
P j
1 The transmit power is allocated by the cell to
each user varies with service demodulation threshold, propagation path loss and the interference received by the user
The downlink transmit power of the cell is
shared by all the users in the cell
We generally use the simulation method to analyze the downlink interference.
Page 52
Downlink Channel Code Resources
The WCDMA network use the codes whose SF is 4~512. The smaller the SF is, the higher the supported data rate will be.
In the code tree, the allocable codes should meet the follow ing conditions:
No codes on the path from this code to the root node of code tree are allocated
No codes in the sub-tree whose root node is this code are allocated
Try to reserve the code words whose SF is small, so as to improve the utilization efficiency
C4,0 C2,0 1
1
1
1
C1,0
1
C4,1 1
1 -1 -1
1
C4,2 C2,1
1 -1
1
SF = 2
Page 53
1 -1 C4,3
1 -1
SF = 1
1
-1
-1
SF = 4
1
Downlink Channel Code Resources Following is an example of code resources allocation SF
4
8
16
32
64
128
256
512
┏━●C(256,0):PCPICH 2 0 ┫ ┗━●C(256,1):PCCPCH 3
┏ ┃ ┫ ┃ ┗ 1
┏ 0 ┃ ┃ ┗━○1
┏ 0 ┃ ┫ ┗━○1
┏ 0 ┃ ┃ ┃ ┫ ┗━○1
┏ 0 ┃ ┫ ┃ ┗ 1
┏ 0 ┃ ┏━●C(256,2): AICH 6 ┃ ┫ ┃ ┗━●C(256,3): PICH 10 ┫ ┗━●C(64,1):SCCPCH 8 ┏━●C(64,2):SCCPCH 9 ┫ ┗━○3
┏━○2 ┃ ┗━○3
Page 54
Channel Element (CE)
The Channel element the quantitative data that measures the resources logically occupied for service processing
The resource occupied by the service processing is mainly related to the spreading factor of this service. The smaller the SF is, the greater the data traffic will be, and more resources will be occupied
The SF of typical services are: AMR12.2kbps
SF=128
CS64kbps
SF=32
PS64kbps
SF=32
PS144kbps
SF=16
PS384kbps
SF=8 Page 55
Channel element (CE)
If we define the resources required for processing AMR 12.2kbps services as a channel processing unit, the number of channel processing units occupied by other services is: Uplink
AMR12.2kbps
Downlink
1
1
CS64kbps
3
2
PS64kbps
3
2
PS128kbps
5
4
PS144kbps
5
4
PS384kbps
10
8
Page 56
Iub Interface Capacity
The contents transmitted on the Iub interface include:
The user data encapsulated in the AAL2 format (common channel and dedicated channel)
Signalling data encapsulated in the AAL5 format
BTS operation & maintenance data
Page 57
Iub Interface Capacity
Factors to be considered when estimating the interface capacity:
Frame coding efficiency. Through segmentation and encapsulation of the application data at each layer, the data quantity at the bottom layer will be increased to different extents compared with the application data at the upper layers
Traffic. More users will generate more data traffic
Maintenance efficiency. Certain bandwidth is required in the background maintenance for BTS data transmission
Page 58
Contents 4 Multi-service capacity estimation 4.1 Network capacity restriction factors 4.2 Typical capacity design methods
Page 59
Erlang-B Formula (I)
Erlang-B formula is used for estimating the peak traffic that meets certain call loss rate when the average traffic (Erlang) is given
Erlang-B formula is only used for
Circuit switched services
Single service
The WCDMA system provides CS and PS domain multi-services
Page 60
Erlang-B Formula (II)
The prerequisite of the Erlang-B is the requests of resources take on a Poisson distribution, namely, its variance is equal to its mean value
If, when a service establishes a link, the service requires the resources which are more than the unit resources, the resource request is no longer equal to its mean value, and the Erlang-B formula is not applicable in this case
Comparison of multi-service capacity estimation methods :
Post Erlang-B
Equivalent Erlangs
Campbell’s Theorem
Page 61
Post Erlang-B（一）
By summing up the capacities required for different services, we obtain the capacities required for the combined services
No consideration of the resource efficiency of different services
X Erl data
capacity Y Erl voice
Page 62
Post Erlang-B (II)
Consider that two services share resources
Service 1: 1 unit resource/connection.12 Erlang
Service 2: 3 unit resources/connection.6 Erlang
Calculate capacity required for each service
Service 1: 12 Erlangs require 19 connections (19 unit resources), meeting the 2% blocking rate
Service 2: 6 Erlangs require 12 connections (equivalent to the 36 unit resources of service 1), meeting the 2% blocking rate
Total 55 unit resources
Page 63
Post Erlang-B (III)
Consider that two services use the same resources
Service 1: 1 unit resource/connection.12 Erlang
Service 2: 1 unit resource/connection.6 Erlang
Calculate capacity required for each service
Service 1: 12 Erlangs require 19 connections, meeting the 2% blocking rate
Service 2: 6 Erlangs require 12 connections, meeting the 2% blocking rate
Total 31 unit resources
However, the reasonable results should be: 18 Erlangs require 26 connections for meeting the 2% blocking rate 2
Post Erlang-B overestimates the capacity requirements! Page 64
Equivalent Erlangs (I)
By converting the bandwidth from one service to another service, combine different services and then calculate the required capacity
Selecting different services as the measurement benchmark will lead to different capacity requirements
Page 65
Equivalent Erlangs (II)
X Erl Erl data
capacity Y Erl Erl voice
Voice service to data service
Page 66
Equivalent Erlangs (III)
X Erl Erl data
capacity Y Erl Erl voice
data service to voice service
Page 67
Equivalent Erlangs (II)
Consider that two services share resources
Service 1: 1 unit resource/connection.12 Erlang
Service 2: 3 unit resources/connection.6 Erlang
If using service 1 as measurement benchmark, the two services are equivalent to 30 Erlangs in total
30 Erlangs require 39 connections (39 unit resources), meeting the 2% blocking rate
If using service 2 as measurement benchmark, the two services are equivalent to 10 Erlangs in total
10 Erlangs require 17 connections (equivalent to 51 unit resources of service
1), meeting the 2% blocking rate The predication results are not unique! Page 68
Campbell’s Theorem (I)
X Erl data Erl dummy service traffic Z
Y Erl voice
Dummy service channel Basic service capacity
Page 69
Campbell’s Theorem (II)
The Campbell theorem sets up a combined distribution Capacity
(C i ai )
OfferedTra ffic
c
c Erlangs a v
Erlangs ai
c
2
i
i
Here:
i
ai is service amplitude, namely, the channel resources required for a single link of the service.
is the mean value, v is the variance .
Page 70
Campbell’s Theorem (III)
Consider that two services share resources
Service 1: 1 unit resource/connection.12 Erlang
Service 2: 3 unit resources/connection.6 Erlang
The system mean value is
v
Erlangs ai 112 3 6 30
The system variance is
c v
2
Erlangs ai 12 12 6 32 66
66 30
2.2
The capacity factor c is 1 Page 71
Campbell’s Theorem (III)
Combined traffic is:
OfferedTra ffic
c
30 2.2
13.63
The number of connections for meeting the blocking rate of 2% is 2
For the target services that meet the same GoS, the capacity
required is (calculated on the basis of the unit resource of service 1)
Goal is service 1: C1 = (2.2×21) +1 =47
Goal is service 2: C2 = (2.2×21) +3 =49
For different services, the same GoS requires different capacities. For the given capacity, the GoS of different servic es will differ sligh tly.
Page 72
The comparison of the different capacity method
Post Erlang-B
Service 1 (1 unit resource/connection, 12Erl) and service 2 (3 unit resources / connection, 6Erl), requiring 55 unit resources in total
Equivalent Erlangs
Calculated according to benchmark of service 1 (1 unit resource/connection, 12Erl), a total of 39 unit resources are required
Calculated according to benchmark of service 2 (3 unit resources/connection, 6Erl), a total of 51 unit resources are required
Campbell’s Theorem
In the same conditions, 47~49 unit resources are required in total.
Page 73
Summary
This chapter deals with the three methods of estimating the multi-service capacity
The detailed process of using the Campbell theorem to calculate the capacity is described
Page 74
Contents 1 Traffic Model 2 Uplink capacity analysis 3 Downlink capacity analysis 4 Multi-service capacity estimation 5 Network estimation procedure 6 Capacity enhancement technologies
Page 75
Network Estimation Procedure User density
Assumption of cell load and carrier number
Cell radius
Cell area
Number of user per cell
Compare
Balance between Yes capacity and coverage dimension?
Uplink capacity dimension , downlink capacity dimension
No
Service message
Adjustment of cell load and carrier number
Page 76
over
Contents 1 Traffic Model 2 Uplink capacity analysis 3 Downlink capacity analysis 4 Multi-service capacity estimation 5 Network estimation procedure 6 Capacity enhancement technologies
Page 77
Transmission Diversity －TxDiv
Transmission diversity can enhance the downlink capacity and coverage
Conclusion of capacity enhancement of transmission diversity
STTD mode: Capacity increase of 17 ~ 24%
TxAA(1) mode: Capacity increase of 16 ~ 23%
TxAA(2) mode: Capacity increase of 31 ~ 37%
Page 78
Sectorization
In the dense urban areas and the normal urban areas with high traffic, increasing sectors of the BTS is a method of improving the capacity
6-sectors BTS generally use the antenna whose horizontal lobe is 33º
The capacity of a 6-sector BTS is 1.67 times that of a 3sector BTS
The capacity of a 3-sector BTS is 2.77 times that of a omniBTS
Page 79
