Professional Documents
Culture Documents
UMTS PS Service Analysis Guide
UMTS PS Service Analysis Guide
Guide
UMTS PS Service Analysis Guide
1 Introduction
Research on PS data transmission performance belongs to the end-to-end (UE <-> data
service server) category. Every constitutional part of the network may affect the data
transmission performance. When doing data transmission performance optimization,
mostly we focus on RAN side (RNC-NodeB-UE) and other system parts (CN, FTP server
and internet etc.). We usually use exclusive method to judge in which NE the service is
interrupted and in which NE the packet is lost, which leads to the rate decrease. Capture
packets at the interfaces of NEs or use NE tracing function to specify the NE, and then
locate the problem based on NE features or experiences.
As shown in the above figure, data transmission optimization involves multiple NEs.
Therefore collecting complete information at different interfaces is important for
2
UMTS PS Service Analysis Guide
optimization and problem location. Data collection mainly includes OMC traffic statistics
collection, DT data collection, background tracking of NEs and end-to-end data capture
on the user plane.
Currently the DT/CQT is usually carried out using CNT software (connected to the UE,
Scanner or GPS). Thus information such as pilot coverage, signaling flow, DL BLER and
UE transmission power can be obtained. Combining with measurement tracking of users
in OMC-R, UL BLER, DL code field transmission power, UL receiving power and
signaling flow at RNC side can be acquired. Use CNA processing software to make
comprehensive analysis based on data collected by CNT, QXDM, and tracking record in
OMC-R operation and maintenance console.
3(62)
UMTS PS Service Analysis Guide
Traffic statistics, DT/CQT and user complaints help to find faults. Apart from traffic
statistics and DT/CQT, other tools and their functions can be combined to analyze and
locate faults. Functions such as connection performance measurement and cell
performance measurement of RNC background tracking, alarm query, and status query
of NEs, and tools such as FlashGet, DU Meter can all help to locate faults. In PS service
test, to reduce the effect brought by TCP receive window of the application layer, multi-
thread downloading tool such as FlashGet is recommended and the thread count is
generally set to 5. For UL data transmission, multiple FTP processes can be used for
uploading.
Data
Data
Viewing/
Data Collection Function Remark
Analyzing
Tool
Tool
DT/CQT QXDM+UE/
QCAT/CNA
testing data CNT+UE
Connectivity Combining DT/CQT
OMCR/ measuring data and
measurement
, cell analysis RNC tracing data,
make analysis from the Refer to online
performance 1.22.26/sign
perspectives of call help on RNC
OMCR alingCalDat
measurement process and coverage O&M
eRa
and RNC etc. console.
e/runsignal
signaling
UESigStat
tracing
OMCR(OM OMCR(OM View the alarms and
Alarms
CB) CB) find the abnormal NE.
Upload or Multi-thread PS service
FlashGet
download null downloading to get the test assisted
etc.
software stable throughput rate. tool.
Rate DU Meter null Make real-time PS service
monitoring monitoring of the test assisted
software throughput rate of the tool.
application layer, and
record total
throughput, average
throughput rate and
peak throughput rate
(the result can only be
recorded through
4
UMTS PS Service Analysis Guide
Data
Data
Viewing/
Data Collection Function Remark
Analyzing
Tool
Tool
screenshots).
When problems occur, make tracing of relevant equipment to make the problem recur,
and get the relevant information:
3. Take low rate problem for example, it is typical and hard to be located. The problem
usually occurs when the PC accesses the network through the UE to perform the
FTP download or VOD services etc. The procedures to locate the problem are
shown below:
i. Start user tracing in the SGSN and GGSN and monitor and eliminate obvious
faults in signaling plane and user plane. Run packet capture tool (for example,
Ethereal) in the PC (connected to the UE), Gi interface or the server.
ii. After the service is stopped, analyze the captured PPP packets at Gi interface
and UE side. Monitor the UL and DL data flow and confirm whether there is
packets loss, retransmission, or out-of-order etc, which affects the rate. By
data flow comparison at different interfaces, judge which part brings the
problem.
5(62)
UMTS PS Service Analysis Guide
The above procedures are targeting at the low rate problems; the procedures to locate
data transmission problem are similar.
Using the above exclusive method, the NE can be located. As described above, PS
networking (not including access network) mainly involves the following equipment: the
SGSN, GGSN, switch, router, and firewall. Use the following method to locate the NE.
Of the equipment listed above, the SGSN is the relatively complicated equipment with
many interfaces, and the data transmission function of the user plane has its own
features. Generally, lu interface between the RNC and SGSN is the ATM port, and the
interface between the Gn and Gp is IP Ethernet port (FE or GE). The data packets of the
UL user plane is transmitted as below: the RNC sends the packets of the user plane to
the forwarding engine of the specified UHPU via the ATM (the IPOA to the RNC has
6
UMTS PS Service Analysis Guide
been created in the forwarding engine). If the IP address of this forwarding engine is the
same with that of the destination of the data packet, the data packet will be processed
directly; otherwise, the forwarding engine will search for the route based on the IP
address of the data packet and send it to the destination forwarding engine (PDP context
is created on this forwarding engine).
The procedures to handle the PS service problems relevant to the SGSN are:
2. Confirm that PDP has been correctly created on the user plane.
If it is suspected that the packets are lost at GGSN, check it based on packet loss
statistics of the NP layer. The NP problem is caused by the product design; therefore
contact the product engineer to solve the problem.
Generally the current data communication equipment supports making packet loss
statistics. Execute commands to query the interface status, check the number of packets
received and sent to see whether there is any packet lost. It should be noted that the
packet loss of the data communication equipment does not indicate the equipment is
abnormal. For example, when filtering rules are configured in the firewall, unreasonable
packets and attacking packets will be discarded. The configuration of the equipment may
lead to the loss of the correct packets. There are many data communication equipment,
please refer to their product command manual to see the command to view the packet
loss statistics.
3 Data Analysis
The data transmission problems of WCDMA PS service falls into three categories in the
aspect of phenomenon: access failure (or dialing connection setup failure), successful
access without the traffic flow, and data transmission with low rate or great fluctuation.
Different problems require different analysis and handling processes.
7(62)
UMTS PS Service Analysis Guide
In the optimization test, the method of PC+UE is most commonly used. During the
DT/CQT test, the drive test tool software CNT runs on a personal computer, which is
usually a portable one. And this method is shortly called as CNT+UE. If the UE fails to
launch the PS service directly, the engineer can use the method of CNT+UE for further
verification to acquire more information. Therefore, the following problem analysis
focuses on CNT+UE.
If the UE launches the PS service directly, the analysis flow of access failure is as
follows:
8
UMTS PS Service Analysis Guide
The process of launching the PS service directly on the UE and the process of launching
the PS service in the method of PC+UE are the same in the signaling flow, but they have
different APNs (name of the access point), and setting approaches of the service visiting
address. If the UE fails to launch the PS service directly, the engineer can find the cause
of the problem by taking the following steps.
1. Make verification in the method of PC+UE. If the PS service is normal, then the
system works normally. The engineer should check and modify the settings of APN,
service visiting address, Proxy and password of the UE.
2. If the access failure of PS service launched in the method of PC+UE happens, the
engineer should analyze and locate the problem according to section 3.1.1.2.
If the UE is used as the Modem of the PC, the analysis flow of the access failure is
shown as follows:
9(62)
UMTS PS Service Analysis Guide
10
UMTS PS Service Analysis Guide
The engineer should check the Port Configuration in the CNT, and make sure
the port No. corresponds to the Com port and the Modem port in the Windows
Hardware Manager.
Wrong driver installation and abnormal termination of the drive test tool during
the test may cause the status of the port mapped by the UE to become
abnormal in the Windows Hardware Manager. For example, the yellow
exclamation mark appears.
Solution: Reinstall the driver, insert and extract the data cable or data card of
the UE.
iii. The port is not closed after abnormal termination of the software
11(62)
UMTS PS Service Analysis Guide
During the test with the drive test tool such as the CNT, the software
terminates abnormally and may not close the corresponding port.
Solution: The engineer can try to restart the CNT software. If the problem still
exists, restart the personal computer.
It is necessary to reinstall the driver, and this problem usually happens when
the personal computer is connected with the UE for the first time.
In this condition, the failure is usually caused by the signaling flow, and the analysis
should focus on the signaling flow during the access process. The analysis flow is
shown as follows:
12
UMTS PS Service Analysis Guide
Trace the NAS and RRC signaling in CNT or trace the single-user signaling in the
OMCR, and analyze the problems according to the standard data service signaling
flow. The problems may be:
[Description]
[Analysis]
If the engineer finds that the UE does not send the RRC Connection Request
message from the RRC Message traced by the CNT, the possible causes may
be:
a) The Modem port is not selected during port configuration in the CNT.
If the UE does not receive any responses, or receives the RRC Connection
Reject message after sending the RRC Connection Request message, the
possible causes are:
a) Poor coverage
For the users, PDP activation can be successful, and the actual
maximum rate is the maximum rate the UE can support. However, if
even the minimum guaranteed bit rate required by the QoS setting in the
UEs PDP activation request is beyond the UEs capability, although the
network has negotiated a lower rate to accept the PDP activation request
of the UE, the UE will send a request of deactivating PDP when it finds
13(62)
UMTS PS Service Analysis Guide
out that the rate negotiated by the network in the PDP activation accept
request is lower than the minimum guaranteed bit rate, and finally, the
PDP activation cannot be completed.
[Description]
[Analysis]
Some UEs can be set to support CS, PS, or CS+PS. If the UE is set to
support CS, the PS service cannot be set up. And then it is necessary to
check the UE setting, and modify it to PS or CS+PS.
Viewing from the signaling flow, the engineer finds that UE receives the
Attach Reject message from the network side after sending the Attach
Reject message. The CN side engineer should check whether the PS
service is supported in USIM card subscription.
[Description]
The signaling flow between NAS signaling Authentication AND Ciphering REQ
and RRC signaling Security Mode Complete is abnormal.
[Analysis]
The CN side engineer should check whether the authentication switch in the
PS domain of the core network is opened, and whether the cryptographic
algorithm and integrity protection algorithm of CS domain, PS domain and
RNC of the core network are consistent.
[Description]
The UE sends the Activate PDP Context Request message, but receives the
Activate PDP Context Reject message.
14
UMTS PS Service Analysis Guide
The problems fall into two categories. One is incorrect setting of APN and rate
limitation at the UE side and the other is the core network problem.
If there are other cause values, and the APN and rate limitation setting at
the UE side is correct, the problem may lie in the core network, for
example, some interfaces of the core network are not through. The
engineer can work with other engineers in the PS domain of the core
network to locate the problem. What is more, if the PS service is
debugged for the first time, the problem may be caused by the case that
the subscription APN in the HLR and that used in the GGSN are
different. The engineer can ask the personnel in charge of the PS
domain of the core network for confirmation.
v. RB setup failure
[Description]
After the Activate PDP Context Request message is sent, the RB setup
message such as the Radio Bearer Setup message is not received, but the
release message is received instead.
vi. Others
15(62)
UMTS PS Service Analysis Guide
The PS service is set up successfully on the signaling plane, but there is no traffic flow
on the user plane. The problem analysis flow is shown as follows:
The successful setup of connection means there is traffic flow on the signaling plane.
But there is no traffic flow on the user plane, which may be caused by the TRB reset at
the RAN side. Especially for HSDPA, the service is borne on the HS-PDSCH, and the
signaling is borne on the associated DCH. When the HS-PDSCH does not have enough
power, there is traffic flow on the signaling plane, but no traffic flow on the user plane.
The following analysis is made from the two aspects of the PS service on the DCH and
the PS service on the E-DCH.
1. DCH Bearer
16
UMTS PS Service Analysis Guide
Trace the service cell pilot RSCP and Ec/Io in CNT+UE mode, and judge
whether this place is with poor coverage. If the RSCP is less than -100 dBm or
the Ec/Io is less than -18 dB, data service can hardly be finished.
Solution: If the RSCP is poor, improve the coverage angle to optimize it. If
the RSCP is fine, but the Ec/Io is very poor, then the following checks are
necessary:
a) Check the pilot pollution, and optimize the seriously polluted pilot.
b) Check the power configuration of the pilot channel, and the power is
always configured as 33 dBm.
17(62)
UMTS PS Service Analysis Guide
iii. Trace and measure the throughput rate and bandwidth of uplink/downlink.
Through monitoring the change of access layer rate and non-access layer rate
of current connected uplink/downlink data transmission, the engineer can
analyze the dynamic channel configuration function and the changing
characteristic of the service source rate conveniently.
iv. Others
Check the problem at the CN side. Please refer to Section 3.2.3, Comparison
and Analysis of Operation Type for the operation method, and compare the
relative parts one by one to narrow the problem range. Then, make a
feedback about the problem.
2. HSPA bearer
When the PS service is be borne on the HSPA, and the engineer can check the
following items.
Check the alarms record to see whether there is anything abnormal at the
time point of the problem, and provide the diagnostic information.
ii. Deactivate the HSDPA feature, and set up the PS service on DCH.
With DEA CELLHSDPA, the engineer activates the HSDPA feature of the cell,
and builds dial-up connection to set up the PS service on DCH. If the data
transmission of the PS service on DCH fails, please refer to the earlier section
for the handling method. If the data transmission of PS service on DCH is
normal, the problem locates at HSDPA, and the engineer should continue with
the following steps.
iii. Check the CQI, HS-SCCH scheduling success rate, SBLER and whether GBR
is configured.
Trace and record the UE reported CQI, HS-SCCH scheduling success rate
and SBLER reported by the UE.
a) CQI
The UE estimates and reports the UE reported CQI on the base of Ec/Nt
of PCPICH. If the UE reported CQI is zero, the NodeB will not send any
data to the UE. If the parameter configurations of pilot Ec/Io,
CellMaxPower, PcpichPower, and MPO constant are normal, but the CQI
18
UMTS PS Service Analysis Guide
is poor, the engineer can try to change the personal computer. PCs with
different model have different noise floors, which will influence the
reported CQI at certain degree.
If the users HS-SCCH Success Rate is about zero, then the use's data
transmission rate is about zero, which means there is no traffic flow on
the user plane.
The possible reasons why the HS-SCCH Success Rate is about zero:
The CQI reported by the UE is too low, which causes that the NodeB
does not schedule this user.
c) SBLER 100%
SBLER is the HS-DSCH block error rate. The above WCDMA HSDPA
Decoding Statistics dialog box shows the SBLER and retransmission
conditions under different TB, and in WCDMA HSDPA Link Statistics
window, the HS-DSCH SBLER-Delta and HS-DSCH SBLER-Average
are shown, in which Delta is an instantaneous value, and Average is an
average value. The reason for the SBLER being 100% is that the HS-
PDSCH Ec/Nt is too low, while the fundamental reason is that the
HSDPA power is not enough. If this parameter is not less than the largest
transmission power of the cell, this power configuration is the dynamic
configuration. In the case of dynamic configuration, HS-PDSCH available
power = largest transmission power of the cell - power margin - R99
downlink load (including common channel load) - HS-SCCH power; in
the case of static configuration, HS-PDSCH available = power of HS-
PDSCH and HS-SCCH HS-SCCH power.
From the above two formulate, in the case of dynamic configuration, the
high power margin configuration, or R99 downlink overload or high HS-
19(62)
UMTS PS Service Analysis Guide
SCCH power configuration may cause the problem that the HS-PDSCH
available power becomes too low.
Solution: Correct the parameter configuration. If the R99 load is too high, the
engineer should consider carrier addition to solve the problem.
iv. Check the available bandwidth, the occupied bandwidth and the assigned
bandwidth of lub.
If the transmission data configuration is wrong, and the IMA group Nos. of
AAL2PATH (for HSDPA) of the NodeB and the RNC are not corresponding to
each other, then the data transmission will not be through.
The problem at the CN side may be caused by the service server with problem, the
incorrect user name and password and so on.
20
UMTS PS Service Analysis Guide
The engineer can make sure that the service software server and the service software
runs normally through other access networks (or LAN).
1. LAN
Through the personal computer on the LAN, the engineer can perform the FTP or
HTTP service to make sure that the services are normal and verify the accessible
user name and password.
21(62)
UMTS PS Service Analysis Guide
the problem at the RAN side. If the engineer finds that the service server works
abnormally, the problem must lie in the CN.
i. HLR
One SIM card can be configured with multi APNs, and each APN corresponds
to a highest rate. When there is no maximum rate limit at the UE side, the
RAB assignment request message delivered by the core network carries the
subscription rate of the user. If there is no limit on the power, code resource
and other resources at the RNC side, the assigned rate will be sent to the UE
through the Activate PDP content Accept message in the NAS signaling, and
the rate in the PDP activation acceptance can be viewed in the QXDM or
other drive test tools.
ii. GGSN
To modify the users QOS parameter on GGSN, the engineer should set the
downlink bit rate and guaranteed rate which are 384 kpbs by default. Set the
maximum downlink rate as 2048 kpbs, so that the CN allows the HSDPA
downlink maximum rate to be 2 Mpbs.
iii. SGSN
The SGSN uses SET 3GSM to modify the users downlink maximum rate and
guaranteed rate to 2 M.
iv. Summary
The checking result shows that the cells have been established. Therefore
problems at the RAN side can be excluded. Then the engineer should check
the HLR subscription rate and the users QoS parameter of the SGSN and the
GGSN at the CN side.
22
UMTS PS Service Analysis Guide
two protocols (RLC/TCP) have great influence on the rate. If the parameters are
improperly configured, or packet error and packet loss happen during the transmission,
the data rate may be reduced. When observing the service quality, the engineer always
uses the application program on the computer with the UE as the MODEM to judge the
quality, then the performance of the computer and the server is involved. Therefore,
there are many factors affecting the PS data transmission performance, and these
factors fall in the categories of the access network problem, the core network equipment
problem, and the application and the service software problem. In this guide, the
application and service software problem and the core network equipment problem are
generalized as the problem at the CN side; and the access network problem is referred
as the problem at the RAN side.
23(62)
UMTS PS Service Analysis Guide
GGSN, LANSWITCH, ROUTER, FIREWALL and other NEs at the CN side should be
checked. The alarms of the clock problem, the transmission errors, the instantaneous
interruption of transmission and other abnormal conditions of the equipment may affect
the data transmission. If the engineer cannot locate the problem by alarms of NEs, he
should make comparison and analysis of the operation type, and try to screen out the
affecting factors and narrow the range of the problem. If the cause of the PS data
transmission problem can be determined to be at the RAN side, the engineer should
analyze the problem at the RAN side; If the cause of problem can be determined to be at
the CN side, the engineer should analyze the data transmission problem at the CN side.
If the cause the problem cannot be fixed, the engineer should analyze from both the
RAN side and the CN side.
The Iub interface transmission error, delay jitter, and lub bandwidth problem will affect
the data transmission. The analysis flow is shown as follows:
Analysis of the factors affecting the data transmission at the lub interface
The engineer can check whether the problem exists by checking the transmission
alarm and the clock alarm.
24
UMTS PS Service Analysis Guide
The methods of checking whether lub congestion exists are shown as follows:
Comparison and analysis of the operation type is to find out the NE on which the
problem happens, and tell where the cause of the problem is: the core network or
service software or the access network. The comparison includes the following
operations:
Changing the USIM card, the cell phone/data card and the PC;
Changing the networks in the same server, for example, 2G or other 3G networks
25(62)
UMTS PS Service Analysis Guide
What is more, for further troubleshooting, the engineer can send the data from the upper
level equipment to the UE for comparison.
26
UMTS PS Service Analysis Guide
If the engineer still cannot find out where the problem lies after making the comparison
and analysis according to the above operation step by step, he should analyze the
problems step by step in the sequence of the RAN side and the CN side.
27(62)
UMTS PS Service Analysis Guide
If the PS service is on the DCH, the factors affecting the data transmission at the Uu
interface mainly include the allocated channel bandwidth, the state migration occurrence
and the error condition at Uu interface. The analysis flow is shown as follows:
28
UMTS PS Service Analysis Guide
1. DCH Bandwidth
When the PS service is on the DCH, RNC allocates certain channel bandwidth for
every accessed UE. The size of the bandwidth is decided by the spreading factor
and the coding method of the code resource.
If the measurement shows that the UEs allocated bandwidth is lower than that in
expectation, there are two possible causes. One is that the channel with higher rate
cannot be allocated to the UE because of the congestion and other abnormal
reasons, which is an abnormal condition. The other is the DRBC algorithm of RNC.
If the parameter of the DRBC algorithm is reasonable, this rate decrease is normal.
Based on the on-site configured parameter, taking the algorithm into account, the
engineer should check whether the current bandwidth allocation and adjustment is
reasonable, whether there is anything abnormal and whether it is necessary to
adjust the parameter to solve the problem. For the low bandwidth caused by
congestion and other abnormal conditions, the engineer can make single-user
tracing, check the downlink load and code resource allocation and the size of the
lub available bandwidth, and acquire the corresponding association log to check
the abnormal printing information, then the clue of the problem can be found.
29(62)
UMTS PS Service Analysis Guide
is normal. Otherwise, it is necessary to analyze the factors causing the error at the
Uu interface.
Power control and coverage are the main factors affecting the uplink/downlink
BLER.
The engineer should make sure that the RNC external loop power control
switch is open.
ii. Coverage
The engineer should make sure that the uplink/downlink power is limited in the
area with poor UL BLER and DL BLER.
iii. UE performance
The engineer can try to change the UE, or compare the current UE and the
UE of other types.
3. In-sequence delivery
Setting the In-Sequence Delivery to TRUE or FALSE will affect the rate and
fluctuation of uplink data transmission. If the In-Sequence Delivery is set to TRUE,
RLC will keep the transfer order of the high-level PDUs. If it is set to FAUSE, RLC
entity which receives the data will allow the SDUs to be transferred to the high-level
in the order different from that of the sender, and then the uplink data transmission
rate will be low, with great fluctuation. It is suggested to set the In-Sequence
Delivery to True, and the setting can be modified in HLR of CN.
The engineer can acquire the throughputs on APP level and RCL level through DT/CQT
test.
If Throughput on APP level/ Throughput on RLC level is lower than the normal range of
theoretical analysis, it means the TCP/IP retransmission costs too much overheads.
TCP receiving window and MTU setting can be checked and modified.
2. Possible Reasons:
30
UMTS PS Service Analysis Guide
ii. The data transmission is interrupted after the handover from 3G to 2G;
iii. The state migration from CELL_DCH to CELL_FAC and CELL_PCH happens
during the data transmission. When the data transmission is recovered, the
systems cannot migrate back to CELL_DCH status, because the resource is
not enough. And the data transmission will be affected.
3. Analysis: The engineer analyzes this problem from the aspects of alarms, signaling
flow.
i. Alarms
The engineer checks the alarms of CN and RAN NE, understands the
abnormal condition of the current system, and guide the problem analysis and
troubleshooting. The engineer can find out the problems such as the
transmission interruption, the cell clock asynchronizaiton and the NE
congestion through the alarms.
31(62)
UMTS PS Service Analysis Guide
signaling: one is to use CNT+UE to collect the signaling sent and received at
the UE side, the other way is to use the OMCR single-user tracing to collect
the signaling sent and received at the RNC side. Through the comparison of
the signaling collected by the two ways, the engineer can check whether there
is message loss caused by the air interface with poor quality. What is more,
the engineer can take the association log into account to locate the problem.
a) Call drop
After the channel status is migrated to the common channel, the channel
cannot migrate back to the CELL_DCH status. The engineer can check
the signal to check whether the problem is caused by the downlink
overload or the lub bandwidth congestion. To solve the problem, the
engineer can add the carriers or the transmission resources.
HSDPA schedules power and code resources among multiple users in code-division or
time-division mode. Under the single-user condition (that is when there is only one
HSDPA user in the cell), the factors that affect data transmission rate are available
power for HSDPA, the number of HS-PDSCH codes (only one HS-SCCH channel is
needed under the single-user condition), the UE category (the maximum number of
codes supported by the UE and the availability of 16QAM support function), radio signals
at the location of the UE, the UEs subscription rate, Iub bandwidth, and the maximum
rate supported by the RNC, NodeB, GGSN, and SGSN. Under the multi-user condition,
apart from the above factors, the algorithm used in the NodeB, the number of HS-
SCCHs configured for the cell will also affect the data transmission rate.
32
UMTS PS Service Analysis Guide
1. The UE reports CQI on the HS-DPCCH, and the NodeB acquires the channel
quality of the location of the UE.
3. The NodeB sends the HS-DSCH parameters on the HS-SCCH, and then sends the
parameters on the HS-DSCH after 2 slots.
4. The UE monitors the HS-SCCH to see whether there is any information for itself. If
there is, then the UE starts receiving data from the HS-DSCH and then cache the
data.
6. The UE demodulates the data received from the HS-DSCH, and sends the
ACK/NACK in response to the uplink HS-DPCCH according to the CRC results.
7. If the NodeB receives NACK, it will resend the data until it receives ACK from the
UE or until the number of resending times reaches the maximum value.
Judge whether the services are based on HSDPA from the following aspects:
1. Whether the cell supports HSDPA: Check whether the HSDPA cell at the RNC side
has been activated, and whether the attribute of the local cell at the NodeB side
has been configured to support HSDPA.
2. HSDPA services access failure will also cause the RNC to reconfigure HSDPA
services to 384 kbps as R99 services. Check whether the following configurations
are reasonable: the uplink and downlink load of R99 services, downlink code
resource, Iub transmission resource, the number of HSDPA services users, the
total rate threshold in the HSDPA cell, the guaranteed rate threshold and the
guaranteed power threshold of stream services.
3. The HSDPA threshold of downlink BE services is too high. The HSDPA threshold of
downlink BE services determines the rate decision threshold of the PS domain
Background/Interactive services on the HS-DSCH. Only when the maximum
downlink rate of the PS domain Background/Interactive services is not less than
this threshold, can the services be borne on the HS-DSCH; otherwise, they will be
borne on the DCH.
33(62)
UMTS PS Service Analysis Guide
During the NodeB scheduling process, TB size is decided by CQI, code, and power. And
TB size/2 ms is the scheduled rate. In the normal condition, scheduled rate and the UE
reported CQI are in a mapping relationship (which depends on the NodeB CQI mapping
table in practical use). Strictly speaking, because the NodeB will filter and correct the UE
reported CQI, in fact the scheduled rate may be in a mapping relationship with the
NodeB scheduled CQI, rather than with the UE reported CQI. Therefore, the following
table shows the corresponding relation between CQI and the reference TB size
according to the protocol 25.214, and the relation between CQI and the scheduled rate
can be worked out.
34
UMTS PS Service Analysis Guide
1 137 1 QPSK 0
2 173 1 QPSK 0
3 233 1 QPSK 0
4 317 1 QPSK 0
5 377 1 QPSK 0
6 461 1 QPSK 0
7 650 2 QPSK 0
8 792 2 QPSK 0
9 931 2 QPSK 0
10 1262 3 QPSK 0
11 1483 3 QPSK 0
12 1742 3 QPSK 0
13 2279 4 QPSK 0
14 2583 4 QPSK 0
15 3319 5 QPSK 0
16 3319 5 QPSK -1
17 3319 5 QPSK -2
18 3319 5 QPSK -3
19 3319 5 QPSK -4
20 3319 5 QPSK -5
21 3319 5 QPSK -6
22 3319 5 QPSK -7
23 3319 5 QPSK -8
24 3319 5 QPSK -9
35(62)
UMTS PS Service Analysis Guide
The factors that affect the scheduled rate are CQI, HSPDA cell available power, and
HSDPA cell available codes. The analysis can be made from the following perspectives:
1. CQI
If the UE downlink rate is low, check whether the UE reported CQI is too
low and at the same time check the RSCP and Ec/lo of the PCPICH in
the current cell. The problem may be caused by:
36
UMTS PS Service Analysis Guide
ii. Solutions:
a) For the poor coverage problem, optimize RF or add sites to improve the
UE reported CQI.
If the available power for the HSDPA cell is too low, it will affect the TB size during
the NodeB scheduling process.
The available power of HSDPA cells can be affected by the following factors:
The HS-PDSCH MPO constant can be modified in the RNC OMC. Under the
condition of HSDPA, the UE reports CQI, and according to the reported CQI,
the NodeB can judge the quality of the current radio link and adjust the TB
size and the power. When the UE reported CQI is less than 5, the NodeB will
not send data to this UE. And the MPO constant can adjust the UE reported
CQI, which can be calculated based on the pilot Ec/Nt.
37(62)
UMTS PS Service Analysis Guide
i. When the code resource is allocated in dynamic mode, usually the OMC will
assign the initial number of HS-DSCHs, the maximum number of HS-DSCHs
and the minimum number of HS-DSCHs. The number of HS-DSCHs occupied
by users should be between the maximum one and the minimum one. If more
R99 CS users need to access the network, they may take the HSDPA code
resource.
ii. HSDPA accompanies the DCH code resource allocation. When a user applies
the high speed PS services, the system will bear the services on HSDPA. This
user occupies the HS-SCCH and the HS-PDSCH, and meanwhile when the
services are being established, the user will be assigned with an associated
DCH (A-DPCH) to transmit signaling at 3.4 kbps. This channel is a dedicated
downlink channel with SF256.
iii. If the codes allocated to HSDPA users are too few, the TB size scheduled by
the NodeB will also be affected.
4. HSDPA UE capacity
ii. In the message RRC Connection Setup Complete the UE will report its
capacity. The IE (information element) of HSDSCH physical layer category
reflects the UEs ultimate capacity.
5. The data volume that can be scheduled by the UE is smaller than the largest TB.
The TB size scheduled by the NodeB depends not only on the UE available power
and codes, but also on the data volume that can be sent by the UE. If the data
volume sent by the UE is smaller than the largest TB, then the physical layer rate
will be lower than the expected value. This problem usually occurs when there is
still data in the NodeB buffer area, but the data volume is smaller than the largest
TB that can be scheduled.
38
UMTS PS Service Analysis Guide
One RAB can be mapped only to the HS-DSCH of one cell, which means that it does not
support soft handover. Therefore data transmission will be interrupted when the serving
cell is changed.
Data transmission interruption happens when the serving cell remains unchanged or no
handover happens, and the possible causes are as follows:
Check the alarms of different NEs of the CN and RAN to see the operating status of
the system. This can guide the troubleshooting. Check the alarms for transmission
interruption, clock synchronization failure, and NE congestion.
During the data transmission, if the data transmission is interrupted for a long
period and cannot be recovered, checked whether the FTP download is completed.
Analyze the signaling of the RNC and UE to check for call drops, H2H serving cell
changes, or H2D/D2H handovers. For the transmission interruption caused by call
drops, analyze the reasons of call drops.
Use Ethereal to capture the TCP/IP packets, and analyze the transmission interval
between different TCP/IP packets.
39(62)
UMTS PS Service Analysis Guide
the data transmission. When testing and optimizing the HSUPA data transmission
performance, we usually focus on the impact from the RAN side (RNC-NodeB-UE).
Impacts brought by other components of the system, such as the CN side (SGSN,
GGSN), data service server, and external network, should be eliminated before the
testing. Then, we can focus on the optimization of the radio network.
From the throughput rate aspect, when the data transmission performance is poor, the
rate is instable, low, or fluctuates greatly. From the QoS aspect, the image of the
streaming media blurs, the watching may be blocked, and the webpage opens slowly.
1. The UE sends the Scheduling Information through the E-DPDCH to request for the
data transmission. The Scheduling Information is determined by UE Power
Headroom (UPH) and _Q (Data volume to be transmitted).
2. The NodeB will determine the grant level based on SI sent by the UE, measured
RoT, and the Happy Bit sent by the UE (carried by E-DPCCH). Then the NodeB
would send T/P through E-AGCH and the tuning command (+1, 0, -1) through E-
RGCH.
3. The UE sends the data based on the grant level, and sends SI for the next frame
and Happy Bit about the grant level through E-DPDCH.
4. After the NodeB receives the demodulation data, it returns ACK/NACK through E-
HICH to feed back whether the data reception succeeds.
40
UMTS PS Service Analysis Guide
The services will be borne on DCH if they are not borne on E-DCH, and then the service
rate will be that for R99 services, which is usually 384 kbps or lower.
If the service is not borne on E-DCH, analyze this problem from the following aspects:
2. Check whether the MBR in the UL subscription information is normal, and check
whether the rate threshold of the E-DCH is too high. If the MBR assigned by the CN
has not reached the rate threshold of E-DCH, the service will be borne on DCH.
Serving Grant Rate Throughput is the Scheduled Rate. To locate this problem, compare
the current Serving Grant Rate Throughput and the maximum allowed ETFC based on
the ETFC of the current SG and the Happy/Unhappy Bit sent by the UE.
Note that when the UE sends the Happy Bit does not necessarily mean that the users
are happy with the current rate. Analyze the reason that the UE sends the Happy Bit.
The Serving Grant Rate Throughput is abnormal. To confirm this, check whether the UE
reports Happy Bit or Unhappy Bit.
If the UE sends the Happy Bit, while the rate cannot reach MBR, the possible causes are
as follows:
2. UE TX power restriction
If the UE sends the Unhappy Bit, and the user rate cannot reach MBR, the possible
causes are as follows:
2. RAN resource restriction, such as the air interface load, Iub bandwidth, and CEs
41(62)
UMTS PS Service Analysis Guide
As specified in the protocol, the UE shall indicate that it is unhappy if the following
criteria are met:
The first criterion is always true for a deactivated process and the ratio of the third
criteria is always 1 for 10ms TTI.
Principle:
Observing method:
Generally, the UEs that support HSUPA will support HSDPA. In other
words, the HSPA carries the user services as a whole. The following part
would check whether the UE supports the HSPA and the specific HS-
DSCH/E-DCH physical layer category based on the RRC messages.
42
UMTS PS Service Analysis Guide
Firstly, the UE will report whether it supports HSDPA and HSUPA in the
RRC_CONN_REQUEST message. As shown in following picture, we
can see that the UE supports HS-DSCH and E-DCH.
43(62)
UMTS PS Service Analysis Guide
When the services are constructed, the RNC sends the RL RECFG
PREP message to the NodeB, which includes the maximum spreading
factor that the UE can use. The corresponding cell is maxSet-E-
DPDCHs. In following picture, maxSet-E-DPDCHs = 9 indicates that the
UE supports two SF4+2SF2 spreading factors.
Solution:
Principle:
The UE can calculate the TB size based on the current available TX power. Then
the UE will compare the TB size supported by the TX power and that supported by
SG, and then choose the small one as the actual TB size. Though the UPH of
44
UMTS PS Service Analysis Guide
different UEs is the same, their TB sizes may be different. The following factors
affect the TB size:
b) The UL load of the cell is heavy (the UE is not at the cell edge);
c) The UE is performing hybrid services, and the DCH services occupy too
many resources, and the remaining power for E-DCH is limited.
Solution:
b) If the UL load of the cell is too high, lower the cell load (if possible);
Principle:
Insufficient data volume in UE RLC buffer leads to low MAC-e PDU Non-DTX Rate.
Solution:
Try to use the UL packet delivery method to exclude the impact of TCP. If this still
does not work, check whether the packets are lost at certain link.
Sometimes the application software of the PC may affect the data transmission.
Change a portable computer. If the problem still exists, use the tool to capture the
packets to locate the abnormality between the PC and the UE.
If the actual load of the cell is heavier than the target value, reduce the UE SG to
reduce the cell throughput rate, when the actual load of the cell is less than the
target value, increase the SG of the UEs reporting Unhappy Bit to increase the cell
throughput rate.
45(62)
UMTS PS Service Analysis Guide
The UE upgrades the SG information through AG and RG (UP, Hold, Down) sent by
the NodeB. The UE would refer to the SG when deciding the actual TX rate, so the
actual TX rate will be lower than or equal to that specified in SG.
ii. RAN resource restriction, such as the air interface load, Iub bandwidth, and
CEs
The resource restriction at the RAN side will leads to the fact that the UE SG
assigned by the NodeB is small, which will further lead to that the UE reports the
restriction of SG.
Solution:
Solution:
Solution:
Add CEs.
Solution:
To increase the throughput rate, the subscription information must be modified, and
a larger MBR should be configured.
iv. UE demodulation error, for example, the AG CRC verification error leads to
the failure of the SG upgrade, or the incorrect UE RG demodulation.
46
UMTS PS Service Analysis Guide
Cause one: The CRC verification error of the AG value leads to the failure of
the SG upgrade.
Solution:
As to cause one, increase the E-RGCH power offset based on the baseline
parameters and the coverage at cell edges.
The following picture shows the analysis flow for poor data transmission performance at
the CN side.
47(62)
UMTS PS Service Analysis Guide
The packet losses at the CN side will lead to RTT timeout, which will trigger congestion
avoidance. The TCP can provide reliable transmission layer. One of its functions is to
acknowledge the message sent by the other end. However, both the data and
acknowledge messages may be lost. To solve this problem, the TCP will trigger a timer
when sending the data. If it fails to receive the ACK message after the timer expires, it
will retransmit the data. The TCP TX end will measure the RTT value, which is the length
of time it takes for a signal to be sent plus the length of time it takes for an
acknowledgment of that signal to be received, and will maintain an RTT timer. If the
measured RTT has timed out, the TCP will think that the network is congested, and will
trigger congestion avoidance. Thus, the data transmission rate is affected. The IP packet
losses at the CN side will lead to RTT timeout.
48
UMTS PS Service Analysis Guide
Rate is also related with the performance and the operating system of the computer, as
well as the installed application software, because the inner algorithm of the application
software and the TCP parameters of the operating system may have great influence on
the performance. Similarly, the PS data transmission rate of the computer with WIN2000
operating system is better than that with WIN98 operating system. Therefore it is
recommended that the computer and server should be installed with WIN2000 Pro and
WIN2000 Server respectively or high versions. Because the portable computer is usually
installed with WINXP, the performance issue caused by the operating system does not
exist. However, the server should use WIN2000 Server rather than WINXP, otherwise
the data transmission will be seriously affected. The computer working as the UE
background must be able to provide better performance. It is verified that the IBM
provides better performance when demonstrating the VOD. If the CPU usage of the
portable computer reaches 100% at the RX end, the TCP receive window will be full.
When the TCP receive window is full, the RX end will inform the TCP TX end, and then
the TCP TX end will stop the data transmission. As a result, RLC BO is zero, the NodeB
cannot perform the data scheduling, and the data transmission performance is affected.
Meanwhile, the performance of the server may also affect the services, and this is
another factor that we must consider.
For services using the TCP protocol, such as VOD and FTP, the sizes of the TCP
windows on the test computer (as the Client) and the server have great influence on the
service. The window size is usually set to a large value to ensure good performance,
and the window sizes on the Client and the Server should be the same. Theoretically,
the TCP receive window should be larger than the product of bandwidth and delay.
The 64-K window is sufficient for the HSDPA CAT12 1.6-MHz bandwidth, but it is
insufficient for the HSDPA CAT6 3.6-MHz bandwidth. Especially when the delay is larger
than 200 ms, the TCP window may be easily fully occupied. Then we will observe that
RLC and NodeB buffer is 0.
Solution:
If there is one data packet to be transmitted at the IP layer, with its length larger than that
of the MTU at the link layer, then the IP layer will divide the packet into different
segments. Each segment is shorter than the MTU. To improve the transmission
efficiency, we should avoid the IP segmentation and regrouping and use a longer MTU
49(62)
UMTS PS Service Analysis Guide
(usually does not exceed 1450 bytes). To modify the MTU length, modify the MTU at the
server and the portable testing computer. After the PS services are connected, the
Server and Client will negotiate about the MTU length. Then, the shorter MTU is used.
1. FTP
When selecting the FTP software, the commercial FTP is preferred. It usually
provides better services than the FTP software of the operating system. Besides,
the FTP uses the binary mode to download the data. Thus, you are advised to use
the multi-thread download software, such as FlashGet. If the upload rate is low, you
can use multiple FTPs to perform the transmission, or use the software to send the
packets at a fixed rate to test the bottom layer.
2. VOD
The maximum transmission rate of the RealPlay should be larger than 384 K, and
the buffering time should be short, for example, 3 s. As to some portable computers
with poor graphics cards, frequency hopping may appear. You can set the
resolution ratio to 800 x 600 pixels. If the problem still exists, replace the graphics
card.
3. NetTV
When the rate at the bottom layer is reduced, the performance of NetTV is hard to
recover. You must pay special attention to this point.
4. Video conference
The output rate of the video conference should be slightly less than the rate at the
bottom layer, otherwise packet losses will occur. For example, the rate for the
NetTV of one vendor is set to 128 K initially, which will be increased by 64 K every
time. Actually the recommended value for this NetTV is 320 K. If this rate is too low,
the bandwidth cannot be fully used; otherwise, a rate higher than 320 K, such as
384 K or higher, may lead to packet losses because the bottom layer cannot satisfy
the demand. Then the effect of video conference will be influenced. The lightning
bolt on the upper right corner of the screen indicates that there are some error
codes or some packets are lost during the transmission process.
50
UMTS PS Service Analysis Guide
Problem analysis: Performance statistics shows that the traffic of this cell is relatively
heavy. Iub bandwidth is 1 M and the band is fully occupied. If the user logs on to the
internet and there is continuous data transmission, the status is stable. If the user
browses a webpage and there is no data transmission for a period of time, the UE will
switch to idle state based on the DRBC algorithm. In this case, if data transmission is
required, the resources should be applied again; because the resources are used by
other users, the application will fail, the connection will be broken, and call drops
happen. Only when the user number is reduced can the service be successfully
established again. The problem is caused by too many users, therefore try the following
methods to solve the problem:
HSDPA downloading rate of one site was low, which was 4 M; HSUPA downloading rate
of that site was even lower, which was about 400 K. On the same conditions, uploading
and downloading tests of R99/PS 384 K were conducted, and the services were found to
be normal.
Fault location:
Based on UE logs and LMT measurement, it was initially concluded that it was caused
by UL packet loss. The Uu interface rate was obviously higher than the RLC rate.
Packets of the site were captured and analysis was made. It was found that there was
packet loss from the NodeB to the RNC and packet loss based on the data capture of
the CCI board of the NodeB. Special attention was paid to packet loss based on the data
capture of the CCI board of the NodeB, but no fault was located. Then the problem of
the packet loss from the NodeB to the RNC was checked. BPO equipment room
engineer checked the GIPI boards and no exception was found. Packets were pinged
from the CCI board of the NodeB to Switch 6513 of the Iub interface, and 8 out of 100
51(62)
UMTS PS Service Analysis Guide
packets were lost. The operation was repeated three times, and the results were the
same. The Ethernet cable was used to connect the SP directly (skip Switch 2818) to
BBUB and to ping Switch 6513, none of the 100 packets was lost. Therefore it was
concluded that there was packet loss between Switch 2818 and BBUB. Considering the
Ethernet cable between BBUB and Switch 2818 was self-made, the engineer changed
the cable to ping packets to Switch 6513, and there was no packet loss. Therefore the
problem must be caused by this self-made cable. The following picture compares the
application layer rates with the abnormal cable and the normal cable.
Application layer rate comparison with the abnormal Ethernet cable and the normal
Ethernet cable
Solution: Change the Ethernet cable between Switch 2818 and BBUB.
The HSDPA downloading rate of one site was low, which was about 3.5 M.
Fault location:
Investigation showed this site was adjacent to another testing site which was not
configured as its neighbor cell and thus brought strong interference to this site.
Therefore the testing site should be blocked during the HSDPA downloading.
The following picture shows the comparison before and after the testing site was
blocked. Please pay attention to the change of CQI.
52
UMTS PS Service Analysis Guide
CQI comparison before and after the testing site was blocked with QXDM
After the testing site was blocked, the HSDPA rate increased to 4.5 Mbps, which was
relatively low. Further investigation indicated that there was another exception:
Attempts to ping packets from the CCI board of the NodeB to Switch 6513 were
conducted.
When the cell had no HSDPA downloading service, none of the 100 packets were lost.
When the cell had HSDPA downloading service, 3 to 8 of the 100 packets were lost.
SP was connected directly (skip Switch 2818) to the NodeB. Data was captured in
Switch 2818. The result showed that packets had been sent to Switch 6513, but no
packet was returned, as shown below:
53(62)
UMTS PS Service Analysis Guide
Data capture of ping packets to Switch 6513 on the CCI board of the NodeB
Therefore it was concluded that there was packet loss from Switch 2818 to Switch 6513.
FST-2802 device was used to test this Iub interface and there was a serious problem of
packet loss.
Solution:
After the SP has repaired the transmission, HSDPA downloading becomes normal.
Suggestions:
There are two factors affecting the HSDPA downloading rate: signaling quality and Iub
interface transmission. The two factors should be checked one by one.
Problem description:
The HSDPA downloading rate was low and the service was unstable.
Fault location:
Seen from the map, the site was very close to another site.
54
UMTS PS Service Analysis Guide
Site location
55(62)
UMTS PS Service Analysis Guide
This area was seriously polluted. Engineers blocked one site and the HSDPA
downloading rate became normal, as shown below:
56
UMTS PS Service Analysis Guide
Solution:
Perform radio network optimization to solve the pollution problem and the HSDPA
downloading rate becomes normal.
4.2 CN Cases
One user found that his MR632 inserted in one computer was not able to perform the PS
service.
Problem analysis:
The user complained that MR632 was normal, but one day it did not work. The problem
must be caused by the computer. Investigation showed that this user had newly installed
the firewall. After the firewall was uninstalled, the service became normal.
57(62)
UMTS PS Service Analysis Guide
Problem analysis:
Tests and analysis show that the access network seemed to be problematic. Based on
the fault location concept, engineers traced detailed user signaling, and collected
statistics of the user plane, information of Uu-interface quality and information of packet
loss on the Iub interface, nothing was found abnormal. The network planning engineers
arrived at the site, and signaling was found to be normal. The engineer changed the PC,
and HSDPA service became normal. Then the problem must be caused by the driver of
the public portable computer. After the portable computer was changed, the problem was
solved.
Problem description:
Engineers from the operator complain that the service can not be normally stopped
when downloading a file longer than 10 minutes in the 3G network. In the networks of
other operators or in 2G networks, there is no such problem.
Problem analysis:
Analysis on the FTP messages captured in ethereal shows that the data session had
been transmitted, but the control session had no interaction completion process. There
was no 221-Goodbye character but the downloaded file could be normally opened. As
shown in following picture, the file had been downloaded and could be used normally.
58
UMTS PS Service Analysis Guide
To make it clear, a comparison between the signaling messages respectively for normal
FTP downloading and abnormal FTP downloading is provided here.
The following messages appeared during the normal FTP downloading process.
59(62)
UMTS PS Service Analysis Guide
60
UMTS PS Service Analysis Guide
Carefully compare the two figures and the problem will be found. Engineers used
Qualcomm 6250 to activate PS service downloading, and captured packets at RNC and
SGSN sides respectively to make analysis.
1. The engineer used the FTP tool of the operator to download a 3.5-M file for about
12 minutes, and the problem occurred.
2. The engineer executed the FTP command of the Windows operating system to
download a 3.5-M file for about 30 minutes, typed bye to quit, and the problem also
occurred. During this process, outlook was transferring data in the background, which
might affect the service. The service was disconnected abnormally a long time after the
transfer was completed, which was also abnormal.
3. The engineer executed the FTP command of the Windows operating system to
download a 0.4-M file for about two minutes, typed bye to quit, and there was no
problem.
4. The engineer used the FTP tool of the operator to download a 0.4-M file for about
two minutes and there was no problem.
5. Same as Step 2, the engineer disconnected the downloading service in the outlook,
and the result was the same.
Then it could be concluded that the problem was related to time instead of the file size. A
lot of statistic analysis shows that the FTP data transmission was normal and the content
was correct and usable, except that the signaling could not be normally closed. The
engineer used the method of changing the NE (UE or server) to locate the problem.
Conclusion:
The engineer used different UEs, and FTP servers to make the test, and the same
problem occurred. Therefore the problem must lie in the network. The engineer
performed tests in the 3G network, 2G network, and conducted the 3G-to-2G handover
test.
The result indicated that the problem must lie between GGSN and FTP server. The on-
site engineers performed tests and found that no problem occurred if the server was not
installed with the firewall. Data analysis shows that the FTP data transmission at the port
was normal but was blocked after 10 minutes. Only the firewall on the system can limit
the time for the ports. If it is set in the firewall that the connection will get broken if there
is no data interaction for ten minutes at one port, the problem will occur. The firewall
professionals give the following explanation to this problem:
The FTP session will generate two session entries in the firewall: one is used for the
FTP control channel, and its default aging time is ten minutes; another is used for the
FTP data channel, and its default aging time is four minutes. If the detect ftp command
was not configured in the domain the aging time will not be updated during data
transmission in the data channel, and the control channel will be invalid after ten minutes
61(62)
UMTS PS Service Analysis Guide
and then be deleted. Then the problem occurs. If the detect ftp command is configured
in the domain, the aging time of the control channel will be updated when data is
transmitted. Thus it can be concluded that the problem has nothing with the radio access
network. The process and result of this fault location are meaningful and needs careful
consideration. The test method of changing the NE is really useful.
62