Beruflich Dokumente
Kultur Dokumente
Voice Quality
Issue 01
Date 2017-02-20
(FOR INTERNAL USE ONLY. DO NOT SEND THIS DOCUMENT OR COMPLETE
CONTENTS IN THE DOCUMENT TO CUSTOMERS)
and other Huawei trademarks are trademarks of Huawei Technologies Co., Ltd.
All other trademarks and trade names mentioned in this document are the property of their respective
holders.
Notice
The purchased products, services and features are stipulated by the contract made between Huawei and
the customer. All or part of the products, services and features described in this document may not be
within the purchase scope or the usage scope. Unless otherwise specified in the contract, all statements,
information, and recommendations in this document are provided "AS IS" without warranties, guarantees or
representations of any kind, either express or implied.
The information in this document is subject to change without notice. Every effort has been made in the
preparation of this document to ensure accuracy of the contents, but all statements, information, and
recommendations in this document do not constitute a warranty of any kind, express or implied.
Website: http://www.huawei.com
Email: support@huawei.com
1.1 Definition
Mean Opinion Score (MOS) is a key performance indicator (KPI) for evaluating the voice
quality of a communication system. Generally, the voice quality is good if the MOS is greater
than or equal to 4, and the voice quality is acceptable if the MOS is greater than or equal to 3.
The MOS defined by China Mobile is a MOS that is obtained from the drive test (DT) and is
calculated using the AMR WB-based POLQA algorithm.
The MOS defined by China Mobile is described as follows:
China Mobile only has a MOS test standard for the voice service, but does not have a
universal MOS test standard for the video service. Therefore, the VoLTE MOS test currently
applies only to the voice service.
In the current scenarios of China Mobile, AMR-WB is used for calls between VoLTE
subscribers to achieve high definition (HD) voice experience, and AMR-NB (same as the
codec used in the CS domain) is used for CS services between VoLTE and 2G/3G subscribers.
Therefore, in a MOS test, calls between VoLTE subscribers are initiated to test the voice
quality when AMR-WB is used as the voice codec.
The MOS defined by China Mobile is a MOS that is obtained from the DT and is calculated
using the P.863 algorithm. China Mobile requires that the MOS test tools be supplied by
DingLi and CDS. In the current phase, HTC M8T is used as the test UE, and Huawei Probe
Assistant (PA) can be used for routine optimization.
Local Test IMS DRA PCRF HSS EPC Averag Averag RTP Avera
Networ Equipmen e RSRP e SINR Packe ge
k t (MO) (MO) t Loss MOS
Rate
(%)
(MO)
Beijing CDS Huawe Huawe Nokia Nokia Nokia -89.881 13.862 0.51% 3.631
Huawei i i Siemen Siemens Siemens
s
Beijing CDS Huawe Huawe Nokia Nokia Huawei -88.544 14.353 0.66% 3.606
ZTE i i Siemen Siemens
s
Changsha DingLi ZTE ZTE ZTE ZTE Huawei -89.321 16.728 0.38% 3.85
Huawei
Changsha DingLi ZTE ZTE ZTE ZTE ZTE NA NA 0.27% 3.83
ZTE
Shenzhen CDS Huawe Huawe Huawei Ericsso Ericsso -90.569 13.621 1.14% 3.661
Huawei i i n n
VQI Description
L.Voice.VQI.UL.Excellent.Times Number of times the uplink voice quality is Excellent
L.Voice.VQI.UL.Good.Times Number of times the uplink voice quality is Good
L.Voice.VQI.UL.Accept.Times Number of times the uplink voice quality is Accept
L.Voice.VQI.UL.Poor.Times Number of times the uplink voice quality is Poor
L.Voice.VQI.UL.Bad.Times Number of times the uplink voice quality is Bad
L.Voice.VQI.DL.Excellent.Times Number of times the downlink voice quality is
Excellent
L.Voice.VQI.DL.Good.Times Number of times the downlink voice quality is Good
VQI Description
L.Voice.VQI.DL.Accept.Times Number of times the downlink voice quality is Accept
L.Voice.VQI.DL.Poor.Times Number of times the downlink voice quality is Poor
L.Voice.VQI.DL.Bad.Times Number of times the downlink voice quality is Bad
L.Voice.E2EVQI.Excellent.Times Number of times E2E VQIs are Excellent
L.Voice.E2EVQI.Good.Times Number of times E2E VQIs are Good
L.Voice.E2EVQI.Accept.Times Number of times E2E VQIs are Accept
L.Voice.E2EVQI.Poor.Times Number of times E2E VQIs are Poor
L.Voice.E2EVQI.Bad.Times Number of times E2E VQIs are Bad
As shown in the preceding figure, two VoLTE terminals are connected to a PC where the
DT software is installed. One MOS box (no model is specified in the China Mobile test
specifications) is connected to the PC.
Huawei provides the following test tools: Probe (DT software), Mate7/HTC M8, and
MOS box.
The Probe license can be applied internally, but the POLQA license used for the MOS
test must be purchased and bound to the PC.
How to check the MOS in the Probe
Among the terminal DT data, pay close attention to "MOS Average". To check the MOS,
start the Probe, choose View > MOS > Speech Quality Evaluation. To check auxiliary
KPIs, choose LTE > VoLTE Parameters.
1.5.2 CN Tools
1.5.2.1 SmartTestKit
The SmartTestKit is a VoLTE terminal dialing test tool developed by the CN VoLTE
integrated service team.
The SmartTestKit introduction and related product document are available in the following
path at www.support.huawei.com:
Support > Product Support > Core Network Common > Core Network Common > CSI
The figure below shows the voice quality test interface of the SmartTestKit:
The SmartTestKit uses the P.862 algorithm to evaluate the voice quality, and currently does not support
P.863. Therefore, if calls use the AMR WB codec, the voice quality evaluation result may be inaccurate.
The test result is for reference only, and cannot be used for acceptance.
1.5.2.2 CN Nastar
After VoLTE3.2, which passed GA in the first half year of 2016, was released, the CN
NaStar2.0 is released for the CCE product. The CN Nastar collects and summarizes in real
time the call history records (CHRs) generated by NEs in the IMS domain, analyzes CHRs,
and then outputs reports.
For the VoLTE3.2, a CHR records the media plane quality detected on the SBC and the MOS
of each VoLTE session. Based on data in these CHRs, the CN Nastar implements the
following functions:
Take statistics and analyze MOSs of sessions by location or time segment.
Query the media plane quality and MOSs recorded in CHRs based on the specified
subscriber number.
For details about the analysis method, see the product documents after a version (CCE
V100R003C30 or later) is released.
Product documents of the CCE are available in the following path at
www.support.huawei.com:
Support > Product Support > Core Network Common > Core Network Common > OSS >
CCE
To use the CN Nastar, the CCE must be deployed and the CN Nastar license is available. The CN Nastar
applies to a site where the CCE is deployed.
The GA phase of Discovery PS was passed in quarter 1 of 2016, and Discovery PS applies to
UGW9811V900R013C00 or a later version.
For details about how to use Discovery PS to analyze the VoLTE voice quality, see the
reference document [1] PS PSS V100R007C00 VoLTE Evaluation and Optimization Guide.
The OMStar and Nastar support automatic analysis on the VoLTE voice quality problems on
the wireless network side. The OMStar is an offline tool, and the required data must be
imported after users collect the data. The Nastar is an online tool and can automatically
import data after subscription. The two tools are applicable to different scenarios.
The VoLTE voice quality optimization solution of the OMStar can be deployed to calculate
KPIs (such as the number of times uplink/downlink VQIs are poor and packet loss rates),
identify cells with poor voice quality from different dimensions, and analyze the air interface
and resource capacity problems that exist on the network. (In the self-adaptive scenario, users
can customize the KPI calculation formulas.) The table below lists the data that should be
imported to the OMStar.
The OMStar associates internal and external CHRs and traffic measurement data, and output
the evaluation and analysis reports for Top N cells. The evaluation report also contains the
evaluation result of the poor VoLTE voice quality problem, basic auditing result, problem
demarcation result, problem locating resulting, and optimization suggestions.
Figure 1-1 Network assessment problem analyzing and resolving process of VoLTE voice quality
The SEQ analyzes low-MOS subscribers by CDR, demarcates problems, and then determines where
problems occur:
1. If problems are demarcated as problems below the S1 interface, problems occurs in the cell.
2. If problems are demarcated as EPC problems, problems occur on the gateway.
3. If problems are demarcated as IMS problems, problems occur the SBC.
4. If codec problems are demarcated, demarcate problems on the IMS.
Is the CN Nastar No
The problem analysis is deployed?
Does Discovery
the same as that without
PS exist?
the SEQ deployed.
Yes
Yes
Does the SEQ collect and Analyze the SBC according to CN Nastar obtains the CHR and MR data of the corresponding cell
to locate and analyze CDRs indicating poor quality (the capability of Collect external and internal
aggregate CDRs indicating poor Yes Is the CN Nastar deployed? Yes the assessment procedure for using internal CHRs for deep problem location on the RAN side is CHRs, traffic statistics, alarms,
voice quality to a top SBC? the network without the SEQ. Associate the Discovery PS CDRs with provided in 2016 September): and xml and use OMStar to
SEQ CDRs (to be implemented). 1. Exception events (such as, reestablishment, handover failures, demarcate problems and deeply
and frequent handovers) locate problems on the RAN side
2. Packet loss or jitter due to causes above the S1 interface (for details, see the optimization
3. Packet loss due to poor quality over the air interface guide to traffic statistics voice
4. Packet loss due to poor quality inside the base station quality).
No 5. Whether the problem occurs on the top UE
Ask forntline engineers
for help to conduct a
Submit a trouble ticket to the test again and collect
GTAC. CellDT tracing data on
Are the
The SEQ assessment result shows The SEQ assessment result shows that the problems located No sites, and submit the
that the packet loss rate above the packet loss rate above the S1-U interface in and resolved? data to the second line
Existence of exception events problem handling
S1-U interface in the uplink is high. the downlink is high.
engineers to locate
causes for poor-voice-
Based on the SEQ and Discovery PS CDRs, Discovery PS Yes quality cells.
calculates the following: Packet loss due to poor quality over the air interface
a. Total number of packets on the right of the S1-U interface Based on the SEQ and Discovery PS CDRs, Discovery
b. Number of packets lost between the S1-U probe and P- PS calculates the following:
Drill down and analyze the IP
GW a. Total number of packets on the right of the S1-U End
Does the SEQ collect and c. Number of packets lost on the right of the P-GW, including interface
Yes Is the ping operation normal? Yes Is the CN Nastar deployed? Yes address status according to the the P-GW b. Number of packets lost between the S1-U probe and P- Packet loss due to problems inside the base station
aggregate CDRs indicating d. Number of packets lost on the right of the Gm interface
assessment procedure for the GW, including the P-GW
poor voice quality to a top Discovery PS also calculates the range where packet loss c. Number of packets lost on the right of the P-GW
network without the SEQ. occurs.
IP address? b/a > 0.4%: Packet loss occurs between the S1-U interface
b/a > 0.4%: Packet loss occurs between the S1-U
interface and the P-GW, including the S-GW and P-GW.
and the P-GW, including the P-GW. You need to check You need to check packet loss on the S-GW or P-GW.
packet loss on the P-GW. c/(b –a) > 0.4%: Packet loss occurs on the right of the P-
c/(b –a) > 0.4%: Packet loss occurs on the right of the P- Top terminals: Obtain the terminal type and contact the
No No GW, excluding the P-GW. You need to check packet loss
GW, including the S-GW. You need to check packet loss on between the P-GW and S-BC. terminal vendor to further analyze problems.
the S-GW. Demarcation based on the number of lost packets
(c –d)/(b –a) > 0.4%: Packet loss occurs between the P- depends on the SEQ results (time and subscriber range),
GW and Gm interface, including the P-GW. You need to and demarcation based on the packet loss rate also exists
check packet loss on the P-GW. (to be verified).
Demarcation based on the number of lost packets depends
Submit the problem to R&D on the SEQ analysis (time and subscriber range), and
Check the transmission status. demarcation based on the packet loss rate also exists (to be
No engineers. verified).
Is the problem a codec Does the SEQ signaling Does the SEQ signaling Is the high transmission
problem? Can the SEQ Yes and tracing data analysis result show Yes and tracing data analysis Yes codec rate function enabled on Enable the high transmission
result show that codec No codec rate function on the UMG
demarcate the problem? that the IM-MGW is involved? conversion is performed the UMG and SBC?
on the SBC or UMG? and SBC.
No No No Yes
Check the uplink codec status Coordinate with the CS side Check whether the codec rate
End in the uplink on the RAN side of engineers to handle the is adjusted downward on the
the peer end. problem. RAN side of the local end.
Figure 1-2 User complaint and DT problem analyzing and resolving procedure of VoLTE voice
quality
Subscriber complaint scenarios (mobile
number, time, and location)
1. Check whether exception events (such as, reestablishment, Does the problem occurs above the
Yes
handover failures, and frequent handovers) exist. Gm interface?
2. Packet loss due to air interface quality
3. Packet loss due to poor quality inside the base station or the
terminal problem
Yes
Yes
End
End
The method is as follows: Use the test data of the SEQ Analyst and terminal to assess voice
quality and check packet loss, latency, and jitter problems. Problem isolation and demarcation
depend on the SEQ Analyst platform, and further problem location depends on log statistics,
traffic statistics, and tracing data of the IMS, gateway, and eNodeB. If the SEQ Analyst
platform is unavailable, create an eNodeB tracing task to isolate problems.
The SEQ Analyst platform is used to detect subscribers' CDR-level MOS and assess and
present the MOS. The platform also determines whether problems occur above or below an
interface for each interface (such as, S1-U and Gm interfaces) by analyzing RTCP and RTP
packet loss, latency, and jitter. Observe each interface to preliminarily isolate the NE range
where problems occur, as shown in the following figure.
Yes
End
Packet loss rate exception Packet loss rate Packet loss rate
between the UE and S-GW exception of the S- exception between the
GW or P-GW P-GW and SBC/UE
No
Step 1 Parameter demarcation: Check whether the E2E packet loss rate (UL Packet Loss Rate)
exceeds the threshold (the default value is 1%).
If it does, go to Step 2.
If it does not, the problem is not located.
Step 2 Check whether the uplink packet loss rate between the UE and P-GW exceeds the threshold.
If it does, perform operations in "Packet Loss Rate Exception Between the UE and
P-GW, and go to Step 3.
If it does not, go to Step 4.
Step 3 Obtain the voice service APN used by the carrier and check whether the uplink packet loss
rate of the S-GW or P-GW exceeds the threshold.
If it does, the packet loss rate of the S-GW or P-GW is abnormal.
If it does not, the packet loss rate between the UE and S-GW is abnormal, and you need
to perform operations in "Packet Loss Rate Exception Between the UE and S-GW".
The uplink packet loss rate exception is assessed based on the following UGW9811
counters:
− 136316009 SGW uplink user traffic discarded in packets (APN)/136316001 SGW
incoming uplink user traffic in packets (APN)
− 138413089 PGW uplink user traffic discarded in packets (APN)/138413066 PGW
incoming uplink user traffic in packets (APN)
Step 4 Obtain the voice service APN used by the carrier and check whether the uplink packet loss
rate of the S-GW or P-GW exceeds the threshold.
Issue 01 (2017-02-20) Huawei Proprietary and Confidential 14
Copyright © Huawei Technologies Co., Ltd.
Optimization Guide to VoLTE Voice Quality 1 VoLTE MOS Optimization Guide
If it does, the packet loss rate of the UGW9811 is abnormal, and you need to perform
operations in "Packet Loss Rate Exception of the S-GW or P-GW".
If it does not, the packet loss rate between the UGW9811 and SBC/UE is abnormal, and
you need to perform operations in "Packet Loss Rate Exception Between the UGW9811
and SBC/UE".
The uplink packet loss rate exception is assessed based on the KPIs in Step 3.
----End
NOTE
IPPM applies only to Huawei eNodeBs. If the peer NE is not a Huawei eNodeB, the problem cannot be
demarcated.
Yes
End
No No
Segmented demarcation: Packet l oss rate Segmented demarcation: Segmented demarcation: Packet l oss
excepti on betw een the UE and S-GW Packet l oss rate excepti on rate excepti on betw een the P-GW
of the S-GW or P-GW and SBC/UE
No
Step 5 Parameter demarcation: Check whether the E2E packet loss rate (DL Packet Loss Rate)
exceeds the threshold (the default value is 1%).
If it does, go to Step 6.
If it does not, the problem is not located.
Step 6 Check whether the downlink packet loss rate between the SBC/UE and P-GW exceeds the
threshold.
If it does, perform operations in "Packet Loss Rate Exception Between the SBC/UE and
P-GW".
If it does not, go to Step 7.
Step 7 Obtain the voice service APN used by the carrier and check whether the downlink packet loss
rate of the S-GW or P-GW exceeds the threshold.
If it does, the packet loss rate of the S-GW or P-GW is abnormal.
If it does not, the packet loss rate between the S-GW and UE is abnormal, and you need
to perform operations in "Packet Loss Rate Exception Between the S-GW and UE".
The downlink packet loss rate exception is assessed based on the following UGW9811
counters:
NOTE
IPPM applies only to Huawei eNodeBs. If the peer NE is not a Huawei eNodeB, the problem cannot be
demarcated.
1. Is an alarm related to the media Yes The SBC clears the alarm related
plane generated on the SBC? to the media plane.
Yes
Y
Step 1 Check whether an alarm related to the media plane in section Error! Reference source not
found."Error! Reference source not found." is generated on the SBC.
Step 2 Check whether the average packet loss rate of "ABCF Access Network IP QoS Measurement"
on the SBC is greater than 1%.
If the rate is greater than 1%, you need to coordinate the RAN side and bearer network to
exclude the packet loss cause.
Step 3 Check whether the RTT of "ABCF Access Network IP QoS Measurement" on the SBC is
longer than 100 ms.
If the rate is longer than 100 ms, you need to coordinate the RAN side and bearer network to
exclude the packet loss cause.
Step 4 If the solution mapping for the IMS side on the live network is VoLTE 3.2, and CN Nastar 2.0
of CCE is deployed on the live network, use CCE to identify the locations with low MOS and
coordinate the RAN side to check the air interface quality of the corresponding cell.
----End
Test software
statistics problems
Changi ng the
terminal to check
whether problems
exist
No
Step 1 Check whether the test software is commissioned and optimized, whether the overall
maximum MOS value is small, and whether the software statistics are normal.
If problems are solved after the DT software is replaced, the DT software causes the problems
that must be resolved by the software vendor.
Step 2 Isolate problems to check whether problems occur on the terminal. Change the test device or
the test area to check whether problems are resolved. If the problems are resolved after the
test device is changed, problems occur on the terminal. For terminal problems, check the
aspects, such as terminal software version and terminal capability. If these aspects cannot be
checked, contact the terminal vendor.
Step 3 Isolate air interface problems and analyze the DT data to check events, such as the RSRP,
SINR, interference, and exception.
If the threshold conditions are not met, optimize the air interface.
Step 4 Check the eNodeB status, fault alarm information, eNodeB version, parameter settings in
scenarios without air interface condition exceptions to check whether problems occur on the
eNodeB.
If problems cannot be resolved after the check for all affecting factors is complete, submit a
trouble ticket to resolve the problems.
----End
Network–>UE indicates the downlink RTP packets received by the UE from the
network. If the downlink sequence number is inconsecutive, packet loss occurs on the
network side. You need to simultaneously check the eNodeB and NEs above the
eNodeB to determine the packet loss position. UE–>Network indicates the uplink
RTP packets received by the network from the UE. If the uplink sequence number is
inconsecutive, problems occur on the UE.
− Latency: The E2E latency is one of the most important factors affecting the quality of
interactive voice communication. A latency may cause a situation that the two parties
speak at the same time because one of them cannot hear the voice of another party in
time. This situation affects user experience, and therefore, the E2E latency must be
limited to a proper range. Experience indicates that when the E2E latency is in the
Issue 01 (2017-02-20) Huawei Proprietary and Confidential 20
Copyright © Huawei Technologies Co., Ltd.
Optimization Guide to VoLTE Voice Quality 1 VoLTE MOS Optimization Guide
range of 175 ms to 200 ms, the MOS jitters and decreases severely, and the MOS
jitter and decrease continue as the subsequent latency increases.
− Jitter: is also defined as latency variation. Jitter refers to the difference in the arrival
time of all data packets during an IP call. When sending a data packet, the sender
adds a time stamp to the RTP packet header. The receiver adds another time stamp
upon receiving the data packet. The transmission duration of this packet can be
determined by calculating the two time stamp difference.
Using the CellDT data to isolate and analyze upstream and downstream eNodeB data
− Packet loss: The TTI tracing (CellDT 138 tracing) of the eNodeB is deployed at the
PDCP layer on the eNodeB side and can detect the downlink packet loss caused by
the CN (S1 link).
In the following figure, the TTI is from 4486447 to 4492787, there are 207 packets
lost in the data received by the eNodeB from the CN (S1 link), and the RTP SN is
from 0x057a to 0x0649.
− Latency: The VoLTE voice service is performed between UE 1 (CRNTI is 7183) and
UE 2 (CRNTI is 6058). The difference between the time when an uplink packet of
UE 1 is sent to the CN and the time when UE 2 receives the packet from the CN to
check the CN latency. In the following figure, the data packet with the RTP SN being
0x88e5 is sent to the CN on the source side with the uplink TTI being 209191805 and
is received on the target side with the downlink TTI being 209191809. The duration
of sending and receiving the packet shows that the CN latency is only 4 ms.
Jitter
Interval
− Jitter: CellDT 138 tracing can also determine whether jitter occurs when the eNodeB
receives packets from the CN (S1 link). For example, the following figure shows that
slight jitter occurs when the eNodeB receives packets from the CN.
VoLTE is deployed on the XX site. A network-wide DT is conducted. The test result indicates
that the number of calls re-initiated by UEs is large and UEs need to be re-connected to the
network after a large number of call re-initiation failures.
[Handling process]
Analyze signaling as follows:
Traced signaling indicates that the UE consistently reports the A3 event, but the eNodeB does
not send a handover command, as shown in the figure below.
The UE initiates a call again. The cell where the call is re-initiated does not have the context.
Therefore, the re-initiated call fails.
After the call re-initiation failure, the UE attempts to access the network again. From
22:00:31.406 when the UE starts to search the system message to 22:00:31.878 when the UE
is finally connected to the cell, there is an interval of 470 ms during which voice packets
cannot be processed.
Issue 01 (2017-02-20) Huawei Proprietary and Confidential 23
Copyright © Huawei Technologies Co., Ltd.
Optimization Guide to VoLTE Voice Quality 1 VoLTE MOS Optimization Guide
[Root cause]
Some adjacent cells of the eNodeB are not configured. Consequently, the handover fails and
the call is re-initiated.
The problem of missing adjacent cell configuration does not affect the intra-frequency and
inter-frequency handover success rates in the system. It has minor impacts on KPIs related to
the data service, such as the handover success rate and throughput, but great impacts on KPIs
of VoLTE voice service, such as the delay, jitter, and packet loss rate. The problem of missing
adjacent cell configuration seriously affects the voice quality. About 20% of VoLTE voice
service problems detected in Hangzhou are caused by missing adjacent cell configuration.
[Solution]
The VoLTE voice service raises stricter requirements for the adjacent cell configuration.
Before deploying the VoLTE voice service, check the adjacent cell configuration on the entire
network.
If the RSRP is good, the MOS does not decrease significantly when a handover occurs.
Therefore, a handover does not necessarily cause a sharp decrease of the MOS. If a handover
occurs in an area where the RSRP is good, the MOS decreases by 0.1 to 0.5. If a ping-pong
handover occurs, the MOS decreases by 0.5 to 1.5.
[Root cause]
A handover that occurs during the MOS test affects the MOS. A common handover does not
seriously affect the MOS, but a ping-pong handover may cause sharp decrease of the MOS.
[Solution]
1. Adjust RF parameters.
2. Adjust the CIO between two cells.
result, the bit error rate increases, and the MOS decreases to about 1. The MOS decreases by
0.5 to 1.5.
The UE consistently sends the MR, but the handover cannot be implemented, but the eNodeB
does not send any handover command because no adjacent cell exists.
[Root cause]
The cell with super-distance coverage does not have any adjacent cell. When the wireless
signal of the UE is poor, the UE initiates a call request again, and then is re-connected to the
network.
[Optimization suggestion]
Method 1: Adjust RF parameters to resolve the overshoot coverage problem.
Method 2: If method 1 does not work, reduce the power.
Method 3: If the preceding two methods cannot resolve the problem, it is recommended that
adjacent cells be configured for the overshoot coverage area.
[Conclusion]
The MOS decreases by 0.4 to 1.0 due to MOD3 interference.
[Optimization suggestion]
Adjust the PCI to avoid MOD3 interference between cells with the same remainder after the
PCI is divided by 3.
The POLQA algorithm is used for VoLTE voice quality evaluation, and applies to WB
services. The 2G services are NB services. Therefore, it is inappropriate to use this algorithm
to evaluate the 2G MOS, and the MOS will drop by 1.5 to 2.0 if this algorithm is used.
[Conclusion]
it is inappropriate to use the POLQA algorithm to evaluate the 2G MOS.
[Optimization suggestion]
1. Optimize the VoLTE coverage, and adjust the eSRVCC threshold.
2. After the UE is handed over to the 2G network due to the weak coverage, an appropriate
MOS evaluation algorithm can be automatically selected.
[Problem analysis]
1. The signal is poor. Therefore, the UE sends the MR, and the eNodeB sends the
inter-frequency 37900 measurement control.
The frequency 37900 adopts the A3-based inter-frequency handover, which is triggered when
the sum of offset and HY3 is greater than or equal to 4 dB. The sum of offset and HY3 in the
MR sent by the UE is 3 dB, which is smaller than the threshold 4 dB. Therefore, the UE
cannot be handed over to the frequency 37900.
2. The signal is poor. Therefore, the UE sends the MR, and the eNodeB sends the
inter-frequency 39250 measurement control.
A4-based inter-frequency handover is adopted between the macro base station and indoor
distributed base station. The A4 threshold is -101 dBm. It is easy to meet this condition due to
leakage of the indoor distributed base station. After the UE is handed over to the indoor
distributed base station, the signal remains poor. The threshold for a handover from the indoor
distributed base station to the macro base station is -105 dBm, which is relatively low.
Therefore, it is difficult to trigger a handover of the UE from the indoor distributed base
station to the macro base station. In mobile state, the signal quickly attenuates. When the
basic signal strength is about -120 dBm, the UE is handed over to cell 272.
[Conclusion]
The UE cannot be handed over to an appropriate cell due to leakage of the indoor distributed
base station.
[Optimization suggestion]
1. Adjust the CIO of the indoor distributed base station so that it is difficult for a UE to be
handed over to a cell of the indoor distributed base station when the UE is located on a road.
Issue 01 (2017-02-20) Huawei Proprietary and Confidential 32
Copyright © Huawei Technologies Co., Ltd.
Optimization Guide to VoLTE Voice Quality 1 VoLTE MOS Optimization Guide
2. Adjust the A4 threshold for handover from the macro base station to the indoor distributed
base station. It is recommended that the A4 threshold be set to -95 dBm. In the case of leakage
of the indoor distributed base station, this setting prevents the UE from being handed over to a
cell of the indoor distributed base station when the UE is located on a road.
In addition, the RTP packet loss rate is not closely associated with the weak coverage. The
average RTP packet loss rate is 0.42%, which is lower than the protocol requirement of 1%.
35 0
30 -20
25
-40
20
-60
15 CRS SINR
-80 CRS RSRP
10
-100
5
0 -120
0 10 20 30 40 50 60 70
-5 -140
rtp plr(%)
When the MOS audio file of grid 66 is played back, the background electronic noise is
detected. The following attachment is the audio file recorded when the MOS is 3.96.
According to the fixed-point test result on the HL-2 site, the average RTP packet loss rate is
0.01%, and the average MOS is 3.976. The MOS is normal.
Comparison between the network-wide DT and the second MOS test indicates that the
average MOS in the second MOS test reaches 3.85, the RTP packet loss rate is 0.26%, and the
highest MOS is 4.18. This proves that the test equipment is not properly debugged in the
previous CDS test.
Log
Average MOS
[Root cause]
Analysis on the MOSs of grid 66 indicates that the MOS test result is closely associated with
debugging of the test equipment. In the subsequent MOS tests, the recommended CDS
configuration method must be used to prevent the inaccurate MOS test results.
[Solution]
Before the MOS test, debug the test equipment at fixed place to ensure that the highest MOS
reaches 4.1. If the MOS is extremely low, check whether the UE volume and attenuation of
the CDS equipment are inappropriately configured and whether the vehicle-mounted power
supply is not properly connected.
1. Connect the MOS box to the USB Hub connector during the test.
2. Set the volume of the UE receiver to the second largest volume.
3. Connect the HTC UE.
In the DT of a site, when HL-4 located in Xintuanjiehu, Chaoyang is occupied, the MOS is 4
and no packet is lost. When HL-6 located in Changhongqiao, Chaoyang is occupied, the MOS
is 1.12 and the packet loss rate is 80%. The MOS and packet loss rate recover only after the
UE is handed over to HL-3 located in Shunfengjiulou, Chaoyang. The problem lasts about 2
minutes, from 10:11:09.161 to 10:13:01.512.
After the UE is handed over from HL-6 to HL-3, the RTP packet loss rate and BER become
normal again.
[Problem analysis]
The eNodeB status and alarms are checked, and no exception is found. KPIs of the cell
indicate that the average number of subscribers in the cell during the day time exceeds 300,
the uplink RBLER is relatively high, and the uplink QCI1 PDCP packet loss rate (20%) is
also high.
The eNodeB data is traced and the uplink scheduling information is analyzed. 41484 indicates
the uplink scheduling failure caused by insufficient CCE resource. The proportion of 41484 is
as high as 12%. It is preliminarily determined that the uplink CCE resource is insufficient due
to the large number of subscribers in the cell. Consequently, uplink scheduling fails.
To further obtain the VoLTE subscriber tracing data, a long call test is conducted in the early
morning. The tracing result indicates that the uplink PDCP SNs of the eNodeB are continuous
and no VoLTE packet is lost.
It can be concluded that the uplink CCE resource of the site is insufficient due to the heavy
traffic. Consequently, the uplink RBLER and VoLTE packet loss rate are high. Upon VoLTE
SR and data service SR, the uplink CCE resource is insufficient. The site does not allocate the
uplink CCE resource preferentially to VoLTE SR. In the eRAN 8.1, VoLTE SR takes
precedence over data service SR, but related configuration parameters are not available on
interface yet. The related MML command is as follows:
MOD eNBCellRsvdPara: LocalCellId=1-6, RsvdSwPara3=RsvdSwPara3_bit1-1;
When a second test is conducted after this command is executed, the QCI1 packet loss rate
becomes normal again.
In addition, more uplink CCEs can be added to resolve the problem of insufficient uplink CE
resource. The following command can be executed to increase the ratio of uplink CCEs to
downlink CCEs:
MOD ENBCELLRSVDPARA:LocalCellId=3,RsvdPara52=10;
This command indicates the ratio of uplink CCEs to downlink CCEs of subframes 3 and 8 is
changed to 10:1 to ensure that CCEs used for uplink scheduling are sufficient.
[Root cause]
The number of subscribers of the site is 200, which is large. Upon VoLTE SR and data service
SR, the uplink CCE resource is insufficient. The site does not allocate the uplink CCE
resource preferentially to VoLTE SR.
[Solution]
This problem affects sites with heavy traffic. Two solutions are available:
Solution 1
Run the following command on the entire network to adjust the scheduling optimization
parameters, ensuring that the priority of the VoLTE SR is higher than that of the data
service SR:
MOD eNBCellRsvdPara: LocalCellId=1-6, RsvdSwPara3=RsvdSwPara3_bit1-1;
Solution 2
Run the following command to change the ratio of uplink CCEs to downlink CCEs of
subframes 3 and 8 to 10:1 to ensure that CCEs used for uplink scheduling are sufficient:
MOD ENBCELLRSVDPARA:LocalCellId=3,RsvdPara52=10;
[Problem analysis]
1. The voice quality deteriorates because the uplink packet loss rate increases. Therefore,
check the quality of the air interface. The result indicates that the uplink interference does not
increase and the BER on the air interface is normal. This means that increase of the packet
loss rate is not caused by quality of the air interface.
As shown in the preceding figure, the IBER and RBLER do not increase when the QCI1
packet loss rate increases.
2. Capture and analyze packets on the UGW.
Capture packets to check reception of voice packets.
The single subscriber tracing result on the UGW indicates that the interval at which VoIP
packets are received sequentially from the eNodeB side is not always 20 ms. At some time
points, multiple RTP packets are concurrently received.
In the downlink direction, packets are sent at a regular interval, as shown in the figure below.
The UE BSR fluctuates seriously when packets are not sent continuously, or a large BSR is
reported continuously. On the UE side, second scheduling may be abnormal or the BSR may
not be reported properly.
The BSR retransmission timer (retxBSR-Timer) expires and the UE has data available for
transmission on a certain logical channel.
The retxBSR-Timer helps the UE to jump out of the deadlock. For example, the UE sends 0
BSR and then other BSRs other than 0 BSR, but the sequence in which the eNodeB receives
BSRs may be reversed. That is, sometimes that eNodeB may receive 0 BSR last. Therefore,
the eNodeB determines that the UE does not have data to send and no longer implements UE
scheduling. If the UE always does not have data with higher priority, the UE cannot generate
a regular BSR or send the SR to the eNodeB. Consequently, the UE data always cannot be
sent. In this case, the UE can generate a regular BSR by starting the retxBSR-Timer.
The periodic BSR timer (periodicBSR-Timer) enables the UE to trigger the BSR in time and
helps the eNodeB with scheduling. The priority of the BSR is higher than that of common
subscriber data. Therefore, when the UL Grant is available, the BSR generally can be sent.
BSR triggering conditions are independent of each other. Therefore, multiple triggering
conditions can be met at the same time. Triggering a BSR is not equivalent of generating a
BSR. A BSR is generated upon MAC packet assembly. One MAC PDU can contain at most
one BSR.
Based on the preceding protocol description and data analysis, the following problem
triggering scenario can be inferred:
The uplink resource is not requested in time for RTP3 packets when the following two
operations are performed concurrently: (1) The UE is going to send RTP2 packets to the
eNodeB; (2) RTP3 packets are being added to the buffer queue at the RLC layer.
From the eNodeB point of view, in the preceding low-probability triggering scenario, the RLC
buffer on the UE side changes from empty to nonempty, and theoretically, the regular BSR
should be triggered according to the protocol. However, the UE determines that its buffer still
has data and no data that belongs to a logical channel with higher priority needs to be
transmitted. Therefore, the UE does not generate a regular BSR, and instead generates a
regular BSR after the periodic BSR timer expires. After this regular BSR is generated, the
BSR is not assembled to the MAC PDU of the RTP2 packet due to abnormal internal
processing of the UE. Just in the subsequent period of time, the eNodeB does not have extra
pre-scheduling resource that can be allocated to the UE. Consequently, the BSR cannot be
reported, and the RTP3 packet is dropped because it stays in the RLC buffer queue for more
than 100 ms.
According to the PDCP layer tracing result when No.138 tracing mode is used, some uplink
RTP packets arrive simultaneously at some time points. As shown in the figure below, packets
numbered 483 to 487 arrive at the same time TTI.
BSRs are classified into short BSRs and long BSRs. The following provides the BSR
reporting information.
//Tracing item 9 is a short BSR, where the data0 column is displayed in an accumulative
manner. Tracing item 10 is a long BSR, where the data0 column is displayed in an
overwritten manner.
Information about reporting of the short BSR is as follows:
Frm=1007/SubFrm=2; Frm=1022/SubFrm=2
From "Frm=1007/SubFrm=2" to "Frm=1022/SubFrm=2", there is a period of about 160 ms,
during which the MCE of the short BSR is not received.
The eNodeB detects that the UE sends the SR at "Frm=1008/SubFrm =7". The eNodeB sends
the UL Grant to the UE at the subframe "Frm=1009/SubFrm=3". The UE sends part of the
data at the subframe "Frm=1009/SubFrm=7", and sends the remaining BSR (a periodic BSR,
the period of which is set to 10 ms by default) in the RLC buffer. Based on the BSR reported
by the UE, the eNodeB sends the UL Grant to the UE at the subframe "Frm=1010/SubFrm=3".
The UE sends the PUSCH data at the subframe "Frm=1010/SubFrm=7". Theoretically, the UE
should send the BSR of the data that is newly added to the RLC buffer and should be
transmitted (at this time, the periodic BSR expires). However, the UE does not generate the
BSR in time due to abnormal processing of the periodic BSR timer on the UE. Therefore, the
BSR record does not exist at the subframe "Frm=1010/SubFrm=7".
At this time, the eNodeB determines that the UE does not have any data for scheduling. The
UE cannot trigger the regular BSR because the BSR is not 0, and determines that the eNodeB
needs to continue scheduling. Consequently, the UE can trigger the BSR only after the
retxBSR-Timer expires. In the preceding example, before the retxBSR-Timer expires, the
uplink resource is generated due to pre-scheduling so that uplink scheduling can continue, and
the uplink data transmission becomes normal again. However, the intermediate period
exceeds the duration of the packet loss timer (100 ms). Therefore, some packets are lost.
[Root cause]
In some scenarios, the UE may fail to report the BSR in time due to abnormal processing of
the periodic BSR. As a result, uplink scheduling cannot be implemented in time, causing loss
of uplink packets. The UE chip supplier, XX, admits that their UEs have bugs in BSR
generation.
[Solution]
To prevent this problem, run the following MML commands to increase the threshold of the
pre-scheduling resource ratio and enable smart pre-scheduling:
MOD CELLULSCHALGO: LOCALCELLID=XXX, PREALLOCATIONBANDWIDTHRATIO=50;