See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/308848233

VoLTE SRVCC optimization as interim solution for LTE networks with coverage
discontinuity

Conference Paper · October 2015

DOI: 10.1109/ICTC.2015.7354531

CITATIONS READS
3 2,154

6 authors, including:

Myleen Villaluz
Huawei Technologies
3 PUBLICATIONS   5 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Underwater communications View project

All content following this page was uploaded by Myleen Villaluz on 14 February 2018.

The user has requested enhancement of the downloaded file.

 VoLTE SRVCC Optimization as Interim Solution
for LTE Networks with Coverage Discontinuity
Myleen Dosado Villaluz , Ragil Putro Wicaksono, Adrian Dan Ebora Atienza, Seiji Kunishige, Kwangrok Chang,
Jennylou Banzon Caasi
MOTiV Research Co.
Tokyo, Japan
firstname.lastname@motiv-research.com

Abstract—This paper is focused on the mobile networks with during VoLTE is the main topic of this paper. It is presented
deployment of LTE offering Voice over LTE (VoLTE) services similar suggestion to optimize the SRVCC thresholds [4]
and existing WCDMA layer – in which voice call continuity is however, the method to obtain the apt thresholds are not
ensured through the Single Radio Voice Call Continuity discussed.
(SRVCC) feature. SRVCC feature enables handover of voice call
in LTE to WCDMA layer when users enter poor LTE coverage It is said that LTE layer is designed to offer similar or better
area. In this paper, it is envisaged that quality of call measured voice quality than existing CS network layers [5]. The voice
through Mean Opinion Score (MOS) is always better in LTE call quality in LTE is higher than in 2G or 3G system. Hence,
compared with WCDMA layer. Due to better call quality, the optimization policy in this paper is motivated by the
maximizing the voice call duration in LTE layer is recommended strategy of leveraging LTE layer for voice services since it can
through SRVCC serving cell threshold optimization. In order to offer better call quality and minimize the drop call rate in LTE
properly set the optimal SRVCC threshold setting, radio through SRVCC. The aim is to identify the minimum serving
condition at which uplink block error rate (UL BLER) increasing cell condition to be set as threshold with consideration on voice
higher than 10% and the acknowledgement rate (ACK Rate) call quality and transmission performance.
degrading below 90% is evaluated. The Transmission Time
Interval (TTI) bundling feature improving the uplink VoLTE It is also considered in this paper a network with and
coverage is also taken into consideration. This paper proposes without deployment of TTI bundling feature together with
serving cell SRVCC threshold RSRP = -108dBm for networks VoLTE. TTI bundling is known to effectively enhance the
without TTI bundling and RSRP = -117dBm for networks with uplink coverage performance for LTE Frequency Division
TTI bundling deployment. Duplexing (FDD) [6]. In SRVCC threshold evaluation with
and without aid of TTI Bundling, uplink performance is taken
Key words—Voice over LTE (VoLTE); Single Radio Voice Call into consideration through Physical Uplink Shared Channel
Continuity (SRVCC); MOS, TTI bundling; mobility optimization (PUSCH) BLER and ACK rate.
I. INTRODUCTION The outline of this paper is given as follows. Section II
Call continuity is essential for networks offering voice call provides the underlying principles for SRVCC optimization
services. Voice call set up in Long Term Evolution (LTE) which includes working mechanism of SRVCC, overview of
network via VoLTE is purely Internet Protocol (IP) based. call quality, uplink transmission performance and TTI
Hence, when VoLTE user enters poor LTE coverage area, Bundling principles. Methods and results are presented on
handover to existing inter-RAT layer is necessary to continue Section III – which contains the call quality evaluation result
the voice call. Several voice call continuity solutions are through POLQA MOS evaluation, uplink transmission
available in LTE layer such as Circuit-switched Fallback evaluation through UL BLER and ACK rate and SRVCC
(CSFB), SRVCC, Voice over LTE via Generic Access serving cell threshold decision. Field data used in this paper is
(VoLGA) and other third party application/services [1]. This obtained in Tokyo Metropolitan area wherein LTE
paper focuses on SRVCC— defined by 3GPP as a voice call discontinuity is observed. Finally, paper conclusion is
continuity between IP Multimedia Core Network (IMS) over presented in Section IV.
Packet Switched (PS) access and Circuit Switched (CS) access
II. UNDERLYING PRINCIPLES FOR SRVCC OPTIMIZATION
layers for VoLTE [2]. It is a practical solution when UE moves
out of LTE coverage area – the possibility of dropped call due A. SRVCC Working Mechanism
to poor coverage is reduced with handover to an existing inter- SRVCC is a voice call continuity feature that enables users
RAT target cell. The handover procedure from LTE layer to with VoIP services to perform smooth handover from E-UTRA
circuit-switched (CS) inter-Radio Access Technology (RAT) layer to inter-RAT layer. This feature is essential for networks
layer is supported with the help of IP Multimedia Subsystem which still have LTE coverage discontinuity and offers VoLTE
(IMS) network [3]. service to ensure call continuity. Practically, SRVCC
The SRVCC procedure is mainly subdivided into two parts: thresholds are configured where LTE coverage is poor such as
radio condition part and the core network part. Radio condition to ensure better call quality using the LTE layer.

978-1-4673-7116-2/15/$31.00 ©2015 IEEE 212 ICTC 2015

 condition. The event B1 measurement report contains the target
inter-RAT cell’s condition (i.e., RSCP and EcNo for WCDMA
layer). On the other hand, the event B2 measurement report
contains serving LTE cell condition and target inter-RAT cell
condition. Depending on the operator, SRVCC can be set to
either B1 or B2. The capability to easily perform inter-RAT
handover is also dependent on the target cell threshold. Based
on these mobility procedures, this paper provides optimization
principle on how to optimize the SRVCC serving cell RSRP
threshold to leverage LTE layer yet considering minimal
dropped calls. Basis of SRVCC threshold optimization is
studied in this paper – call quality and BLER are taken into
account.
MME receives handover request from E-UTRAN to
Fig. 1. Measurement reporting flow for SRVCC UTRAN indicating that this is for SRVCC handover then
triggers the MSC server to initiate handover procedure. Then,
SRVCC is mainly initiated by measurement reporting by the session transfer procedure is initiated by the IMS server and
the User Equipment (UE) for target neighbor cells as shown in coordinates with the target UTRAN for CS handover.
Fig. 1. UE’s measurement reports include the Reference Signal Handover response from the MME contains the necessary CS
Received Power (RSRP) and Reference Signal Received HO command information for the UE to successfully access
Quality (RSRQ). For UE that has single radio capability, it the UTRAN layer.
requires measurement gaps which support the mobility
procedures. The eNodeB is responsible for delivering the B. Evaluation of POLQA MOS
measurement gap (MG) information through ITU-T P.863 defines the POLQA as an algorithm for voice
RRCReconfigurationMessage. Depending on the vendor, call quality assessment [8]. MOS value is the rational number
mobility parameters can be set per Quality of Service Class value output for the POLQA algorithm which ranges from
Identifier (QCI). score of 1 to 5. Table 1 below shows the interpretation of MOS
values. The algorithm does not take into consideration the
Measurement Gap (A2):
delay caused by network and transmission, it only compares
•൅›•൏Š”‡•Š (1) the reference input audio signal with the recorded degraded
signal. Hence, it is possible to obtain similar high MOS value
Inter-RAT B1 Entering Condition: at good and poor radio condition – also stated in [8]. Thus, in
Mn+Ofn-Hys>Thresh (2) the data analysis of field tests, average MOS value is evaluated.

Inter-RAT (B2) Entering Condition: TABLE I. POLQA MOS INTRERPRETATION

Ms+Hys<Thresh (3) MOS Value Interpretation

5 Excellent
Mn+Ofn-Hys>Thresh (4)
4 Good
Where: 3 Fair
Ms = Serving cell measurement result 2 Poor
Hys = Hysteresis parameter for each event 1 Bad
Ofn = Frequency-specific offset of the frequency of the neighbor cell
Thresh = Threshold parameter for each event C. UL BLER and PHICH Ack rate
Mn = Neighboring cell measurement result UL BLER refers to the ratio of erroneous blocks with the
total data blocks sent. Its value is directly related to the user’s
UL condition and is essential for voice services to be kept
UE needs to fulfill and send measurement report A2 MG in minimum. The BLER value can be directly related to the
order to receive the inter-RAT B1/B2 parameters. Event A2 overall call quality. 3GPP has set acceptable target BLER
MG reported by the UE contains the serving cell’s RSRP and condition at 10% [9]. Average 10% UL BLER against DL
RSRQ condition. The possibility to perform inter-RAT pathloss condition is used for evaluation in this paper.
handover earlier is dependent on how early the UE enters MG The UL BLER condition is calculated in the eNode-B side.
and fulfills inter-RAT B1/B2. Upon fulfillment of MG defined
Through the PHICH, the acknowledgement rate of UL
in Eq. (1) based on 3GPP [7], UE starts to measure the
transmitted data, UL BLER is known by the UE. PHICH
frequencies as stated in the reconfiguration message delivered
by the eNodeB. It is defined in 3GPP standard the events B1 carriers the HARQ acknowledgements (ACK/NACK) for
and B2 as SRVCC measurement events – B1 entering uplink data transfers. There is no specified error rate for
condition defined in Eq.(2) and B2 entering condition defined PHICH acknowledgment and it is up to network operators to
in Eq. (3) for serving cell condition and Eq.(4) for target cell decide on the value [10]. Hence, UL BLER should ideally be

213
equal to 100 – PHICH ACK rate. Using the 10% threshold for Using TEMS Pocket as testing tool for call quality, MOS
UL BLER, PHICH ACK rate is set to 90%. samples are collected in various RF condition. TEMS
Discovery Device 10.0.6 [12] is used to process collected data
D. TTI Bundling Feature to analyze the radio condition and corresponding call quality.
In normal data transmission in uplink, the first time data is With SRVCC’s working procedure described previously in
transmitted is tagged with redundancy version (RV) 0. The Section II, RSRP and RSRQ being included in the
uplink data transmission is through the PUSCH. The network, measurement report becomes the motivation for comparison
upon reception, will try to decode the PUSCH data and send with MOS value. The RSRP and RSRQ conditions in LTE are
ACK/NACK via the Physical Hybrid-ARQ Indicator Channel averaged per MOS sample which is reported at fixed interval of
(PHICH). The ACK/NACK indication is to be received by the 10 seconds. It is previously known that LTE can offer better
UE N+4 TTI from the first data transmission. When the UE call quality than WCDMA layer [5] – this conclusion is
receive NACK indication, it will retransmit the data with verified in this paper as follows.
different RV number. The data is to be retransmitted at N+13 LTE
RSRP Condition (dBm)
to N+16 TTI after the initial data transmission, as shown in -120 -117 -114 -111 -108 -105 -102 -99 -96 -93 -90 -87 -84 -81 -78 -75
-17 1.19
Fig. 2. -16 1.24
TTI bundling is an uplink scheduling feature which helps -15 3.72 2.35 3.76 3.68 3.47
to imp rove the UL coverage gain for real-time voice services

RSRQ Condition (dB)

-14 3.30 3.67 3.90 3.75 4.15 4.09
-13 2.99 3.69 3.66 3.50 3.60 3.95 3.64 3.24
[11]. TTI bundling’s mechanism is to transmit similar data in -12 2.77 3.60 3.47 3.74 2.66 3.51 3.56 3.95 4.01 4.13
four consecutive TTI with different RV as shown in Fig. 2. -11 3.45 4.01 3.35 3.78 3.76 3.67 3.86 3.59 3.81 3.77 3.43 2.27
When UE receive NACK indication from the eNode-B N+4 -10 3.63 3.92 3.68 3.83 3.79 3.76 3.89 3.77 3.85 3.91 3.77 3.74 3.73 3.90
-9 4.12 3.73 4.00 3.30 3.74 3.80 3.71 3.94 3.83 3.75 3.38 3.55 3.73 3.82 3.56
TTI after the last RV of initial transmission, it will retransmit -8 3.94 4.04 3.78 3.85 3.83 3.82 3.87 3.91 3.85 3.73 3.66 3.65 3.85 3.85
at N+13 to N+16 TTI after last RV of initial transmission. -7 4.10 4.16 3.79 3.96 3.82 3.84 3.89 3.72 3.75 3.77 3.77 3.58 3.93 3.64
In high path loss – power limited area, Signal to -6 3.63 3.85 3.63 3.92 3.84 3.85 3.88 3.79 3.63
-5 4.08 4.11
Interference Noise Ratio (SINR) is low, and thus, the decoding
performance is lower. Hence, more retransmission is needed – Fig. 3. VoLTE MOS – RSRP, RSRQ matrix
which in return, contributes to delay affecting the call quality. RSCP Condition (dBm)
WCDMA
With TTI bundling, the retransmission rate will be lower due -120 -117 -114 -111 -108 -105 -102 -99 -96 -93 -90 -87 -84 -81 -78 -75
-17
to a more robust data transmission. TTI bundling helps to -16
reduce transmission delay which is suitable for real-time -15 2.12
-14
services like VoLTE.
Ec/No Condition (dB)

-13
-12
-11 2.17 2.37
-10 2.74
-9 2.38 2.40 2.83 2.21 2.28
-8 2.83 2.60 2.07 2.39 2.56 2.52 2.49
-7 1.82 2.82 2.66 2.31 2.44 2.41 2.42 2.53 2.73
-6 2.66 2.18 2.41 2.49 2.42 2.46 2.48 2.51 2.61
-5 2.48 2.33 2.38 2.48 2.54 2.47 2.47
-4 2.67 2.40 2.51 2.53 2.52 2.41 2.52
-3 2.39 2.36 2.68 2.45 2.60

Fig. 4. WCDMA MOS – RSCP/EcNo matrix

Fig. 2. Normal retransmission versus bundled retransmission

Fig. 3 shows the VoLTE MOS versus RSRP, RSRQ
The activation of TTI bundling is triggered by UE’s uplink condition – it is evident that across a wide range of RF
limited power condition [10] – UE reports power headroom condition, LTE provides MOS > 3 with no obvious trend
(PHR) and PL condition. The UE’s PHR and PL report serves correlation for a wide range of RSRP and RSRQ values. Only
as a triggering indication for the eNode-B to activate TTI until reaching poor RSRP/RSRQ condition, MOS value starts
bundling. Once activated, the UE performs similar data to become unstable – RSRP < -117dBm and RSRQ < -12dB.
WCDMA MOS condition against RSCP and EcNo range is
transmission for four consecutive TTI.
shown in Fig. 4. From these results, the previously cited
III. FIELD TEST RESULTS conclusion from [5] that LTE can offer better call quality is
verified.
Data collection for MOS evaluation and uplink
transmission performance is done in Tokyo Metropolitan area
wherein discontinuity in LTE coverage is evident. The results
evaluation is divided into call quality and uplink transmission
analysis.
A. Call Quality Evaluation through POLQA MOS
In this paper, MOS condition during VoLTE is obtained
and compared with voice call done in the WCDMA layer.

214
 LTE/WCDMA Average MOS at Varying DL Pathloss Condition UL BLER/ACK Rate - DL Pathloss (dB)
-60 5 100 100
90 95
-70 4.5
80 90
Average RSRP/RSCP (dBm)

PHICH ACK Rate %

4 70 85

PUSCH BLER %
-80
3.5 60 80

Average MOS
(UL BLER)
-90
50 (ACK RATE) 75
3
-100 40 Poly. ((UL BLER)) 70
2.5
30 65
-110
2 20 11 12 11
60
10 11 10
7 8
-120 1.5
10 4 2 2 1 2 2 2 2 3 3 4 5 5 55
0
0 50
-130 1

80
81
84
87
90
93
96
99
102
105
108
111
114
117
120
123
126
129
132
135
138
141
144
87 90 93 96 99 102 105 108 111 114 117 120 123 126 129 132 135 138 141 144
DL Pathloss (dB)
Pathloss (dB)
LTE MOS WCDMA MOS Avg RSRP Avg RSCP
Fig. 6. PUSCH BLER, PHICH ACK rate versus DL PL condition
Fig. 5. LTE, WCDMA average MOS comparison against DL pathloss
In Fig. 6, PUSCH BLER condition start to reach 10%
LTE MOS is higher than WCDMA at the same PL threshold at DL PL = 126dB (RSRP = -108dBm) – higher
condition. PL is calculated using formula given in Eq. (5) condition compared with previously mentioned LTE
where RSpower refers to reference signal power and RSRPs is the instability point. RSRP = -108dBm is proposed to be the
UE measured serving cell RSRP based on Fig. (5). It is noted SRVCC serving cell threshold. However, if network has
that in the live network test scenario, RS power for LTE cells is deployment of TTI Bundling feature with VoLTE, it is
set to 18dBm and for WCDMA cells set to 32dBm. Hence, if expected that SRVCC serving cell threshold is lower.
network operators want to maintain call quality on best
condition, the strategy to keep UE to stay longer in LTE layer C. Uplink Performance Improvement by TTI Bundling
is recommendable. From the field tests, TTI bundling starts to be utilized at
DL path loss = RSpower - RSRPs (5) high PL condition where there is not enough power headroom
for UE – DL Pathloss = 120dB. It is noticeable that TTI
Average LTE MOS’s condition against RSRP and RSRQ is bundling improves the UL BLER which increases the UL
observed to be flat and no obvious trend line correlation can be coverage as well as the PHICH ACK rate by approximately
observed from Fig. 5. With strategy to keep users in LTE layer 5% in Fig. 7 below.
to obtain better call quality, the optimal threshold condition for
serving LTE cell must be equal to the instability point: RSRP < UL BLER/ACK Rate - DL Pathloss (dB)
-117dBm and RSRQ < -12dB. Although it is also possible to 100 100
set SRVCC threshold to higher value, it is not recommended to 90 95
do so since user may experience call quality degradation upon 80 90

PHICH ACK Rate %

handover to WCDMA. Using MOS value only to evaluate 70 85
PUSCH BLER %

(UL BLER)

SRVCC thresholds is not a reliable metric. In this paper, LTE 60 (ACK Rate)
Poly. ((UL BLER))
80
50 75
instability point is already known and from this point,
40 70
transmission performance is evaluated. 30 65
22

B. Uplink Transmission Performance without TTI Bundling 20 12 15 60

10 1 2 4 4 4 4 55
In uplink transmission performance, PUSCH BLER and 0 50
PHICH ACK rate condition against DL PL is evaluated. In
80
81
84
87
90
93
96
99
102
105
108
111
114
117
120
123
126
129
132
135
138
141
144
DL Pathloss (dB)
Section III.A, the proposed threshold is set in high DL PL
condition (RSRP < -117dBm, RSRQ < -12dB) in order to Fig. 7. PUSCH BLER, PHICH ACK rate versus DL PL condition with TTI
leverage LTE’s better call quality. 10% UL BLER evaluation bundling feature deployed
is used to obtain the optimal threshold for SRVCC. In actual Considering that the network implemented TTI bundling
network test scenario, it is to be noted that reference signal feature in line with VoLTE, at DL PL = 138dB, the UL BLER
power in LTE is set to 18dBm. condition starts to degrade abruptly above 10% which is the
limit UL BLER for data transmission. It is concluded that
beyond this PL condition, the quality of voice call is degraded
as indicated by the UL BLER and UL data ACK rate. At PL =
135dB equivalent to RSRP = -117dBm –LTE MOS instability
point obtained in Section III.A, the UL BLER condition is still
below 10% safe line. In this paper, it is proposed that RSRP of
-117dBm is to be used as serving cell SRVCC threshold. With
the improvement in transmission performance due to TTI

215
 bundling, it is also recommended to deploy TTI bundling 117dBm with TTI bundling deployment. This paper mainly
feature along to aid in VoLTE performance improvement. focused on SRVCC threshold setting for RSRP. In the
implementation phase, the RSRQ threshold is set to value
IV. CONCLUSION equal to the call quality evaluation where MOS becomes
This paper provided serving cell SRVCC threshold unstable.
optimization approach for networks offering VoLTE service
with discontinuous LTE coverage. Since SRVCC is mainly REFERENCES
triggered by user’s measurement reporting, optimal RSRP [1] Paisal V., "Seamless Voice over LTE",2010 IEEE 4th
values are studied with operator’s motivation to leverage LTE International Conference on Internet Multimedia Services
Architecture and Application, Bangalore, 2010
layer with careful consideration to minimize call drop. [2] 3GPP TS 23.216, Single Radio Voice Call Continuity
Through voice call tests performed in live networks, MOS [3] Koshimizu, T. et al, "Improvement on the VoLTE (Voice
samples are collected to be related with RSRP and RSRQ over LTE) Domain Handover with Operator's Vision",
values. Additionally, comparison of serving LTE layer and World Telecommunications Congress, Miyazaki, 2012
target WCDMA layer MOS quality are compared. In [4] Tingting, Y. et al, "Performance Evaluation and
conclusion, LTE layer provides better call quality versus Optimization of SRVCC Handover Scheme", 2014 IEEE
International Conference on Computer and Information
WCDMA and flat MOS scores against a wide range of RF Technology, Xian, 2014
condition. Hence, it is concluded that it is best to keep UE in [5] Holma, H. et al, “Voice Over LTE”, UK, 2012
LTE layer. The MOS condition only becomes unstable upon [6] Jing H. and Haiming W., "Principle and Performance of
reaching very poor RF condition – where RSRP < -117dBm TTI Bundling for VoIP in LTE FDD Mode", Wireless
and RSRQ < -12dB. Communications and Networking Conference, Budapest,
However, SRVCC thresholds cannot be purely based on 2009
MOS values since scores have abnormalities in varying RF [7] 3GPP TS 36.311, Radio Resorce Control Protocol
Specification
condition. Hence, evaluation of UL data BLER and ACK rate
[8] ITU-T P.863, Perceptual Objective Listening Quality
is taken into consideration. It can be concluded that the Assessment
optimal SRVCC RSRP threshold may be set at condition [9] 3GPP TR 36.824, LTE Coverage Enhancements
where PUSCH BLER is above 10% and ACK rate starts to [10] Dahlman, E. et al, “4G LTE/LTE Advanced for Mobile
degrade steeply. In relation to this, the implementation of TTI Broadband,” MA, USA, 2011, pp. 177
Bundling feature helps to improve the UL coverage and [11] Elnashar A. et al, “Design, Deployment and Performance
minimizes the block error rate. This paper provided live of 4G LTE Networks: A Practical Approach,” Malaysia,
2014, pp. 488-489
network performance improvement with TTI bundling.
[12] TEMS Discovery Device 10.0.6 Manual
Finally, the SRVCC serving cell threshold is proposed to
be RSRP = -108dBm for no TTI bundling and RSRP = -

216

View publication stats