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

net/publication/326069853

VoLTE Performance Analysis and Evaluation in Real Networks

Conference Paper · June 2018

CITATIONS READS
0 1,380

3 authors, including:

Gentian Bytyqi Driton Statovci

University of Prishtina TU Wien
2 PUBLICATIONS   0 CITATIONS    36 PUBLICATIONS   180 CITATIONS   

SEE PROFILE SEE PROFILE

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

Copper-Optical Network Evolution for ultra-broadband access (CONE) View project

VoLTE Performance Analysis and Evaluation in Real Networks View project

All content following this page was uploaded by Driton Statovci on 25 July 2018.

The user has requested enhancement of the downloaded file.

 Second International Balkan Conference on Communications and Networking Podgorica, Montenegro,
June 6-8,2018

VoLTE Performance Analysis and Evaluation in Real

Networks
Bujar Krasniqi Gentian Bytyqi Driton Statovci
Faculty of Electrical and Computer Faculty of Electrical and Computer Institute of telecommunications
Engineering Engineering TU WIEN
University of Prishtina University of Prishtina Vienna, Austria
Prishtina, Kosovo Prishtina, Kosovo driton.statovci@nt.tuwien.ac.at
bujar.krasniqi@uni-pr.edu gentian.bytyqi@uni-pr.edu

Abstract—Moving voice traffic from circuit-switched domain services, such as High Definition (HD) voice. To reliably
to packet-switched domain seems to be the only rational way to deliver voice calls over all IP network, mobile operators have
improve quality of experience. Voice over Long-Term Evolution adopted, deployed, and recently started with service called
(VoLTE) is a technology that enables transmission of voice calls Voice-over-LTE (VoLTE). Until the end of 2016, around 156
over LTE network by using IP Multimedia Subsystem (IMS). mobile operators in 73 countries have invested in VoLTE
VoLTE is bandwidth efficient and offer better speech quality including 102 operators that have launched HD voice service
than legacy technologies such as Universal Mobile using VoLTE [1]. The VoLTE solution introduces the voice
Telecommunications System (UMTS) and Global System for functionality in the LTE network by using new IP Multimedia-
Mobile communication (GSM). In this paper, we analyze the
Subsystem (IMS) framework [2].
impact of radio propagation conditions, codec type
implementation, and mouth-to-ear delays on VoLTE service A performance analysis and optimization of CSFB from
performance. In particular, we consider drive test, stationary LTE to UMTS is performed by authors in [3]. Differently, the
test, and train test measurements in real VoLTE network authors in [4] have analyzed the VoLTE performance of user
deployments. Testing results indicate that when both end users equipment (UE) using carrier aggregation. A deeper analysis
are using VoLTE technology, the obtained speech quality is and evaluation of CSFB and VoLTE in terms of end-to-end
impressive compared to the case when one user is using legacy assessment of call setup delay under different radio conditions
technologies (UMTS or GSM) or performing Circuit Switched
has been done by authors in [5]. Our main contributions in this
Fall Back (CSFB) to legacy. Furthermore, results show that
VoLTE call setup time and mouth-to-ear delays are shorter while
paper are the analysis and evaluation of CSFB performance,
call reliability is comparable with legacy technologies. However, VoLTE call setup, VoLTE speech quality, end-to-end delay
the percentage of VoLTE dropped calls is high and this indicates and call reliability for all possible scenarios such us drive test,
a potential for further optimization in the current stationary test, train test and overall.
implementation of VoLTE. The rest of the paper is organized as follows. In Section II
we shortly introduce VoLTE technology and the key
Keywords—VoLTE, LTE, CSFB, SRVCC, UMTS, GSM,
performance indicators to analyze the performance of VoLTE.
handover
In Section III we perform a detailed performance analysis of
VoLTE under different test scenarios. Conclusions of our work
I. INTRODUCTION are drawn in Section IV.
The cellular communications industry has witnessed
extensive growth since the mid of 1990’s. The demand for II. VOICE OVER LTE
higher speeds and better Quality of Experience (QoE) in
The absence of CS network in LTE has led the industry and
mobile communications network is continually increasing.
standardization bodies to propose various solutions to support
Long-Term Evolution (LTE) supports only packet-switched
voice over LTE network. The CS fallback solution is defined in
network across an all-IP system, whereas previous cellular
[6] and supports voice services in LTE systems by reusing the
networks GSM or UMTS support both packet- and circuit-
existing GSM/UMTS network. Mobile users connected to a
switched network. In the beginning of LTE deployment, all
LTE network for data services are obligated to fall back on a
voice traffic is handled by legacy Circuit-Switched (CS)
legacy network when they make or receive a voice call. A CS
networks, while data traffic is handled by LTE packet-switched
fallback enabled mobile device connected to LTE, may use
networks. Some solutions have been proposed in order to
UMTS/GSM network to connect to the CS domain. Thus, users
deliver voice services in LTE. In practice mostly it is used the
have to perform both registrations in LTE and GSM/UMTS
so-called Circuit Switched Fall Back (CSFB) approach, which
networks in order to proceed with the call setup. The
enables voice transmission in LTE via legacy networks.
registration is done by mobility management entity.
Even LTE was originally designed as packet-switched
network, its Quality of Service (QoS) and capacity provide
mobile subscribers significant improvement in QoE for voice
A. VoLTE Key Performance Indicators benefit from this codec, the user equipment has to support
The most important performance indicators in VoLTE supper wideband (up to 14 kHz) or full band (up to 20 kHz)
service are: speech quality, call setup time, call setup success audio frequency range [8]. Speech quality is evaluated using
ratio, call reliability, handover success ratio and dropped calls Perceptual Objective Listening Quality Analysis (POLQA)
ratio [7]. Speech quality depends deeply on the voice codec wideband algorithm [9].
sampling rate and the resulting audio bandwidth. Adaptive
Multi-Rate Narrow Band (AMR-NB) provides audio III. VOLTE PERFORMANCE ANALYSIS
bandwidth spectrum from 80 Hz to 3700 Hz while Adaptive In this section, we study the effects of radio propagation
Multi-Rate Wide Band (AMR-WB) extends the audio conditions, codec type implementation, and end-to-end delays
bandwidth spectrum from 50 Hz to 7000 Hz as illustrated for on VoLTE service. Furthermore, we analyze the differences in
particular examples in Fig. 1 and Fig. 2. Higher voice speech quality between end-to-end VoLTE call and a voice call
bandwidth of AMR-WB codec results in better quality and which combines VoLTE with UMTS/GSM or is employing
more natural sound.and therefore the VoLTE users using this Single Radio Voice Call Continuity (SRVCC) and CSFB. The
type of codec will experience better QoE compared to voice in VoLTE speech quality, mouth-to-ear delay, call setup time, and
CS networks. call reliability are analyzed and compared with legacy
technologies.

A. Measurement Methodology
To perceive the end user experience while they are using
VoLTE service, both mobility and stationary test
measurements are conducted gathering real-time data from the
network. Numerical values of relevant parameters and their
statistical distributions have been extracted from the
measurement campaigns carried out in different mobile
operators in Europe where the VoLTE technology have been
already widely deployed. Measurements have been conducted
across the big metropolitan areas, the connecting roads
between those cities and smaller towns close to the connecting
roads. Technical performances of the networks have been
tested from the point of view of the end consumer on VoLTE
capable smartphones. The measurements were conducted
between two independent mobile-to-mobile systems. Both
Fig. 1. An example of audio bandwidth of Adaptive Multi-Rate Wide Band
(AMRWB) codec.
mobiles, A-party and B-party, were calling each other using
Samsung Galaxy S5 LTE Cat.4. Voice test calls were mobile-
to-mobile while the measurement cars were moving and the
duration of each call was 100 seconds.
To perform measurements, different Rohde&Schwarz test-
solutions have been used. Rohde&Schwarz ROMES software
platform in combination with other test and measurement
equipment have been involved in recording and visualization
of test parameters. In addition to signal strength, the Received
Signal Strength Indicator (RSSI) and the Reference Signal
Received Quality (RSRQ) are displayed for every cell as well
as the reference Signal to Interference-plus-Noise Ratio
(SINR).

B. Speech Quality Test Results for Different Technologies

This section presents the differences in the call quality
between end-to-end VoLTE call and a voice call which
combines VoLTE with UMTS and GSM or employing CSFB
and SRVCC. The speech quality is evaluated in real VoLTE
Fig. 2. An example of audio bandwidth of Adaptive Multi-Rate Narrow and GSM/UMTS network deployments. Drive test
Band (AMR-NB) codec.
measurements were performed using two measurement cars. In
Additionally, another type of codec which offers better both cars are located mobile phones, A-party and B-party,
quality than AMR-WB is the evolved HD voice, which is also where both mobiles are calling each other. Speech quality was
called Enhanced Voice Service (EVS). EVS extends the HD measured by playing reference voice samples on talking side
voice service experience with natural voice and music by and recording the transmitted samples on listening side.
supporting the full human voice frequency range. In order to POLQA wideband algorithm was applied to derive the average
Mean Opinion Score (MOS) values. For each single call eight used in VoLTE should follow the limitations on codec mode
voice samples were recorded and total number of samples is changes in UMTS and GSM networks.
showed in Table 1.
The difference in quality between VoLTE drive and
TABLE I. NUMBER OF RECORDED SAMPLES FOR DIFFERENT stationary tests measurements is shown in Fig. 4. The total
SCENARIOS number of recorded voice samples for drive test was 13594
Different call Technologies while for stationary test was 6717 voice samples.
scenario VoLTE-VoLTE VoLTE-UMTS VoLTE-GSM
Nr. Of recorded
1394 1998 330
samples

The cumulative distribution of speech quality between end-

to-end VoLTE and VoLTE to UMTS/GSM calls is shown in
Fig. 3. The results indicate that when both end users are using
VoLTE technology, the obtained speech quality is impressive
comparing with the case when one mobile is using legacy or
performing CSFB to legacy. The small number of samples for
VoLTE to GSM call scenario may not accurately quantify the
real-word impact and can be further improved with more
recorded samples. In VoLTE to VoLTE call scenario the
average of speech quality is 4.11 MOS; while in VoLTE to
UMTS/GSM is 3.35 MOS and 2.77 MOS respectively. Voice Fig. 4. CDF of VoLTE speech quality for drive tests and stationary tests.
calls involving multiple technologies may result in using
codecs in tandem. For example, in the case of VoLTE to GSM Measurement results indicate that the speech quality is lower
call, the output of AMR-NB codec voice data for purpose of when both VoLTE UEs are in mobility. For drive test
interworking must beconverted into another format e.g., Pulse measurements the probability that speech quality will take
Code Modulation (PCM), which further degrades voice values less than 4.0 MOS is 39 %, while for stationary test it is
quality. 29 %. This is due to many factors; one of them is the higher
probability for making handover in mobility scenario. VoLTE
handover is SRVCC, which seamlessly handovers the VoLTE
call session to GSM/UMTS networks. If the path handover
interruption is high then the speech frames will be lost and this
will impact the speech quality. Finally, the degradation of
speech quality during and after SRVCC handover is
represented in Fig. 5.

Fig. 3. Cumulative Distribution Function (CDF) of speech quality for

different call scenarios.

For the VoLTE to VoLTE call scenario only AMR-WB

codec is used and PCM based transport cannot be used with the
wideband codec because PCM only applies to a narrowband
voice. Therefore, AMR-WB codec use Transcoder Free
Operation (TrFO) [10]. TrFO is a solution where encoded Fig. 5. CDF of speech quality: VoLTE-VoLTE and SRVCC-VoLTE
speech is transmitted through the network as a packet, without
the need to decode it into PCM format, in this way the quality Only drive test measurements were conducted for two
of mobile-to-mobile calls is improved. different call scenarios. In the first scenario, both mobiles are
using VoLTE without SRVCC handover. In the second
It is important to mention that in VoLTE to UMTS call scenario during the VoLTE call mobiles in car 1 were
scenario, AMR-WB and AMR-NB codecs were used in both performing SRVCC handover to UMTS/GSM technology,
sides, but in 85% of the calls the codec rate was less than while mobiles in car 2 were using VoLTE and were not
12.65 kbps. Whereas, for VoLTE to GSM only AMR-NB was performing SRVCC handover.
used with highest codec rate of 12.2 kbps. The codec operation
 The drive test results indicate that SRVCC handover to VoLTE. Fig. 8 presents mouth-to-ear delay when one mobile
UMTS or GSM has a large impact on speech quality. The path is using VoLTE the other one is using legacy or when both
interruption time (voice interruption) due to handover was on mobiles are using legacy.
average 150 ms. For voice services such as VoLTE, the size of
packets is small, and the inter-arrival time is 20 ms. The path
interruption of 150 ms due to SRVCC handover will cause
packet loss, which degrades the speech quality. From the
results in Fig. 5 can be concluded that the average of speech
quality is 4.11 MOS without handover, while 3.9 MOS when
the SRVCC is employed. Both inter- Radio Access
Technology (RAT) handover and session transfer in the core
network contribute in the voice interruption time. To minimize
the voice interruption, the SRVCC initiates the inter-RAT
handover and session transfer simultaneously so they can run
in parallel.

C. VoLTE End-to-End Delay Test Results

Fig. 7. Mouth-to-ear delay in VoLTE-VoLTE.
VoLTE end-to-end delay (mouth-to-ear delay) is one of
the most important metric when we measure VoLTE
performance. Mouth-to-ear delay is latency between the
speaker utters a word and the listener actually hears it [11],
containing both one-way latency in the network and the time
spent on encoding and decoding audio packets. The maximum
mouth-to-ear delay in good quality communication should be
lower than 250 ms [12]. According to ITU-R and 3GPP
standard, the 153 ms mean mouth-to-ear delay of VoLTE
suggests excellent user satisfactory. To calculate the delay
budget the assumption from [12] is used. The speech coder
takes 20 ms of speech samples and encodes into a speech
frame. The voice encoding delay is assumed to be 30 ms,
including 20 ms frame size. The uplink transmission takes
50 ms, while Evolved Node B (eNB) processing take 5 ms and Fig. 8. Mouth-to-ear delay in VoLTE-Legacy and Legacy-Legacy.
gateway processing 1 ms. The transport delay is assumed to be
10 ms, while downlink transmission takes 50 ms. Decoding From the test results shown in Fig. 7, one can see that
and processing delay in the receiver UE is assumed to be VoLTE calls have low mouth-to-ear delay. The average of
5 ms. Under these assumptions the mouth-to-ear delay is mouth-to-ear delay for VoLTE is found to be 154 ms. This
approximately 150 ms and it is illustrated in Fig. 6. value is 4 ms greater than the ITU-R requirements for one-
way VoLTE end-to-end delay to experience high quality voice
call. In another hand when users are performing CSFB to
legacy calls or both users are using legacy (cf. Fig. 8) the
average of mouth-to-ear delay is 280 ms and 410 ms,
respectively. This will have large impact on the user
experience. Radio frequency conditions will also impact the
mouth-to-ear delay. Fig. 9 represents the RSSI impact in
mouth-to-ear delays.

Fig. 6. Average speech quality: VoLTE-VoLTE and SRVCC-VoLTE [12]

VoLTE uses dedicated bearer offering Guaranteed Bit

Rate (GBR) for transporting voice media to minimize delays.
In another hand, eNB may decrease the GBR rate for UEs in
undesirable network conditions. To proof the QoS guarantee
provided by dedicated bearer, VoLTE calls are performed
under various channel conditions. To calculate the mouth-to-
ear delay, a connection box (GPS-time synchronized) is used. Fig. 9. Mouth-to-ear delay versus RSSI values.
Fig. 7 shows mouth-to-ear delay when both mobiles are using
Results indicates that when the RSSI is greater than –78 dBm VoLTE shows an improvement in call setup time, while
the mouth-to-ear delay is constantly less than 200 ms, while high number of dropped calls leaves a room for further
for RSSI lower than –78 dBm the mouth-to-ear delay might optimization. In the case of mobility tests, VoLTE shows
reach up to 480 ms. higher number of dropped calls, especially for train tests.

D. VoLTE Call Setup Time and Reliability IV. CONCLUSION

The equal number of call attempts and the same mobility In this paper, we presented a study about Voice over LTE
route was taken for VoLTE and legacy (GSM/UMTS or (VoLTE) with focus on performance analysis in real network
CSFB). Calls were mobile-to-mobile and both mobiles were in deployments. We considered both mobility and stationary test
mobility or stationary, except for railway test, one mobile was measurements to gather real-time data from the network. From
in train and the other one in highway. Fig. 10 shows the call the test results, we found that that for good radio channel
setup time for VoLTE and legacy for mobility and stationary conditions the best possible speech quality is delivered using
tests. Adaptive Multi-Rate Wide Band (AMR-WB) with coding rate
of 23.85 kbps, while for bad radio conditions (Received Signal
Strength Indicator (RSSI) < –85 dBm) the best speech quality
is achieved using AMR-WB with coding rate 12.65 kbps. The
average mouth-to-ear delay was 154 ms. In another hand, when
RSSI < –78 dBm the mouth-to-ear delay has been increased
drastically. Finally, the call reliability and call setup time for
VoLTE and legacy calls was analyzed. Test results indicate
clearly that VoLTE shows an improvement in call setup time
while the number of dropped calls is higher than in legacy
technologies.

V. REFERENCES
[1] Juniper Research “VoLTE - Voice over LTE: Key Market Movements
2017,” tech. rep., Juniper, 2017.
Fig. 10. Call setup time for mobility and stationary tests. [2] Ayman ElNashar, Mohamed El-saidny, Mahmoud Sherif. Design,
Deployment and Performance of 4G-LTE Networks: A Practical
VoLTE call setup time shows an impressive improvement Approach, May 2014
comparing with call setup time in legacy. The mean VoLTE [3] J. Bautista, S. Sawhney, M. Shukair, I. Singh, V. Govindaraju and S.
call setup time is 2.39 s, which is much better than legacy calls. Sarkar, “Performance of CS Fallback from LTE to UMTS,” IEEE
Employing CSFB while making a call lead a higher call setup Comm. Magn. September 2013
time and indicates a potential for optimization. A call is [4] J. Sonkusare, and S. Dhage, “LTE UE Performance Evaluation for
Carrier Aggregation and Voice-over-LTE,” IEEE Applied and
defined as a successful call if the call is successfully Theoretical Computing and Communication Technology (iCATccT),
established and maintained for at least 100-seconds during the 2015 International Conference on. October 2015
test, so the call reliability can be measured as the probability of [5] A. Elnashar, M. El-Saidny and M. Mahmoud, “Practical Performance
making a successful call. The comparison results of call Analyses of Circuit-Switched Fallback and Voice Over LTE,” IEEE
reliability for VoLTE and legacy in both stationary and Transactions on Vehicular Technology, Vol. 66, No. 2, February 2017
mobility cases are shown in Fig. 11. [6] 3GPP 3rd Generation Partnership Project; Technical Specification
23.272, "Circuit Switched (CS) fallback in Evolved Packet System
(EPS)", Stage 2 (Release 13).
[7] J. rankin, A Costaiche, J. Zeto and K. O’Neil “Validating VoLTE A
Definitive Guide to Successful Deployments”, IXIA 2014.
[8] Ericsson white paper, “Evolved HD Voice for LTE”, October 2014.
[9] International Telecommunication Union, ITU-T P.863, Perceptual
Objective Listening Quality Assessment, September 2014.
[10] C. Birkehammar, S. Bruhn, P. Eneroth, K. Hellwig and S. Johansson,
“New high-quality voice service for mobile networks”2006.
[11] C.Agastya, D.Mechanic, and N.Kothari.”Mouth-to-ear latency in
popular voip client”.2009.
[12] 3GPP 3rd Generation Partnership Project; Technical Specification
23.216, “Single Radio Voice Call Continuity (SRVCC)”, Stage 2
(Release 13)
[13] Shyam Chakraborty, Tomas Frankkila, Janne Peisa and Per Synnergen
“IMS Multimedia Telephony over Cellular Systems”, 2007.

Fig. 11. Call reliability for mobility and stationary tests.

View publication stats