As LTE networks start to mature across the world, more and more carriers are looking introduce voice services on LTE networks using a technology called Voice over LTE (VoLTE). What is not always understood is that voice services on LTE are very different from those on 2G and 3G networks. An appreciation of the differences leads us to the conclusion that we must think and act differently for 4G. This article addresses the uniqueness of VoLTE and briefly explains the technical features required to make VoLTE work. Most importantly, we discuss the implications this has on making sure VoLTE and next-generation wireless voice services work as expected.
Voice Quality and Call Performance Can't Go Backward
Wireless subscribers are accustomed to a certain experience when making voice calls on 2G and 3G wireless networks, and those expectations do not change when carriers offer VoLTE on their 4G networks. In fact, they probably don’t know, or care, that instead of getting a dedicated voice bearer, their call is being funneled through a giant data pipe along with all the other IP traffic generated by wireless devices in the area. They don’t care that delivering the same quality of service they are used to in this all-data world is very, very hard to do on wireless networks. It just has to work—as well, or better, than it did in the past. If not, users of 4G smartphones will either turn off 4G on their phone or return their devices in favor of something that works better.
Making sure this does not happen means focusing on the aspects of VoLTE call performance captured in Table 1.
Table 1 VoLTE Call Performance
There is nothing new about these metrics—they apply equally to 2G and 3G. However, with VoLTE, voice services are delivered very differently than those of the past, requiring us to think differently about the way we ensure quality for VoLTE and other next-generation voice services on all-IP wireless networks.
Voice Services on LTE Are Unlike Anything in the [Wireless] Past
All that matters for 2G/3G voice today is the robustness of layers 1-3 in the mobile device and radio access network (RAN). The dedicated voice bearer needs to be reliably setup and maintained—not always an easy task, but performance issues almost always come down to problems with the physical and access layers. However, LTE voice services are based on data bearers and VoLTE is delivered via IP Multimedia Subsystem (IMS) networks. For this to work, the entire stack (all seven layers) must function and interoperate correctly in devices and in the network. The dependency on interworking between the application layers of devices and core network elements is new in the wireless world. In the wired world, landline VoIP took a while to mature and be viable, even though it had the advantage of being able to solve many problems just by increasing bandwidth. In wireless, the spectrum is limited, bandwidth is more expensive, and radio interface performance is very unpredictable. This presents challenges never before faced by wireless network operators, mobile device vendors, or network equipment manufacturers.
Overcoming the Challenge
Fortunately, networks and devices are being upgraded with new features that will help ensure that high quality LTE voice services become a reality. They can be broken down into the following areas: RAN, IMS Core, Codecs, and QoS (see Table 2).
Table 2 New Features Needed to Make VoLTE Work
The new RAN features help minimize packet latency and jitter by reducing signaling overhead and ensuring packets can be transferred in all radio conditions. Packet latency and jitter have an impact on the mouth-to-ear delay of speech, so if the RAN features fail to work properly, users will experience annoying delays in their conversations.
VoLTE is enabled by IMS, and all speech packets must get routed through the IMS core network. IMS is the key to providing all the voice and supplementary services expected by users, such as call waiting, call forwarding, hold, and conferencing. Interoperability between IMS clients on devices and the IMS service is crucial to functional voice services, especially when roaming outside of a user’s home network.
While not exclusive to VoLTE, the addition of wideband adaptive multi-rate (AMR-WB) codecs promises to enhance the user’s voice experience. AMR-WB codecs operate at 6.6, 8.85, or 12.65kbps which results in higher voice quality than traditional narrowband voice. Services offering AMR-WB codecs are often labeled as “HD Voice,” and the primary metric for measuring the voice quality is mean opinion scoring (MOS).
Sample results from speech quality testing of two devices, comparing WB-AMR and Standard NB codecs, can be seen in Figure 1.
Figure 1 Speech Quality MOS Distribution for WB-AMR vs Standard NB Codec (results collected with Spirent’s NomadHD)
Quality of Service (QoS)
VoLTE has several mechanisms to ensure that VoLTE voice services get the priority needed to meet minimum QoS requirements. First, a dedicated data bearer separates speech traffic from other data traffic so it can be delivered with a better QoS level (see Figure 2). Second, a quality of service class identifier (QCI) indicates the maximum acceptable delay and packet error rate so that network functions can make appropriate priority and resource allocations. Finally, the eNB dynamic scheduler grants resources to VoLTE data bearers, taking QCI into account. The QoS functions play a key role in ensuring packet loss and delays don’t exceed thresholds that would cause perceptible issues in speech quality, latency, or jitter.
Figure 2 Dedicated VoLTE Data Bearer
Making Sure VoLTE and Next Generation Wireless Voice Services Work as Expected
Because voice services such as VoLTE can encounter issues on all-IP networks at so many layers, the place to start when looking to ensure a good end-user experience is at the endpoints. Measuring the KPIs in Table 1 in an end-to-end system ensures that QoE can be completely characterized, and it also provides a basis for measuring improvements and benchmarking different components.
Speech quality is one of the key metrics to be measured at the endpoints and is particularly important for VoLTE. Due to the inclusion of WB-AMR codecs (see Table 2) in VoLTE solutions, speech quality measurement systems should implement the Perceptual Objective Listening Quality Analysis (POLQA) algorithm defined in ITU-T P.863. POLQA uses a full-reference model where known undistorted audio speech is sent by one device over the network and received by the second device. The received audio is typically degraded due to noise, compression, network impairments, and other factors. It is then compared to the original reference voice signal by simulating human hearing. The result is an objective score on a scale of 1 to 5, which has 97% correlation with the subjective results from human listeners. Example results are shown in Table 3.
Table 3 POLQA MOS Score Comparison for Four VoLTE-Enabled Devices
A measurement system that can accurately measure speech quality, call setup time, call initiation rate, call drop rate, and mouth-to-ear latency at the endpoints (example reports shown in Figure 3 & Figure 4) is very valuable in evaluating performance, both on live networks as well as in a lab environment. Because VoLTE is so new and deployed on very few commercial networks to date, much of the testing needs to be done in the lab. Ideally, the same call performance measurement system can be used for lab and live network testing to ensure that all KPIs are measured the same way and to make data analysis more valuable.
Figure 3 Sample Call Initiation and Call Setup Time Reports
Figure 4 Sample Call Drop Rate Report
Using network and channel emulators, lab testing has the added benefit of fully controlling the RF and network environments. For example, when benchmarking the speech quality and call performance of multiple devices, lab testing allows the exact same scenario to be played back over and over again — ensuring a true “apples to apples” comparison.
Once end-to-end measurement solutions are in place for voice services, strategies can be developed for diving down into the lower layers of mobile device and network element implementation to help debug and diagnose problems when they occur. This requires a way to capture and analyze data at various layers so that root causes can be isolated and fixed. This “top down” approach of verifying voice service performance is very different from what is typically done for 2G and 3G, but it is absolutely essential for VoLTE and next generation LTE voice services. All-IP wireless data networks are a new territory for voice services, and require us to think differently in order to make sure the mobile experience remains the same or [hopefully] improves.
For more information on the challenges of and solutions for VoLTE and next-generation voice service quality, please review the VoLTE Technology Page  on Spirent’s web site  — where you with find white papers, posters, application notes, and videos related to the topic.
About the Author:
Brock Butler is currently marketing director for mobile devices at Spirent Communications. After joining Spirent in 1999 as a development engineer, Brock has held various positions in product development, product management, and marketing—including five years leading Spirent’s wireless location technology product segment. He regularly writes for industry publications, speaks at events, and uses digital media to promote the understanding of wireless technology and its impact on testing. Brock holds a BSc in Electrical Engineering from Villanova University, Pennsylvania, USA. Connect with Brock on Linkedin  or follow him on Twitter at @brockbutler .
As LTE networks start to mature across the world, more and more carriers are looking introduce voice services on LTE networks using a technology called Voice over LTE (VoLTE). What is not always understood is that voice services on LTE are very different from those on 2G and 3G networks. An appreciation of the differences leads us to the conclusion that we must think and act differently for 4G.