A debate is currently raging around whether LTE can use packet switching to transmit voice. End user expectations cannot be ignored because, after more than 20 years of GSM, users have come to expect good voice quality.
How will voice transmission be implemented in LTE and how can the functionality and quality built in during the development phase remain evident during real life network operation? Transmitting voice over LTE at acceptable quality requires the right mechanisms and architectures in the radio and core networks. This affects LTE networks and conventional 2G and 3G networks, since it cannot be assumed that universal LTE coverage will be available. Another consideration is subscriber mobility; a voice call initiated on LTE must not be dropped when the subscriber moves into a 2G or 3G network. Voice over LTE (VoLTE) assumes the IP multimedia subsystem (IMS) will be available. The IMS reference architecture specified in 3GPP TS 23.228, including the interfaces to existing networks and to other IP based multimedia systems, shows the system's flexibility as well as its complexity. However, voice transmission requires only a subset of these functions and functional units (fig 1). An LTE UE must register with the IMS system before a voice call is set up. The LTE system uses different 'bearers' for signalling and for voice data. Signalling bearers, which have very low loss ratios, ensure control commands are received reliably. Voice bearers, on the other hand, have low latency and a low latency variability, so provide good voice quality. The session initiation protocol (SIP) is used for signalling, while voice packets are transmitted using the real time transport protocol (RTP). In addition, specialised packet allocation algorithms ensure specific requirements for voice transmission are met. One example is semi persistent scheduling (SPS), which provides a nearly static allocation of frequency and time resources on the air interface. It must be noted that the LTE system transmits voice and data services simultaneously via the same radio channel. Whenever possible, a UE will use only one air interface when handing over to existing networks. The procedures in the network and on the UE side are specified as single radio voice call continuity. Introducing this feature, the IMS system controls the handover of the voice service, and any data service running in parallel, from the packet switched (PS) domain within the LTE system to the 3G networks' circuit switched (CS) domain for voice or to the 3G network's PS domain for data. IMS, specified years ago, has been expanded, even though it has been implemented in only a few mobile networks. This is why alternative solutions have been sought and recent implementations include: simultaneous voice and LTE (SV-LTE), with voice over the existing 2G/3G networks and data over LTE; and CS fallback (CSFB), which terminates the signalling connection in LTE and sets up a voice connection via 2G/3G networks. A UE based on SV-LTE uses two independent air interfaces that can be used simultaneously for different services. Voice is transmitted over 2G/3G and LTE is used for data transmission. However, this solution uses more power and affects battery life. UEs equipped in this manner also cost more. The CSFB solution, on the other hand, avoids simultaneous operation of two air interfaces. If an LTE UE is in an LTE cell and either receives or initiates a voice call, the call is set up via a 2G or 3G network and not via LTE. It is assumed that, wherever there is LTE coverage, a 2G or 3G network will also be available. If not, the call is either forwarded to voice mail or will not be set up. The advantage is obvious; users receive the voice quality they have come to expect from GSM networks and none of the LTE network capacity is used for voice calls. But this process takes longer because the signalling connection has to be terminated in LTE and voice service reestablished in the 2G/3G network. In addition, a data service in place when the call is placed usually cannot be continued in paralle. Fallback to a 3G network makes it possible to support the data service by means of a handover. But this option is not available for fallback to 2G because the required dual transfer mode is not implemented. There are several variations of CSFB. For example, a distinction is made between RRC reconnection release with redirection and PS handover. Of these, handover is considered to be more reliable. With handover, the UE receives dedicated signalling indicating which cell it should access in a 2G or 3G system. This is not the case for RRC reconnection release. Here, the UE selects the new cell independently. Before the call is set up, measurements must be taken and the relevant access information decoded from the broadcast channel. This is a major reason for additional latency during call setup. The CSFB procedure can be speeded, based on RRC reconnection release, by transmitting all necessary access information to the UE via the LTE network before the transition to the 2G/3G network. Voice in network operation A series of tests is needed in order to ensure that functionality and performance are available during use, see fig 2. The CMW500 simulates all mobile network functions required for the test, including LTE, 2G and 3G. If CSFB is being tested, the procedure can be reproduced for fallback to UMTS, for example. In this case, the test instrument sets up an LTE signalling connection to the UE. An incoming call is initiated and the UE is instructed to fall back to 3G. The UE sets up the signalling and the voice channel in the 3G network is simulated by the test instrument. All information exchanged between the test instrument and the UE on the various signalling layers can be logged. This makes it possible to verify that the UE is functioning correctly and implementation errors can be detected and eliminated. When testing VoLTE, the IMS functionality is implemented on the PC that controls the test instrument, as well as in the test workflow. Again, the correct functioning of the relevant procedures must be verified in the UE. And, just with layer 2 and 3 protocols, every message can also be logged and analysed. Once the UE is functioning correctly, the next step is to assess voice quality. In LTE, voice transmission is exclusively packet switched: analogue voice signals are digitised, put into packets, transmitted via the LTE air interface, converted back into analogue voice signals on the receive side and output to a loudspeaker. Voice quality is thus affected not only by the technology used on the air interface, but also by the voice codecs, the microphone and the loudspeaker. The perceptual evaluation of speech quality (PESQ) method is used to evaluate the entire transmission chain. The PESQ method permits a quantitative assessment with results that range from -0.5 (poor) to 4.5 (very good). Author profile: Meik Kottkamp is LTE/LTE Advanced technology manager in Rohde & Schwarz' strategic marketing group.