Voice and LTE

Back when LTE first got started, all IP networks were all the rage in the cellular industry. People envisioned that all voice services would be provided over the IP network. So long as the IP network and its underlying transport provided acceptable QoS, VoIP would fit the bill.

But if it was all this simple, why hasn’t VoIP replaced the entire PSTN years ago? The problem is a combination of legacy, quality and availability. You can’t just move to VoIP without some bridge to the existing PSTN. Robust solutions to VoIP/PSTN interworking have only recently been deployed despite the fact that VoIP has been around in its present basic form for 15 years. End-to-end quality with VoIP over best effort links is still spotty and there are no end-to-end QoS mechanisms deployed in the Internet. Finally, IP is still not available in all places the PSTN is available.

Similar issues exist with trying to replace the Circuit Switched (voice) cellular network technology with an all IP solution. On the issue of legacy, the existing provider core networks are not even all IP. There is still plenty of legacy ATM gear being used to transport both voice and data traffic between network elements. Most of this equipment is likely to be in the network for years to come. Secondly, the move to an all IP infrastructure will require the deployment of a new generation of cellular to PSTN interworking gateways. Such gateways must re-implement all the voice mobility functions that exist in the current generation of gear. This is a huge investment.

One “solution” to the legacy issue is something called “VoLGA”. VoLGA is the application of UMA technology to LTE. UMA was invented to support cellular voice over WiFi. The idea is that given an IP transport between the handset and the cellular core network, all the legacy cellular telephony protocols can be tunneled over the IP transport. A UMA gateway is then used to remove the IP transport and inject the telephony signaling into the core network. If vulgar doesn’t come to mind it should. VoLGA is not a compelling voice architecture and only in the sickest of naturally selected environments does it represent Evolution. It will certainly be looked at by future observers as one of the ugliest spandrels of all time.

Considering UMA’s lack of success with WiFi, there is no reason to believe it is likely to be any more successful over LTE. Issues regarding voice quality are likely, and it is not merely a question of network QoS. Real-time mobility in harsh environments is at issue here. There is one existing solution to cellular voice that address voice quality issues due to RF environment and mobility, is widely deployed and supported on all legacy handsets. Specifically 3GPP R99 (WCDMA) voice. Continuing to support R99 voice in LTE deployments allows operators to leverage their existing equipment as they slowly migrate to a native voice over IP solution. R99 voice also has some interesting features that are hard to emulate with basic packet switching.

One feature is soft-handover. Soft-handover is a combination of make-before-break plus simultaneous transmission and reception of voice data from multiple radio nodes. With soft-handover, mobile phones receive voice data simultaneously from multiple radio nodes and reconstruct the voice data from the total RF energy received. It is not a simple matter of receiving two distinct packets and selecting the best. By reconstructing the voice signal from multiple sources, the likelihood of reconstruction the original packet is improved by up to a factor of 2.

R99 voice also uses an encoding scheme that is resilient to bit errors. For a given voice frame, multiple encodings are transmitted with progressively lower susceptibility to bit errors. These multiple encodings give voice quality a smoother degradation as conditions deteriorate, rather than a complete dropout, which would be the case with a single VoIP packet transmission.

Basic LTE data services provide some capabilities that when leveraged for voice could give better performance. LTE has the concept of MIMO, which is a more advanced form of the multiple, simultaneous senders functionality found in R99. LTE also has a Hybrid ARQ protocol running between the radio node and and the handset. This provides more reliable delivery at the RF interface. The radio node should aggressively reduce the bit error rate for any radio links carrying voice traffic to ensure reliable transport and avoid the introduction of jitter. This implies that radio links should be CoS aware not only from a scheduling perspective but also from a bit error rate perspective. Finally, LTE provides much faster handover times, which should be sufficient to maintain acceptable voice quality during handover by keeping jitter due to handover within reasonable bounds.

Nevertheless, it is likely that R99 voice and VoIP/LTE will have to coexist for some time. Therefore, handover from one technology to the other will also be a concern both at the handset and in the mobile core. It will take some time for all of this technology to get widely deployed. This may give parasitic voice operators such as Skype an opportunity to leverage high speed mobile data links in conjunction with their existing packet redundancy and bandwidth adaptation schemes.

R99 voice might be one of those technologies, like 64Kbps circuit switched voice, that is going to be very hard to displace and may be with us for years to come. And when you’re trying to make that 911 call from the trunk of a car parked in a tunnel on a stormy night, you might be thankful.

  • Thanks, Jim, for this overview of why we’ll see LTE used for data first and voice later. On the 3GPP2 side, operators now operating with the cdma2000 family of technologies will likely do the same thing: keep cdma2000 1x and its circuit-switched-channel voice service until LTE networks are sufficiently deployed to allow at least the same quality of voice service — perhaps for years, as you suggest. There has been a lot of work on interworking standards in both 3GPP and 3GPP2 to make this all play together, and there are some good multimode chips as well.

    In addition to air interfaces and core networks, backhaul latency has to be at an acceptable level for VoIP on wide-area networks, and in some of these networks it isn’t today.

  • Pingback: High Tech Forum Welcomes new Contributor Jim Murphy :: Burning With The Bush()