Sunday 23 November 2008

Solving the LTE voice dilemma

Continuing the discussion from LTE World Summit, this is something that has been discussed in the past by myself and other blogs as well. We know that there is no out of the box solution for voice calls in Release 8 but there are some solutions that are being standardised for this problem. Dr. Howard Benn, Director of Cellular Standards, Motorola Mobile Devices gave an interesting presentation on this topic titled, "Voice –how to talk over LTE". Here is the summary of his presentation along with some more information:
As we know, IMS was introduced in Rel 5 but even till today, there has been no major IMS rollouts. There are some operators working on deploying the IMS solution but in reality its not been as successful as it should have been. If IMS is available then the problem of voice call on LTE goes away. The problem can be solved using Voice Call Continuity or VCC. Infact there is a bunch of specifications on IMS Centralized Services (ICS) and network Centric VCC for solving this and other similar problems.

So with IMS not being available, the first alternative for this problem is Circuit Switched Fallack (CSFB). In this, as can be seen from the MSC above, the user is attached to an LTE network. MSC can send Paging to the UE and if the user accepts the voice call then he is handed over to 2G/3G network. The big problem with this approach is additional time required to establish the voice call and the PS services might get disrupted, depending on how its handled.

The second solution is to have a Generic Access Network (GAN... previously known as UMA) based solution. This is similar solution to the ones used by some Femtocells. This would mean that the UE's would require GAN chipsets and GAN is known to be power hungry so it can impact the battery life significantly.

China Mobile's, Bill Huang in a recent interview mentioned that “We could carry voice over UMA” and “We will have an LTE network that supports voice…”. He was referring to this approach mentioned above.

Finally there are always proprietary options like Skype that can be used along with the data services to solve the voice problem.

Infact a service like Vonage, modified for mobiles, can solve this problem easily. You can connect a VoIP client from your phone or device to Vonage and you are given a landline number that you can pass to others. When calls are received on this number, the client in the mobile rings and you answer the call normally.

Nick Yamasaki from KDDI mentioned that KDDI will roll out LTE with CS fallback option for voice initially but then SRVCC (Single Radio VCC) solution will be adopted in future.


Dean Bubley said...

Hi Zahid

VCC won't actually help define the voice application for LTE - it'll just enable continuity between VoIP and circuit as the phone moves out of LTE coverage back to 2G/3G.

There still needs to be some form of SIP VoIP service or application defined in IMS as being "mobile voice version 2.0". It's possible to re-use ordinary fixed NGN VoIP applications, but ideally you'd have something optimised for mobile voice.


Anonymous said...

Hi Zahid,

Thanks for the LTE World Summit coverage on GAN over LTE, very insightful. I'll post my additional thoughts as well, soon.


Anonymous said...

am sorry, but i do not understand why someone would use UMA over LTE. expect if it is kineto proposing it...i understand that you are suggesting reuse of the msc infrastructure. wouldn't it be a lot simpler to just upgrade the msc servers to handle sip as well. i will be surprised if the nsn and ericsson rel 4 msc server products don;t already have this capability....

Steve said...

Hi Zahid,

I’m with Kineto and I wanted to jump in on this post.

First, there is no such thing as a “GAN chipset”. GAN, or “Generic Access Network” is simply a software protocol for tunneling 2G (A/Gb) and/or 3G (Iu) services over IP networks.

The idea that GAN, a software protocol, is power-hungry is simply not true.

Second, there may be some confusion about what was being explained. This is not a traditional GAN use case where a dual-mode Wi-Fi phone or even a femtocell is used. What the presenter from Motorola was showing was a way to use GAN to tunnel the existing circuit services over the LTE IP access network.

Third, from the anonymous comment above, Kineto is certainly supportive of GAN for voice in LTE. But as you can clearly see, the comments about voice over LTE via GAN were from Motorola and China Mobile. I don’t believe Kineto was even at the LTE Summit.

However, the idea of putting a SIP client in front of an R4 MSC seems to add a layer of complexity to the problem.

If the goal is to get voice over LTE, a light-weight tunneling protocol like GAN which uses the existing signaling for voice (TS 24.008) seems pragmatic.

Putting a protocol translation engine between the RAN and service core to convert from SIP to/from 24.008 does not seem practical. In addition, there is a significant impact in complexity to the handset. The handset will always need to support 24.008 for the 2G/3G networks. Now the handset would need to add a second SIP voice signaling stack when on LTE with the *exact* same functionality.

With voice over LTE via GAN, the handset continues to use the existing 2G/3G voice signaling protocol. When the phone connects via LTE, the TS 24.008 is simply packaged and tunneled over the LTE/IP access network. If/when the handset moves out of LTE coverage and back to the 2G or 3G network, the same telephony protocol is already in use. Given that every major handset manufacturer already supports GAN, this is a very low risk way to bring voice services forward.



Zahid Ghadialy said...

Steve, Thanks for the detailed explanation.

Anonymous said...

steve, this is anonymous...

good to see kineto reacting. i meant it as a complement. associate uma with kineto....:)

msc-servers maintain call states and sip is just another protocol. i see no reason why they cannot support sip directly. i see no need to convert things between sip and the "A" interface.

this is what the big guys mean by ims anyway. anyway at some point people need to offer some more services than what can be offered using a traditional msc. other wise users will prefer to use their lte enabled devices to run skype.

Anonymous said...

btw. last i checked uma was still only a method for tunneling 2.5G traffic and protocols over ip. has the Iu version being approved? femto seems quite different to me...

Steve said...

Per the last anonymous comment, the GAN specification (TS 43.318) was enhanced to support 3G (Iu) interfaces on the GAN-C. I believe it was completed in November of 2007 (about a year ago). Iu support first appeared in the ver 8.1 of the spec.


Dean Bubley said...

One additional factor to bear in mind: the network side is the easy bit.

I've lost count of the times that people assume things like you can just "upgrade the MSC servers to handle SIP". Or deploy 3G, IMS, WiFi, UMA, femtocells etc - without thinking about the impact on the handset side.

Getting a full mobile VoIP implementation *right* on a handset is exceptionally difficult, and certainly beyond most of the current IMS-type approaches.

And the difficulty isn't about hard measurable things like standards, it's about things like user interface design, integration with other applications, battery consumption, and a million other things.

The potential advantage of the UMA-type approach is that it can use the standard dialler, phonebook, SMS, call log and so forth. These are all highly mature, whereas even standalone IP softphones are still highly variable in user experience and reliability.

IMS handset frameworks with VoIP clients are nowhere close - especially if they have to hand calls over to circuit outside LTE coverage.

This will take many years to solve. UMA-type solutions may be a necessary stepping stone.

To reiterate: the network is the easy bit. The phone is much harder. It took Apple to get even mobile web browsing "right". VoIP is an order of magnitude trickier.


Anonymous said...

GAN is simply a software protocol for tunneling 2G (A/Gb) and/or 3G (Iu) services over IP networks.

In my view, the choice of GAN for realtime service delivery over LTE is no more than an quickfix interim option. Mobile operators should support development of multimedia services and not merely replicate the traditional CS network services. GAN based approach is not a future proof solution and will lead to sunk investments . It is no comparison to flexibility of Multi-Service IMS Architecture with open and standard interfaces and SIP based Clients for multimedia support. With maturity of ICS and SC standards in 3GPP R8, IMS Centralized Service Architecture provides the best option for future network evolution of full service operators. ICS/SC solutions will be ready in time and therefore a short term quickfix solution should be avoided.

1. Contrary to popular perception, investments in IMS are small, and certainly represent a very small fraction of the LTE/SAE investments. Furthermore, operator can build small scale systems and scale based on service uptake, thus minimizing risks. Thus, I disagree with you on the risk with IMS path.

2. I disagree with the remarks on different complexity for GAN client and IMS Client support on handsets. There is no additional complexity in providing support for IMS. Recent initiatives of IMS RCS (Rich Communication Suite), Standardization of IMS Service APIs (JSR 289, JSR 325) in JCP will solve any remaining issues. I don't agree that support for IMS will add any delay to the availability of LTE Handsets, potentially delaying LTE rollout. In anycase, what is the business driver for the operators to focus on voice in the initial launch of LTE?

3. By deploying LTE alone, mobile operators face a significant risk of becoming a fat dumb pipe, and loosing large part of revenues from messaging and even voice services - to over the top players. The solution lies in building a flexible service delivery architecture with IMS. I agree with the re-use of CS Core infrastructure and leveraging these investments. The re-use of MSC-S as ICS Enhanced MSC Servers or MGCFs or IBCFs can provide some investment protection. Operators can also avoid recreating all CS services in IMS domain, by adopting a smooth service driven migration from 2G/3G to LTE. This approach will focus on offerring subscribers a new multimedia telephony experience and not merely a replication of CS voice services

I believe that there will be co-existence of different approaches and much will depend on how the business case of different operators works out in their specific competitive and regulatory environment.