LTE Smartphones At CES – The Way Forward?

At the recent CES a number of companies seem to have shown the first CDMA/LTE smartphones to be launched at some point later in the year. From a telecoms point of view the interesting thing is how these phones will do voice calls while connected to an LTE network. The issue here is that LTE doesn’t have a built in voice service like GSM and UMTS. While most of the press mostly commented around the design, screen size, processor and other features, this topic was completely overlooked.

Steve, over at the Voice over LTE blog, however, comes up with some interesting answers on how voice service is implemented. According to his post, these LTE phones will be connected to LTE and the CDMA network simultaneously so IP based traffic can use the LTE network while voice calls and text messages can be handled simultaneously by the CDMA network.

From a 3GPP GSM/UMTS engineer’s point of view this is, well, lets say, outmost sub-optimal. In the world of GSM and UMTS a lot of care is taken that voice calls and data sessions can move seamlessly between the two technologies. Ongoing calls will be handed over between the technologies when running out of UMTS coverage, for example, and the IP address and session contexts will also be preserved when hopping between the two radio networks. Voice and IP services even run concurrently, something that doesn’t work in the CDMA world at all. For LTE, the same mechanisms have been standardized for IP sessions and a number of different solutions exist to hand over ongoing LTE voice calls to GSM and UMTS.

So why are those phones not using these solutions? For one thing, they are designed for CDMA 2.5 / 3G network operators and from what I can tell they can’t use CS fallback, because that’s specified for fallback to GSM and UMTS only (please correct me if I’m wrong here). Also, Voice over LTE via GAN (VOLGA) is not an option for them as it is based on GSM and UMTS technology. That leaves them with VoLTE, which is an IMS solution with a specified feature set. Talked about in the industry for pretty much a decade it’s still nowhere to be seen and the announcement of dual-active-radio phones for LTE indicates that showtime for that technology won’t be anytime soon. Hence, their move to dual-active-radio phones.

Apart from the opinion of a 3GPP engineer, is a dual-active-radio device a solution that the customer could like? After all, the customer isn’t interested in sleek designs under the hood, but whether the device works well or not. So from my point of view, I as a customer want my smartphone to do two major things: It has to provide me with a highly reliable voice service anytime, anywhere and it has to provide me with a fast and reliable Internet connectivity for my IP based applications such as email, web browsing, Twitter, etc. etc.

So the question is, will those dual-active-radio phones provide that? From what I can tell the answer is yes, with some restrictions. The first one is the potentially higher power requirements to keep two radios listening to two networks at the same time. However, if I get a full day of use out of the device then I, as a user, don’t care to how many networks the device is connected to simultaneously. If those devices can do that is something yet to be seen. Secondly, voice service. Yes, the CDMA network layer provides that and as the radio is turned on all the time, my voice service in terms of availability and call setup time should be in the same ballpark as current 2G/3G phones. Furthermore, I wonder if the first generation of the CDMA/LTE networks and the devices themselves can switch the IP context  between the different networks. If not, then applications will loose their connection to the network every time a switch between LTE and CDMA is done for data services.

Loosing the IP context between CDMA and LTE might come in as a big argument against such a solution. But is it really an issue? Today, most smartphones can use both 3G and Wi-Fi for internet connectivity and automatically switch between the two radio technologies. Here, IP connectivity is also lost and applications have to re-connect to their server on the Internet each time the switch it made. Not very pretty from a design point of view but it seems to work in practice already today without anyone loudly complaining.

Oh, yes, and before I receive some comments on the Skype implementation on the phones that was reported, yes, I’ve seen that, too and I very much like it as an added goodie. However, I’d never rely on it as my main telephony service for a number of reasons. The most important one is that I often make calls while moving and often implicitly use, what I  think is the network operators biggest asset, the capability to hand over ongoing calls from 3G to 2G.

I find the thought around Wi-Fi very interesting. In effect, if the IP context can’t be taken over from LTE to CDMA, then the LTE network can be seen to do the same thing Wi-Fi already does today in 3G phones. So instead of having only a Wi-Fi chip, LTE smartphones have an LTE chip in addition that can also provide IP connectivity. And that, I think, is a crucial point in the discussion and way of thinking of whether it’s a good idea to have dual-active-radio LTE phones, which in fact will be triple-active-radio phones, as Wi-Fi is switched on all the time as well.

So the final train of thought for this post is whether this approach could and should work in the GSM and UMTS world as well!? Let’s first look at the “should”. From what I can tell, CDMA operators are keen on launching LTE because they depend on it as an upgrade path from their CDMA network technology that is not evolving anymore and is capable, from what I hear, to deliver around 1.5 MBit/s but not much more. Also, from a capacity point of view, it’s much more limited to what’s currently going on in UTMS networks. On the UMTS front, there’s a healthy evolution program in place. Many networks are already upgraded for speeds up to 7.2 or 14.4 MBit/s in the downlink, with 21 MBit/s in the downlink and dual-carrier 42 MBit/s on the way to reality. Also, network capacity keeps rising with more advanced devices coming to the market all the time that make better use of the resources, use of several carriers and densification of the network, all while preserving backwards compatibility with GSM and concurrent voice and IP based operation. So why go for the compromises of a GSM/UTMS/LTE phone that has to have two radios switched on at the same time? You can have it all on the evolution of UMTS for smartphones, while LTE is being used for non-voice devices such as netbooks, LTE USB sticks, pad computers, etc. For me, a GSM/UMTS/LTE phone only makes sense once voice over LTE works as good as voice over GSM/UMTS, including, and that’s most important, handover to GSM and UMTS. As an operator, anything less doesn’t cut it, Skype and others are here today. Whether an operator bets on IMS VoLTE or VOLGA, or a combination of both will be sorted out by competition.

And finally for today the “could” part. Running GSM/UMTS and LTE separately at the same time is going to be a bit of an engineering challenge. As per design, a single SIM card can only supply the credentials to be active in one radio network at the same time. So to be active in GSM and LTE at the same time would require two IMSIs (International Mobile Subscriber Identities). You can’t do that with standard SIM cards today. Of course one could think about solutions such as one SIM card having two IMSIs and two secret keys but why work on this when the real issue is to have Voice on LTE with seamless interworking with GSM and UMTS?

In summary, I think the answer for CDMA networks is to have dual-active-radio CDMA / LTE phones until they’ve got the voice question solved as they don’t have many other alternatives to evolve their networks and live with the downsides for the moment. For UMTS network operators, I think the equation is different as the strong evolution path of UMTS and using LTE for non-voice centric devices makes dual-active-radio designs not very attractive.

Comments

[Benjamin Lovell]

So one of the things I keep wondering about doing voice on LTE is, once they finally get there(3 years, maybe 5), will anyone care? Granted Skype and 100 like them have not yet taken over the world but they are increasing rapidly and at cost points plus feature velocity that no carrier could hope to match.

I don't expect to see any large US or international carriers offering data only plans for smartphones to encourage this next month or even next year but I think it's starting to look more and more likely as I read about subscriber and feature growth with services like Skype along with the technical difficulties, cost, and time involved in moving over to something like IMS. Some statements I have seen in trade rags about the Skype and Verizon deal earlier this year included Verizon setting up a detected bearer channel with QoS guaranties for Skype calls.

Is there any reason a carrier could not simply use a detected data bearer channel with QoS guaranties for voice traffic. Wired IP networks simply promise to not drop packets marked with DSCP EF for voice. I could see a data plan with two services levels. One for bulk data traffic and another for realtime traffic. Your realtime allotment would equate to some amount of voice or video. Granted there would be very little chance to up sell value added services in this model but they're going to lose that battle anyway and might be willing to take what they can get once their voice revenue starts declining.

Just some thoughts from someone who came up in a packet switched world and has a startling lack of respect for legacy business models. ;)

[przemobe]

Regarding to LTE and CDMA IP context: From my best knowledge RAN vendors correctly have or almost have solutions for LTE to HRPD handover in their portfolio. So depending only on operators and UE vendors there is no worry about dropping data sessions when switching between these technologies.

[GutEQ]

A very rough thought on IP context issue you mentioned: over time, if no satisfactory solution is provided by vendors and operators, application providers might develop their own replacement, application level soft-handover: having multiple IP sessions over multiple transmission media (LTE/EVDO/UMTS/WiFi/other) established at the same time as these media become available to the mobile terminal, with multiple (yet logically associated to the same 'application subscriber' logical entity at the application level) data streams being sent over these media and with "stream selection" (P2P-alike) logic built into application on both ends, client (mobile) and server (service). This kind of solution might also provide superior use experience due to data redundancy (multiple streams) at the expense of excessive network resources usage (redundancy again).

As for the problem of single IMSI preventing simultaneous idling in two radio networks, the not-very-elegant-but-probably-effective solution that comes to mind, would be a UICC with single GSM/UMTS-only identity and LTE IMSI provisioning OTA e.g. based on STK.

[Brent Bischoff]

Regarding the mention of using CSFB for CDMA, have you any thoughts on CS Fallback to 1xRTT? I would be very interested on your technical analysis and which CDMA operators are considering this as a temporary LTE voice solution.

[David Boettger]

The dual-radio solution solves some problems that are unique to the CDMA/EVDO community and thus probably shouldn't be thought of a potential voice solutions for pure 3GPP operators. CS fallback to CDMA (which *is* supported, see TS 36.331) would require two tricky issues to be solved. First, the CDMA (voice) and LTE core networks would have to be able to communicate, such that paging and SMS notifications could be sent from the CDMA net to LTE. Second, as you point out, it's sort of pointless temporarily to move a data session from LTE to CDMA since CDMA can't handle simultaneous voice and data. So dual radios solve these problems.

As przemobe notes, the HRPD specs enable data session mobility from (usually) EVDO to LTE and back, but the primary use case for this functionality is a "layer cake" situation where the EVDO coverage footprint is much bigger than the LTE coverage footprint, which will be the case for a long time.

Note, too, that there are some U.S. operators with relatively odd network configurations which find dual-radio handsets particularly appealing. MetroPCS, for example, has CDMA and LTE but no EVDO.

[Jamncl4]

As stated in previous comments, there are multiple standard methods for providing data (IP) mobility between CDMA and LTE, e.g., eHRPD, among other techniques. Handover isn't just a radio thing when it comes to IP. Your core must be able to support the handoff for seamless session continuity.