Speaking Engagements & Private Workshops - Get Dean Bubley to present or chair your event

Need an experienced, provocative & influential telecoms keynote speaker, moderator/chair or workshop facilitator?
To see recent presentations, and discuss Dean Bubley's appearance at a specific event, click here

Wednesday, December 07, 2011

What is the cost of VoLTE? Is it a huge strategic error for the mobile industry?

I'm sitting at the Layer123 conference on EPC and LTE in London.

I've just heard a joint presentation on VoLTE from the GSMA and the MSF. The latter was about the recent interop trial, and the former was a general flag-waving / propaganda pitch. Apparently "everyone" is now agreed on VoLTE deployment. (I assume that comes from the same GSMA dictionary as "ubiquitous" used in discussions on RCS / RCSe). The subsequent presentation is from Germany consultancy Detecon, and the former one was from BT Wholesale.

Some outputs so far on LTE Voice:

- General agreement that CSFB still reduces QoE by increasing call setup time and dropping the LTE data connection
- "Pure" VoLTE should be available without handoff when the user moves out of LTE coverage in late 2012 / early 2013, after trials in early/mid 2012
- E911 support is a bit of a hassle for the US and a short-term cludge will be needed
- VoLTE with SRVCC (for handoff to 2G/3G at the edge of coverage) is late and complex
- Massmarket VoLTE with SRVCC "should" be available in 2014. I think that's both (a) delayed from initial discussions, and (b) very optimistic. I think 2016 is more likely when you take into account various other factors like devices

But the elephant in the room is the growing realisation in the industry that mobile voice telephony revenues are declining. A survey of strategists at the London Telco 2.0 event I attended a few weeks ago suggested 20-30% revenue erosion for operators on telephony in the next 3 years. The Asian equivalent in Singapore last week was even more bearish. Private conversations I've had with telco voice and strategy executives also demonstrate pessimism - sometime to the extent that they see telephony revenues evaporating almost entirely, under the onslaught from MVNOs, Google, Skype and the regulators.

So let me get this straight:

- Telephony revenues are declining because of external (and unstoppable) factors
- CSFB gives a worse user experience than circuit telephony
- VoLTE is only useful where there's actually LTE coverage
- Without SRVCC, VoLTE will drop calls (worse QoE than circuit)

Now in theory, VoIP means more calls can be squeezed into a Hz of frequency - so there are efficiencies on the radio and on backhaul capacity. But all the complexity of policy management, QoS, legacy interworking needs to be factored in.

Can anyone honestly say they believe that VoLTE will offer a lower "production cost" for telephony than GSM? Or are the costs per unit about to go up, just when the revenues are going down? And CSFB probably means higher costs AND lower quality than today's GSM.

That doesn't make economic sense - what operator will want to invest in more-expensive technology in order to provide a worse service in a declining market?

EDIT - a couple of people have pointed out that CS voice infrastructure isn't that cheap either, and neither are the legacy billing/OSS systems. True, but modern developing world 2G/3G networks show that CS can be really cheap if deployed today. I think a CS refresh (or full OTT-style service with a chance of Future of Voice type new revenue streams) may be a better bet than VoLTE.


robrobstation said...

First off what I don't understand, if my Wi-Fi radio can maintain a connection while I am on the phone then why should LTE drop if I make a CS phonecall over GSM or 3G? Surely this could be remedied.

Second thing is: GSM and 3G have PLENTY of capacity for voice and SMS traffic and will for years to come through tech such as UMTS 900 and innovations like Ericsson VAMOS.

LTE is a standard mainly developed for delivering data over IP. GSM/3G works pretty well at delivering voice at the ever shrinking price point with little major capital investment needed anymore. Why can we not run two standards that each do a particular job well?

Steve said...

If only there was a technology which could use the existing 3G/GSM service core to deliver voice over LTE…

- It would have the exact same features as a voice call over 3G/GSM.
- It would actually have a *faster* call setup times than UMTS.
- It would require no added investment in the existing 3G/GSM RAN or core.
- It would deliver what people want - a simultaneous voice & data over LTE experience.
- It would dramatically simplify in-call handover between LTE and 3G/GSM.
- It would be available today (actually it was working at MWC 2010).

Yes, it’s VoLGA. It’s ready, waiting, and extremely cost effective.

Dean Bubley said...

Rob - Yes, I agree. I'm an advocate of the "dual radio" approach (keeping both LTE and 2G/3G radios up simultaneously) as an alternative to CSFB.

The incremental battery use - and some complexities with registering the subscriber to two networks at the same time - do not outweigh the convenience & usability gains.

Steve - I agree, and it was interesting that VoLGA was referenced a couple of times yesterday almost wistfully, including by a major operator. Seems to have been a victim of timing and "standards body politics". I suspect it is due for reincarnation. Hopefully with a better name this time.

Alex - yes, you're missing the simultaneous dual-radio (sometimes "dual-standby") approach as well. It's already used for LTE/CDMA phones although the first devices had early chipsets & poor battery life. I believe that is fixable though with next-gen silicon and better-designed RF elements.

Also you're missing a possible reincarnation of VoLGA (or something similar) which is essentially "CS-over-IP"

CSFB has many flaws - eg longer call setup times, and dropping the LTE data connection when you make/receive a phone call.

Sami Jokinen said...

Remember a time when data was carried over CS connections, even in fixed networks? It was partially because data was only a fraction of the traffic generated by voice. Now, imagine a (not so distant) future where voice is only a fraction of all the data carried in networks, including wireless networks. How long will it be cost efficient to maintain a dedicated network (including CS core) for voice service?

You seem to imply that the delay introduced by CSFB automatically leads to bad, or very bad, user experience – please quantify and justify your position. Also, I find it interesting that you have faith in engineers solving issues caused by dual-radio, and yet do not trust CSFB or SR-VCC can be made to deliver good user experience.

GSM networks will be there for quite some, and those who do care for the benefits of LTE may continue using GSM handsets. I’m sure someone still prefers POTS. In short term, for some the benefits of LTE data override potential service degradation caused by CSFB.

These comments are based on my personal views, and may not reflect those of my employer.

Dean Bubley said...


Thanks for your comments.

In fixed networks, the profitability argument is moving against CS, although it has taken 10-15 years to get to 15% penetration of VoIP on the access network replacing PSTN.

Mobile will take *much* longer to transition to VoIP as the networks are newer and better-optimised than fixed CS ever was. (Think of networks in India making money at $3 ARPU).

There are certainly open questions about the rate of decline of mobile telephony revenues - and so yes, even GSM may one day be unprofitable, especially with older networks and European opex costs.

However, VoLTE certainly doesn't solve that issue - and CSFB probably makes it worse by adding more cost & reducing quality. Both GSMA and Telefonica mentioned CSFB's problems (eg 1-2 sec extra latency) at the event yesterday. The setup latency may be fixable, but I can't see how the data dropping can be - and that's much more important.

Dropping a fast data connection just because a phone call comes in is a 50-year throwback. How many calls are worth interrupting app usage for?

Dual-radio already exists for CDMA/LTE and chipsets are moving to the 2nd generation. I generally have more faith in handset engineers understanding the nuances & pyschology of user experience than network engineers.

The optimum solution is an always-on LTE data connection, using a full and flexible VoIP platform (beyond just legacy telephony) if it is feasible, and using *simultaneous* CS if it's not.


Sami Jokinen said...

I think IP trunking is more relevant reference in fixed networks than VoIP penetration.

"The setup latency may be fixable, but I can't see how the data dropping can be - and that's much more important."
I don’t quite see what the issue is in CSFB to WCDMA; WCDMA/HSPA has multiRAB and with it simultaneous voice and data. Or do you consider HSPA+ not fast enough for apps during voice call?

How does “full and flexible VoIP platform” solve declining operator voice revenue?

Dean Bubley said...


No, I disagree about the SIP trunking analogy. End-to-end VoIP is if much easier in fixed networks than mobile - you can use an ATA to connect an old handset, the broadband is pretty predictable, no mobility, low expectations for user experience / UI.

Yes, perhaps HSPA+ speeds are OK for data concurrent with telephony. But that's dependent on coverage - and assumes that the 3G network isn't congested. It also needs extensive testing to make sure it doesn't break applications because of IP addressing effects, or take extra seconds to reconnect. It might work, but I'd need to see proof & ongoing service assurance.

> How does “full and flexible VoIP platform” solve declining operator voice revenue?

No easy answers. As an operator said at a conference recently "Revenue rebalancing will happen anyway". But done well, with vision and innovation (and execution) there may be new voice services beyond telephony that add generate revenues, or which can add value/loyalty to a bundle. I talk about the sources of value in voice communication in depth in my "Future of Voice" workshops and private advisory sessions.

Dean Bubley said...

One other thing.... what happens to RCS/RCSe during CSFB events?

Presumably, that's the "most likely" data application to be running concurrently with a call, if you believe the RCS hype.

In particular, I'd assume that dropping to CS with an interruption in data connection could cause various problems with video-sharing (eg audio/video being out of sync). We'd also have the weird situation of using CS for voice and PS for video...

I'm assuming that someone has thought about this, but I haven't seen anything obvious.

Fraydog said...

I don't think the dual radio model as done by Verizon and AT&T here in the US is all it's cracked up to be either. In Verizon's case the glue that connects the old EV-DO network to LTE is eHPRD. To say it's janky is an understatement. Occasionally, you will get the wonderful nationwide outage because of the eHRPD glue coming undone. Fun times!