Columnists

TI Developer Conference: The future of mobile and fixed VOIP

Conference sessions take on Asterisk, voice quality, power consumption for Skype over Wi-Fi.

By Brian Dipert, Senior Technical Editor -- EDN, 3/12/2007

After many years spent attending vendor-specific technical conferences, I've determined that you can tell a lot about a company (both its current health and its future aspirations) from a perusal of the course catalog.

Take Intel, for example. For several years after the company acquired Level One (along with its other communications-centric acquisitions that predated the dot-com Bubble Burst), its then-twice-yearly Developer Forums featured a notable dearth of traditional microprocessor content. The visibility of the company's flash-memory portfolio has similarly waxed and waned over the years, reflective of the amount of cumulative fab network capacity the group was allocated at any particular point in time, coupled with its degree of success in filling that capacity with existing customer demand.

Or take Microsoft. Back in the days when it was battling Netscape for control of users' browser experiences, Internet Explorer-related technical sessions dominated conference agendas. For the last several years, Windows Vista has controlled the stage. And, I suspect, Web-based services will garner a disproportionate share of presenter attention in this and future years' Microsoft conferences that I attend, reflective of the growing Google threat to Microsoft's market dominance.

So what did I discern about Texas Instruments' aspirations from its Developer Conference (which I'm in the midst of flying home from as I write this, after a several-day layover in Las Vegas at PMA) focus?

A lot of what I encountered at the TIDC was predictable, given the company's longstanding DSP heritage. Audio, communications, industrial, systems, tools, and video-and-imaging tracks were robustly staffed with courses and personnel. One class categorization pleasantly surprised me, though. TI devoted Thursday to a series of presentations on QOE ("quality of experience"), and VOIP was the dominant application discussed there.

VOIP also leaked into other tracks; a session on Asterisk in the communications track on Wednesday afternoon, for example, along with one on convergence in that same category Thursday afternoon and a tools class on Wednesday. VOIP, both wired and wireless, is an application that I've closely followed over the past several years, so I appropriately focused my schedule attention in that area.

My VOIP experience at TIDC started off on the wrong foot when, in the Asterisk session, Mark Spencer (the Chairman and CTO of Digium, and the initial developer of Asterisk) claimed that cellular phones are the best thing that could have ever happened for VOIP, since they pre-condition consumers to tolerate varying audio quality, long delays between speaking and receiving a response, and occasional service interruptions.

While I buy his argument for wireless VOIP (a technology with which I've had quite a bit of personal experience of late, both from an infrastructure standpoint and with respect to equipment, and which I aspire to relate to you in the near future in my blog), I strongly feel that "landline" VOIP must mimic the generally-reliable POTS experience as closely as possible if it is to garner substantial interest from potential customers.

Fortunately, presenters at the next-day convergence session scaled back the overly broad Asterisk enthusiasm, agreeing with my stance. My opinion is a reflection of the oft-touted Drucker Rule (attributed to famous business consultant Peter Drucker), which suggests that an emerging technology must be 10 times better than the incumbent approach in order to have a reasonable likelihood of replacing that incumbent. Wired VOIP is substantially cheaper than POTS, particularly for international calls, thereby satisfying the 10× rule, but degraded voice quality versus POTS would similarly degrade its Drucker advantage.

With respect to wireless VOIP, which was the primary focus of the convergence session (and others at TIDC), I further extrapolated the Drucker Rule with my Q&A question to the panelists. Here the incumbents are CDMA and GSM, and the relatively long battery life with cellular voice (several hours) versus Wi-Fi'd Skype (several dozen minutes) on my dual-mode iMate SP5m has left me with no shortage of skepticism regarding the practical feasibility of the wireless VOIP upstart. TI's Michael Yonkers, director of technology strategy for the Wireless Terminals Business Unit, didn't attempt to dissuade me from my doubts in the near term. But, as I'd already predicted in my head (but didn't share with the panelists) he assured me that he saw a clear path to wireless VOIP being more battery-efficient than cellular voice by the end of this decade.

The main three avenues to this flip-flop of today's status quo should come from the ascendance of the industry-standard SIP protocol (which Yonkers claimed is inherently less CPU-demanding than Skype, even when fully implemented in software), the predictable migration of portions of the SIP processing pipeline from software into function-dedicated and power-optimized hardware on the company's planned VOIP-optimized SOCs, and continued power-cognizant evolution of, along with more intelligent use of, the Wi-Fi transceiver.

I can certainly relate to Yonkers' claims regarding Skype's CPU demands; with early revisions of the Skype client for Windows Smartphones (and before that, a hacked Pocket PC-intended client), I had to substantially overclock my phone's 200-MHz OMAP processor (thereby further clobbering battery life) in order to obtain a glitch-free Skype experience. But I'm also struck by the significant market penetration that Skype has cultivated over the past few years, first with voice-only communications and now with video-enhanced sessions.

I got a consistent "no-comment" response whenever I asked various company representatives about their plans for more comprehensive Skype support. TI's Voice-over-IP and DaVinci-fueled video-over-IP groups might prefer to focus their attentions on SIP, thereby precluding the need to take a Skype license. But I'll wager that sooner or later, they'll also need to pull out their wallets and finance a more feasible Skype experience for their systems development partners, as well as for end system users.