Intel opened the doors on its radio work yesterday, allowing us an early glimpse into what it has promised will be a decade long project. Labs leaders said the culmination of integrating analog and radio into silicon would take that long, or at least - as they assured us several times - that they would have retired or left the company by the time it would be achieved.
No other semiconductor company can afford to invest as much in R&D, and when so many companies view research as a short-term cost rather than an investment, this kind of long term investment is to be applauded.
With the goals so distant, it was a commendably relaxed and open session, and when we spoke to Intel Fellow Kevin Kahn about the far-horizon work we saw from the Labs in the context of the nascent WiFi Bubble, he acknowledged claims were getting out of hand.
"There's never been a technology that hasn't been overhyped," he told us. "Is it being overhyped? Of course. Do we contribute to overhyping it? On occasion, probably yes. But there are real things in engineering terms to be achieved too."
Indeed, Intel didn't fudge the scale of the task ahead, and we'll go into these challenges in a moment. But Intel has decided it needs more spectrum to work with, and here Kevin's the man.
As well as overseeing all Intel's RF work, Kahn serves on the FCC Technological Advisory Council, chairs the National Science Foundation's Engineering Advisory Council, and is Intel's point man on regulatory issues.
He told us he had found Powell far more conducive to suggestions that the FTC regards the spectrum scarcity as 'artificial'. But he didn't find the fanatical deregulation lobby - the all-or-nothing 'Open Spectrum' ideologues - particularly helpful either, although they did raise awareness of the issue. (Here's an FAQ summarising the Open Spectrum lobby case.)
"Some people have a quasi-religious belief in it, ours is more pragmatic," he told us.
He thought it "reasonable" that incumbents - broadcasters, wireless carriers - should be respected, which is nice, but the case had to be made for new applications. At one point Kevin referred to the objectors to spectrum deregulation as "antibodies", and said that it wasn't practical "to wipe the slate clean" - presumably wiping away those antibodies.
But probably far more useful to the cause than 'Open Spectrum' is Module Certification. As Jeff Schiffer, co-director of Intel's wireless interconnect lab pointed out, allowing certify-once and use-anywhere would be pretty useful to vendors.
Kahn has been taking his case to the Europe and Japan, participating in ETSI proceedings and talking to the European Commission.
The endgame, says Intel, is "reconfigurable, intelligent CMOS radios". The first step will be an integrated radio chip that's cheap enough to be appealing. Radios are very, very low cost already thanks to dedicated baseband processor and DSPs. Intel thinks it will be possible to produce an integrated chip that's cheaper, which is probably the claim that was met with the most skepticism yesterday.
How cheap? Nathan Brookwood, principal analyst at Insight 64, pointed out that "analog is not amenable to Moore's Law", and radios are already very cheap indeed: the disposable handset is almost upon us.
The technical challenges fall into two strands (regulation is cited as a third). Which is, integrate the multiple frequency bands onto one die; and create a re-usable baseband architecture. The latter is currently performed by discrete DSPs, and explains why DSP pioneer Texas Instruments has its own roadmap to a one-chip wireless device.
There's a lot of research going into smart antennas and into analog to digital conversion. For the former, Intel showed a demonstration of a four-antenna laptop which effectively doubled the range of an 802.11 signal.
These will be pretty big chips, but Intel thinks it can achieve the flexibility of a FPGA at very low cost. In, for example, what Intel calls RCA (Reconfigurable Communications Architecture) the processor is designed to configure itself to run different protocols on demand - essential given the plethoria of 802.nn wireless standards. We look forward to a closer look a the technology. This was the first time much of this had been unveiled, and Intel seems keen to let us take a closer peek.
Cost remains the ultimate challenge. Even one-chip baseband seems to meet with a lot of industry skepticism: perhaps the sheer volumes simply do not favor Intel, here. Economics and recent history suggest that putting all you can on a chip - maths co-processors, peripheral interfaces - is inevitable, but that's only the case if the integration costs create a real saving. With ubiquitious DSPs, this isn't necessarily the case.
Actually, we reckon there are some technologies that achieve mass acceptance without much hype. The combustion engine was one, and the humble cellphone was another. Both reached global consumer acceptance - thanks to Ford in the first case, and to co-operation between vendors and regulators in the second - and the 2G phone has only recently been hyped, in retrospect.
Intel's biggest problem - it realizes this - is persuading the regulators that something needs to be done.
The most telling moment of the 'Open Spectrum' Summit organized by Lawrence Lessig in February (notice how the words "Free" or "Open" are obligatory buzzwords) was when John Gilmore stood up to remind the conference that no matter what was decided in the room, nine-tenths of the world would ignore it. That's because in much else of the world, there isn't perceived to be a spectrum problem, so there isn't a demand for a fix. (This is where the words "Free" and "Open" play their part).
Given all the obstacles, you might surmise that Intel is crazy to pursue radio. Then again, given the facts that volumes will be with personal and M2M (or machine-to-machine communications); that it can invest for the very long term, has very good engineers and R&D often brings about unintended consequences, it would be pretty crazy not to try. ®