Son of WiMax: the IEEE tries to clean up its committees
by Guy J Kewney | posted on 28 December 2006
As the WiMax hype machine moves into high gear ahead of next month's 802.20 working group meeting in London, UK, workers on the various wireless standards are anxious to calm down negative reports about the problems they might have on making this family of wireless work in true mobile fashion.
One of the problems facing WiMax is that Intel - a major sponsor - wants to see it used on the move, at high speed in moving vehicles, not just in coffee shop hotspots. Unfortunately, the WiMax standard, 802.16d doesn't cover moving nodes, and the mobile version, 802.16e, has issues with rapid hand-off from cell to cell.
One possible solution to the hand-off problem is the 802.20 standard proposal. The credibility of that proposal has bene under something of a cloud, partly for technical reasons (many engineers remain sceptical about it) but mainly for political reasons.
The politics got so bad that at one point, the IEEE actually suspended the committee that was working on the standard.
Exactly what happened, is in dispute. The official IEEE magazine went on the record with a full analysis which makes ugly reading. However, the committee is now back in action, and due to meet in London - and is hoping to redeem its reputation.
Little of what has been written by outside commentators has pleased the members. For example, the IEEE magazine, The Institute, wrote:
The working group, IEEE 802.20, was formed in 2003 as an offshoot of the IEEE 802.16 standard activity, often referred to as WiMax, the technology enabling fixed wireless broadband access as an alternative to cable and DSL. The proposed IEEE 802.20 standard would do that but with a twist: it would support broadband wireless for laptop computers and other devices used in fast-moving vehicles such as cars and trains.
It turns out that this is one possible interpretation of events. Committee members find it misleading. They also feel aggrieved at the number of reports of the suspension of the committee and the excessive excitement over the sorry details of how, and why, the committee was suspended.
The sorry details are, indeed, simply summarised. To quote The Institute again:
Intel, Kyocera, Motorola, Qualcomm, and other giants with stakes in the market all had representatives on the new working group, which numbered roughly 175 people. Members of working groups are expected to vote as individuals and not represent their companies’ interests.
But charges flew from the very beginning that members’ votes were driven by company loyalties. A disputed 2003 election of officers for the group led to allegations that consultants who had failed to disclose their affiliations with major industry players had participated.
In the third quarter of 2005, individuals affiliated with Intel and others feared that the group’s decision to cut the technology submission phase from six months to one month would not allow them sufficient time to prepare their proposals. What’s more, when they tried to get the group to consider their proposals they were repeatedly voted down.
By last January, the working group had narrowed its deliberations to a joint proposal from Qualcomm and Kyocera that could become a competitor to IEEE 802.16e, in which several companies, including Intel, have a big stake. IEEE 802.16e, an amendment to 802.16, addresses mobility and calls for operation at just below 6 gigahertz, while 802.20 supports access at bands around 3.5 GHz.
Employees of Intel and Motorola on the working group filed appeals with the Standards Board, challenging the group’s procedures. Qualcomm officials in turn accused Intel of using procedural maneuvers to delay the adoption of the standard.
In short, politics, inevitable with such powerful corporations all seeking to control the new technology.
But behind the politics, there lurk some genuine technical issues. Prime is the question of cellular hand-off.
This is - according to the Committee itself - not a problem. David Brunnen Managing Director of ABFL Groupe Intellex picked up on a comment piece I wrote recently, suggesting that it was a problem; and Brunnen fiercely denied it, with evidence: "You may have picked up the story that the recent Mobi commercial launch in Beirut means that The Lebanon is now catching up fast – fully mobile IP access indoors and out and on the move," he told NewsWireless.
He continued: "The technology platform for this is the same 802.20 candidate as used commercially (not trials) in countries such as Australia, South Africa, and Norway and has already been standardised by ATIS/ANSI as HC-SDMA. Extensive trials are underway in many other countries."
The committee takes the view that things are back on track, the technology has proved itself in the field, and the management of the standard is now whiter than white; so what's all the excitement about? It must be bad press.
We can expect more along these lines in two weeks, when the committee meets in London, and committee members have promised to clarify matters. But the idea that this is all down to press hostility is unlikely to hold water. As an IEEE member put it, off the record: "Any technical issue which involves Qualcomm is going to generate controversy."
Qualcomm's problem with the rest of the world is very simple. It owns - and does not share - a lot of technical patent assets.
In the wireless world, the normal way of managing intellectual property is that it gets traded. You have 20 patents, and I have 25 and Motorola has 40 and Intel has 50; and we all share them, and no money changes hands. Sometimes, the deal is quite simple. Other times it's the basis of a cartel, as with the GSM patents.
The exception is when Qualcomm is involved. It has a general rule (say its rivals) of hiring as many lawyers as technologists; it will let you use its IP, but only on a strict royalty basis. And the amount it charges is seen as prohibitive - to the point that the GSM spec was drawn up almost entirely with the objective of avoiding any technology owned by Qualcomm.
The result is that if standards bodies find themselves faced with the choice of developing a new technology, or buying something from Qualcomm, they will go to a lot of trouble to avoid the big Q option. That trouble will include glossing over quite serious problems with the alternative, say engineers.
For example, there's a press release, jointly issued by Korea's mobile operator, SK Telecom and Interdigital, relating to WiBro.
WiBro is closely related to WiMAX. Exactly how close you think it is, seems to depend on which giant electronic IP-owning corporation you work for. But, as the press release shows, it shares one technical feature with WiMax: it hasn't yet solved the cellular hand-off problem.
To quote my anonymous IEEE engineer: "I'm sure Moore's Law will make this problem manageable in some future time-scale, but right now, it's a serious issue because of the amount of processing power it takes to keep track of two cellular transmitters as you cross from one to the next."
Strangely, the problems are worse for data than for voice. You might imagine that a real-time stream like a voice call, which has to be synchronised accurately, would be more demanding; apparently, this isn't the case, for a number of reasons. Amongst them: a voice stream is very little data; it is buffered heavily, and there's a latency built into it to cope with interruptions. By comparison, a data stream tends to be required for high-data-rate applications, where we're looking at megabits per second (voice is happy with ten kilobits) and re-transmissions are to be avoided.
The 802.20 committee seems content with its progress on this technology issue. The WiBro authorities seem equally relaxed about how much work needs to be done. And the standard WiMax gang (the Intel lobby, at any rate) is staunchly optimistic that 802.16e is on target, and will Work As Designed within the expected time-scale.
Ultimately, all these problems are likely to be solved simply by the power of the typical processor chip shipping in the 2015 time frame. But sceptics say that the market may not wait that long.
The question that isn't easy to answer is this: "What problem are we trying to solve?"
As far as Intel is concerned, that's easy: laptop computing must be able to pick up the Internet anywhere. The WiMax dream is to have a system that does everything Centrino WiFi does today, without all the inherent problems of a licence-exempt wireless. WiMax won't be something any heating engineer or kindergarten supervisor or second viola player or recycling operative can buy off the shelf and stick up on a pole, creating neighbourhoods with 20 different transmitters all on Channel 6.
In the UK, the comms regulator, Ofcom, has already announced that it expects to make a bit of money auctioning off this "limited and valuable" resource to the WiMax operators. It will be a managed spectrum.
The problem is that Intel's requirements are not the requirements of the rest of the market. Intel provides comms for laptop computers. It's easy to see that a technology that suits a laptop PC user, will be no good at all for the mobile phone user; just look (for example) at the size of the PC display! - something that big and power hungry will never be used on a mobile phone.
And WiMax technology, today, is big and power hungry. It's all very well saying, as committee members do, that the problem will be solved, but the mobile phone market can't wait forever. And they have an alternative: the femtocell
The problem the mobile phone world is facing is a big one: it needs more bandwidth, but it can't afford to provide it with big cellular masts.
The cost of putting up UMTS cells is huge, and fraught with bureaucratic obstacles, as anxious neighbourhoods contest the erection of the masts on the grounds of possible health issues, or the effect on property values.
But there's the extra problem of just how many users you can support on a single mast; and it was always a problem which engineers knew they'd have to face, when the UMTS standards body decided to pick the 2.1 GHz frequency. It had many advantages, but the serious drawback of 2.1 GHz is that it's really not good at going through brick walls. In-building penetration is a real issue, and there's only one fool-proof solution to the problem: put the masts inside the buildings.
Femtocells are wireless access points. The only difference between them and ordinary WiFi access points, is that they transmit on mobile phone frequencies, and are licensed to mobile phone operators. Otherwise, they are high-speed wireless data in the home (or the office) for the user of the householder alone.
If you were Vodafone, and had a real problem in extending your access from current 3G levels, and you were told "You can use WiFi, or WiMax, or WCDMA femtocells" you would almost certainly take WiMax seriously, because of the option of owning the licence. You'd take WiFi seriously, if only because there are already a lot of mobile phones which can switch from UMTS to WiFi automatically; but ultimately, you'd really want to go for a solution that you can install in 2007, and which you'd own. And that's the femtocell.
WiFi is not as power hungry as WiMax will be; but even there, a WiFi smartphone runs a third as long on a single charge, as it does if you turn WiFi off and stick to standard cellular. A WiMax phone, built in 2007 (if there wre such a thing) would probably last half the time the WiFi phone did - maybe less. And it would be -crucially - no better at moving from one cell to the next than the WiFi phone is ... which is to say, crap at it.
Intel has made it clear that it doesn't think the phone market is going to buy WiMax. It put its money where its heart was, not where the mouth was. It owned one of the biggest, most powerful mobile phone processor designs in XScale; and it sold X-Scale off, abandoning all ambitions to own a share of the market.
Intel can make all the excuses it likes for that; bottom line is that it lacked faith in that technology for phones, and that if it had had faith, it would never have been content to let someone else reap the rewards of that faith. It baled out.
If and when the cellular hand-off problem is solved, and operators can see a network smoothly and swiftly allowing high-speed data users to drive a 140 kmph car while downloading large files reliably, onto a pocket-sized terminal that carries a full day's battery charge, they will be interested.
And at that point, they'll start looking at how many WiMax cells they can install, how man they need per square mile, what it costs per user, and how many users extra it will give them in the first five years.
Against that, they have to look at the number of households that already have broadband. Each of those households can be converted into a femtocell for around $50, with a payback period measured in months, maybe even weeks. They'll be encouraged to use their UMTS phone as the universal data gateway if they have that, meaning more paid-for traffic on their public masts. And the handsets will be very, very much cheaper and simpler to design, operate and fix, than multiple-radio devices which have to do WCDMA, GSM, WiFi, Bluetooth and WiMax.
Can the 802.20 committee find a way of avoiding that morass, and getting a short-cut to a UMA style solution that works for mobile phones as well as for PC laptops?
In another decade, we'll know what happened. Right now, the only thing that's certain is that Intel doesn't seem to think so.
First published at The Register
Technorati tags: WiMax IEEE WiBro 802.20
All will be revealed? - You can discuss this article on our discussion board.