Fujitsu chip reflects growing complexity of WiMAX network architecture

Fujitsu’s recently-announced WiMAX Mobile Wave-2 MAC/PHY chip reflects in its architecture the growing complexity of the Mobile WiMAX network problem—not so much on the surface as in the details.

On the block diagram level, the chip seems fairly straightforward. The SoC includes a baseband processor supporting the necessary MAC and PHY functionality, plus the data converters and driver/receiver pins to attach an external RF module. The chip is intended for 2-by-2 MIMO operation, but without beam-forming functionality. It has an added feature of a GPS receiver to implement both network organization functions (more on that in a moment) and whatever location-based functionality regulators may impose on the network one they notice it.

Two points hidden in the fine print are in a way more interesting than the datasheet information, however. One is scalability, and the other is that GPS block. The scalability, according to Fujitsu Microelectronics marketing manager for wireless solutions Dean Chang, is provided so that users may implement any size base station from microcells to picocells to femtocells with the chip. Femtocells?

The underlying issue here is that at the frequencies the WiMAX people have selected, in the 2-3 GHz range–almost every medium short of free space is highly absorptive. That means the familiar scheme of cellular base station antennas scattered around the neighborhood on big towers is of marginal value. It’s probably realistic to say that Mobile WiMAX will only deliver interesting bandwidth on line-of-sight connections over short distances, like hundreds of meters, if you happen to live in a humid or polluted climate. That, in turn, means that WiMAX can’t get by with the sort of regular geometric cell structure used by today’s cell phones. (This may be equally true of LTE, but that’s another discussion.) So the network architects have proposed an hierarchy of base stations.

You start out with a cellular-like topology, complete with the big towers. Then you sniff around in the vicinity—or simply wait for complaints to pour in from subscribers—to identify shadow areas. Into shadow areas you put microcell base stations, capable of supporting a moderate number of clients in a rather limited area.

But one of the best absorbers of 2 GHz seems to be building materials. So once you have augmented your cell architecture with microcells, you now have only covered the out-of-doors. You still have to put even more-localized base stations inside any structures—each floor of an office building, for instance—where you want subscribers to have indoor coverage. Those are the picocells.

Picocells, though, are too powerful and expensive for home use. So if you want your subscribers to actually use their devices at home—a concept that seems to come very hard to today’s 3G service providers—you need to put yet another level of small base stations in subscribers’ homes. Those, in case you haven’t been keeping count, are the femtocells.

Now the big problem here is that cell sites can interfere with each other. And, at these frequencies, their coverage patterns will depend on weather and the locations of large moving objects such as cars, bicycle commuters, or people in their living rooms. So to avoid interference, the service provider will have to install the first three levels of base stations—cell through picocell. Current thinking is that femtocells can be user-installed, much as 802.11 hubs are today.

The catch is that, because transmission conditions vary so much, and because the end-users’ femtocells are basically portable, the network will have to be dynamically self-organization. If it clouds over and a microcell starts interfering with a nearby cell, the network may have to reduce the microcell’s power, or even shut it down. If an individual takes his femtocell hub out into the back yard for a barbeque, the network may simply turn it off. Mind you that while the network is shape-shifting, it must also support mobile clients without dropping broadband connections.

Hence the need for GPS. The self-organizing network may need to know the precise location of all of its cell sites, including the femto ones. And each base station must be agile in terms of power, frequency bands, and modulation scheme. So a little femtostation may suddenly find itself instructed to handle a quite compute-intensive baseband load.

This represents an interesting challenge for the chip architects. Fujitsu believes it has put all of the base hardware capabilities and software scalability into the PHY/MAC chip to make such a network implementable. And it is seeing interest from users who want to use the chip as part of a development vehicle, as well as for early deployments.

But it represents another order of problem altogether for the network architects. Chang says that members of the Service Provider Working Group and others are working now—one would think rather frantically—on finding algorithms that could permit a network to self-organize dynamically on this sort of scale. It seems like an odd choice of problem to leave until last.