Will 4G wireless ever really happen?

In some respects, it’s a fair question. In others, it’s a definite yes.

The definite yes part comes from the nature of marketing. Some people are already starting to refer to “3.5G” service, and before long someone is going to trot out a half-baked extension to EV-DO or HSPA and call it 4G—not because the service has anything to do with what the global telecommunications industry is thinking of as 4G, but because some other product marketing manager will have already used 3.5G.

But given the potential impact that real 4G service could have on SoC design, the question is pertinent, and deserves a better answer. There are at least two legitimate ways to define 4G in order to make an answer possible. One is the way the 3GPP does it for LTE: in terms of 100 Mbit/s peak data rate for fully mobile devices and 1 Gbit/s for nomadic devices. Unfortunately, this definition lets nearly everyone except the end-user off the hook, by carefully avoiding the question of sustained data rates. It’s perfectly possible for a handset to have a theoretical peak rate of 100 Mbits/s, and yet to show no better performance than today’s 3G devices in actual use. It’s almost certain that networks meeting these peak criteria will be deployed.

The other way to define 4G—let’s call it Real 4G–is in terms of user expectations—the kinds of new applications that will work on a 4G handset. Here there is no room for deception: if an interactive video application falls behind, drops frames or loses its connection, it is not working. Such applications are insensitive to peak data rates, but are critically sensitive to sustained data rates. It is much less obvious that networks and devices conforming to this notion of 4G will happen within the next decade.

Much of the problem is technical. Real 4G requires about a half order of magnitude increase in the number of bits per second per Hertz of channel bandwidth compared to existing services. This, in turn, requires better noise floors, better channel equalization, better linearity, better data recovery, better compression, and after all of that, probably still spatial-division multiplexing with several antennas. None of those things will come easily—especially the antenna part. Just how do you put six antennas, far enough apart for MIMO and large enough to use in the bands carriers actually own, on a handset anyone would be willing to carry?

But the more serious problem may be economic. In a recent talk, iSuppli senior analyst Francis Sideco pointed out that Smart Phones—the high-end segment where new capabilities are introduced—are under pressure from less expensive Feature Phones, and their sales have been disappointing to vendors. The initial success and now price cuts on the iPhone—a device with Feature-Phone features and Smart-Phone pricing—are not going to help that picture much. Further, service providers have been under heavy price pressure to reduce the cost of high-access service packages, reducing the revenue they can recover from monthly fees, even for their full set of network features.

So are service providers going to spend a fortune to build out a complete new 4G network, incompatible in just about every respect with their existing networks, and maybe an additional fortune on new spectrum, all in support of phones that have a record of not selling well and services they have trouble billing for? I have a feeling that 4G is going to have to be presented to the service providers as a complete package–applications available and working, software on the shelves and equipment ready to roll in and plug in–before they will even look at it. And that scenario creates more chicken-and-egg problems than an industrial poultry farm.