The attraction of SDR stems from two quite different value claims, one pretty easy to accept and one more challenging. The first claim is that if the function of the radio transceiver is defined by software, which is easily alterable, then the function of the transceiver itself is easily alterable. This is a more complex statement than it might appear on the surface.

There are applications in which changing the functionality of a radio is a marketing convenience. For instance, a software-defined radio could presumably add features, such as security, support for data transmission or media data types, by simply altering the software, assuming that the right modules in the transceiver chain were in fact defined by software and that they had sufficient performance. That's a nice benefit for consumer-level appliances.

Great benefits are offered
But this ability to alter functionality looms much more important in telematics, wireless networking and cellular communications when one considers the almost total failure of global standards efforts. Even for a very mature standard such as GSM cellular, it is necessary to work in four different frequency bands in order to have a single handset that covers the global market. Add in the variations in center frequencies, bandwidths, modulation schemes and network protocols for all of the world's current and proposed 2G and 3G "standards," and you have a nightmare. And that's before even thinking about adding non-cellular functionality like location, networking or Bluetooth connectivity to the device.

Clearly there is great benefit to a manufacturer of base stations or handsets in being able to quickly switch between quite different configurations. But there is also huge value to the user, who might someday not have to carry a briefcase full of little plastic things just so he can make phone calls and communicate with the central office over the entire geographic range of his territory.

There are other applications where the ability to shift functionality is not just a great convenience, but a necessity. One example would be in secure military radios, where not just code hopping and frequency hopping but rapidly shifting between modulation schemes and network topologies is necessary. Another, less dramatic example is the ugly future of existing cellular and wireless services. There is growing concern that as the number and diversity of wireless sources will continue to increase. Hence designers are looking at extreme measures, ranging from multiple antennas to beam forming to adaptive modulation schemes that try to find some usable spectrum anywhere in the band. None of these ideas require a change to the external functionality of the transceiver, but all of them are predicated on the availability of great flexibility and very substantial processing power all along the transceiver chain.

Wrestling with a definition
Which brings us to the question of definition again. Just how much of the functionality of the radio has to be defined by software in order for it to be an SDR? Said Doug Grant, director of business development for wireless products at Analog Devices, Inc. (Norwood, Mass.): "You hear definitions all over the map. But the purist would say that for a radio to qualify for SDR, all the tasks would have to be either under software control or actually done by software."

The "or" in that statement conceals the other great ambiguity in SDR. Certainly some tasks will be executed entirely in software. As CPUs-especially configurable CPUs-and DSP cores increase in performance and adapt to the tasks, the number of such tasks will increase, gradually spreading back toward the antenna. But other tasks will remain, for the near future, "under software control."

That covers quite a range of ideas. For instance, in today's 3G base stations, chip-rate processing-the pre-filtering and rake filtering operations that prepare signals for symbol-rate operations-are often done in dedicated, hardwired datapaths. But these datapaths are configurable, and the configuration is done via software.

The actual radio-frequency circuitry is another difficult problem. No one has figured out how to implement an antenna in software yet. And while it is theoretically possible to design low-noise and power amplifiers, RF filters and switches that could be moved from band to band under software control, the only solution in use today is to duplicate the RF components-and usually the antenna itself-for each band being used.