Practical Chip Design

IMEC’s software-defined radio (SDR) research team has announced a new phase in its research program. The effort to date has passed some very significant milestones: a C-programmable VLIW core tightly coupled to a configurable ALU array, a configurable analog front end and digital front end set, an ultra-low-power baseband ADC, and an integration of all these cores into a compact reference design for multi-standard mobile terminals and handsets. With tape-out of the final chip in this set in the near future, IMEC will have essentially completed the 3-antenna, MIMO, multi-standard terminal reference design that was the goal of the current project.

Now, according to wireless-systems science director Liesbet Van der Perre, it’s time to move on to the next step: from configurable radio to self-configuring, cognitive radio. Van der Pere explains that with the average number of radios in a mobile terminal projected to reach six by 2009, SDR has become almost a necessity. But the problem isn’t stopping there.

Van der Perre observed that the number of standards and the number of terminals will continue to increase. Both spectrum and physical space will get increasingly crowded, until the likelihood of a single-mode handset being able to achieve a connection at all will become too low. In this environment, in order to get and sustain a connection, a terminal will have to be agile: able to sense any service in available spectrum and use it. That, at a very general level, defines cognitive radio.

There are new technical challenges, Van der Perre said, primarily related to cognitive radio’s need to continuously search a broad slice of spectrum for signals, isolate, and categorize them. This process requires an always-on detection front-end, and a fair amount of continuous baseband processing—much of which would in a normal handset only occur at the inception of a connection.

These requirements, in turn, complicate the greatest challenge in the current SDR design: energy conservation. Even SDR gives up a great deal of energy efficiency in exchange for software programmability—energy that must be recaptured through exacting dynamic power-management algorithms. Greater use of parallel computing, even more aggressive power management, and tricks such as a special low-power spectrum monitor circuit, will be very much in order.

Even the semiconductor process will be a challenge. Van der Perre says the 45nm process necessary to achieve the program’s performance, density, and power targets is not an ideal medium for implementing the 60GHz analog front end the project requires. RF designers are already busy on that one. But with such problems, cognitive radio will make an excellent platform on which to stress IMEC’s process, integration, and tool development programs, and so Van der Perre’s group is likely to remain a vital test bed for the research consortium’s other efforts.