DARPA HEALICS: A Self-Healing Mixed-Signal Baseband Processor for Cognitive Radios

Professors B. Razavi, D. Markovic, A. H. Sayed, D. Cabric, J. Woo

The congestion in the frequency spectrum continues to rise as more users access wireless networks. Cognitive radios (CRs) offer an approach to alleviating this issue: they continually sense the spectrum and detect and utilize unoccupied channels. While present efforts in CR design have focused on the TV bands below 1 GHz, it is expected that CRs will eventually operate from tens of megahertz to about 10 GHz.

The DAPRA HEALICS initiative has provided a framework for the development of cognitive radios with the aid of "self-healing" circuits. Such circuits incorporate innovative techniques to overcome device and system imperfections, targeting an order of magnitude improvement in the performance.

A challenging task in cognitive radios is spectrum sensing, i.e., determining which channels are not occupied. In the presence of "shadowing," the system must identify signals that are as much as 20 dB below noise. Channel-by-channel sensing greatly eases the design of the baseband analog-to-digital converter (ADC) but it requires an inordinate amount of time for sensing a large number of channels.

A more practical method is to take a spectral snapshot of a block of channels. Unfortunately, in the presence of numerous occupied RF channels, the ADC must achieve a wide dynamic range and a high bandwidth. In fact, the probability of finding an available channel is directly proportional to both the resolution and the sampling speed of the ADC: downconversion of a broader block of channels raises the probability but demands a proportionally higher ADC resolution because the number of high-power channels is also potentially larger. It is therefore desirable to maximize the dynamic range and speed of the ADC while maintaining low power consumption.

Shown in the figure, the mixed-signal cognitive radio baseband processor envisioned in this program consists of a high-speed, low-power analog-to-digital converter (ADC), a phase-locked loop (PLL), and a spectrum sensing processor (SSP). The ADC and PLL employ self-healing techniques to achieve unprecedented performance, and the SSP implements new algorithms to enable wideband sensing and orchestrate the self-healing of the ADC and the PLL.

This project is led by Professor Behzad Razavi (PI) and by the following co-PI faculty members: Professors Danijela Cabric, Dejan Markovic, Ali H. Sayed, and Jason Woo.