Adaptive RF Front-end Design via Self-discovery: Using Real-time Data to Optimize Adaptation Control

It has been established in prior research that significant power can be saved by dynamically trading off the performance of individual RF modules for power consumption across changing channel conditions. It has also been shown that the control law that reconfigures the RF front end must take into account the process corners from which the RF devices are selected in order to trade off performance for power in an optimal manner (minimum energy/bit at prescribed data throughput). Design of such an optimal control law is virtually intractable due to the complexity of simulating the RF front end across all possible channel and device process conditions. Hence, existing control algorithms are based on a coarse sampling of the channel-process space, suffer from modeling inaccuracies and are inherently sub-optimal from a performance vs. power perspective. In contrast, in this paper, we propose a RF front end control methodology that is optimized during real-time operation, does not require upfront simulation across all channel process conditions and is not susceptible to simulation inaccuracies. This results in far more robust/optimal control as opposed to current practice. A simulated annealing (SA) based framework for process optimization is proposed along with the use of built-in sensors for monitoring of performance and power. Simulation results and hardware data are presented to show the feasibility of the proposed approach.