Error-aware power management for memory dominated OFDM systems

Mobile wireless systems are typically designed assuming worst-case propagation scenarios, a situation that occurs infrequently in real-life. To reduce power, we exploit the dynamic and statistical nature of wireless signals to aggressively scale the supply voltage using Voltage over Scaling (VoS) on buffering memory modules during run-time. VoS introduces a controlled amount of spatially random errors in embedded memory modules. This paper presents a novel approach to model VoS induced noise as Gaussian noise that can be analytically described to obtain the system performance in terms of BER and PER. Finally, a power management policy is presented that utilizes the available slack in the received SNR with the goal of minimizing power consumption, while keeping performance within normative margins.