Ultralow-Power and Robust Embedded Memory for Bioimplantable Microsystems

Bioimplantable microsystems, such as pacemaker and cochlear implant, interface with internal body parts to monitor and/or control their activity. These systems typically record biological signals, analyze them in real time, and then transmit them to outside world or take appropriate corrective action. They require ultralow-power miniaturized electronics for long-term reliable operation using on-board battery. Embedded memory used to temporarily store the recorded data, forms an integral and important part of these systems. In this paper, we explore the design space and propose an optimal design of embedded memory for implantable applications. First, we compare a conventional super-threshold implementation of memory with a sub-threshold design with respect to energy efficiency. Next, we propose a super-threshold static random access memory (SRAM) design operating at a frequency much higher than the sampling frequency. We show that it can achieve very low energy dissipation by taking advantage of extensive power gating. Moreover, compared to a sub-threshold memory, it provides significantly better area and higher robustness of operation, both of which are important requirements for implantable systems. As a case study, we consider a neural control system that records and analyzes neural spikes. Simulation results for 45nm CMOS process using pre-recorded neural data from sea-slug (Aplysia californica) show that the proposed design can lead to significant energy reduction, without compromising the robustness and performance, compared to its sub-threshold counterparts.