For the control of complex robotic prostheses, the direct recording of action potentials (APs) is the only kind of brain machine interface (BMI) solution to provide sufficient temporal and spatial resolution. When microelectrode arrays are implanted to record APs, however, scar tissue forms. This tissue severely degrades the recording signal-to-noise ratio (SNR) over time. Ideally, a solution would eliminate interface cables and use a wireless link to transfer power and data. The implant also should be small and light enough to float freely in brain tissue, eliminating friction with the dura or skull. Much research is being devoted to such solutions, such as the wireless neural sensor created by the University of California’s William Biederman, Daniel Yeager, Nathan Narevsky, Aaron Koralek, Jose Carmena, Elad Alon, and Jan Rabaey.
To enable an electrode-sized implant to float in brain tissue, the researchers realized that they needed a fully integrated system-on-a-chip (SoC) device with an order-of-magnitude reduction in active circuit area. Compared to prior state-of-the-art wirelessly powered systems, they achieved a 10x reduction in area and a 58x reduction in power per channel. In just 0.125 mm2, their 65-nm CMOS integrated circuit (IC) integrates four 1.5-μW amplifiers (6.5 μVrms input-referred noise measured in a 10-kHz bandwidth).
Both power-conditioning and communication circuitry are included. The multi-node backscatter frequency locks to a wireless interrogator using a frequency-domain multiple-access communication scheme. The full system, which was verified with in vivo recordings, consumes 10.5 μW while operating within a 1-mm range in air. It boasts 50 mW transmit power. A subcranial interrogator both powers and communicates with an array of implanted, free-floating AP sensors through the brain’s dura (the outermost membrane) layer. The sensor nodes are implanted lengthwise, which enables the four electrodes to extend deep enough to reach relevant neurons.
Four data-acquisition channels amplify and digitize the sensed neural potentials in an 800-kb/s data stream via 10-b, 20-kHz analog-to-digital converters (ADCs). A single receive (Rx) coil on the sensor couples perpendicularly to a superdural transmit (Tx) coil, achieving power and data transmission simultaneously. See “A Fully-Integrated, Miniaturized (0.125 mm2 10.5 μW Wireless Neural Sensor,” IEEE Journal Of Solid-State Circuits, April 2013, p. 960.
