Abstract
We present an algorithm and, for the first time, a chip to predict the neural theta oscillation phase, for closed-loop phase-locked stimulation triggering. The performance, assessed on test data recorded from human hippocampus, achieves high precision and accuracy in targeting any theta phase with 25%, 50%, and 75% of the predictions falling within +/- 13, +/- 28, and +/- 53 degrees from the desired target phase, respectively. Design interleaving channel-depth optimization achieves 41% energy and 58% area savings compared to a single-channel design. The 32-channel chip consumes a 216 nW per channel and occupies a core area of 0.011 mm(2) per channel in a 40nm low-power technology, making it suitable for implantable devices.