Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
410803 | Neurocomputing | 2007 | 10 Pages |
We propose an analog current-mode subthreshold CMOS circuit implementing a piecewise linear neuromorphic oscillator. Our circuit was derived from a piecewise linear oscillator model proposed by Matsuoka, well known as a building block for constructing a robot locomotion controller. We modified Matsuoka's oscillator to be suitable for analog current-mode integrated circuit implementation, and designed and fabricated it as an analog current-mode circuit. Through circuit simulations and experimental results on a fabricated chip, we demonstrate that our neuromorphic oscillator generates a stable oscillation, and the amplitude and frequency of the oscillation can be controlled by tuning bias currents over a wide range. Further, we propose a compensation for device mismatch in the neuromorphic oscillator through feedback from a coupled physical system.