Subthreshold analog CMOS (silicon neurons and synapses)

Operating transistors in the subthreshold (weak-inversion) regime makes their drain current an exponential function of gate voltage, mirroring the exponential kinetics of ion channels and letting a handful of transistors implement a biophysically-styled silicon neuron or synapse at nanoamp currents. This is the classical Mead-style neuromorphic approach: physics does the computation directly, at very low power, in continuous time.

The price is sensitivity to transistor mismatch, temperature, and noise, so subthreshold analog cores need calibration and are hard to reproduce exactly. Modern BCI-oriented designs often blend a compact analog front-end (amplification, filtering, spike detection, feature extraction) with digital or mixed-signal decoding to get efficiency without giving up robustness.