In a neuromorphic sensing device, the non-idealities at the sensing circuitry manifest as spike timing errors at neuronal output. In this paper an on-chip adaptive spike timing based offset cancellation scheme for neuromorphic sensing devices is presented. A spike-time dependent learning algorithm suited to analog VLSI implementation is developed for weight adaption. An analog VLSI circuit implementation of the offset cancellation scheme is presented. Circuit simulations are performed using offset specifications drawn from a neuromorphic olfaction sensor to demonstrate the effectiveness of the scheme. An adaptive offset cancellation chip fabricated in analog VLSI using 0.35 ?m CMOS technology is currently under test.