A 16-channel, 359 μW, parallel neural recording system using Walsh-Hadamard coding

Application of an algebraic coding to a multichannel parallel neural recording system is presented. The Walsh-Hadamard coding enables back-end hardware sharing between recording channels, using a linear and orthogonal superposition of the analog inputs. Moreover, this technique preserves the temporal information of the channels in contrast to the conventional architectures which use time-multiplexed ADC. In the proposed architecture a single ADC operates on a superposed signal, thereby the dynamic range of the ADC is effectively shared between channels benefiting from the sparsity characteristics of the channels. A 16-channel parallel recording system is implemented as a proof of concept. The system is implemented in a 0.18 μm CMOS technology and occupies 1.99 mm² of silicon area. The input-referred noise of a single channel integrated from 10 Hz to 100 kHz equals to 4.1 μV_rms, and the effective power consumption of each channel is measured at 22.4 μW from a 1.2 V power supply, which results in a system level NEF of 5.6.