Adaptive Resolution ADC Array for an Implantable Neural Sensor.

This paper describes an analog-to-digital converter (ADC) array for an implantable neural sensor which digitizes neural signals sensed by a microelectrode array. The ADC array consists of 96 variable resolution ADC base cells. The resolution of each ADC cell in the array is varied according to neural data content of the signal from the corresponding electrode. The resolution adaptation algorithm is essentially to periodically recalibrate the required resolution and this is done without requiring any additional ADC cells. The adaptation implementation and results are described. The base ADC cell is implemented using a successive approximation charge redistribution architecture. The choice of architecture and circuit design are presented. The base ADC has been implemented in 0.13 μm CMOS as a 100 kS/s SAR ADC whose resolution can be varied from 3 to 8 bits with corresponding power consumption of 0.23 μW to 0.90 μW achieving an ENOB of 7.8 at the 8-bit setting. The energy per conversion step figure of merit is 48 fJ/step at the 8-bit setting. Resolution adaptation reduces power consumption by a factor of 2.3 for typical motor neuron signals while maintaining an effective 7.8-bit resolution across all channels.