A Multichannel Recording Frontend for MEA

Glass-substrate Multi-electrode Arrays (MEAs) have become a valuable tool in neurophysiological research and drug screening. A typical setup for MEA recording is based on discrete-element preamplifiers located close to the MEA device and a multi-wire cable that conducts the pre-amplified analog signals to a data acquisition card. With the increase in the number of MEA sensing sites, this approach may turn impractical. We investigate the use of an integrated circuit as a mixed-signal analog/digital front-end for a MEA. This work presents a 0.35μm IC implementing a 12-channel mixed signal integrated front-end for neuronal recording successfully employed with a MEA system. 1 Conventional MEA Recording Glass-substrate Multi-electrode Arrays (MEAs) have become a valuable tool in neurophysiological research and drug screening [1-4]. A typical setup for MEA recording (Fig. 1) is based on discrete-element preamplifiers located close to the MEA device and a multi-wire cable that conducts the pre-amplified analog signals to a data acquisition system. Fig.1 Laboratory MEA setup With the increase in the number of MEA sensing sites this approach may turn impractical, due to the large number of pre-amplifying channels, data acquisition channels, and, perhaps the most important, very thick wires. A novel approach, based on integrated circuit for MEA recording front-end, is called for. The immediate advantage is the tight integration of the analog signal conditioning and data acquisition: A single IC can handle tens of MEA channels providing a digitized data stream in a suitable format. A less obvious advantage is associated with the integration of data processing on the same die with data acquisition: A 100channel MEA with channels sampled at 50KSps at 10bit precision would generate a data stream of 50MSps that can be handled by a dedicated PC. A thousand MEA channels would already present a serious problem, especially if the system has to be responsive, i.e. generate stimuli as a reaction to the recorded signal. An integrated front-end with data processing capabilities can perform level detection and reduce the datarate by sending only the segments with suspected activity. Moreover, an advanced integrated front-end may even perform spike sorting and communicate only the times and the sources of the firing events, reducing the data-rate even further. 2 CMOS Multichannel Neuronal Recording Frontend