Event Abstract Back to Event In-Column Cross-Talk Suppression in High-Density CMOS-MEAs Gabriel Bertotti1, Günther Zeck2, Karl-Heinz Boven3 and Roland Thewes4* 1 Technische Universität Berlin, Chair of Sensor and Actuator Systems, Germany 2 NMI at the University Tuebingen, Neurochip Research, Germany 3 Multi Channel Systems MCS GmbH, Germany 4 Technische Universität Berlin, Chair of Sensor and Actuator Systems, Germany Motivation The high-density CMOS-MEA system introduced in [1], enabling simultaneous stimulation and recording [2], enables neural imaging at high spatial resolution thanks to its circuit topology in the readout path. However, this architecture also leads to a kind of “in-column cross-talk" if many sites belonging to the same column record highly-correlated signals which may occur under specific operating conditions. This is e.g. the case during electrical stimulation applied to a large portion of the entire active chip area or during system calibration by means of applying a calibration signal to the bath directly. In this work, we present a simple approach to model this type of cross-talk and use this model to post-process the recorded data in order to efficiently suppress cross-talk-related artifacts. Materials and Methods The high-density CMOS-based MEA used here [1, 3] has 4225 purely capacitively-coupled recording sites, which are organized in 65 rows and 65 columns as schematically shown in Fig. 1. Every sensing site consists of a sensor transistor, whose gate is capacitively coupled to the electrolyte through a metal electrode at the chip surface covered by a thin high-k dielectric. Local variations of the electrolyte potential induced by neural activity are converted into current signals by the sensor transistors. In order to achieve full imaging capability (readout from all available recording sites) multiplexing at column level is performed. Hence, only one column of sensor transistors is active within a given time frame. On this basis, the simplified small-signal equivalent circuit depicted in Fig. 2 can be derived. For the sake of simplicity, transistor parameter variations, channel length modulation-related effects, and parasitic resistances related to metal interconnects are neglected here. Resistor r_S models the resistance of the switch transistor, which connects the sensor transistor's source to bias voltage V_1. Resistor r_S leads to a so-called source degeneration which translates into in-column cross-talk. The following relation applies: --------> Here Formula #1 with --------> Here Formula #2 Moreover, g_S = 1/r_S and g_m stands for the sensor transistors' transconductance. This means that, for k=1...65, output current i_k depends on the signals recorded by all sensor transistors in the respective column (v_1 ... v_65 ), i.e. the considered output signal i_k is affected by cross-talk. It can be shown that matrix A_C is invertible, hence the original (cross-talk free) recorded signals can be recovered by multiplication of the recorded data by the inverse matrix A_C^-1. Since both g_m and g_s are operating point dependent and may vary from chip to chip, a calibration procedure is developed, which estimates the parameter y on the basis of a simple and straight-forward measurement. Results Two examples of successful in-column cross-talk artifact suppression are reported in Figs. 3 and 4, respectively. In both cases, in order to force the required signals to the different sensor transistors we take advantage of the programmable stimulation capability of the CMOS-MEA system (for details, see figure captions and [1, 3]). It is worth to note that both measurements are performed with the same MEA chip operated in the same operating point, so that only a one-time determination of matrix A_C^-1 is required for both signal post-processing procedures. Conclusion In this work, in-column cross-talk effects are investigated of the high-density CMOS-based MEA suggested in [1, 3]. A post-processing method for compensation of related cross-talk artifacts is proposed as well. Two exemplary experiments are reported, where successful compensation of cross-talk-related artifacts is achieved.
Read full abstract
Jun 1, 2018
Yun Miao + 1
Open Access