Design for good matching in multichannel low-noise amplifier for recording neuronal signals in modern neuroscience experiments

https://doi.org/10.1016/S0026-2714(03)00217-8Get rights and content

Abstract

The paper reports on the design of a prototype 16-channel ASIC for readout of signals from live neuronal systems. Single channel comprises a low-noise amplifier and a low-frequency pass-band filter. The amplifier design optimised with respect to noise and power consumption is discussed. The design of a continuous-time high-pass filter with lower cut-off frequency as low as 20 Hz, which is suitable for realisation in a CMOS process, is presented. Special attention is paid to uniformity of analogue parameters in the multichannel IC. Channel matching is evaluated by Monte Carlo simulation. The design has been fabricated in a 0.7 μm CMOS process and measurements of basic parameters and characteristics have been performed for the prototypes. Good agreement between the simulation and measurements has been achieved for single channel parameters as well as for channel matching. The obtained results on channel matching are discussed with respect to design reliability and production yield.

Introduction

Experimental neurobiology is a research area which needs high performance analogue circuit for recording electrical signals from live neuronal systems. Neuronal systems consist of thousands of electrically active cells and in order to understand the structure and functionality of such complex systems one needs to record simultaneously signals from as many cells as possible. Therefore, a general trend in modern neurobiology experiments is to employ multielectrode systems. The signals from neurone cells are small, especially in the extracellular measurements, and such systems require low noise front-end electronics to extract the signals before they can be digitised. The microelectrode arrays are high-density micropattern structures with typical spacing of the order of tens of microns. In order to record neuronal signals simultaneously from tens or hundreds of such microelectrodes the front-end electronics must be highly integrated, which can be achieved only by developing multichannel ASICs.

The development described in this paper is driven by a project to study retina tissues [1], [2]. A prototype ASIC NEURO32 has been developed and is used successfully in a retina experiment employing an array of 61 microelectrodes [3]. Here we report on the design exploring a new concept of the front-end amplifier and the filter amplifier. The basic requirements concerning the features and the parameters of the new design are summarised below:

  • Multichannel architecture that results in particular requirements with respect to the layout of the ASIC and matching performance.

  • Low noise, because amplitudes of extracellular neural signals are typically in the range of 50–500 μV with the frequency spectrum distributed from 20 to 2000 Hz. A low noise preamplifier and a low frequency band-pass filter are required.

  • The inputs of the preamplifiers should be AC coupled to the electrodes and the AC coupling circuits should be preferably integrated on the IC. This requirement is driven by experimental observations that small neuronal signals are superimposed on much larger DC offsets.

  • The input of the preamplifier must be differential since the signals from the readout electrodes are measured with respect to a common reference electrode. The differential structure of the input stage has, on one hand, a drawback concerning noise of the input transistors increased by a factor of √2 compared to a single-ended stage. On the other hand, this is a preferable solution due to improved immunity of the differential configuration to noise of power supplies, switching noise and external noise sources. The differential configuration also improves the linearity and increases the output signal swing so the dynamic range is considerably improved.


The requirements concerning the parameters and characteristics of single channel are very demanding and to meet them one has to push the design to the limits of the technology. This is contradictory with the requirement of low sensitivity of the design to variation of the process parameters that is essential for obtaining good matching in a multichannel IC. In the paper we present first the design of single channel and then we discuss the matching issues. The simulations and measurements of the channel matching are presented and discussed.

The prototype ASIC comprises 16 identical channels and bias circuits that are common for all channels. The number of 16 has been chosen as a compromise between the cost of the prototype and the requirement to obtain statistically valid data on channel matching.

Section snippets

Single channel design

The single channel is built of an AC coupling circuit at the input, low noise preamplifier, filter amplifier and the output source follower. The whole signal chain is fully differential. The differential output signals are foreseen to be connected later to an analogue multiplexer. The structure of single channel is shown in Fig. 1.

Matching in multichannel ASIC

An aspect that is very specific for multichannel ASICs and has to be taken into account at an early stage of the design is matching of analogue parameters of all channels. Unavoidable spread of parameters of the components used (resistors, capacitors and transistors) results in spread of electrical parameters of nominally identical channels in single IC. Channel-to-channel spread of parameters in a multichannel IC can be affected by systematic offsets and random mismatch of parameters.

Conclusions

A multichannel prototype ASIC for readout of signals from neuronal systems has been designed in a 0.7 μm CMOS technology. The main design aspects taken into account are low noise performance and channel matching. We have demonstrated that the matching can be modelled reasonably well by Monte Carlo simulation, however, some additional margins should be applied for long distance mismatch effects, which are not modelled. Some spread of parameters for a multichannel ASIC is unavoidable but

Acknowledgements

This work was supported by the Polish State Committee for Scientific Research––project no. 7 T11B 054 21.

References (13)

There are more references available in the full text version of this article.

Cited by (0)

View full text