Scalable Technologies for High Bandwidth Neural Interfacing

High-bandwidth neural interfaces seek to establish a long-lived information channel between the electrical activity of large numbers of neurons (ranging from thousands to as many as millions) in highly parallel nervous systems and the predominantly serial, human-made electronics. This thesis has explored the possibility of designing one such device using readily available off-the-shelf components and low-cost microfabrication processes while choosing those technologies that maximize scalability (i.e. more interfaced neurons while shrinking the device size) following future investments into more advanced lithography techniques. Following a comprehensive state-of-the-art review with some interesting neurobiology insights, the first experimental portion of the thesis deals with the design and microfabrication of polyimide-based neural probes with high-density (100 µm pitch) direct-to-chip interconnects. This was followed by an investigation of a novel, area-saving neural amplifier topology – the neural transimpedance amplifier (NTIA) – which was adopted later on in three iterations of 350 nm chip designs for neural recording, where it served, in conjunction with area-miniaturized delta-sigma modulators, as a basis of the “digital pixel” paradigm that promises more homogenous layouts that eliminate long-range routing and multiplexing of analog signals. Ultimately, one of the fabricated chips (2 mm x 2 mm die area) sporting 128 differential recording channels was integrated, along with all required auxiliary components, into a miniature (3.5 mm x 5.5 mm footprint) printed circuit board, where it was interconnected with 256-site, gold-bumped polyimide neural probes with the help of anisotropic conductive adhesive. The measured noise figure of the fabricated 128-channel chip was 1.79 µV rms (5 µA input stage current, 5 channels) at a total chip power of 20.1 mW at 1 Mhz sampling rate. The completed microsystem (overall current consumption 15 mA at >3.5 V supply voltage) was tested via multi-channel spike recordings in insect preparations. While the developed neural probes and interconnect systems may already prove useful for shorter-term neuroscience experiments, further work is required to ensure sufficient probe insulation lifetime and develop long-term hermetic encapsulation suitable for clinical implantable devices. Similarly, future chip designs should improve further upon neural amplifier (e.g. linearized feedback) and ADC (e.g. MASH topology) design techniques while investing in more on-chip test, calibration, and measurement circuitry. Learning from these developments, the thesis finale imagines some futuristic possibilities – an axon-oriented intracerebral interface, maintaining a living brain outside of the body, and an in-silico mouse brain simulation.