John Lachapelle

An implantable 64-channel neural interface with reconfigurable recording and stimulation

Next generation implantable medical devices will have the potential to provide more precise and effective therapies through adaptive closed-loop controllers that combine sensing and stimulation across larger numbers of electrode channels. A major challenge in the design of such devices is balancing increased functionality and channel counts with the miniaturization required for implantation within small anatomical spaces. Customized therapies will require adaptive systems capable of tuning which channels are sensed and stimulated to overcome variability in patient-specific needs, surgical placement of electrodes, and chronic physiological responses. In order to address these challenges, we have designed a miniaturized implantable fully-reconfigurable front-end system that is integrated into the distal end of an 8-wire lead, enabling up to 64 electrodes to be dynamically configured for sensing and stimulation. Full reconfigurability is enabled by two custom 32×2 cross-point switch (CPS) matrix ASICs which can route any electrode to either an amplifier with reprogrammable bandwidth and integrated ADC or to one of two independent stimulation channels that can be driven through the lead. The 8-wire circuit includes a digital interface for robust communication as well as a charge-balanced powering scheme for enhanced safety. The system is encased in a hermetic package designed to fit within a 14 mm bur-hole in the skull for neuromodulation of the brain, but could easily be adapted to enhance therapies across a broad spectrum of applications.

Calibration of Delta-Sigma Data Converters in Synchronous Demodulation Sensing Applications

Delta-sigma converters used in synchronous demodulation sensing applications like tuning fork gyroscopes or accelerometers typically contain zeroes in the noise transfer function for aggressive noise shaping to get higher resolution; however, analog component errors typically result in misplacement of the notch frequency resulting in lower SNDR and reduced sensor sensitivity. The error detection technique proposed here corrects for the analog errors in the digital domain by injecting a single tone directly at the input of the quantizer inside the loop and monitoring the RMS output in the digital domain. Expecting the tone to vanish due to the presence of the notch, a closed loop least mean squares (LMS) algorithm adjusts either the component values (capacitor tuning), or the sampling frequency of the modulator until the tone strength is reduced to a minimum, which corresponds to the accurate placement of the notch. Simulation results have demonstrated an improvement of 20 dB for a Single-Loop, Third-Order CIFB (Cascade of Integrators, Feedback Form), single-bit quantizer, delta-sigma modulator with an oversampling ratio (OSR) of 128 and a notch filter coefficient error of 50%.

An implantable, designed-for-human-use peripheral nerve stimulation and recording system for advanced prosthetics

Complex suture prostheses that deliver sensory and position feedback require a more sophisticated integration with the human user. Here a micro-size active implantable system that provides many-degree-of-freedom neural feedback in both sensory stimulation and motor control is shown, as one potential human-use solution in DARPAs HAPTIX program. Various electrical and mechanical challenge and solutions in meeting both sensory /motor performance as well as ISO 14708 FDA-acceptable human use in an aspirin-size active implementation are discussed.Complex suture prostheses that deliver sensory and position feedback require a more sophisticated integration with the human user. Here a micro-size active implantable system that provides many-degree-of-freedom neural feedback in both sensory stimulation and motor control is shown, as one potential human-use solution in DARPAs HAPTIX program. Various electrical and mechanical challenge and solutions in meeting both sensory /motor performance as well as ISO 14708 FDA-acceptable human use in an aspirin-size active implementation are discussed.

A new GaN HEMT nonlinear model for evaluation and design of 1–2 watt power amplifiers

A large-signal model for GaN HEMT devices is presented for use in medium power (1-2 W), S-band PA applications. The emphasis of the model is on quick extraction from standard measurements to facilitate research in this new operating regime for GaN HEMT devices. The entire model is extracted using small-signal S-parameter measurements under a small variety of bias conditions and DC IV characteristics without the need to sacrifice devices in the process. The validity of the model is demonstrated by the design and fabrication of both a class AB PA (achieving P1dB = 30.7 dBm and PAE = 64%) and class E PA (achieving max PAE = 64.4% at P0= 30.1 dBm) based on the model described.

A layered approach to artifact rejection for improved neural recording

Small biological signals can be resolved in presence of large artifacts using various methods. A layered approach to artifact rejection is presented here, using multiple techniques in a cascade fashion including blanking on the second gain stage. Results show superior performance of a layered approach including fast-settle blanking compared to other methods of rejection.

Physical constraints in design of a lowest-phase noise SiGe VCO

Low phase noise oscillators are critical to signal recognition and separation in poor reception conditions. Typically the choice is between a large, high performance hybrid oscillator or a lower performance integrated solution. It is possible to use integrated circuit components to construct a high performance oscillator, but only with great attention to detail. Every IC component has to be optimized in parallel to balance the constraints of spiral inductors, metal-insulator-metal (MIM) capacitors, somewhat lossy varactors, and active devices with breakdown limitations. Physical implementation is critical; some locations are insensitive to trace resistance, other areas are sensitive to even tiedown placement. This paper is intended to discuss, and demonstrate the last mile of integrated circuit VCO design for robust systems, which is less discussed, but maybe just as important as innovative circuit approaches.

A Neural Stimulation System Model to Enhance Neural Integrated Circuit Design

Lacking in the progression of neural stimulation probe, module, and integrated circuit design has been a system-to-device design approach, resulting in less than satisfying in-vivo results. The goal of this paper is neural integrated circuit (IC) design, including time-domain interaction with the electrode, tissue, and neural interfaces, which demonstrates system optimization prior to fabrication. The result of this work is a demonstration of a working CMOS multiplexer for neural stimulation, as well as a model of the multiplexer-electrode-neuron interface.

Ultra-high density packaging technology for injectable medical devices

Future implantable medical devices will be highly miniaturized and almost certainly leverage die-level electronics miniaturization and packaging. Here, an integrated ultra-high density packaging platform is proposed to enable a new class of medical devices. Dense modules are obtained by interconnecting existing ASICs and discrete components using a process which achieves the highest packaging densities available.