<p>A clinically viable brain-machine interface (BMI) requires a fully-implanted wireless neural acquisition system to limit the impediments of percutaneous connections. For an implanted system with an appreciable telemetry range, and where significant</p><p>neural signal processing is performed continuously, a major obstacle for clinical application is the need for a power source. Existing battery technology and wireless power delivery systems have not addressed the need for a mid-range power supply, capable</p><p>of 1-3 W delivery, that limits both induced noise and temperature rise. These factors are crucial for the succesful operation of a fully-implanted neural acquisition system. This work seeks to fill this void, and presents both a wireless power solution suitable for a neural recording device, and a system capable of real time monitoring of tissue temperature rise.</p><p>During this research, a 2 W transcutaneous energy transfer system (TETS) was designed, built and tested. The TETS was designed specifically for a 96-channel implanted neural data acquisition system, which requires continuous power. The major design constraints were tolerance to coil misalignment, low induced noise,</p><p>and reasonable efficiency. The design of the primary circuit consists of an H-bridge switching network driving a planar spiral Litz wire primary coil. The primary also incoporates a novel circuit for detecting the presence of the secondary. The implanted secondary components include a complimentary planar spiral coil connected to a voltage doubling rectifier. The key approach to mitigating axial coil misalignments was the use of step-down switching regulators in the secondary. With this approach, link efficiency remained nearly constant at 40%, for axial coil displacements of up to 2 cm.</p><p>Noise in the recorded neural signals was minimized using two techniques. First, the 250 kHz operating frequency of the system was tuned, such that the aliased harmonics of the switching frequency lay above the bandwidth of the amplifier used for neural recording. The second approach was to limit the impact of induced displacement currents in the body by physically separating the recording front end from the power supply components. A large titanium enclosure was used to house some of the secondary electronics, and provided a low impedance return path for further</p><p>reduction of current-induced noise.</p><p>Limiting the temperature rise of internal components was also a critical design constraint. The need for real time temperature information led to the design of a six channel temperature measurement system and incorporation of the temperature data into the acquisition system data transmission scheme. This system consisted of bead thermistor temperature transducers, and an off-the-shelf microcontroller with a built-in instrumentation amplifier.</p><p>The TETS and temperature system was fully tested in an ovine model during several acute studies. Recorded temperature rise was limited to approximately 5.5&deg C when the system was implanted at an adequate depth in muscle. The TETS was able to successfully power the 2 W neural acquisition system during a data processing task. Received rectified voltage in the secondary ranged from 14.86 V to 20.2 V, while link efficiency remained virtually constant. Acquired neural data was examined for TETS switching noise. The measured RMS noise increased by less than 1 &mu V, averaged over several experiments. These results demonstrate the first mid-range TETS solution for powering a fully implanted neural acquisition system.</p> / Dissertation

Design techniques for heterogeneous three-dimensional (3-D) integrated circuits are developed in this thesis. Heterogeneous 3-D integration is a platform for multifunctional, high performance, and low power electronics. For the advancement of heterogeneous 3-D ICs, contactless solutions are investigated to implement inter-tier communication between tiers manufactured with disparate processes and heterogeneous technologies. Two challenges for the development of contactless inter-tier communication are addressed, the design of energy efficient, heterogeneous inductive link transceivers and the impact of crosstalk noise due to the on-chip spiral inductors. Inter-tier communication between circuits fabricated with disparate technologies requires transceivers capable of operating at dissimilar voltages. A low power transceiver design methodology is proposed exploiting the difference in the core voltage between disparate manufacturing processes in a 3-D system in package. A transceiver is designed to provide inter-tier communication between a sensing layer, designed in a commercial 0.35 Âμm process and a processing layer, designed in an advanced 65 nm process. A significant gain in the power consumed by the transceiver is shown compared to equivalent state-of-the-art prototypes, profiting by the tradeoff between the core voltage and sensing ability of the transceiver circuit in each process. Due to their wireless nature, however the use of inductive links introduces crosstalk noise due to the coupling between the on-chip inductor and on-chip interconnects in the vicinity of the inductor. The noise caused by the inductor on the power distribution network of an integrated system is explored, analysed, and modelled through electromagnetic simulations. The spatial distribution of the noise is described for several power distribution topologies to determine the preferred placement solution for the power and ground network in the vicinity of the inductor, considering the impact on other sources of noise, such as the resistive drop. Depending upon the power distribution network topology, the induced noise can be reduced up to 70% when the additional noise caused by the inductive link is considered by the routing algorithm. Additionally, a methodology utilising an analytic model is proposed for the evaluation of the crosstalk noise without resorting to electromagnetic simulations. A closed-form magnetostatic model is developed to assess the mutual inductance between the on-chip inductor and the power distribution network. Utilising the mutual inductance model, the crosstalk noise is evaluated with SPICE simulations. A signifcant benefit in speedup is achieved, up to four orders of magnitude for determining the mutual inductance and up to 4.7× for the assessment of the crosstalk noise. The accuracy of the model is within 10% of the electromagnetic simulation.

The objective of the proposed research is to advance the power efficiency of wireless neural stimulating systems in inductively powered implantable medical devices (IMD). Several innovative system- and circuit-level techniques are proposed towards the development of power-management circuits and wireless neural stimulating systems with inductive power transmission to improve the overall stimulation power efficiency. Neural stimulating IMDs have been proven as effective therapies to alleviate neurological diseases, while requiring high power and performance for more efficacious treatments. Therefore, power-management circuits and neural stimulators in IMDs should have high power efficiencies to operate with smaller received power from a larger distance. Neural stimulating systems are also required to have high stimulation efficacy for activating the target tissue with a minimum amount of energy, while ensuring charge-balanced stimulation. These features provide several advantages such as a long battery life in an external power transmitter, extended-range inductive power transfer, efficacious and safe stimulation, and less tissue damage from overheating. The proposed research presents several approaches to design and implement the power-efficient wireless neural stimulating IMDs: 1) optimized power-management circuits for inductively powered biomedical microsystems, 2) a power-efficient neural stimulating system with adaptive supply control, and 3) a wireless switched-capacitor stimulation (SCS) system, which is a combination structure of the power-management circuits and neural stimulator, to maximize both stimulator efficiency (before electrodes) and stimulus efficacy (after electrodes).

With the advancement of biomedical instrumentation technologies sensor based remote healthcare monitoring system is gaining more attention day by day. In this system wearable and implantable sensors are placed outside or inside of the human body. Certain sensors are needed to be placed inside the human body to acquire the information on the vital physiological phenomena such as glucose, lactate, pH, oxygen, etc. These implantable sensors have associated circuits for sensor signal processing and data transmission. Powering the circuit is always a crucial design issue. Batteries cannot be used in implantable sensors which can come in contact with the blood resulting in serious health risks. An alternate approach is to supply power wirelessly for tether-less and battery- less operation of the circuits.Inductive power transfer is the most common method of wireless power transfer to the implantable sensors. For good inductive coupling, the inductors should have high inductance and high quality factor. But the physical dimensions of the implanted inductors cannot be large due to a number of biomedical constraints. Therefore, there is a need for small sized and high inductance, high quality factor inductors for implantable sensor applications. In this work, design of a multi-spiral solenoidal printed circuit board (PCB) inductor for biomedical application is presented. The targeted frequency for power transfer is 13.56 MHz which is within the license-free industrial, scientific and medical (ISM) band. A figure of merit based optimization technique has been utilized to optimize the PCB inductors. Similar principal is applied to design on-chip inductor which could be a potential solution for further miniaturization of the implantable system. For layered human tissue the optimum frequency of power transfer is 1 GHz for smaller coil size. For this reason, design and optimization of multi-spiral solenoidal integrated inductors for 1 GHz frequency is proposed. Finally, it is demonstrated that the proposed inductors exhibit a better overall performance in comparison with the conventional inductors for biomedical applications.

Link prediction is a crucial task in many downstream applications of graph machine learning. Graph Neural Networks (GNNs) are a prominent approach for transductive link prediction, where the aim is to predict missing links or connections only within the existing nodes of a given graph. However, many real-life applications require inductive link prediction for the newly-coming nodes with no connections to the original graph. Thus, recent approaches have adopted a Multilayer Perceptron (MLP) for inductive link prediction based solely on node features. In this work, we show that incorporating both connectivity structure and features for the new nodes provides better model expressiveness. To bring such expressiveness to inductive link prediction, we propose LEAP, an encoder that features LEArnable toPology augmentation of the original graph and enables message passing with the newly-coming nodes. To the best of our knowledge, this is the first attempt to provide structural contexts for the newly-coming nodes via learnable augmentation under inductive settings. Conducting extensive experiments on four real- world homogeneous graphs demonstrates that LEAP significantly surpasses the state-of-the-art methods in terms of AUC and average precision. The improvements over homogeneous graphs are up to 22% and 17%, respectively. The code and datasets are available on GitHub*. / Att förutsäga länkar är en viktig uppgift i många efterföljande tillämpningar av maskininlärning av grafer. Graph Neural Networks (GNNs) är en framträdande metod för transduktiv länkförutsägelse, där målet är att förutsäga saknade länkar eller förbindelser endast inom de befintliga noderna i en given graf. I många verkliga tillämpningar krävs dock induktiv länkförutsägelse för nytillkomna noder utan kopplingar till den ursprungliga grafen. Därför har man på senare tid antagit en Multilayer Perceptron (MLP) för induktiv länkförutsägelse som enbart bygger på nodens egenskaper. I det här arbetet visar vi att om man införlivar både anslutningsstruktur och egenskaper för de nya noderna får man en bättre modelluttryck. För att ge induktiv länkförutsägelse en sådan uttrycksfullhet föreslår vi LEAP, en kodare som innehåller LEArnable toPology augmentation av den ursprungliga grafen och möjliggör meddelandeöverföring med de nytillkomna noderna. Såvitt vi vet är detta det första försöket att tillhandahålla strukturella sammanhang för de nytillkomna noderna genom en inlärningsbar ökning i induktiva inställningar. Omfattande experiment på fyra homogena grafer i den verkliga världen visar att LEAP avsevärt överträffar "state-of-the-art" metoderna när det gäller AUC och genomsnittlig precision. Förbättringarna jämfört med homogena grafer är upp till 22% och 17%. Koden och datamängderna finns tillgängliga på Github*.

Wireless operation, near-field or far-field, is a core functionality of any mobile or autonomous system. These systems are battery operated or most often utilize energy scavenging as a means of power generation. Limited access to power, expected long and uninterrupted operation, and constrained physical parameters (e.g. weight and size), which limit overall power harvesting capabilities, are factors that outline the importance for innovative low-power approaches and designs in advanced low-power wireless applications. Low-power approaches become especially important for the wireless transceiver, the block in charge of wireless/remote functionality of the system, as this block is usually the most power hungry component in an integrated system-on-chip (SoC). Three such advanced applications with stringent power requirements are examined including space-based exploratory remote sensing probes and their associated radiation effects, millimeter-wave phased-array radar for high-altitude tactical and geological imaging, and implantable biomedical devices (IMDs), leading to the proposal and implementation of low-power wireless solutions for these applications in SiGe BiCMOS and CMOS and platforms.