Ising Machines (IMs) provide a novel approach for solving combinatorial optimization problems efficiently. This paper investigates the practical implementation of CMOS oscillator-based IMs, focusing on two architectures: Duffing oscillator and pump-based parametric oscillator networks. Both architectures, scaled up to 5-node networks, have been designed at transistor-and behavioral-level, respectively, using a 180nm CMOS process. Simulation results show that the Duffing oscillator IM achieves 100% accuracy with convergence times of 2 to 4.5 mu s, while the pump-based parametric oscillator IM achieves 98% with convergence times of 10 to 15 mu s, when solving 5-node Max-Cut problems at 22 MHz. A comparative analysis reveals critical insights into the trade-offs between accuracy, complexity, and feasibility of these CMOS oscillator-based IMs.

This paper presents a low-noise bioimpedance (bio-Z) spectroscopy interface for electrical impedance myography (EIM) over the 1 kHz to 2 MHz frequency range. The proposed interface employs a sinusoidal signal generator based on direct-digital-synthesis (DDS) to improve the accuracy of the bio-Z reading, and a quadrature low-intermediate frequency (IF) readout to achieve a good noise-to-power efficiency and the required data throughput to detect muscle contractions. The readout is able to measure baseline and time-varying bio-Z by employing robust and power-efficient low-gain IAs and sixth-order single-bit bandpass (BP) ΔΣ ADCs. The proposed bio-Z spectroscopy interface is implemented in a 180 nm CMOS process, consumes 344.3 - 479.3 μ W, and occupies 5.4 mm 2 area. Measurement results show 0.7 mΩ/√Hz sensitivity at 15.625 kHz, 105.8 dB SNR within 4 Hz bandwidth, and a 146.5 dB figure-of-merit. Additionally, recording of EIM in time and frequency domain during contractions of the bicep brachii muscle demonstrates the potential of the proposed bio-Z interface for wearable EIM systems.

This paper presents a 16-channel analog frontend (AFE) for wearable multi-channel surface electromyography recording. The proposed architecture adopts frequency-division multiplexing (FDM), enabling extensive sharing of signal conditioning circuitry and requiring only three cables for transmission to an external back-end for digital signal processing. Benefiting from the up-conversion process, the FDM signal exhibits robustness against cable motion artifacts and mains interference, allowing short-distance communication without compromising signal quality. This work achieves a channel-to-cable ratio of 16:3, dramatically reducing system complexity. Implemented in a 180 nm CMOS technology, the proposed 16-channel AFE occupies only 6.42 mm2 and consumes 40.1 µW per channel from a 1.5 V supply. Measurements on a forearm demonstrate the capability to simultaneously detect muscle activation from various muscle groups, showcasing its potential for gesture recognition applications.

Neuromuscular electrical stimulation (NMES) is a self-directed home based therapeutic tool in early rehabilitation for musculoskeletal (MSK) conditions. However, the effectiveness of traditional NMES is fundamentally constrained by muscle fatigue. To address this limitation, this work proposes a detection system, which simultaneously records multifrequency electrical impedance myography (EIM) and surface electromyography(sEMG) in real time for time-frequency analysis of muscle activation, contraction, and fatigue. To demonstrate the ability to monitor these muscle physiological states, two experiments involving weightless and weighted dynamic contractions of the biceps brachii muscle were performed. Results from these experiments show synchronous changes in sEMG and EIM spectra during contractions, and clear trends in sEMG’s mean power frequency (MPF) and EIM spectra with fatigue progression. Additionally, the configurable 4-channel NMES has been electrically evaluated for clinical use, demonstrating the feasibility of the proposed system for closed-loop stimulation. This work showcases the potential of sEMG and multi-frequency EIM to enhance the effectiveness of NMES for MSK conditions by capturing the behavior of distinct mechanisms of muscle fatigue.

Multi-Frequency surface Electrical Impedance Myography (MF-sEIM) is a technique that provides valuable electrophysiological information of muscles. This technique measures bio-Z spectroscopy of muscles, and applies Ohm's law by injecting a small-amplitude and high-frequency current into tissues and measuring the voltage response. As biological tissues are electrolytic conductors, this technique provides information on fundamental dielectric properties of tissues, which are an objective biomarker of neuromuscular disorders and have practical value in several muscle healthcare applications. Due to its versatility, simplicity, and ease of integration, this technique is a great candidate to complement stand-alone surface electromyography (sEMG) in wearable devices for continuous monitoring of muscle health. Nonetheless, wearable MF-sEIM imposes challenging requirements on the bio-Z spectroscopy interface. Previously reported solutions fall short of meeting these requirements, or do so with low power efficiency.

This thesis focuses on the research and development of a fully integrated bio-Z spectroscopy interface, which complies with the challenging requirements of wearable MF-sEIM in a power-efficient way. The electrophysiological mechanisms of MF-sEIM and the system-level requirements for clinical relevance were investigated, as MF-sEIM is a relatively novel technique, which requires standarization. From this established set of requirements, the main building blocks of the \mbox{bio-Z} spectroscopy interface, i.e., the current signal generator (CSG), and voltage readout, were developed. The CSG generates pseudo-sine waves through direct digital synthesis (DDS) to obtain the required linearity with high power efficiency. A high-linearity full current-mode CSG was also proposed to comply with the stricter bio-Z accuracy requirements of clinical diagnostics. The voltage readout is based on a low-IF quadrature (I/Q) demodulation architecture and features a pseudo 2-path bandpass (BP) Delta-Sigma ADC to achieve high precision and power-to-noise efficiency. A mixer-first analog front-end (AFE) was also proposed to enable bio-Z spectroscopy measurements employing dry electrodes. The bio-Z interface was integrated with a 16-Channel sEMG AFE and a 4-Channel neuromuscular stimulator (NMES) in an application-specific integrated circuit (ASIC). Experimental results show that the implemented bio-Z spectroscopy interface achieves a comparable performance with the state of the art, while being capable of detecting the large baseline and the time-varying impedances of muscle. A proof-of-concept system, based on the multi-modal ASIC, was developed. This system demonstrates the potential of combining real-time monitoring of MF-sEIM and sEMG for detecting muscle fatigue, enabling efficient closed-loop NMES.

Multifrekvent ytelektroimpedansmyografi (MF-sEIM) är en teknik som ger värdefull elektrofysiologisk information om muskler. Denna teknik mäter bioimpedans-spektroskopi (bio-Z-spektroskpi) av muskler och tillämpar Ohms lag genom att injicera en högfrekvent småsignalström i vävnader och mäta spänningssvaret. Eftersom biologiska vävnader är elektrolytiska ledare, ger denna teknik information om grundläggande dielektriska egenskaper hos vävnader, som är en objektiv biomarkör för neuromuskulära störningar och har praktiskt värde i flera tillämpningar för muskelsjukvård. På grund av dess mångsidighet och enkel integration är denna teknik en utmärkt kandidat för att komplettera fristående ytelektromyografi (sEMG) i bärbara enheter för kontinuerlig övervakning av muskelhälsa. Icke desto mindre ställer bärbar MF-sEIM utmanande krav på bio-Z-spektroskopigränssnittet. Tidigare rapporterade lösningar uppfyller inte dessa krav, eller gör det med låg energieffektivitet.

Denna avhandling fokuserar på forskning och utveckling av ett helt integrerat bio-Z-spektroskopigränssnitt, som uppfyller de utmanande kraven för bärbar MF-sEIM på ett energieffektivt sätt. De elektrofysiologiska mekanismerna för MF-sEIM och kraven på systemnivå för klinisk relevans undersöktes, eftersom MF-sEIM är en relativt ny teknik som kräver standardisering. Från de definerade kraven på systemnivå utvecklades huvudbyggstenarna för bio-Z-spektroskopigränssnittet, dvs. strömsignalgeneratorn (CSG) och spänningsavläsningen. CSG genererar pseudo-sinusvågor genom direkt digital syntes (DDS) för att erhålla den erforderliga linjäriteten med hög energieffektivitet. En CSG i fullströmsläge med hög linjäritet föreslogs också för att uppfylla de striktare noggrannhetskraven på bioimpedansmätning för klinisk diagnostik. Spänningsavläsningen är baserad på en demoduleringsarkitektur med låg IF-kvadratur (I/Q) och har en pseudo 2-vägs bandpass (BP) Delta-Sigma ADC för att uppnå hög precision och energi till brus effektivitet. En mixer-framför analog front-end (AFE) föreslogs också för att möjliggöra bio-Z-spektroskopimätningar med användning av torra elektroder. Bio-Z-gränssnittet integrerades med en 16-kanals sEMG AFE och en 4-kanals neuromuskulär stimulator (NMES) i en applikationsspecifik integrerad krets (ASIC). Experimentella resultat visar att det implementerade bio-Z-spektroskopigränssnittet uppnår en prestanda som är jämförbar med den senaste tekniken, samtidigt som den kan detektera storsignalsimpedansen hos muskler och dess tidsvariation. Ett prototyp, baserat på den multimodala ASIC, utvecklades. Detta system visar potentialen i att kombinera realtidsövervakning av MF-sEIM och sEMG för att upptäcka muskeltrötthet, vilket möjliggör effektiva NMES med återkoppling.

The ability to perform accurate continuous glucose monitoring without blood sampling has revolutionised the management of diabetes. Newer methods that can allow measurements during longer periods are necessary to substantially improve patients’ quality of life. This paper presents an alternative method for glucose monitoring which is based on electrical impedance spectroscopy. A battery-less implantable bioimpedance spectroscope was designed, built, and used in an in vivo study on pigs. After a recovery period of 14 days post surgery, a total of 236 subcutaneous bioimpedance measurements obtained from intravenous glucose tolerance tests, with glucose concentration ranges between 77.4 and 523.8 mg/dL, were analyzed. The results show that glucose concentrations estimated by subcutaneous bioimpedance measurements correlate very well to the blood glucose reference values. The pigs were clinically healthy throughout the study, and the postmortem examinations revealed no signs of adverse effects related to the sensor. The implantation of the sensor requires minor surgery. The implant, being externally powered, could in principle last indefinitely. These encouraging results demonstrate the potential of the bioimpedance method to be used in future continuous glucose monitoring systems.

This paper presents a low-distortion current-mode sinusoidal signal generator for bioimpedance spectroscopy measurements. The proposed full current-mode operation enables linearity enhancement and potential savings in silicon area and power consumption. Programmability in the low-pass filter and current driver enables impedance measurements from 0.2 Ω to10 kΩ over a wide frequency range from 1 kHz to 1 MHz.The current generator, designed in a 0.18 μm CMOS process, consumes between 736 μW at the lowest frequency and gain, and 1.70 mW at the highest frequency and gain, and occupies 1.76 mm² silicon area. Post-layout simulation results show a spurious-free dynamic range larger than 40 dBc over the entire frequency range, which enables bioimpedance measurements with errors below 1%, as it is required for wearable devices evaluating neuromuscular disorders.

This paper presents a high input impedance, low-noise, and low-distortion analog front-end (AFE) for bioimpedance (bio-Z) spectroscopy measurements targeting neuromuscular health assessments. The proposed 8-phase quadrature mixer-first architecture achieves a high input impedance through passive mixers driven by non-overlapping clocks. The 8-phase signals are recombined to extract the real and imaginary parts of the bio-Z, while rejecting unwanted harmonics to improve linearity. Programmability of the AFE enables accurate bio-Z measurements up to 10 kΩ for 11 logarithmically spaced frequencies, in the 1 kHz to 1 MHz frequency range. The AFE, designed in a 0.18 μm CMOS process, consumes 245.99 μW at the lowest gain and 300.56 μW at the highest gain, and occupies 2.4 mm2 silicon area. Post-layout simulation results show that the input impedance is always higher than the electrode impedance by more than 10x. The AFE achieves a sensitivity of 7.7 mΩrms, and a maximum SNDR of 103.87 dBFS over a 61 Hz bandwidth. These results demonstrate that the proposed AFE enables bio-Z measurements, using dry electrodes, with errors below 1%.

Objective: Electrical impedance myography (EIM) measures bioimpedance over muscles. This paper proposes a circuit-based modelling methodology originated from finite element analysis (FEA), to emulate tissues and effects from anthropometric variations, and electrode placements, on EIM measurements. The proposed methodology is demonstrated on the upper arms and lower legs. Methods: FEA evaluates impedance spectra (Z-parameters), sensitivity, and volume impedance density for variations of subcutaneous fat thickness (tf), muscle thickness (tm), and inter-electrode distance (IED), on limb models over 1Hz-1MHz frequency range. The limbs models are based on simplified anatomical data and dielectric properties from published sources. Contributions of tissues to the total impedance are computed from impedance sensitivity and density. FEA Z-parameters are imported into a circuit design environment, and used to develop a three Cole dispersion circuit-based model. FEA and circuit model simulation results are compared with measurements on ten human subjects. Results: Muscle contributions are maximized at 31.25kHz and 62.5kHz for the upper arm and lower leg, respectively, at 4cm IED. The circuit model emulates variations in tf and tm, and simulates up to 89 times faster than FEA. The circuit model matches subjects measurements with RMS errors < 36.43 and < 17.28, while FEA does with < 36.59 and < 4.36. Conclusions: We demonstrate that FEA is able to estimate the optimal frequencies and electrode placements, and circuit-based modelling can accurately emulate the limbs bioimpedance. Significance: The proposed methodology facilitates studying the impact of biophysical principles on EIM, enabling the development of future EIM acquisition systems.