Classifying stress in firefighters poses challenges, such as accurate personalized labeling, unobtrusive recording, and training of adequate models. Acquisition of labeled data and verification in cage mazes or during hot trainings is time consuming. Virtual Reality (VR) and Internet of Things (IoT) wearables provide new opportunities to create better stressors for firefighter missions through an immersive simulation. In this demo, we present a VR-based setup that enables to simulate firefighter missions to trigger and more easily record specific stress levels. The goal is to create labeled datasets for personalized multilevel stress detection models that include multiple biosignals, such as heart rate variability from electrocardiographic RR intervals. The multi-level stress setups can be configured, consisting of different levels of mental stressors. The demo shows how we established the recording of a baseline and virtual missions with varying challenge levels to create a personalized stress calibration.

A point-of-load (PoL) voltage regulator module (VRM) with high efficiency, conversion ratio and power density can enable dense integration of CPUs in data centers, and subsequently increase the overall data center energy efficiency. State-of-the-art VRMs are already quite small and achieve high peak efficiencies. However, given the variable nominal load currents of these VRMs in response to changing computational loads of the data centers, a flat high-efficiency response is more desirable. This paper presents a programmable VRM with a flat < 90 % efficiency response over a wide load dynamic range. The proposed programmable VRM is based on multiphase interleaved synchronous buck topology with phase shedding. An extensive exploration of different power switch technologies has been performed in order to obtain the most suitable MOSFET technology for this topology. Our design space exploration also included sweeping the voltage conversion ratio, passive components and switching frequency. The finalized programmable VRM, using discrete Si-MOSFETs, performs a 12-1 V DC-DC conversion with an operating load range of 10 -200 A and a switching frequency of 200 kHz. Experimental characterization of the VRM demonstrates an efficiency of 92.2 % across the wide load range of 10 -200 A, confirming design simulations with a maximum relative error of 0.7 %.

Miniaturization of efficient integrated voltage regulators is critical in order to fully benefit from dynamic voltage and frequency scaling. A major challenge to this is the high parasitic loss of inductor cores at switching frequencies in the MHz range. In this study, we report on a novel soft magnetic core material with anisotropic magnetic properties. This composite core is fabricated using magnetic field directed assembly of magnetic particles in an epoxy matrix. Various experimental conditions were explored, and subsequent image analytics of the resulting samples was used to optimize the process parameters, in order to yield the desired morphology. Magnetic measurements revealed improved magnetization, lower coercivity and higher permeability along the percolation axis of the magnetic particles when compared to isotropic composites with the same particle loading. In addition, the permeability of the anisotropic composites remained close to constant up to 100 MHz. Finally, a funicular “necking” structure with nano-sized particles assembled around contact points of micro-sized particles in the composite was achieved via a bi-modal particle distribution, resulting in further improvements in magnetic properties. Our anisotropic composite core materials are attractive candidates to establish highly efficient power inductors.

This chapter presents various novel heat removal approaches to compile a roadmap toward true 3D integration. A topology change from single side to dual side and volumetric heat removal is a more disruptive option to reduce the thermal constraints on 3D chip stacks. Designing a thermal underfill is a multivariable problem with targets for mechanical, thermomechanical, and electrical properties, besides thermal aspects. Therefore, the complete set of properties needs to be studied in order to guide the experimental work regarding the choice of materials and geometries. Effective properties of percolating and neck-based thermal underfills can be computed using finite element methods. Bringing a liquid coolant closer to the heat source changes the package topology. The chapter overviews the consequences of such a topology change. Using liquid coolant within the package relaxes the thermal requirements on some package parts - such as rendering Cu lids obsolete - but adds requirements on loads, sealing, replaceability, and reliability.

Wearables that acquire relevant vital and contextual parameters improve work safety as well as quality of life of elderly citizens or patients with chronic diseases. A scalable architecture connects wearables via a hub to the cloud and combines edge with cloud computing to provide optimal user interaction and allow analytics on multi-stream data. The functionality was expanded to enable demonstrations of physiological and psychological stress classification in firemen and mobile health interventions in patients with lung diseases. Following an initial table-top edge demonstrator a hemi-spherical display improves emotional contact to users. A first use case tested an integrated acquisition and inference system that was trained to differentiate physical and emotional stress. The system measured stress in firemen during training in a cage maze and in hot training locations and provided functions to acquire expert labels. A second use case focused on mobile-health intervention for patients suffering from Chronic-Obstructive-Pulmonary-Disease (COPD), to improve their quality-of-life. Patient-physician conversations are extended through a communication channel and a virtual assistant provides disease related information, reminders, and alerts.

Stress detection has a huge potential for disease prevention and management, and to improve the quality of life of people. Also, work safety can be improved if stress is timely and reliably detected. The availability of low-cost consumer wearable devices that monitor vital-signs, gives access to stress detection schemes. Heart rate variability (HRV), a stress-related vital-sign, was derived from wearable device data to reliably determine stress-levels. In order to build and train a deployable stress-detector, we collected labeled HRV data in controlled environments, where subjects were exposed to physical, psychological and combined stress. We then applied machine learning to separate and identify the different stress types and understand the relationship with HRV data. The resulting C5 decision tree model is capable of identifying the stress type with 88% accuracy, in a 1-minute time window. For the first time physical and psychological stress can be distinguished with a 1-minute time resolution from smoke-divers, firefighters, who enter high-risk environments to rescue people, and experience intense physical and psychological stress. To improve our model, we created an integrated system to acquire expert labels in real-time from firefighters during their training in a Rescue Maze. A next goal is to transfer the algorithms into generic systems for monitoring and coaching high-risk professionals to improve their stress resilience during training and reduce their risk in the field.

In this work we perform a study of various unsupervised methods to identify mental stress in firefighter trainees based on unlabeled heart rate variability data. We collect RR interval time series data from nearly 100 firefighter trainees that participated in a drill. We explore and compare three methods in order to perform unsupervised stress detection: 1) traditional K-Means clustering with engineered time and frequency domain features 2) convolutional autoencoders and 3) long short-term memory (LSTM) autoencoders, both trained on the raw RRI measurements combined with DBSCAN clustering and K-Nearest-Neighbors classification. We demonstrate that K-Means combined with engineered features is unable to capture meaningful structure within the data. On the other hand, convolutional and LSTM autoencoders tend to extract varying structure from the data pointing to different clusters with different sizes of clusters. We attempt at identifying the true stressed and normal clusters using the HRV markers of mental stress reported in the literature. We demonstrate that the clusters produced by the convolutional autoencoders consistently and successfully stratify stressed versus normal samples, as validated by several established physiological stress markers such as RMSSD, Max-HR, Mean-HR and LF-HF ratio.

Pumped two-phase cooling utilizing an interconnect compatible dielectric fluid is an enabling technology to fully optimize the benefits of the improved integration density possible with three-dimensional (3D) stacking, but is faced with significant developmental challenges, including the need for high fidelity modeling. In the present work, a Eulerian multiphase model developed for predicting two-phase flow and heat transfer behavior in parallel micro-channels and micro-pin fields has been extended to radial expanding channels populated with micro-pins. The model was used to design the cooling channel structures in and to predict the thermal behavior of an embedded two-phase liquid cooled microprocessor module. A detailed model validation showed that this model can predict the chip junction temperature to within two degrees of the experimental data.

Inductors are a key component for voltage regulators and its on-chip integration could significantly improve their performance and scalability capabilities. However, innovative microfabrication processes have yet to be developed in order to miniaturise the inductors while maintaining their overall performance. In that regards, thin-film magnetic materials on glass substrates, which amplifies inductance values, should ease on-chip inductor integration. Using Ni45Fe55 and Co80P20 magnetic materials, we successfully fabricated 2D core and coreless copper inductors, on both glass and silicon substrates for comparison purposes. Inductors on glass exhibit better performance than the ones fabricated on silicon. An inductance of 10.3 nH (14.7 nH/mm(2)) at 100 MHz and a quality factor of 5 were reached. We also propose a process flow and morphological characterizations of a 3D inductor architecture on glass substrate featuring through glass vias.

Integrated Flow-Cell Arrays (FCAs) represent a combination of integrated liquid cooling and on-chip power generation, converting chemical energy of the flowing electrolyte solutions to electrical energy. The FCA technology provides a promising way to address both heat removal and power delivery issues in 3D Multiprocessor Systems-on-Chips (MPSoCs). In this paper we motivate the benefits of FCA in 3D MPSoCs via a qualitative analysis and explore the capabilities of the proposed technology using our extended PowerCool simulator. PowerCool is a tool that performs combined compact thermal and electrochemical simulation of 3D MPSoCs with inter-tier FCA-based cooling and power generation. We validate our electrochemical model against experimental data obtained using a micro-scale FCA, and extend PowerCool with a compact thermal model (3D-ICE) and subthreshold leakage estimation. We show the sensitivity of the FCA cooling and power generation on the design-time (FCA geometry) and run-time (fluid inlet temperature, flow rate) parameters. Our results show that we can optimize the FCA to keep maximum chip temperature below 95 $^\circ$ C for an average chip power consumption of 50 W/cm2 while generating up to 3.6 W per cm2 of chip area.

On-chip integration of voltage regulators could significantly improve performance of 3D ICs. However, integration of voltage regulators is difficult due to the size of inductors. To reduce inductor size, thin-film magnetic materials on glass substrates, which amplifies inductance values, should help on-chip inductor integration. Using Ni45Fe55 and Co80P20 magnetic materials, we successfully fabricated 2D core and coreless copper inductors, on both glass and silicon substrates for comparison purposes. Inductors on glass exhibit better performance than the ones fabricated on silicon. An inductance density of up to ~30 nH/mm 2 was reached.

The performance and power efficiency of high-end servers benefit from dense system integration. Accordingly, we introduce a scalable packaging platform supporting high-performance chip stacks, as a continuation in server-system density scaling.

Wearables that continuously acquire medically relevant parameters can reduce duration of hospitalizations and derive treatment optimizations for individual patients thus improving the quality of medical treatments. We demonstrate an architecture that includes wearables, edge and cloud computing to provide optimal user interaction and analytics of multi-parameter wearable data to accomplish this goal. We also explore the trade-offs of on-wearable processing versus raw data transmission and the use of commercial location monitors to acquire indoor location data.

Novel heat removal and power delivery topologies are required to enable `extreme 3D integration' with cube-sized compute nodes. Therefore, a technology roadmap is presented supporting memory-on-logic and logic-on-logic in the medium and long-term, by (i) dual-side cooling and integrated voltage regulators, and (ii) interlayer cooling and electrochemical power delivery.

Interlayer cooling utilizing pumped two-phase flow of a chip-to-chip interconnect-compatible dielectric fluid is an enabling technology for future high power 3D (three-dimensional) chip stacks. Development of this approach requires high fidelity and computationally manageable conjugate thermal models. In this paper, a conjugate heat transfer model developed for simulating two-phase flow boiling through chip embedded micron-scale channels is described. This model uses a novel hybrid approach where governing equations for flow-field and convection in the single-phase flow regions (e.g. inlet plenum) as well as that for heat conduction in solids is solved in detail (i.e., full-physics) while in the two-phase flow regions (e.g. micro-channels), a reduced-physics approach is used. Extensive model validation using data from several experiments was performed to quantify the accuracy of this model under different operating conditions.

Integrated Voltage Regulators (IVRs) have become a viable solution for microprocessor's power delivery. The active parts of the most recent IVRs are built in deep-submicron CMOS technologies and use stacked transistors to allow for the use of advanced low voltage devices with superior switching performance compared to the higher voltage long-channel devices. This paper evaluates three different topologies of CMOS half-bridge converters with respect to efficiency, implementation effort, suitability for on-chip integration, and multiphase applications: the conventional half-bridge converter, the half-bridge converter with conventional Active Neutral Point Clamping (ANPC), and a half-bridge converter with a modified circuit to achieve ANPC. In-depth analysis of the transient processes during switching for all three converters, based on Cadence simulations, reveal that both half-bridge converters with ANPC achieve proper balancing of the blocking voltages of the main transistors and are capable to attain similar efficiencies of 93% at an output power of 200 mW, input and output voltages of 1.6 V and 0.8 V, respectively, and a switching frequency of 150 MHz, which is 1% higher than the one attained with the conventional half-bridge converter. Of the two ANPC half-bridge converters, however, the proposed topology allows to completely turn off its entire power stage or parts of it, features less efficiency sensitivity to variations of dead-time, and achieves the peak efficiency at relatively higher dead-time values. These qualities render the proposed topology particularly suitable for multiphase systems and low load operation.

Chip embedded two phase evaporative cooling is an enabling technology to provide intra-chip cooling of high power chips and interlayer cooling for 3D chip stacks. Utilizing an interconnect-compatible dielectric fluid provides a cooling solution compatible with chip to chip interconnects for future high power 3D chip stacks. However, lack of high fidelity and computationally manageable conjugate thermal models limits the development of this technology. To address that, a thermal model for fast and accurate prediction of thermal and electrical behavior of an embedded two-phase liquid cooled micro-processor module is described in this paper. This model consists of a state-of-the-art conjugate heat transfer model for two-phase flow boiling through chip embedded micron-scale channels and a physics-based empirically tuned electrical model of the microprocessor. Extensive model validation using data from several experiments was performed to quantify the accuracy of this model under different operating conditions (including various chip operating frequencies and coolant mass flow rates). Results showed that this model can predict the electrical behavior as well as two-phase flow and heat transfer characteristics with very good accuracy. Overall, the chip junction temperature predictions were within two degrees of the experimental data and the temperature-dependent chip power predictions were within 10%.

Wearables that continuously acquire vital and other medically relevant parameters facilitate treatment optimizations for individual patients and reduce the duration of hospitalizations - thus improving the patients' quality of life. To accomplish this, we demonstrate a scalable architecture that connects wearables through a hub to the cloud, combines edge and cloud computing to provide optimal user interaction, and allows analytics on multi-stream data from those connected devices.