Review of Bumpless Build Cube (BBCube) Using Wafer-on-Wafer (WOW) and Chip-on-Wafer (COW) for Tera-Scale Three-Dimensional Integration (3DI)

Bumpless Build Cube (BBCube) using Wafer-on-Wafer (WOW) and Chip-on-Wafer (COW) for Tera-Scale Three-Dimensional Integration (3DI) is discussed. Bumpless interconnects between wafers and between chips and wafers are a second-generation alternative to the use of micro-bumps for WOW and COW technologies. WOW and COW technologies for BBCube can be used for homogeneous and heterogeneous 3DI, respectively. Ultra-thinning of wafers down to 4 μm offers the advantage of a small form factor, not only in terms of the total volume of 3D ICs, but also the aspect ratio of Through-Silicon-Vias (TSVs). Bumpless interconnect technology can increase the number of TSVs per chip due to the finer TSV pitch and the lower impedance of bumpless TSV interconnects. In addition, high-density TSV interconnects with a short length provide the highest thermal dissipation from high-temperature devices such as CPUs and GPUs. This paper describes the process platform for BBCube WOW and COW technologies and BBCube DRAMs with high speed and low IO buffer power by enhancing parallelism and increasing yield by using a vertically replaceable memory block architecture, and also presents a comparison of thermal characteristics in 3D structures constructed with micro-bumps and BBCube.

Multichannel ADC IP Core on Xilinx SoC FPGA

Sensing of analogue signals such as voltage, temperature, pressure, current etc. is required to acquire the real time analog signals in the form digital streams. Most of the static analog signals are converted into voltage using sensors, transducers etc. and then measured using ADCs. The digitized samples from ADC are collected either through serial or parallel interface and processed by the programmable chips such as processors, controllers, FPGAs, SOCs etc. In some cases, Multichannel supported ADCs are used to save the layout area when the functionalities are to be realized in a small form factor. In such scenarios, parallel interface for each channel is not a preferred interface considering the more number of interfaces / traces between the components. Hence, Custom, Sink synchronized, Configurable multichannel ADC soft IP core has been developed using VHDL coding to interwork with multichannel supported, time division multiplexed ADCs with serial interface. The developed IP core can be used either as it is with the SPI interface as specified in this paper or with necessary modifications / configurations. The configurations can be the number of channels, sample size, sampling frequency, data transfer clock, type of synchronization – source / sink, control signals and the sequence of the operations performed to configure ADC. The efficiency of implementation is validated using the measurements of throughput, and accuracy for the required range of input with acceptable tolerances. ZYNQ FPGA and LTC2358 ADC are used to evaluate the developed IP core. Integrated Logic Analyser (ILA) which is an integrated verification tool of Vivado is used for Verification.

Well-Lit: A programmable and customizable assistant for manual multi-well plate pipetting

A very large number of biology and biochemistry laboratory protocols require transferring liquid aliquots from individual containers into individual wells of a multi-well plate, from plates to individual containers, or from one plate to another. Doing this by hand without errors, such as skipping wells, placing two samples in the same well, or swapping sample locations, especially when using plates with 96 wells or more, is difficult and requires enormous operator focus and/or a tedious manual error checking system. We present here a device built to facilitate error-free pipetting of samples from individual barcoded tubes to a multi-well plate or between multi-well plates (both 96 and 384 wells are supported). The device is programmable, modular and easily customizable to accommodate plates with different form-factors, and different protocols. The main components are only a 12.3" touch screen, a small form-factor PC, and a barcode scanner, combined with custom-made parts can be easily fabricated with a laser cutter and a hobby-grade 3D printer. The total cost is between approximately US$550 and US$600, depending on the configuration.

A stroke detection and discrimination framework using broadband microwave scattering on stochastic models with deep learning

Stroke poses an immense public health burden and remains among the primary causes of death and disability worldwide. Emergent therapy is often precluded by late or indeterminate times of onset before initial clinical presentation. Rapid, mobile, safe and low-cost stroke detection technology remains a deeply unmet clinical need. Past studies have explored the use of microwave and other small form-factor strategies for rapid stroke detection; however, widespread clinical adoption remains unrealized. Here, we investigated the use of microwave scattering perturbations from ultra wide-band antenna arrays to learn dielectric signatures of disease. Two deep neural networks (DNNs) were used for: (1) stroke detection (“classification network”), and (2) characterization of the hemorrhage location and size (“discrimination network”). Dielectric signatures were learned on a simulated cohort of 666 hemorrhagic stroke and control subjects using 2D stochastic head models. The classification network yielded a stratified K-fold stroke detection accuracy > 94% with an AUC of 0.996, while the discrimination network resulted in a mean squared error of < 0.004 cm and < 0.02 cm, for the stroke localization and size estimation, respectively. We report a novel approach to intelligent diagnostics using microwave wide-band scattering information thus circumventing conventional image-formation.

RTOB SLAM: Real-Time Onboard Laser-Based Localization and Mapping

RTOB-SLAM is a new low-computation framework for real-time onboard simultaneous localization and mapping (SLAM) and obstacle avoidance for autonomous vehicles. A low-resolution 2D laser scanner is used and a small form-factor computer perform all computations onboard. The SLAM process is based on laser scan matching with the iterative closest point technique to estimate the vehicle’s current position by aligning the new scan with the map. This paper describes a new method which uses only a small subsample of the global map for scan matching, which improves the performance and allows for a map to adapt to a dynamic environment by partly forgetting the past. A detailed comparison between this method and current state-of-the-art SLAM frameworks is given, together with a methodology to choose the parameters of the RTOB-SLAM. The RTOB-SLAM has been implemented in ROS and perform well in various simulations and real experiments.

A Non-Contact Compact Portable ECG Monitoring System

Cardiovascular diseases (CVDs) have been listed among the most deadly diseases worldwide. Many CVDs are likely to manifest their symptoms some time prior to the onset of any adverse or catastrophic events, and early detection of cardiac abnormalities is incredibly important. However, traditional electrocardiography (ECG) monitoring systems face challenges with respect to their scalability and affordability as they require direct body contact and cumbersome equipment. As a step forward from the large-scale direct-contact ECG monitoring devices, which are inconvenient for the user in terms of wearability and portability, in this research, we present a small-sized, non-contact, real-time recording system for mobile long-term monitoring of ECG signals. The device mainly comprises three non-contact electrodes to sense the bio-potential signal, an AD8233 AFE IC to extract the ECG signal, and a CC2650 MCU to read, filter, and transmit them. The device is powered by a 2000 mAh lithium-ion battery with isolation between digital and analog powers on the board using two low-dropout regulators (LDOs). The board’s dimension is 8.56 cm × 5.4 cm, the size of a credit card, making it optimal to be worn in a shirt chest pocket. In spite of its small form factor, the device still manages to achieve a continuous measurement battery life of over 16 h, total harmonic distortion below −30 dB across the interested frequency range, an input-referred noise as low as 1.46 µV for contacted cases and 5.15 µV for non-contact cases through cotton, and clear ECG recording for both contact and non-contact sensing, all at a cost around USD 50.

Mobile Computational Photography: A Tour

The first mobile camera phone was sold only 20 years ago, when taking pictures with one's phone was an oddity, and sharing pictures online was unheard of. Today, the smartphone is more camera than phone. How did this happen? This transformation was enabled by advances in computational photography—the science and engineering of making great images from small-form-factor, mobile cameras. Modern algorithmic and computing advances, including machine learning, have changed the rules of photography, bringing to it new modes of capture, postprocessing, storage, and sharing. In this review, we give a brief history of mobile computational photography and describe some of the key technological components, including burst photography, noise reduction, and super-resolution. At each step, we can draw naive parallels to the human visual system.

Write, Attend and Spell

Text entry on a smartwatch is challenging due to its small form factor. Handwriting recognition using the built-in sensors of the watch (motion sensors, microphones, etc.) provides an efficient and natural solution to deal with this issue. However, prior works mainly focus on individual letter recognition rather than word recognition. Therefore, they need users to pause between adjacent letters for segmentation, which is counter-intuitive and significantly decreases the input speed. In this paper, we present 'Write, Attend and Spell' (WriteAS), a word-level text-entry system which enables free-style handwriting recognition using the motion signals of the smartwatch. First, we design a multimodal convolutional neural network (CNN) to abstract motion features across modalities. After that, a stacked dilated convolutional network with an encoder-decoder network is applied to get around letter segmentation and output words in an end-to-end way. More importantly, we leverage a multi-task sequence learning method to enable handwriting recognition in a streaming way. We construct the first sequence-to-sequence handwriting dataset using smartwatch. WriteAS can yield 9.3% character error rate (CER) on 250 words for new users and 3.8% CER for words unseen in the training set. In addition, WriteAS can handle various writing conditions very well. Given the promising performance, we envision that WriteAS can be a fast and accurate input tool for smartwatch.