Variation Tolerance Research Articles

A Differential Flip-Flop With Static Contention-Free Characteristics in 28 nm for Low-Voltage, Low-Power Applications

A static contention-free differential flip-flop (SCDFF) is presented in 28-nm CMOS for low-voltage and low-power applications. The SCDFF offers fully static and contention-free operation without redundant internal clock transitions with footed differential latches and the same area as a conventional transmission-gate flip-flop (TGFF). The fully static and contention-free operation allows high variation tolerance at a low supply voltage regime, achieving wide-range voltage scalability (1–0.3 V). The measurement results with a test chip fabricated in 28-nm CMOS technology show that power consumption is reduced by 64%/56% with a 0%/10% activity ratio at 1 V compared to that of a TGFF. All 100 dies from five process corners were functional with supply voltage as low as 0.28 V.

Acceptor-based qubit in silicon with tunable strain

Long coherence time and compatibility with semiconductor fabrication make spin qubits in silicon an attractive platform for quantum computing. In recent years, hole spin qubits are being developed as they have the advantages of weak coupling to nuclear spin noise and strong spin-orbit coupling (SOC), in constructing high-fidelity quantum gates. However, there are relatively few studies on the hole spin qubits in a single acceptor, which requires only low density of the metallic gates. In particular, the investigation of flexible tunability using controllable strain for fault-tolerant quantum gates of acceptor-based qubits is still lacking. Here, we study the tunability of electric dipole spin resonance (EDSR) of acceptor-based hole spin qubits with controllable strain. The flexible tunability of LH-HH splitting and spin-hole coupling (SHC) with the two kinds of strain can avoid high electric field at the "sweet spot", and the operation performance of the acceptor qubits could be optimized. Longer relaxation time or stronger EDSR coupling at low electric field can be obtained. Moreover, with asymmetric strain, two "sweet spots" are induced and may merge together, and form a second-order "sweet spot". As a result, the quality factor $Q$ can reach $10^{4}$ for single-qubit operation, with high tolerance for the electric field variation. Furthermore, the two-qubit operation of acceptor qubits based on dipole-dipole interaction is discussed for high-fidelity two-qubit gates. The tunability of spin qubit properties in acceptor via strain could provide promising routes for spin-based quantum computing.

Segmented continuously varying pulse interval sequence design for extremely high-resolution spaceborne sliding spotlight SAR

ABSTRACT Synthetic aperture radar (SAR) is an active microwave remote sensing device used to obtain high-resolution SAR images.With the development of SAR, extremely high-resolution SAR (i.e. resolution reaching 0.1 m) plays an increasingly important role in applications such as target recognition. In extremely high-resolution spaceborne SAR, the variation of slant range history becomes extremely severe, resulting in echoes of interest overlapping with transmitted pulses and nadir echoes when transmitting the pulses at constant pulse intervals (PIs), further causing the loss of echo data. Although a continuously varying pulse interval (CVPI) system can expand the tolerance of slant range variation to some extent, overlap still occurs when the variation in slant range exceeds a certain threshold. Thus, this paper proposes segmented CVPI sequence design methods to solve the data acquisition problem in the situation of severe variation in slant range history. The timing diagram- and strip layering diagram-based segmented CVPI sequence design methods, which are suitable for the acquisition of spaceborne SAR data with any variation of slant range, are given. The characteristics of the two methods are analysed. Finally, several computer simulations indicate that the echoes generated by the proposed methods are not overlapped, which verifies the validity of the proposed methods.

Examination of fuel elements irradiated in the reactor of the World’s First NPP after long-term storage

Examinations of fuel elements with two different fuel compositions, U-Mo+Mg and UO2+Mg, irradiated in the AM reactor after their long-term storage do not reveal any visible defects on the surface of their outer claddings. However, in the fuel elements with U-Mo fuel, an increase in the diameter of the outer cladding is observed. This is most noticeable in the upper part of the fuel element. Storage of the fuel elements with UO2 fuel for 15–22 years does not lead to a change in their diameter within the measurement accuracy. At the same time, metallographic studies have shown that on the external surface of the outer cladding and the internal surface of the inner cladding of the fuel elements with U-Mo+Mg and UO2+Mg fuel compositions, after long-term storage, defects are observed in the form of intergranular and irregular frontal corrosion, pits and pittings up to 20 µm deep. No interaction is found at the points of contact between the fuel claddings and the fuel composition of the layers. There is no noticeable decrease in the thickness of the outer and inner claddings of the fuel elements after long-term storage, nor does the thickness of the claddings at the locations of defects go beyond its minimum initial value, taking into account the technological tolerance for variations in thickness. It is noteworthy, however, that cracks are found in both types of fuel elements both in the fuel grains and in the magnesium matrix. As a result of long-term storage of the fuel elements with U-Mo fuel for 45–55 years, the mechanical properties of their outer claddings gradually degrade, due to which the plasticity of the cladding is significantly reduced.

USE OF PIDGIN LANGUAGE IN COMMUNITY CONVERSATIONS SOUTH KOREA

The purpose of this study is to describe the pidgin language of the people of South Korea by looking at the pidgin in Korean which was formed through the background of Japanese colonial history and Korean colonial history. Similar to Indonesia and Malaysia, South Korea was also under Japanese colonial rule from 1910 to 1945. Against this historical background, Japanese expressions or Japanese writing methods in Korean society are very commonly used today. The use of Japanese expressions and methods can easily be found in legal sentences and official documents used by the government. The data in this study are the formation of a new language from several languages or pidginization. this type of research is qualitative research. Data collection techniques in this study used observation, interview and documentation techniques, in this study the method used was content analysis method. From the research results, it was found that the Pidginization Process in general is a simplification of language form. for example morphologically (word structure) and syntactically (grammatical structure), tolerance for phonological variations (pronunciation), reduction of a number of functions, and extensive borrowing of words from several languages.

SqueezeLight: A Multi-Operand Ring-Based Optical Neural Network With Cross-Layer Scalability

Optical neural networks (ONNs) are promising hardware platforms for next-generation artificial intelligence acceleration with ultrafast speed and low-energy consumption. However, previous ONN designs are bounded by one multiply–accumulate operation per device, showing unsatisfying scalability. In this work, we propose a scalable ONN architecture, dubbed <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">SqueezeLight</monospace> . We propose a nonlinear optical neuron based on multioperand ring resonators (MORRs) to squeeze vector dot-product into a single device with low wavelength usage and built-in nonlinearity. A block-level squeezing technique with structured sparsity is exploited to support higher scalability. We adopt a robustness-aware training algorithm to guarantee variation tolerance. To enable a truly scalable ONN architecture, we extend <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">SqueezeLight</monospace> to a separable optical CNN architecture that further squeezes in the layer level. Two orthogonal convolutional layers are mapped to one MORR array, leading to order-of-magnitude higher software training scalability. We further explore augmented representability for <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">SqueezeLight</monospace> by introducing parametric MORR neurons with trainable nonlinearity, together with a nonlinearity-aware initialization method to stabilize convergence. Experimental results show that <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">SqueezeLight</monospace> achieves one-order-of-magnitude better compactness and efficiency than previous designs with high fidelity, trainability, and robustness. Our open-source codes are available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/JeremieMelo/SqueezeLight</uri> .

Neuronal morphology enhances robustness to perturbations of channel densities

Biological neurons show significant cell-to-cell variability but have the striking ability to maintain their key firing properties in the face of unpredictable perturbations and stochastic noise. Using a population of multi-compartment models consisting of soma, neurites, and axon for the lateral pyloric neuron in the crab stomatogastric ganglion, we explore how rebound bursting is preserved when the 14 channel conductances in each model are all randomly varied. The coupling between the axon and other compartments is critical for the ability of the axon to spike during bursts and consequently determines the set of successful solutions. When the coupling deviates from a biologically realistic range, the neuronal tolerance of conductance variations is lessened. Thus, the gross morphological features of these neurons enhance their robustness to perturbations of channel densities and expand the space of individual variability that can maintain a desired output pattern.

Schmitt Trigger 12T1M Non-volatile SRAM cell with improved process variation tolerance

This paper presents a Schmitt Trigger (ST) based 12T1M non-volatile static random-access memory (nvSRAM) cell which significantly provides tolerance against process variations. The memristor M1 is connected between internal node Q and bitline RBL through read pass transistor. The impact of process variation is studied on margin and delay performances of proposed and existing read decoupled (RD) nvSRAM cells. The proposed cell achieves 1.62X, 1.17X and 1.33X higher write, read and hold margins, respectively, in comparison to existing RD nvSRAM cells. Furthermore, the maximum reduction in write, read and hold failure probabilities by 105×, 104× and 104×, respectively, is observed for proposed cell. The overall Vmin of proposed cell is found as 425 mV which is minimum among the existing RD counterparts. In addition, there is significant reduction in leakage power consumption and write/read delay deviation for the proposed cell. The store/restore performance of the proposed cell is same as most of the considered cells. All the comparison are made using 32 nm CMOS PTM model and at Vdd = 0.6 V. The effect of supply voltage variation is also studied on the performance of proposed and existing RD nvSRAM cells for the sake of completeness.

Platyhelminth fauna of the Island of Sylt: a meta-analysis of distributional patterns and description of 19 new species

Over the past 70 years, numerous authors studied the platyhelminth fauna near the island of Sylt in the eastern North Sea, each with a specific focus on varying aspects of ecology, morphology or systematics, and most of them dealing with a single habitat type. These data are combined with new data to get a comprehensive view of species richness, the distribution of individual species across tidal levels and sediment types and the resulting communities. So far, 398 platyhelminth species have been recorded around Sylt island, plus a still growing number of unidentified or undescribed species, in particular from subtidal waters. The distribution over habitat types (as defined by sediment types and tidal level) is described for the known species. Neighbouring habitat types overlap in species composition, and faunal composition changes continuously over environmental gradients. The number of species recorded per habitat type mirrors the sampling intensity and varied between some 20 species in poorly studied habitats and 150 in the well-studied intertidal sand flats. Corrected for variations in sampling intensity, platyhelminth species richness showed no significant differences between sediment types and only moderate variation over tidal levels. On a larger spatial scale, three faunal assemblages can be differentiated: the supratidal harbours brackish-water species (mesohaline in the upper and polyhaline in the lower supratidal), the intertidal polyhaline-marine species with a wide tolerance of variations in physical factors and the subtidal marine (± stenohaline) species. With respect to sediment type, mud and sand dwellers are well separated in the supra- and subtidal belt but less in the intertidal. Provided these rules are general, I conclude platyhelminth species richness in a given section of coastline mainly depends on the ranges of environmental factors covered. Nineteen new species encountered during this study are described.

A Simulation Study of Tolerance of Breathing Amplitude Variations in Radiotherapy of Lung Cancer Using 4DCT and Time-Resolved 4DMRI.

As patient breathing irregularities can introduce a large uncertainty in targeting the internal tumor volume (ITV) of lung cancer patients, and thereby affect treatment quality, this study evaluates dose tolerance of tumor motion amplitude variations in ITV-based volumetric modulated arc therapy (VMAT). A motion-incorporated planning technique was employed to simulate treatment delivery of 10 lung cancer patients' clinical VMAT plans using original and three scaling-up (by 0.5, 1.0, and 2.0 cm) motion waveforms from single-breath four-dimensional computed tomography (4DCT) and multi-breath time-resolved 4D magnetic resonance imaging (TR-4DMRI). The planning tumor volume (PTV = ITV + 5 mm margin) dose coverage (PTV D95%) was evaluated. The repeated waveforms were used to move the isocenter in sync with the clinical leaf motion and gantry rotation. The continuous VMAT arcs were broken down into many static beam fields at the control points (2°-interval) and the composite plan represented the motion-incorporated VMAT plan. Eight motion-incorporated plans per patient were simulated and the plan with the native 4DCT waveform was used as a control. The first (D95% ≤ 95%) and second (D95% ≤ 90%) plan breaching points due to motion amplitude increase were identified and analyzed. The PTV D95% in the motion-incorporated plans was 99.4 ± 1.0% using 4DCT, closely agreeing with the corresponding ITV-based VMAT plan (PTV D95% = 100%). Tumor motion irregularities were observed in TR-4DMRI and triggered D95% ≤ 95% in one case. For small tumors, 4 mm extra motion triggered D95% ≤ 95%, and 6-8 mm triggered D95% ≤ 90%. For large tumors, 14 mm and 21 mm extra motions triggered the first and second breaching points, respectively. This study has demonstrated that PTV D95% breaching points may occur for small tumors during treatment delivery. Clinically, it is important to monitor and avoid systematic motion increase, including baseline drift, and large random motion spikes through threshold-based beam gating.

Revealing the Phenotypic and Genomic Background for PHA Production from Rapeseed-Biodiesel Crude Glycerol Using Photobacterium ganghwense C2.2.

Polyhydroxyalkanoates (PHA) are promising biodegradable and biocompatible bioplastics, and extensive knowledge of the employed bacterial strain's metabolic capabilities is necessary in choosing economically feasible production conditions. This study aimed to create an in-depth view of the utilization of Photobacterium ganghwense C2.2 for PHA production by linking a wide array of characterization methods: metabolic pathway annotation from the strain's complete genome, high-throughput phenotypic tests, and biomass analyses through plate-based assays and flask and bioreactor cultivations. We confirmed, in PHA production conditions, urea catabolization, fatty acid degradation and synthesis, and high pH variation and osmotic stress tolerance. With urea as a nitrogen source, pure and rapeseed-biodiesel crude glycerol were analyzed comparatively as carbon sources for fermentation at 20 °C. Flask cultivations yielded 2.2 g/L and 2 g/L PHA at 120 h, respectively, with molecular weights of 428,629 g/mol and 81,515 g/mol. Bioreactor batch cultivation doubled biomass accumulation (10 g/L and 13.2 g/L) in 48 h, with a PHA productivity of 0.133 g/(L·h) and 0.05 g/(L·h). Thus, phenotypic and genomic analyses determined the successful use of Photobacterium ganghwense C2.2 for PHA production using urea and crude glycerol and 20 g/L NaCl, without pH adjustment, providing the basis for a viable fermentation process.

CMQ: Crossbar-Aware Neural Network Mixed-Precision Quantization via Differentiable Architecture Search

The RRAM-based accelerators have become very popular candidates for neural network acceleration due to they perform matrix-vector multiplication in-memory with high storage density and low latency. Many related works have used fixed-precision quantization to achieve the model compression and enhance the tolerance of process variation, but these methods still suffer a large accuracy degradation and poor robustness to nonideal effects. In this work, we propose a crossbar-aware mixed-precision quantization scheme, which enables to search for the optimal precision of each part of the network as a way to improve the accuracy and robustness to noise. First, we introduce a group quantization strategy that can flexibly adjust the group size dynamically according to the crossbar size. Then, we propose a detailed mixed-precision search flow to search for the optimal precision set of the network. Finally, we give a noise injection adaption training method to enhance the tolerance of noise. Experimental results show that our proposed method can improve inference accuracy by at least 2.04% compared to the fixed-precision quantization under the same resource cost. The searched architecture with the highest accuracy most accurate (MA) could achieve an accuracy of 92.39% and a resource saving of 93.30% compared to the full precision model. The searched architecture with the biggest resource savings most efficient (ME) could achieve an accuracy of 91.11% and a resource saving of 95.57% compared to the full precision model. And, the average precision of the ME mixed-precision architecture is only 1.4 bits. Besides, the results show the mixed-precision network with noise adaption training is more robust to noise than the fixed-precision network with noise adaption training.

Asymmetric Bidirectional Capacitive Power Transfer Method With Push–Pull Full-Bridge Hybrid Topology

Bidirectional capacitive power transfer technology makes it possible for energy sharing among multiple electronic devices. This article proposes an asymmetrical bidirectional power conversion topology to satisfy different input and output characteristics’ requirements and increase load variation tolerance. A π–T ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">CLC</i> – <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">LCL</i> ) resonant topology is designed for this bidirectional conversion mode with constant output voltage characteristics in both directions. A hybrid power flow regulation strategy is proposed by integrating multiple zero-voltage switching soft-switching operating points switching and phase-shifted mode. This method overcomes the problem of mixing two different type topologies (push–pull and full bridge) with two different power regulation modes (switch soft-switching operating points and phase shift), which provide a way to make two different chargers compatible. Simulation and experimental results verified the proposed method.

Goat kid recognition of their mothers' calls is not impacted by changes in fundamental frequency or formants

AbstractFeatures varying more between than within individuals are considered as potential cues for individual recognition. According to the source‐filter theory of vocal production, the fundamental frequency of mammals' vocalizations depends on the characteristics of vocal folds, while Formants are determined by the characteristics of the vocal tract. Goat mothers and their kids (Capra hircus) display mutual recognition, and both source‐related parameters (f0) and filter‐related ones (Formants) have been shown to be individualized. Here, we aimed to identify if f0 and Formants are used by goat kids to recognize their mother's vocalizations. To do this, we independently modified these parameters in calls of goat mothers to different degrees (within or exceeding the range of natural intra‐individual variability), and we played back these modified calls to their kids. We found no effect of f0 or Formants modification on the kids' reactions. Further analyses revealed that goat kids emitted fewer calls when modifications to f0 resulted in higher values of the first energy quartile, suggesting a role of the distribution of energy in the spectrum in maternal recognition. We propose that either: (i) f0 and Formants are not involved in goats' maternal recognition; (ii) goat kids have a tolerance for variation when recognizing their mother's calls that exceeds the performed shifts in these parameters; or (iii) goat kid maternal recognition is based on a combination of features and might be more flexible than previously thought, such that when one feature is modified, kids focus on other features. The effect of the spectral energy distribution modification on the kid responses, which depends both on f0 and Formant heights, suggests that (iii) is a likely explanation. Our findings support the hypothesis of complex individual acoustic recognition from the early stages of development in ungulates.

Application of machine learning algorithms in landslide susceptibility mapping, Kali Valley, Kumaun Himalaya, India

The study focuses on the preparation of landslide susceptibility maps in the Kali River valley, Kumaun Himalaya using three machine learning algorithms, namely K-nearest neighbour (KNN), random forest (RF) and extreme gradient boosting (XGB). Fifteen landslide conditioning factors (LCFs) were selected and an inventory of 368 landslides was used for the analysis. Multicollinearity analysis using the variation inflation factor, tolerance and Pearson correlation coefficient (PCC) depicted less to no similarity between all factors. Evaluation of variable importance suggests LCFs such as slope, elevation and distance to thrust contributed significantly and consistently for all three models. Model accuracy was determined and compared using the area under the receiver operating characteristic curve and other statistical signifiers like accuracy, sensitivity, F-measure, accuracy, specificity and recall. The results show that the ensemble algorithms, XGB and RF, yield higher accuracy of approximately 85% compared to the KNN model with 81% accuracy.

Estimation of spring barley adaptability to the aridity of the Stavropol Territory according to the trait ‘1000-grain weight’

A necessary condition for the stable improvement of barley production in the country is the extension of crops, varieties which are most adapted to the conditions of the region, capable of providing a high and stable level of productivity under changing climatic conditions. Productivity is a complex quantitative indicator, which is formed by various yield structure elements, one of which is ‘1000-grain weight’. The purpose of the current study was to estimate adaptability of the spring barley samples of various geographical origins to the arid conditions of the Stavropol Territory according to the trait ‘1000-grain weight’ and their further use in breeding work. The objects of the study were 60 spring barley samples. In order to determine stability, there has been used a coefficient of variation (V %), ecological adaptability (bi), homeostaticity (Hom), and stress tolerance (Ymin–Ymax). On average, during 2016–2021, the trait ‘1000-grain weight’ among the varieties ranged from 35 to 50 grams. The maximum values of the trait ‘1000-grain weight’ were identified in the varieties ‘Victor’ (Russia) with 50.1g (+5.0% to the standard) and ‘Nutans 74840’ (Russia) with 49.7g (+4.2%). The varieties of the extensive type (bi = 0.49–0.86) ‘Nutans 74840’ (Russia), ‘Prikumsky 22’ (Russia), ‘Pri­kumsky 14’ (Russia) and the standard variety ‘Strannik’ were characterized with high stability (V = 5.9–9.8 %), stress tolerance (Ymin–Ymax= –8–11) and homeostaticity (Hom = 43.1–101.2). The varieties ‘Adapt’ (Ukraine), ‘Vodogray’ (Ukraine) and ‘Novonikolaevsky’ (Russia) responded to the weather improvement to a greater extent (bi = 1.21–1.59), but significantly reduced 1000-grain weight when it worsened (Ymin–Ymax = –17 –21). The varieties ‘Nutans 74840’ and ‘Strannik’ turned to be the most adaptive ones for the aridity of the Stavropol Territory, and therefore the best breeding material.

Engineering nano-NiS2 embedded in graphitized carbon skeleton in hollow spherical structure as stable anode material for reversible Li+ storage

As a potential anode alternative for Li-ion batteries (LIBs), NiS2 has the advantages of high theoretical capacity (873 mAh g−1), low cost, and good reversibility. However, the application of NiS2 is limited by poor conductivity, sluggish kinetics, and large volume variation. Herein, the hollow spherical NiS2/C-M, or nano-NiS2 embedded in graphitized carbon skeleton, is prepared through construction of the MOF template with hollow cavity, carbonization, and sulfurization. Compared to the aggregated NiS2/C-P particles, the NiS2/C-M composites have more BET surface area to provide more active sites and larger interface between electrolyte and the electrode, greater volume variation tolerance, improved ion and electron conductivity, less active material loss during cycling. As a result, NiS2/C-M exhibits high reversible capacity and excellent rate capability, 1064.2 and 508 mAh g−1 at 0.1 and 2 A g−1, respectively. NiS2/C-M can retain the capacity of 558 mAh g−1 even after 600 cycles at 1 A g−1. The redox mechanism is proposed that NiS2 can reversibly transform to Ni and Li2S via the LixNiS2 intermediate. The formation mechanism of hollow cavity in MOF is explicated to follow the Ostwald’s step rule. This feasible method to fabricate NiS2/C-M may provide a novel approach to develop high-capacity anode for LIBs.

Molecular genetic differentiation of native populations of Mediterranean blue mussels, Mytilus galloprovincialis Lamarck, 1819, and the relationship with environmental variables

Blue mussels of the genus Mytilus are important ecosystem engineers in intertidal and subtidal communities. The distribution of Mytilus mussels is influenced by a number of benthic and pelagic environmental variables (e.g., substratum type and availability, water movement, phytoplankton production, physical disturbance) as well as interactions between these variables. Because of its broad tolerance of environmental variation the Mediterranean species, Mytilus galloprovincialis, has the greatest ability of all blue mussels to colonise new geographic regions. Understanding how population genetic variation is related to, or caused by, environmental variation is important but has long been a challenge. The present study examined the genetic differentiation of native populations of M. galloprovincialis throughout its entire geographic range in the Mediterranean Sea, the Black Sea and the Sea of Azov using 53 single nucleotide polymorphisms (SNP loci). Mussels, in total 1004 individuals collected from 36 locations, were genotyped and combined with existing SNP data for mussels from 11 reference sites. Pairwise comparisons of F ST values, correspondence analysis (CA) and STRUCTURE analysis all revealed four groups of populations: the Atlantic Ocean; the western Mediterranean; the Aegean Sea; and the Azov, Black and Marmara Seas. One population – from Algeria (Oran West) – was intermediate between the two main groups of the Mediterranean Sea and Atlantic Ocean. Seascape genetic analyses using GLM and DistLM analyses were employed to test site-specific genetic variation as a function of 13 environmental variables. The GLM identified five environmental variables that explained variation in site-specific F ST values, whilst in the DistLM best-fit model only four were significant. These analyses suggest that a complex mix of environmental variables contribute to explaining genetic variation of M. galloprovincialis populations within the Mediterranean Sea, which most likely reflects the complex geological history of formation, isolation and reconnection among the regional sub-basins of the Sea.

Variation-Tolerant and Low R-Ratio Compute-in-Memory ReRAM Macro With Capacitive Ternary MAC Operation

A novel Resistive random access memory (ReRAM)-based Compute-in-memory (CIM) macro is proposed to overcome the limited accuracy and throughput of a conventional ReRAM-based CIM macro that results from to the low R-Ratio and large variation of ReRAM. The proposed structure consists of 1T2R1C bit-cells and 4-kb ReRAM-based nvCIM architecture with ternary weight and ternary input. Ternary multiplication is implemented with voltage division between paired ReRAM devices within a bit-cell to make the output voltage variation tolerant and less sensitive to low R-ratios. An accumulation operation is realized with capacitive coupling so that linearity can be guaranteed for a large number of operands, allowing accurate and fast multiply-and-accumulate (MAC) operations. For comprehensive validation of the proposed CIM macro, the Verilog-A models for ReRAM devices with an adjustable R-ratio and adjustable variations are adopted to perform simulation on various R-ratio and variation conditions. With the peripheral circuits designed in 180-nm CMOS technology, the proposed CIM macro is confirmed to have high variation tolerance, high throughput, and less sensitivity to a low R-ratio, resulting in a high ternary DNN accuracy of 99.07% (0.01% drop) for the MNIST and 83.79% (0.38% drop) for the CIFAR-10 data sets with an R-ratio as low as 38 and 20%/40% low/high resistance variation.

A Proposal of Template Matching Based on Pixel Selection for Constraint Rotated Patterns

Image matching technology is an important technology used for positioning and inspection of objects in the factory automation field. One of the most common recognition methods is template matching. However, this method requires matching a large number of templates in order to detect rotated objects in an image. Therefore, it has a problem of computation time. We propose to match only a small number of templates by assigning a rotation variation tolerance to the templates. The rotation variation tolerance is given by selecting pixels from the templates. We propose a new method of matching with only a few templates. Experimental results showed the method achieved a 91.9% recognition rate and a 288 msec processing time.