Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 15
Encontrada(s) 1 página(s)

Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

Simulation study of scaled In0.53Ga0.47As and Si FinFETs for sub-16 nm technology nodes

  • Aldegunde, M.
  • Seoane Iglesias, Natalia
  • Nagy, Daniel
  • Elmessary, M. A.
  • Indalecio Fernández, Guillermo
  • García Loureiro, Antonio Jesús
  • Kalna, K.
We investigate the performance and scalability of III-V-OI In0.53Ga0.47As and SOI Si FinFETs using state-of-the-art in-house-built 3D simulation tools. Three different technology nodes specified in the ITRS have been analysed with gate lengths (LG) of 14.0 nm, 12.8 and 10.4 nm for the InGaAs FinFETs and 12.8 nm, 10.7 and 8.1 nm for the Si devices. At a high drain bias, the 12.8 and 10.4 nm InGaAs FinFETs deliver 15% and 13% larger on-currents but 14% larger off-currents than the equivalent 12.8 and 10.7 nm Si FinFETs, respectively. For equivalent gate lengths, both the InGaAs and the Si FinFETs have the same ION/IOFF ratio (5.9 × 104 when LG = 12.8 nm and 5.7 × 104 when LG = 10.4(10.7) nm). A more pronounced S/D tunnelling affecting the InGaAs FinFETs leads to a larger deterioration in their SS (less than 10%) and DIBL (around 20%) compared to the Si counterparts.
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

A Multi-Method Simulation Toolbox to Study Performance and Variability of Nanowire FETs

  • Seoane Iglesias, Natalia
  • Nagy, Daniel
  • Indalecio Fernández, Guillermo
  • Espiñeira Deus, Gabriel
  • Kalna, Karol
  • García Loureiro, Antonio Jesús
An in-house-built three-dimensional multi-method semi-classical/classical toolbox has been developed to characterise the performance, scalability, and variability of state-of-the-art semiconductor devices. To demonstrate capabilities of the toolbox, a 10 nm gate length Si gate-all-around field-effect transistor is selected as a benchmark device. The device exhibits an off-current ( IOFF ) of 0.03 μ A/ μ m, and an on-current ( ION ) of 1770 μ A/ μ m, with the ION/IOFF ratio 6.63×104 , a value 27% larger than that of a 10.7 nm gate length Si FinFET. The device SS is 71 mV/dec, no far from the ideal limit of 60 mV/dec. The threshold voltage standard deviation due to statistical combination of four sources of variability (line- and gate-edge roughness, metal grain granularity, and random dopants) is 55.5 mV, a value noticeably larger than that of the equivalent FinFET (30 mV). Finally, using a fluctuation sensitivity map, we establish which regions of the device are the most sensitive to the line-edge roughness and the metal grain granularity variability effects. The on-current of the device is strongly affected by any line-edge roughness taking place near the source-gate junction or by metal grains localised between the middle of the gate and the proximity of the gate-source junction
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

Impact of threshold voltage extraction methods on semiconductor device variability

  • Espiñeira Deus, Gabriel
  • Nagy, Daniel
  • García Loureiro, Antonio Jesús
  • Seoane Iglesias, Natalia
  • Indalecio Fernández, Guillermo
This paper presents a study of the impact that several widely used threshold voltage (VT) extraction methods have on semiconductor device variability studies. The second derivative (SD), linear extrapolation (LE) and third derivative (TD) extraction techniques have been compared to the standard method used in variability, the constant current criteria (CC). To estimate the influence of these methods on the results, an ensemble of 10.7 nm gate length Si FinFETs affected by RD variability have been simulated. We have shown that variability estimators like the VT, VT and the VT shift, are heavily affected by the selected extraction methodology, with up to 30% differences in the standard deviation. We have demonstrated that being aware of which VT extraction technique has been used in a variability analysis is crucial to properly interpret the results as they may be heavily method-dependent
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

Drift-Diffusion Versus Monte Carlo Simulated ON-Current Variability in Nanowire FETs

  • Nagy, Daniel
  • Indalecio Fernández, Guillermo
  • García Loureiro, Antonio Jesús
  • Espiñeira Deus, Gabriel
  • Elmessary, Muhammad A.
  • Kalna, Karol
  • Seoane Iglesias, Natalia
Variability of semiconductor devices is seriously limiting their performance at nanoscale. The impact of variability can be accurately and effectively predicted by computer-aided simulations in order to aid future device designs. Quantum corrected (QC) drift-diffusion (DD) simulations are usually employed to estimate the variability of state-of-the-art non-planar devices but require meticulous calibration. More accurate simulation methods, such as QC Monte Carlo (MC), are considered time consuming and elaborate. Therefore, we predict TiN metal gate work-function granularity (MGG) and line edge roughness (LER) induced variability on a 10-nm gate length gate-all-around Si nanowire FET and perform a rigorous comparison of the QC DD and MC results. In case of the MGG, we have found that the QC DD predicted variability can have a difference of up to 20% in comparison with the QC MC predicted one. In case of the LER, we demonstrate that the QC DD can overestimate the QC MC simulation produced variability by a significant error of up to 56%. This error between the simulation methods will vary with the root mean square (RMS) height and maximum source/drain $n$ -type doping. Our results indicate that the aforementioned QC DD simulation technique yields inaccurate results for the ON-current variability
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

Fourier–Mellin registration of two hyperspectral images

  • Ordóñez Iglesias, Álvaro
  • Argüello Pedreira, Francisco
  • Blanco Heras, Dora
Hyperspectral images contain a great amount of information which can be used to more robustly register such images. In this article, we present a phase correlation method to register two hyperspectral images that takes into account their multiband structure. The proposed method is based on principal component analysis, the multilayer fractional Fourier transform, a combination of log-polar maps, and peak processing. The combination of maps is aimed at highlighting some peaks in the log-polar map using information from different bands. The method is robust and has been successfully tested for any rotation angle with commonly used hyperspectral scenes in remote sensing for scales of up to 7.5× and with pairs of hyperspectral images taken on different dates by the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) sensor for scales of up to 6.0×
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Capítulo o parte de un libro (BookPart)

Transformada de Fourier aplicada al alineamiento de imágenes multidimensionales en GPU

  • Ordóñez Iglesias, Álvaro
  • Argüello Pedreira, Francisco
  • Blanco Heras, Dora
La alineación de imágenes es una tarea previa fundamental en numerosas aplicaciones de imágenes. La gran cantidad de información que contienen las imágenes multidimensionales puede ser utilizada para conseguir alineaciones más precisas. Además, la mayoría de los algoritmos de alineación ignoran el tiempo de cómputo. En este artículo presentamos un algoritmo de alineación de imágenes multidimensionales basado en la transformada rápida de Fourier, el análisis de componentes principales y la combinación de mapas log-polar. El algoritmo propuesto hace uso de la correlación de fase para recuperar los parámetros que permiten alinear la imagen. Asimismo, se ha realizado una implementación eficiente sobre Unidades de Procesamiento Gráfico (GPUs). Diferentes técnicas de optimización, como el uso de bibliotecas CUDA optimizadas y el uso eficiente de las diferentes memorias, han sido aplicadas para obtener el mejor rendimiento. Como resultado, el algoritmo es robusto y ha sido probado con diferentes imágenes consiguiendo alineaciones de hasta una escala 6.0x para todos los ángulos, y con una aceleración de la implementación paralela en GPU de hasta 165.4x respecto de la implementación secuencial en CPU
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

Study of basic vector operations on Intel Xeon Phi and NVIDIA Tesla using OpenCL

  • Coronado Barrientos, Edoardo
  • Indalecio Fernández, Guillermo
  • García Loureiro, Antonio Jesús
The present work is an analysis of the performance of the basic vector operations AXPY, DOT and SpMV using OpenCL. The code was tested on the NVIDIA Tesla S2050 GPU and Intel Xeon Phi 3120A coprocessor. Due to the nature of the AXPY function, only two versions were implemented, the routine to be executed by the CPU and the kernel to be executed on the previously mentioned devices. It was studied how they perform for different vector’s sizes. Their results show the NVIDIA architecture better suited for the smaller vectors sizes and the Intel architecture for the larger vector’s sizes. For the DOT and SpMV functions, there are three versions implemented. The first is the CPU routine, the second one is an OpenCL kernel that uses local memory and the third one is an OpenCL kernel that only uses global memory. The kernels that use local memory are tested by varying the size of the work-group; the kernels that only uses global memory are tested by varying the arrays size. In the case of the first ones, the results show the optimum work-group size and that the NVIDIA architecture benefits from the use of local memory. For the latter kernels, the results show that larger computational loads benefits the Intel architecture
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

Landing sites detection using LiDAR data on manycore systems

  • García Lorenzo, Óscar
  • Martínez Sánchez, Jorge
  • López Vilariño, David
  • Fernández Pena, Anselmo Tomás
  • Cabaleiro Domínguez, José Carlos
  • Fernández Rivera, Francisco
Helicopters are widely used in emergency situations, where knowing if a geographical location is adequate for landing is a critical issue, and it is far from being a straightforward task. In this work, we present a method to detect and classify landing sites from LiDAR data in parallel on multi- and manycore systems using OpenMP. Load balancing was identified as the main cause of poor performance because the computational cost depends mainly on the input data. Results for a set of LiDAR point clouds that represent different real scenarios were used as case studies in this work. Balancing strategies for three different multi- and manycore systems were analyzed. The proposed load balancing techniques increase performance up to three times from the unbalanced case.
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

HypeRvieW: an open source desktop application for hyperspectral remote-sensing data processing

  • Suárez Garea, Jorge Alberto
  • Ordóñez Iglesias, Álvaro
  • Blanco Heras, Dora
  • Argüello Pedreira, Francisco
In this article, we present a desktop application for the analysis, reference data generation, registration, and supervised spatial-spectral classification of hyperspectral remote-sensing images through a simple and intuitive interface. Regarding the classification ability, the different classification schemes are implemented by using a chain structure as a base. It consists of five configurable stages that must be executed in a fixed order: preprocessing, spatial processing, pixel-wise classification, combination, and post-processing. The modular implementation makes its extension easy by adding new algorithms for each stage or new classification chains. The tool has been designed as a platform that is open to the incorporation of algorithms by the users interested in comparing classification schemes. As an example of use, a classification scheme based on the Quick Shift (QS) algorithm for segmentation and on Extreme Learning Machines (ELMs) or Support Vector Machines (SVMs) for classification is also proposed. The application is license-free, runs on the Linux operating system, and was developed in C language using the GTK library, as well as other free libraries to build the graphical user interfaces (GUIs)
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

FinFET Versus Gate-All-Around Nanowire FET: Performance, Scaling, and Variability

  • Nagy, Daniel
  • Indalecio Fernández, Guillermo
  • García Loureiro, Antonio Jesús
  • Elmessary, Muhammad A.
  • Seoane Iglesias, Natalia
Performance, scalability, and resilience to variability of Si SOI FinFETs and gate-all-around (GAA) nanowires (NWs) are studied using in-house-built 3-D simulation tools. Two experimentally based devices, a 25-nm gate length FinFET and a 22-nm GAA NW are modeled and then scaled down to 10.7and 10-nm gate lengths, respectively. A TiN metal gate work-function granularity (MGG) and line edge roughness (LER) induced variability affecting OFF and ON characteristics are investigated and compared. In the OFF-region, the FinFETs have over an order of magnitude larger OFF-current that those of the equivalent GAA NWs. In the ON-region, the 25/10.7-nm gate length FinFETs deliver 20/58% larger ON-current than the 22/10-nm gate length GAA NWs. The FinFETs are more resilient to the MGG and LER variability in the subthreshold compared to the GAA NWs. However, the MGG ON-current variability is larger for the 10.7-nm FinFET than that for the 10-nm GAA NW. The LER ON-current variability depends largely on the RMS height; whereas a 0.6-nm RMS height yields a similar variability for both FinFETs and GAA NWs. Finally, the industry preferred (110) channel orientation is more resilient to the MGG and LER variability in both architectures.
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

Benchmarking of FinFET, Nanosheet, and Nanowire FET Architectures for Future Technology Nodes

  • Nagy, Daniel
  • Espiñeira Deus, Gabriel
  • Indalecio Fernández, Guillermo
  • García Loureiro, Antonio Jesús
  • Kalna, Karol
  • Seoane Iglesias, Natalia
Nanosheet (NS) and nanowire (NW) FET architectures scaled to a gate length (L G ) of 16 nm and below are benchmarked against equivalent FinFETs. The device performance is predicted using a 3D finite element drift-diffusion/Monte Carlo simulation toolbox with integrated 2D Schrödinger equation based quantum corrections. The NS FET is a viable replacement for the FinFET in high performance (HP) applications when scaled down to L G of 16 nm offering a larger on-current (I ON ) and slightly better sub-threshold characteristics. Below L G of 16 nm, the NW FET becomes the most promising architecture offering an almost ideal sub-threshold swing, the smallest off-current (I OFF ), and the largest I ON /I OFF ratio out of the three architectures. However, the NW FET suffers from early ION saturation with the increasing gate bias that can be tackled by minimizing interface roughness and/or by optimisation of a doping profile in the device body
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

PRECISION: A Reconfigurable SIMD/MIMD Coprocessor for Computer Vision Systems-on-Chip

  • Nieto Lareo, Alejandro Manuel
  • López Vilariño, David
  • Brea Sánchez, Víctor Manuel
Computer vision applications have a large disparity in operations, data representation and memory access patterns from the early vision stages to the final classification and recognition stages. A hardware system for computer vision has to provide high flexibility without compromising performance, exploiting massively spatial-parallel operations but also keeping a high throughput on data-dependent and complex program flows. Furthermore, the architecture must be modular, scalable and easy to adapt to the needs of different applications. Keeping this in mind, a hybrid SIMD/MIMD architecture for embedded computer vision is proposed. It consists of a coprocessor designed to provide fast and flexible computation of demanding image processing tasks of vision applications. A 32-bit 128-unit device was prototyped on a Virtex-6 FPGA which delivers a peak performance of 19.6 GOP/s and 7.2 W of power dissipation
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

GPU Accelerated FFT-Based Registration of Hyperspectral Scenes

  • Ordóñez Iglesias, Álvaro
  • Argüello Pedreira, Francisco
  • Blanco Heras, Dora
Registration is a fundamental previous task in many applications of hyperspectrometry. Most of the algorithms developed are designed to work with RGB images and ignore the execution time. This paper presents a phase correlation algorithm on GPU to register two remote sensing hyperspectral images. The proposed algorithm is based on principal component analysis, multilayer fractional Fourier transform, combination of log-polar maps, and peak processing. It is fully developed in CUDA for NVIDIA GPUs. Different techniques such as the efficient use of the memory hierarchy, the use of CUDA libraries, and the maximization of the occupancy have been applied to reach the best performance on GPU. The algorithm is robust achieving speedups in GPU of up to 240.6×
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

Caffe CNN-based Classification of Hyperspectral Images on GPU

  • Suárez Garea, Jorge Alberto
  • Blanco Heras, Dora
  • Argüello Pedreira, Francisco
Deep learning techniques based on Convolutional Neural Networks (CNNs) are extensively used for the classification of hyperspectral images. These techniques present high computational cost. In this paper, a GPU (Graphics Processing Unit) implementation of a spatial-spectral supervised classification scheme based on CNNs and applied to remote sensing datasets is presented. In particular, two deep learning libraries, Caffe and CuDNN, are used and compared. In order to achieve an efficient GPU projection, different techniques and optimizations have been applied. The implemented scheme comprises Principal Component Analysis (PCA) to extract the main features, a patch extraction around each pixel to take the spatial information into account, one convolutional layer for processing the spectral information, and fully connected layers to perform the classification. To improve the initial GPU implementation accuracy, a second convolutional layer has been added. High speedups are obtained together with competitive classification accuracies.
RODIN. Repositorio de Objetos de Docencia e Investigación de la Universidad de Cádiz
Publicación Artículo científico (JournalArticle)

GPU Projection of ECAS-II Segmenter for Hyperspectral Images Based on Cellular Automata

  • López Fandiño, Javier
  • Priego Torres, Blanca María
  • Blanco Heras, Dora
  • Argüello, Francisco
  • Duro, Richard J.
Segmentation plays a crucial role in the analysis of multidimensional images, such as those used in remote sensing. Typically, segmentation algorithms for these images start by reducing their dimensionality, which can lead to the loss of potentially important information for the segmentation process. Evolutionary cellular automata segmentation (ECAS-II) offers an alternative by utilizing a cellular automata-based approach that considers all the spectral information in a hyperspectral image, without resorting to dimensionality reduction techniques. This paper introduces an efficient implementation of ECAS-II on a graphics processing unit (GPU) for segmenting hyperspectral land cover images. The proposed method is integrated into a spectral-spatial classification framework based on extreme learning machines (ELM). Experimental results indicate that this approach provides better accuracy for land cover classification compared to other segmentation-based spectral-spatial classification techniques.