arXiv subjects

Tianyang Wang

Publications and source records attributed to Tianyang Wang.

135 records · Page 8Linked to original sources

arXiv2020

BDAQ53, a versatile pixel detector readout and test system for the ATLAS and CMS HL-LHC upgrades

BDAQ53 is a readout system and verification framework for hybrid pixel detector readout chips of the RD53 family. These chips are designed for the upgrade of the inner tracking detectors of the ATLAS and CMS experiments. BDAQ53 is used in applications where versatility and rapid customization are required, such as in laboratory testing environments, test beam campaigns, and permanent setups for quality control measurements. It consists of custom and commercial hardware, a Python-based software framework, and FPGA firmware. BDAQ53 is developed as open source software with both software and firmware being hosted in a public repository.

physics.ins-det↗

arXiv2020

DMAPS Monopix developments in large and small electrode designs

LF-Monopix1 and TJ-Monopix1 are depleted monolithic active pixel sensors (DMAPS) in 150 nm LFoundry and 180 nm TowerJazz CMOS technologies respectively. They are designed for usage in high-rate and high-radiation environments such as the ATLAS Inner Tracker at the High-Luminosity Large Hadron Collider (HL-LHC). Both chips are read out using a column-drain readout architecture. LF-Monopix1 follows a design with large charge collection electrode where readout electronics are placed inside. Generally, this offers a homogeneous electrical field in the sensor and short drift distances. TJ-Monopix1 employs a small charge collection electrode with readout electronics separated from the electrode and an additional n-type implant to achieve full depletion of the sensitive volume. This approach offers a low sensor capacitance and therefore low noise and is typically implemented with small pixel size. Both detectors have been characterized before and after irradiation using lab tests and particle beams.

physics.ins-det↗

arXiv2019

The Monopix chips: Depleted monolithic active pixel sensors with a column-drain read-out architecture for the ATLAS Inner Tracker upgrade

Two different depleted monolithic CMOS active pixel sensor (DMAPS) prototypes with a fully synchronous column-drain read-out architecture were designed and tested: LF-Monopix and TJ-Monopix. These chips are part of a R&D effort towards a suitable implementation of a CMOS DMAPS for the HL-LHC ATLAS Inner Tracker. LF-Monopix was developed using a 150nm CMOS process on a highly resistive substrate (>2 k$Ω\,$cm), while TJ-Monopix was fabricated using a modified 180 nm CMOS process with a 1 k$Ω\,$cm epi-layer for depletion. The chips differ in their front-end design, biasing scheme, pixel pitch, dimensions of the collecting electrode relative to the pixel size (large and small electrode design, respectively) and the placement of read-out electronics within such electrode. Both chips were operational after thinning down to 100 $\mathrmμ$m and additional back-side processing in LF-Monopix for total bulk depletion. The results in this work include measurements of their leakage current, noise, threshold dispersion, response to minimum ionizing particles and efficiency in test beam campaigns. In addition, the outcome from measurements after irradiation with neutrons up to a dose of $1\times10^{15}\,\mathrm{n_{eq} / cm}^{2}$ and its implications for future designs are discussed.

physics.ins-det↗

arXiv2018

Instance-based Deep Transfer Learning

Deep transfer learning recently has acquired significant research interest. It makes use of pre-trained models that are learned from a source domain, and utilizes these models for the tasks in a target domain. Model-based deep transfer learning is probably the most frequently used method. However, very little research work has been devoted to enhancing deep transfer learning by focusing on the influence of data. In this paper, we propose an instance-based approach to improve deep transfer learning in a target domain. Specifically, we choose a pre-trained model from a source domain and apply this model to estimate the influence of training samples in a target domain. Then we optimize the training data of the target domain by removing the training samples that will lower the performance of the pre-trained model. We later either fine-tune the pre-trained model with the optimized training data in the target domain, or build a new model which is initialized partially based on the pre-trained model, and fine-tune it with the optimized training data in the target domain. Using this approach, transfer learning can help deep learning models to capture more useful features. Extensive experiments demonstrate the effectiveness of our approach on boosting the quality of deep learning models for some common computer vision tasks, such as image classification.

cs.CV↗

arXiv2018

Data Dropout: Optimizing Training Data for Convolutional Neural Networks

Deep learning models learn to fit training data while they are highly expected to generalize well to testing data. Most works aim at finding such models by creatively designing architectures and fine-tuning parameters. To adapt to particular tasks, hand-crafted information such as image prior has also been incorporated into end-to-end learning. However, very little progress has been made on investigating how an individual training sample will influence the generalization ability of a model. In other words, to achieve high generalization accuracy, do we really need all the samples in a training dataset? In this paper, we demonstrate that deep learning models such as convolutional neural networks may not favor all training samples, and generalization accuracy can be further improved by dropping those unfavorable samples. Specifically, the influence of removing a training sample is quantifiable, and we propose a Two-Round Training approach, aiming to achieve higher generalization accuracy. We locate unfavorable samples after the first round of training, and then retrain the model from scratch with the reduced training dataset in the second round. Since our approach is essentially different from fine-tuning or further training, the computational cost should not be a concern. Our extensive experimental results indicate that, with identical settings, the proposed approach can boost performance of the well-known networks on both high-level computer vision problems such as image classification, and low-level vision problems such as image denoising.

cs.CV↗

arXiv2018

Depleted Fully Monolithic Active CMOS Pixel Sensors (DMAPS) in High Resistivity 150~nm Technology for LHC

Depleted monolithic CMOS active pixel sensors (DMAPS) have been developed in order to demonstrate their suitability as pixel detectors in the outer layers of a toroidal LHC apparatus inner tracker (ATLAS ITk) pixel detector in the high-luminosity large hadron collider (HL-LHC). Two prototypes have been fabricated using 150 nm CMOS technology on high resistivity (> 2 k$Ω$ $cm^2$) wafers. The chip size is equivalent to that of the current ATLAS pixel detector readout chip. One of the prototypes is used for detailed characterization of the sensor and the analog readout of the DMAPS. The other is a fully monolithic DMAPS including fast readout digital logic that handles the required hit rate. In order to yield a strong homogeneous electric field within the sensor volume, thinning of the wafer was tested. The prototypes were irradiated with X-ray up to a total ionization dose (TID) of 50 Mrad and with neutrons up to non-ionizing energy loss (NIEL) of $10^{15}$ $n_{eq}/cm^2$. The analog readout circuitry maintained its performance after TID irradiation, and the hit-efficiency at > $10^7$ noise occupancy was as high as 98.9 % after NIEL irradiation.

physics.ins-det↗

arXiv2017

Development of depleted monolithic pixel sensors in 150 nm CMOS technology for the ATLAS Inner Tracker upgrade

This work presents a depleted monolithic active pixel sensor (DMAPS) prototype manufactured in the LFoundry 150 nm CMOS process. The described device, named LF-Monopix, was designed as a proof of concept of a fully monolithic sensor capable of operating in the environment of outer layers of the ATLAS Inner Tracker upgrade for the High Luminosity Large Hadron Collider (HL-LHC). Implementing such a device in the detector module will result in a lower production cost and lower material budget compared to the presently used hybrid designs. In this paper the chip architecture will be described followed by the simulation and measurement results.

physics.ins-det↗

arXiv2017

Dilated Deep Residual Network for Image Denoising

Variations of deep neural networks such as convolutional neural network (CNN) have been successfully applied to image denoising. The goal is to automatically learn a mapping from a noisy image to a clean image given training data consisting of pairs of noisy and clean images. Most existing CNN models for image denoising have many layers. In such cases, the models involve a large amount of parameters and are computationally expensive to train. In this paper, we develop a dilated residual CNN for Gaussian image denoising. Compared with the recently proposed residual denoiser, our method can achieve comparable performance with less computational cost. Specifically, we enlarge receptive field by adopting dilated convolution in residual network, and the dilation factor is set to a certain value. We utilize appropriate zero padding to make the dimension of the output the same as the input. It has been proven that the expansion of receptive field can boost the CNN performance in image classification, and we further demonstrate that it can also lead to competitive performance for denoising problem. Moreover, we present a formula to calculate receptive field size when dilated convolution is incorporated. Thus, the change of receptive field can be interpreted mathematically. To validate the efficacy of our approach, we conduct extensive experiments for both gray and color image denoising with specific or randomized noise levels. Both of the quantitative measurements and the visual results of denoising are promising comparing with state-of-the-art baselines.

cs.CV↗

arXiv2017

An ELU Network with Total Variation for Image Denoising

In this paper, we propose a novel convolutional neural network (CNN) for image denoising, which uses exponential linear unit (ELU) as the activation function. We investigate the suitability by analyzing ELU's connection with trainable nonlinear reaction diffusion model (TNRD) and residual denoising. On the other hand, batch normalization (BN) is indispensable for residual denoising and convergence purpose. However, direct stacking of BN and ELU degrades the performance of CNN. To mitigate this issue, we design an innovative combination of activation layer and normalization layer to exploit and leverage the ELU network, and discuss the corresponding rationale. Moreover, inspired by the fact that minimizing total variation (TV) can be applied to image denoising, we propose a TV regularized L2 loss to evaluate the training effect during the iterations. Finally, we conduct extensive experiments, showing that our model outperforms some recent and popular approaches on Gaussian denoising with specific or randomized noise levels for both gray and color images.

cs.CV↗