Read Noise Research Articles

Characterization of Cameras for the COSMO K-coronagraph

Abstract Digital image sensors are ubiquitous in astronomical instrumentation and it is well known that they suffer from issues that must be corrected for data to be scientifically useful. I present discussion on errors resulting from digitization and characterization of nonlinearity and ADC errors of the PhotonFocus MV-D1024E cameras selected for the K-coronagraph of the Coronal Solar Magnetism Observatory. I derive an analytic expression for quantization errors. The MV-D1024E camera has adequate bit depth for which quantization error is not an issue. I show that this is not the case for all cameras, particularly those with deep wells and low read noise. The impact of nonlinearity and ADC errors on science observations of the K-coronagraph is analyzed using a simplified telescope model. Errors caused by the camera ADCs result in systematic errors in the measurement of the polarimetric signal of several times 10−9 B ⊙, which is about an order of magnitude above the desired sensitivity. I demonstrate a method for post-facto data correction using a lookup table and derive parameters from camera characterization measurements that were made with a lab setup. Nonlinearity is traditionally addressed with a global correction. I show through analysis of calibration data that for the MV-D1024E this correction leaves residual systematic errors after dark and gain correction of up to 1% of the signal. I demonstrate that a pixel-wise correction of nonlinearity reduces the errors to below 0.1%. These corrections are necessary for the K-coronagraph data products to meet the science requirements. They have been implemented in the instrument data acquisition system and data reduction pipeline. While no other instruments besides the K-coronagraph or cameras besides the MV-D1024E are discussed here, the results are illustrative for all instruments and cameras.

Effect of the Transition Points Mismatch on Quanta Image Sensors.

Mathematical models and imaging models that show the relationship between the transition points mismatch of analog-to-digital converters (ADCs) and the bit error rate (BER) in single-bit and multi-bit quanta image sensors (QISs) are established. The mathematical models suggest that when the root-mean-square (r.m.s.) of the read noise in jots is 0.15e−, the standard deviation of the transition points should be less than 0.15e− to ensure that the BER is lower than 1% in the single-bit QIS, and 0.21e− to ensure that the BER is lower than 5% in the multi-bit QIS. Based on the mathematical models, the imaging models prove that the fixed-pattern noise (FPN) increases with a stronger transition point mismatch. The imaging models also compare the imaging quality in the case of different spatial oversampling factors and bit depths. The grayscale similarity index (GSI) is 3.31 LSB and 1.74 LSB when the spatial oversampling factors are 256 and 4096, respectively, in the single-bit QIS. The GSI is 1.93 LSB and 1.13 LSB when the bit depth is 3 and 4, respectively, in the multi-bit QIS. It indicates that a higher bit depth and a larger spatial oversampling factor could reduce the effect of the transition points mismatch of1-bit or n-bit ADCs.

ReRAM-Based Processing-in-Memory Architecture for Recurrent Neural Network Acceleration

We present a recurrent neural network (RNN) accelerator design with resistive random-access memory (ReRAM)-based processing-in-memory (PIM) architecture. Distinguished from prior ReRAM-based convolutional neural network accelerators, we redesign the system to make it suitable for RNN acceleration. We measure the system throughput and energy efficiency with the detailed circuit and device characterization. Reprogrammability is enabled with our design, and an RNN friendly pipeline is employed to increase the system throughput. We observe that on average the proposed system achieves $79{\times}$ improvement of computing efficiency compared with graphics processing unit baseline. Our simulation also indicates that to maintain high accuracy and computing efficiency, the read noise standard deviation should be less than 0.2, the device resistance should be at least 1 $\text{M}{\Omega }$ , and the device writes latency should be minimized.

Evaluation of the KASI Detector Performance Test System Using an Andor iKon M CCD Camera

The characterization of detectors installed in space- and ground-based instruments is important to evaluate the system performance. We report the development of a detector performance test system for astronomical applications using the Andor iKon M CCD camera. The performance test system consists of a light source, monochromator, integrating sphere, and power meters. We adopted the Czerny–Tuner monochromator with three ruled gratings and one mirror, which covers a spectral range of 200–9,000 nm with a spectral resolution of ~1 nm in the visible region. Various detector characteristics, such as the quantum efficiency, sensitivity, and noise, can be measured in wide wavelength ranges from the visible to mid-infrared regions. We evaluated the Korea Astronomy and Space Science Institute (KASI) detector performance test system by using the performance verification of the Andor iKon-M CCD camera. The test procedure includes measurements of the conversion gain (2.86 e−/ADU), full well capacity (130 K e−), nonlinearity, and pixel defects. We also estimated the read noise, dark current, and quantum efficiency as a function of the temperature. The lowest measured read noise is 12 e−. The dark current at 223 K was determined to be 7 e−/s/pix and its doubling temperature is 5.3°C ± 0.2°C at an activation energy of 0.6 eV. The maximum quantum efficiency at 223 K was estimated to be 93 % ± 2 %. We proved that the quantum efficiency is sensitive to the operating temperature. It varies up to 5 % in the visible region, while the variation increases to 30 % in the near-infrared region. Based on the comparison of our results with the test report by the vendor, we conclude that our performance test results are consistent with those from the vendor considering the test environment. We also confirmed that the KASI detector performance test system is reliable and our measurement method and analysis are accurate.

Do orienting responses predict annoyance from irrelevant sounds?

Unexpected sounds are distracting and can be annoying but individuals may differ in susceptibility to them. Irrelevant sounds occurring at sparse temporal intervals induce a psychophysiological orienting response reflecting involuntary capture of attention away from the primary task. We hypothesize that the frequency and/or magnitude of individual listeners’ orienting responses to irrelevant sounds will predict annoyance ratings and task performance in distracting noise. Participants read essays while seated in a comfortable chair in a sound-shielded booth facing a semicircular array of 6 speakers located 1.5 m away at 30°, 60°, and 90° to the left and right. Unintelligible background speech (ISTS) played at 60 dB(A) SPL from each loudspeaker (unsynchronized). At 50–70 s intervals one of 12 non-speech sounds (IADS) played for 6 s from one loudspeaker at approximately 70 dB(A) SPL. Order and location of sounds were randomized but each sound played from each speaker exactly once over the experiment (72 trials, ~80 min). Cardiovascular, electrodermal, electro-ocular, and bilateral posterior auricular muscle activity were recorded from participants to quantify orienting response. Behavioral measures of reading comprehension, noise sensitivity, personality traits, and subjective effort, frustration and annoyance were also collected and will be related to physiological measures.

P092 Validation of the automated NGSGO library preparation workflow for HLA typing on the BIOMEK 4000

Aim The GenDx NGSgo workflow for HLA typing using Next-generation sequencing (NGS) consists of four practical steps: amplification, library preparation, sequencing and data analysis. Out of these four steps the library preparation is the most labor intensive and most interspersed by incubations. To decrease hands-on time and reduce the risk of human error, NGSgo library preparation can be automated from start to finish using a liquid handler workstation. Here we describe the results of the automation of the NGSgo library preparation for Illumina on a Beckman Coulter Biomek 4000 workstation. The workflow was fully automated from amplicon pool to finalized library, making it completely hands-off to accommodate overnight runs and allow the operator to step away from the machine. Methods The Biomek setup included a thermal cycler, 96-well plate magnet, two peltiers devices for cooling 96-well plates and 1.5 ml tubes, a gripper tool, reagent reservoirs and liquid waste disposal. Pipetting was performed by two pipet tools. This setup allowed for flexible processing of 4 to 24 samples per run, with any number of loci. Per run only three boxes of pipetting tips were used for 24 samples. Panels with variable numbers of samples and replicate panels were processed with the Biomek. Subsequently, the pooled libraries were sequenced on a MiSeq system and the resulting data was analyzed with NGSengine software (GenDx). Results MiSeq run parameters were indicative of a high-quality sequencing runs. Read depth, mappability and noise levels of the sequence data presented here showed that the automated workflow resulted in high quality reads and accurate typings. Results from the replicate panel showed a 100% typing match. Conclusions – NGSgo library preparation for 4 to 24 samples can be performed on the Biomek 4000 platform, without manual intervention. – The automated workflow results in high-quality reproducible libraries. – Performing automated NGSgo library preparation as described here reduces hands-on time from 1.5 h to 10 min as compared to manual execution. B. Luiken: 5. Employee; Company/Organization; GenDx. E. Tavkin: 5. Employee; Company/Organization; Beckman Coulter. S. Parakh: 5. Employee; Company/Organization; Beckman Coulter. J. Lu: 5. Employee; Company/Organization; Beckman Coulter. A. Jackson: 5. Employee; Company/Organization; Beckman Coulter. S. van Wageningen: 5. Employee; Company/Organization; GenDx. M. Penning: 5. Employee; Company/Organization; GenDx.

Characterization of the SPARC4 CCDs

We present the photometric characterization of the four iXon Ultra 888 CCD cameras of the SPARC4 instrument, which will be installed on the 1.6 m telescope of the Pico dos Dias Observatory in Brazil. We applied experimental methodologies for a systematic characterization of the read noise, electronic gain, dark current, and quantum efficiency of the CCDs. We have analyzed the statistical distribution of the read noise, and also its spatial gradient and temporal variability, where we obtained an average value of the read noise of 6.33 electrons. We applied the Janesick method to determine the electronic gain, where we obtained an average value of 3.35 e-/ADU. We have also obtained an average dark current of 0.00014 e- pix−1 s−1 for CCD internal temperature of −70 °C. We have inspected the dependency of the dark current with temperature and the spatial distribution of the dark current, where we found a variable profile in the CCD 9917. We developed an experiment using a bench-mounted monochromator to obtain the spectral dependency of the quantum efficiency in the spectral range between 350 nm and 1100 nm, where we measured the quantum efficiency for each camera. The camera 9915 presents the highest quantum efficiency of 95.8%. Our results are compared with those from the manufacturer. These experiments allow us to diagnose the performance of these CCD cameras, an important sub-system of the SPARC4 instrument. It also provides a systematic way for monitoring the aging of the CCDs.

Effects of the pre-processing algorithms in fault diagnosis of wind turbines

The wind sectors pends roughly 2200M€ in repair the wind turbines failures. These failures do not contribute to the goal of reducing greenhouse gases emissions. The 25–35% of the generation costs are operation and maintenance services. To reduce this amount, the wind turbine industry is backing on the Machine Learning techniques over SCADA data. This data can contain errors produced by missing entries, uncalibrated sensors or human errors. Each kind of error must be handled carefully because extreme values are not always produced by data reading errors or noise. This document evaluates the impact of removing extreme values (outliers) applying several widely used techniques like Quantile, Hampel and ESD with the recommended cut-off values. Experimental results on real data show that removing outliers systematically is not a good practice. The use of manually defined ranges (static and dynamic) could be a better filtering strategy.

Efficient Convex Optimization for Energy-Based Acoustic Sensor Self-Localization and Source Localization in Sensor Networks.

The energy reading has been an efficient and attractive measure for collaborative acoustic source localization in practical application due to its cost saving in both energy and computation capability. The maximum likelihood problems by fusing received acoustic energy readings transmitted from local sensors are derived. Aiming to efficiently solve the nonconvex objective of the optimization problem, we present an approximate estimator of the original problem. Then, a direct norm relaxation and semidefinite relaxation, respectively, are utilized to derive the second-order cone programming, semidefinite programming or mixture of them for both cases of sensor self-location and source localization. Furthermore, by taking the colored energy reading noise into account, several minimax optimization problems are formulated, which are also relaxed via the direct norm relaxation and semidefinite relaxation respectively into convex optimization problems. Performance comparison with the existing acoustic energy-based source localization methods is given, where the results show the validity of our proposed methods.

Projected Pupil Plane Pattern: an alternative LGS wavefront sensing technique

We have analysed and simulated a novel alternative Laser Guide Star (LGS) configuration termed Projected Pupil Plane Pattern (PPPP), including wavefront sensing and the reconstruction method. A key advantage of this method is that a collimated beam is launched through the telescope primary mirror, therefore the wavefront measurements do not suffer from the effects of focal anisoplanatism. A detailed simulation including the upward wave optics propagation, return path imaging, and linearized wavefront reconstruction has been presented. The conclusions that we draw from the simulation include the optimum pixel number across the pupilN = 32, the optimum number of Zernike modes (which is 78), propagation altitudes h1 = 10 km and h2 = 20 km for Rayleigh scattered returns, and the choice for the laser beam modulation (Gaussian beam). We also investigate the effects of turbulence profiles with multiple layers and find that it does not reduce PPPP performance as long as the turbulence layers are below h1. A signal-to-noise ratio analysis has been given when photon and read noise are introduced. Finally, we compare the PPPP performance with a conventional Shack–Hartmann Wavefront Sensor in an open loop, using Rayleigh LGS or sodium LGS, for 4-m and 10-m telescopes, respectively. For this purpose, we use a full Monte Carlo end-to-end AO simulation tool, Soapy. From these results, we confirm that PPPP does not suffer from focus anisoplanatism.

A 4 + 2T SRAM for Searching and In-Memory Computing With 0.3-V $V_{\mathrm {DDmin}}$

This paper presents a 4+2T SRAM for embedded searching and in-memory-computing applications. The proposed SRAM cell uses the n-well as the write wordline to perform write operations and eliminate the write access transistors, achieving 15% area saving compared with conventional 8T SRAM. The decoupled differential read paths significantly improve read noise margin, and therefore reliable multi-word activation can be enabled to perform in-memory Boolean logic functions. Reconfigurable differential sense amplifiers are employed to realize fast normal read or multi-functional logic operations. Moreover, the proposed 4 + 2T SRAM can be reconfigured as binary content-addressable memory (BCAM) or ternary content-addressable memory (TCAM) for searching operations, achieving 0.13 fJ/search/bit at 0.35 V. The chip is fabricated in 55-nm deeply depleted channel technology. The area efficiency is 65% for a $128 \times 128$ pushed-rule array including all peripherals such as column-wise sense amplifier for read/logic and row-wise sense amplifier for BCAM/TCAM operations. Forty dies across five wafers in different corners are measured, showing a worst-case read/write $V_{\mathrm {DDmin}}$ of 0.3 V.

A dual Vt disturb-free subthreshold SRAM with write-assist and read isolation

This paper presents a new dual Vt 8T SRAM cell having single bit-line read and write, in addition to Write Assist and Read Isolation (WARI). Also a faster write back scheme is proposed for the half selected cells. A high Vt device is used for interrupting the supply to one of the inverters for weakening the feedback loop for assisted write. The proposed cell provides an improved read static noise margin (RSNM) due to the bit-line isolation during the read. Static noise margins for data read (RSNM), write (WSNM), read delay, write delay, data retention voltage (DRV), leakage and average powers have been calculated. The proposed cell was found to operate properly at a supply voltage as small as 0.41 V. A new write back scheme has been suggested for half-selected cells, which uses a single NMOS access device and provides reduced delay, pulse timing hardware requirements and power consumption. The proposed new WARI 8T cell shows better performance in terms of easier write, improved read noise margin, reduced leakage power, and less delay as compared to the existing schemes that have been available so far. It was also observed that with proper adjustment of the cell ratio the supply voltage can further be reduced to 0.2 V.

Optimization and performance of the Robert Stobie Spectrograph Near-InfraRed detector system

At the University of Wisconsin-Madison, we are building and testing the near-infrared (NIR) spectrograph for the Southern African Large Telescope—RSS-NIR. RSS-NIR will be an enclosed cooled integral field spectrograph. The RSS-NIR detector system uses a HAWAII-2RG (H2RG) HgCdTe detector from Teledyne controlled by the SIDECAR ASIC and an Inter-University Centre for Astronomy and Astrophysics (IUCCA) ISDEC card. We have successfully characterized and optimized the detector system and report on the optimization steps and performance of the system. We have reduced the CDS read noise to ∼20 e− for 200 kHz operation by optimizing ASIC settings. We show an additional factor of 3 reduction of read noise using Fowler sampling techniques and a factor of 2 reduction using up-the-ramp group sampling techniques. We also provide calculations to quantify the conditions for sky-limited observations using these sampling techniques.

Accuracy of Shack–Hartmann wavefront sensor using a coherent wound fibre image bundle

AbstractShack–Hartmannwavefront sensors using wound fibre image bundles are desired for multi-object adaptive optical systems to provide large multiplex positioned by Starbugs. The use of a large-sized wound fibre image bundle provides the flexibility to use more sub-apertures wavefront sensor for ELTs. These compact wavefront sensors take advantage of large focal surfaces such as the Giant Magellan Telescope. The focus of this paper is to study the wound fibre image bundle structure defects effect on the centroid measurement accuracy of a Shack–Hartmann wavefront sensor. We use the first moment centroid method to estimate the centroid of a focused Gaussian beam sampled by a simulated bundle. Spot estimation accuracy with wound fibre image bundle and its structure impact on wavefront measurement accuracy statistics are addressed. Our results show that when the measurement signal-to-noise ratio is high, the centroid measurement accuracy is dominated by the wound fibre image bundle structure, e.g. tile angle and gap spacing. For the measurement with low signal-to-noise ratio, its accuracy is influenced by the read noise of the detector instead of the wound fibre image bundle structure defects. We demonstrate this both with simulation and experimentally. We provide a statistical model of the centroid and wavefront error of a wound fibre image bundle found through experiment.

A Low-Noise X-ray Astronomical Silicon-On-Insulator Pixel Detector Using a Pinned Depleted Diode Structure.

This paper presents a novel full-depletion Si X-ray detector based on silicon-on-insulator pixel (SOIPIX) technology using a pinned depleted diode structure, named the SOIPIX-PDD. The SOIPIX-PDD greatly reduces stray capacitance at the charge sensing node, the dark current of the detector, and capacitive coupling between the sensing node and SOI circuits. These features of the SOIPIX-PDD lead to low read noise, resulting high X-ray energy resolution and stable operation of the pixel. The back-gate surface pinning structure using neutralized p-well at the back-gate surface and depleted n-well underneath the p-well for all the pixel area other than the charge sensing node is also essential for preventing hole injection from the p-well by making the potential barrier to hole, reducing dark current from the Si-SiO2 interface and creating lateral drift field to gather signal electrons in the pixel area into the small charge sensing node. A prototype chip using 0.2 μm SOI technology shows very low readout noise of 11.0 e−rms, low dark current density of 56 pA/cm2 at −35 °C and the energy resolution of 200 eV(FWHM) at 5.9 keV and 280 eV (FWHM) at 13.95 keV.

A 45 nm Stacked CMOS Image Sensor Process Technology for Submicron Pixel.

A submicron pixel’s light and dark performance were studied by experiment and simulation. An advanced node technology incorporated with a stacked CMOS image sensor (CIS) is promising in that it may enhance performance. In this work, we demonstrated a low dark current of 3.2 e−/s at 60 °C, an ultra-low read noise of 0.90 e−·rms, a high full well capacity (FWC) of 4100 e−, and blooming of 0.5% in 0.9 μm pixels with a pixel supply voltage of 2.8 V. In addition, the simulation study result of 0.8 μm pixels is discussed.

A Robust Algorithm for State-of-Charge Estimation With Gain Optimization

The charging and discharging procedure of a battery is a typical electrochemical process, which can be modeled as a dynamic system. State of charge (SoC) is a commonly used measure to quantify the charge stored in the battery in relation to its full capacity. Recent efforts of optimizing battery performance require more accurate SoC information. The noise in sensor readings makes the estimation even more challenging, especially in battery-operated systems where the supply voltage of the sensor keeps changing. Traditionally used methods of Coulomb counting and extended Kalman filter suffer from the accumulation of noise and common phenomenon of biased noise, respectively. The traditional approach of dealing with ever-increasing demand for accuracy is to develop more complicated and sophisticated solutions, which generally require special models. A key challenge in the adoption of such systems is the inherent requirement of specialized knowledge and hit-and-trial-based tuning. In this paper, we explore a new dimension from the perspective of a self-tuning algorithm, which can provide accurate SoC estimation without error accumulation by creating a negative feedback loop and enhancing its strength to penalize the estimation error. Specifically, we propose a novel method, which uses a battery model and a conservative filter with a strong feedback, which guarantees that worst-case amplification of noise is minimized. We capitalize on the battery model for data fusion of current and voltage signals for SoC estimation. To compute the best parameters, we formulate the linear matrix inequality conditions, which are optimally solved using open-source tools. This approach also features a low computational expense during estimation, which can be used in real-time applications. Thorough mathematical proofs, as well as detailed experimental results, are provided, which highlight the advantages of the proposed method over traditional techniques.

Photon-number-resolving megapixel image sensor at room temperature without avalanche gain

In several emerging fields of study such as encryption in optical communications, determination of the number of photons in an optical pulse is of great importance. Typically, such photon-number-resolving sensors require operation at very low temperature (e.g., 4 K for superconducting-based detectors) and are limited to low pixel count (e.g., hundreds). In this paper, a CMOS-based photon-counting image sensor is presented with photon-number-resolving capability that operates at room temperature with resolution of 1 megapixel. Termed a quanta image sensor, the device is implemented in a commercial stacked (3D) backside-illuminated CMOS image sensor process. Without the use of avalanche multiplication, the 1.1 μm pixel-pitch device achieves 0.21e− rms average read noise with average dark count rate per pixel less than 0.2e−/s, and 1040 fps readout rate. This novel platform technology fits the needs of high-speed, high-resolution, and accurate photon-counting imaging for scientific, space, security, and low-light imaging as well as a broader range of other applications.

Data reduction pipeline for the CHARIS integral-field spectrograph I: detector readout calibration and data cube extraction

We present the data reduction pipeline for CHARIS, a high-contrast integral-field spectrograph for the Subaru Telescope. The pipeline constructs a ramp from the raw reads using the measured nonlinear pixel response, and reconstructs the data cube using one of three extraction algorithms: aperture photometry, optimal extraction, or $\chi^2$ fitting. We measure and apply both a detector flatfield and a lenslet flatfield and reconstruct the wavelength- and position-dependent lenslet point-spread function (PSF) from images taken with a tunable laser. We use these measured PSFs to implement a $\chi^2$-based extraction of the data cube, with typical residuals of ~5% due to imperfect models of the undersampled lenslet PSFs. The full two-dimensional residual of the $\chi^2$ extraction allows us to model and remove correlated read noise, dramatically improving CHARIS' performance. The $\chi^2$ extraction produces a data cube that has been deconvolved with the line-spread function, and never performs any interpolations of either the data or the individual lenslet spectra. The extracted data cube also includes uncertainties for each spatial and spectral measurement. CHARIS' software is parallelized, written in Python and Cython, and freely available on github with a separate documentation page. Astrometric and spectrophotometric calibrations of the data cubes and PSF subtraction will be treated in a forthcoming paper.

Quantitative performance evaluation of a back-illuminated sCMOS camera with 95% QE for super-resolution localization microscopy.

Scientific Complementary Metal Oxide Semiconductor (sCMOS) cameras were introduced into the market in 2009 and are now becoming a major type of commercial cameras for low-light imaging. sCMOS cameras provide simultaneously low read noise, high readout speed, and large pixel array; however, the relatively low quantum efficiency (QE) of sCMOS cameras has been a major limitation for its application in single molecule imaging, especially super-resolution localization microscopy which requires high detection sensitivity. Here we report the imaging performance of a newly released back-illuminated sCMOS camera (called Dhyana 95 from Tucsen) which is claimed to be the world's first 95% QE sCMOS camera. The imaging performance evaluation is based on a new methodology which is designed to provide paired images from two tested cameras under almost identical experimental conditions. We verified that this new 95% QE sCMOS camera is able to provide superior imaging performance over a representative front-illuminated sCMOS camera (Hamamatsu Flash 4.0 V2) and a popular back-illuminated EMCCD camera (Andor iXon 897 Ultra) in a wide signal range. We hope this study will inspire more studies on using sCMOS cameras in super-resolution localization microscopy, or even single molecule imaging. © 2017 International Society for Advancement of Cytometry.