FENG Jie, LI Yudong,*, FU Jing,3, WEN Lin, GUO Qi
(1.Xinjiang Technical Institute of Physics and Chemistry, Urumqi 830011, China; 2.Xinjiang Key Laboratory of Electronic Information Material and Device, Urumqi 830011, China; 3.University of Chinese Academy of Sciences, Beijing 100049, China)
Abstract: The effects of total ionizing dose (TID) and displacement damage from proton irradiation on an 8-transistor global shutter exposure CMOS image sensor (CIS) within a star sensor were presented to analyze the sources of star sensor performance degradation and the decrease of attitude measurement accuracy. The dark current, dark signal non-uniformity, and photon response non-uniformity versus the displacement damage dose (DDD) were investigated. The star diagonal distance accuracy, and star point centroid positioning accuracy of the star sensor versus the DDD were also analyzed. The influence of space radiation on star sensor performance parameter was analyzed innovatively from a system level point of view. This work lays the foundation for the research of star sensor attitude error measurement and correction technology, and also provides some theoretical basis for the design of high-precision star sensor.
Key words:star sensor; CIS; proton irradiation; performance degradation
Star sensors, important equipment for the attitude determination of spacecraft such as satellites and spaceships, are designed to detect the stars in the sky, create star-parallel-light through the optical components, and pool into a star point. According to the exposure diagram of the image sensor, the data process unit of the star tracker can then determine the pre-process of the image, star point centroid, star identification, and attitude calculation[1-4]. Star sensors generally consist of an optical system, an imaging system, a data processing system, and a data exchange system. The imaging system is an important part of the star sensor, and its performance determines the star sensor detection capability. The imaging system of the star sensor is primarily composed of a charge coupled device (CCD) or complementary metal oxide semiconductor (CMOS) image sensor, which captures the star image. Owing to low power consumption, high levels of integration, and low cost, CMOS image sensors have reached, and even exceeded, the performance level of CCDs[5]. Moreover, CMOS image sensors meet the requirements of space equipment due to their miniaturization and because they are lightweight and have low power consumption. Currently, most star sensors have adopted the imaging system based on 8-transistor (8T) global shutter exposure CMOS image sensors (CIS)[6-7]. High-energy charged particles in a space radiation environment can produce cumulative radiation effects (total ionization dose and displacement damage) and single particle effects on the CIS, resulting in the degradation of performance parameters such as dark current, dark signal non-uniformity noise, light response non-uniformity noise, and even functional failure. Many studies have been dedicated to radiation effects on 4T-CMOS image sensors[8-11], but fewer studies have focused on the radiation effects on an 8T CIS. Le Roch et al. identified the displacement damage defects induced by proton and carbon irradiation in a commercial off-the-shelf (COTS) PPD 8T CIS dedicated to space applications and operating in global shutter mode[7].
In practical applications, there are numerous examples of star sensor performance degradation, due to radiation damage of the CIS. For example, when a certain type satellite passed through the South Atlantic anomaly (SAA), performance degradation of the 8T CIS image sensor was caused by space radiation. As a result, the star sensor generated invalid data in orbit navigation. Counting the coordinate positions of star points corresponding to invalid data, we find that most of invalid data corresponding to the star positions are concentrated in SAA region. The radiation in the SAA area is mainly composed of protons, and the central area of proton irradiation is concentrated near longitude-40° and latitude-30°. When high energy protons are incident into CIS, both ionization and displacement effects occur in the device.
At present, there are many studies on ionization and displacement effects of CIS due to space proton irradiation at home and abroad, but the mechanism of CIS proton radiation effect on star sensor performance has not been carried out. The purpose of this work is to establish the correspondence between space radiation, CIS proton radiation sensitivity parameters, and star sensor performance parameters, the transfer mechanism of CIS parameter degradation to star sensor parameter degradation is revealed. This work is not only the basis for the reliability growth of current star sensors, but also an inevitable requirement for the development of high-performance star sensors in the future.
An 8T global shutter exposure CMV4000-type CIS, produced by CMOSIS of Belgium, was selected for our experiment. The CMV4000 is a high speed CMOS image sensor with 2048 pixel×2048 pixel (one optical inch) developed for machine vision applications. The image array consists of 5.5 μm× 5.5 μm global shutter pixels, which allows exposure during read out while performing correlated double sampling (CDS) operation. The main modules of the CMV4000 are the internal timing generator, serial peripheral interface, temperature sensor, pixel array, analog front-end, and low-voltage differential signal transmission channel.
Fig.1 shows a schematic structure of the 8T CIS pixel unit.
Fig.1 Schematic of 8T CIS pixel unit
Compared to a 4T basic pixel unit, the 8T pixel unit adds transistors, such as sampling tubes (S1 and S2) and a pixel internal pole follower (SF2). Global exposure is realized by storing voltage signals in capacitors C1 and C2.
Irradiations were performed with proton sources at Institute of Heavy Ion Physics, Peking University. The deviation of dosimetry in all sources is within ±5%. All devices were covered before irradiation. All pins of the devices were grounded during irradiation. The irradiation parameters are listed in Table 1.
Table 1 Irradiation parameters
Dark current represents the signal response when a photo-detector is not exposed to light, i.e., the signal measured in the absence of incident photons. It was found that the dark current increased significantly with the DDD as shown in Fig.2. The incident protons interact with the device through coulombic and nuclear interactions, which lead to total ionizing and displacement damages. The ionizing damage occurs in the SiO2and generates positive charges as well as interface states at the Si/SiO2interfaces. However, displacement damage is caused by protons which collide with the silicon atoms within the crystal lattice of the detector array and create vacancy-interstitial pairs. Most of these will recombine after the collision but some will migrate through the lattice and form stable bulk traps with energy levels within the band-gap[12].
Fig.2 PPD CIS dark current versus DDD
The CIS readout images include a series of pixel output and vary from pixel to pixel. The inhomogeneities are no-noise, which makes the output signal vary with time. The inhomogeneities only distribute randomly; therefore, it is better to describe this effect as non-uniformity. There are two basic non-uniformities. First, the dark signal can vary from pixel to pixel. This effect is called dark signal non-uniformity (DSNU). Second, the variation of the sensitivity is called photon response non-uniformity (PRNU). Both DSNU and PRNU of a CIS will be degraded by displacement damage.
In one hand she held a sceptre, and in the other the imperial globe, and on either side of her stood two rows of halberdiers, each smaller than the other, from a seven-foot giant to the tiniest little dwarf25 no higher than my little finger
The DSNU represents the non-uniformity of dark images, which results from the variance of the output from pixel to pixel in the CIS. DSNU is measured according to the EMVA1288 standard. The proton radiation induces an increase not only in the dark signal but also in its pixel to pixel non-uniformity. The change of DSNU with DDD is shown in Fig.3. Proton irradiation increases the pixel-to-pixel non-uniformity. The degradation of DSNU is considered to be due to dark signal fluctuations within the pixels of sensors induced by the various generation states[13-14], and related to the nonionizing energy loss of a particle in elastic and inelastic collisions[15].
Fig.3 Change of DSNU versus DDD
Fig.4 Change of PRNU versus DDD
Photon response non-uniformity (PRNU) is defined as a standard deviation relative to the mean value and presents the spatial standard deviation of the photo response non-uniformity in percent from the mean. The relation between the PRNU and DDD was investigated with light, as shown in Fig.4. We observed that as the DDD increases, the non-uniformity of the light response of the output image becomes worse. The electrons and holes will recombine randomly while migrating, and the random recombination probability of each pixel unit is different, resulting in non-uniformity of the device. It is speculated that the defect energy levels generated by the radiation effect act as the carrier generation-recombination centers, which also increases the electron-hole generation-recombination probability, thereby increasing non-uniformity.
The test system, includes a turntable, collimator, microscope, auto-collimation theodolite, image acquisition system, and static star simulator. Five CMV4000 detectors irradiated by 9×109, 1.35×1010, 1.8×1010, 3.6×1010, and 9×1010cm-2were installed on the test circuit board. The focal length of the optical lens is 24 mm, and the optical field of view is 20°.
First, the calibrations of the turntable and collimator were performed by the auto-collimation theodolite. Then, the relationship between the collimator and installation surface of the turntable was established, and the structure was installed on the turntable. The turntable position was then set to the relative zero position. The camera circuit module, adjusted under the microscope, was installed on the structure. In the experiment, the zero-magnitude and second-magnitude star points were imaged by adjusting the voltage of the static star simulator. The single star point was imaged using the camera, and the optical lens was simultaneously adjusted to optimize star point imaging. The corresponding star point coordinates were used as the main point position (x0,y0). The turntable rotated from approximately -3° to +3°, with a pitch from approximately -3° to +3°, in each direction, according to the cross shape positioning in 1° steps. A total of 13 datasets were collected. The data collection locations are listed in Table 2. Each dataset was stored in 100 frames.
Table 2 Data collection location
Star diagonal distance refers to the angle between the directions of two stars in the geocentric equatorial inertial coordinate system. Considering two stars A and B, their right ascension and declination are (αA,δA) and (αB,δB), respectively. The vectors of stars A and B are given by
(1)
and
(2)
θ=arccosVAVB
(3)
A static star simulator and collimator were used to simulate a star on the celestial sphere. The rotation of the turntable was used to simulate the change of the direction of the star's incident, that is, another star was simulated. The rotation angle of the turntable represented the change angle of the incident direction of the star. The theoretical value of the star diagonal distance can be calculated using the position of the turntable twice. The calculation method is given by
θ=arccosV1V2
(4)
whereV1andV2are the position vectors of the turntable.
The test value of star diagonal distance is obtained by the following method. The star light generated by simulation is imaged in the detector array of the star sensor. First, the star position coordinates (xi,yi) are extracted. According to Eq. (5), the direction of the star point can be calculated by
(5)
where (x0,y0) is the position of the main point calibrated during the experiment, that is, the coordinate position of the star point when the rotation angle and pitch angle of the turntable are equal to 0°. In Eq. (5), (xi,yi) are the coordinates of the star point corre-sponding to the turntable’s rotation and pitch angles (AandE, respectively);fis the focal length of the static star simulator; and
(6)
The star diagonal distance can also be calculated by
θ1=arccosViVj
(7)
The differences between the measured value (θ1) and (θ) theoretical value of the star diagonal distance is equal to the star diagonal distance accuracy. The change of star diagonal distance accuracy of the zero-magnitude star and the second-magnitude star with TID are listed in Table 3. As the proton cumulative fluence increases, the measurement accuracy of the star diagonal distance decreases. When irradiated to 9×1010cm-2, the star diagonal distance of the zero-magnitude star increases by 19.7% compared with the 9×109cm-2value. When irradiated to 9×1010cm-2, the star diagonal distance of the second-magnitude star increases by 106% compared with the 9×109cm-2value. Therefore, the radiation has a great influence on the star diagonal distance.
Table 3 Star diagonal distance accuracy of zero-magnitude star and second-magnitude star versus fluence and DDD
The centroid extraction algorithm, with a threshold, was used to calculate the centroid position of the star. The formula for calculating the centroid position of the star with threshold is given by
(8)
and
(9)
whereI(x,y) is the gray value of the star point at (x,y) andσthis the threshold of star point extraction. Due to the influence of radiation noise and the plug-in detector, the calculated centroid position of the star is different from the theoretical value. This deviation is generally evaluated by the centroid positioning accuracy.
The change of star point centroid positioning accuracy of zero-magnitude and second magnitude stars with TID are listed in Table 4. As the proton cumulative fluence increases, the star point centroid positioning accuracy decreases. When irradiated to 9×1010cm-2, the star point centroid positioning accuracy of the zero-magnitude star increased by 19.3% compared with the 9×109cm-2case. When irradiated to 9×1010cm-2, the star point centroid positioning accuracy of the second-magnitude star increased by 106% compared with the 9×109cm-2case. Therefore, the radiation has a great influence on the star point centroid positioning accuracy. The increase of DSNU results in the change of the gray distribution of the star point, which leads to the shift of the star point centroid position. The proton radiation introduces interface trap charges at the Si-SiO2interface around the PPD. The distribution area of the interface trap charges in each pixel unit and the carrier generation rate are not the same, showing the growth of DSNU across the entire CMOS image sensor. Ionization damage and displacement damage increases the difference of the dark current generation rate between pixels, which leads to the increase of DSNU. The DSNU is generated by the interface trapped charges. Since the star diagonal distance error is a form of star point centroid positioning error, it is also caused by the increase of DSNU.
Table 4 Star point centroid positioning accuracy of zero-magnitude star and second-magnitude star versus fluence and DDD
10 MeV proton radiation effects on an 8T CIS induced by proton radiation were investigated. The parameters of the CIS, such as dark current, DSNU, and PRNU, were measured after irradiation. The internal interface states and trapped positive charges of pixels increase due to irradiation, which lead to the increase of dark current, DSNU, and PRNU.
As the proton cumulative fluence and DDD increases, the star diagonal distance accuracy and star point centroid positioning accuracy decreased. The increase of DSNU results in the change of gray distribution of the star point, which leads to the shift of the star point centroid position. As star diagonal distance error is a form of star point centroid positioning error, it is also caused by the increase of DSNU.
The results provided in this study lay a foundation for the research of star sensor attitude error measurement and correction technology. We have also provided a theoretical basis for the design of a high-precision star sensor. In practical applications, due to the invalid data of star sensor in the SAA region, it is necessary to study further the influence of proton radiation on the performance parameters of a star sensor.