Mechanism of Total Ionizing Dose Effects of CMOS Image Sensors on Camera Resolution

Abstract

The nuclear industry and other high-radiation environments often need remote monitoring equipment with advanced cameras to achieve precise remote control operations. CMOS image sensors, as a critical component of these cameras, get exposed to γ-ray irradiation while operating in such environments, which causes performance degradation that adversely affects camera resolution. This study conducted total ionizing dose experiments on CMOS image sensors and camera systems and thoroughly analyzed the impact mechanisms of the dark current, Full Well Capacity, and quantum efficiency of CMOS image sensors on camera resolution. A quantitative evaluation formula was established to evaluate the impact of Full Well Capacity and quantum efficiency of the CMOS image sensor on camera resolution. This study provides a theoretical basis for the evaluation of the radiation resistance of cameras in environments with strong nuclear radiation and the development of radiation-resistant cameras.

1. Introduction

The nuclear industry is crucial to national security, but the presence of strong nuclear radiation in the environment poses significant risks with regard to the operation, maintenance, and emergency response of nuclear facilities. These radiation environments are extremely harmful to human health; therefore, to ensure the safety of staff and facilities, it is necessary to use remote monitoring equipment and radiation-resistant robots for refined remote control operations [1,2]. However, both remote monitoring equipment and radiation-resistant robots rely on cameras to acquire target information, and the environment contains high levels of neutrons and α-, β-, and γ-rays with high dose rates and total doses. The neutron dose levels are generally low outside the operating reactor, and α and β radiation can be effectively shielded by relatively thin housing. γ-rays have strong penetration ability, and their impact on cameras cannot be ignored. The effects of strong nuclear radiation on electronic systems can cause significant camera performance degradation. The degradation of the camera resolution will lead to the loss of information with potentially disastrous consequences. As such, mitigating the impact of γ-rays on cameras is critical for effective remote monitoring and control in nuclear facilities.

Several studies have investigated the impact of radiation on camera resolution in the past. These include studies by KIM et al. in 2004 and 2007, who evaluated the resolution of scintillator-coupled CMOS sensors under X-rays based on the Modulation Transfer Function (MTF) and its sensitivity to dark signals [3,4]. In 2010, Jie Yu et al. conducted an analysis of the effect of CCD (Charge-Coupled Device) camera transient noise on imaging resolution in neutron photography, taking into account specific shielding requirements [5]. However, these studies mainly focused on the relationship between the radiation dose, ray type, and CMOS image sensor (Complementary Metal Oxide Semiconductor Image Sensor, CIS) noise or total ionizing dose (TID) effects on the CCD system, without a detailed analysis of the mechanism by which radiation-sensitive parameters in the CIS affect camera resolution. Furthermore, these studies did not establish quantitative relationships between CIS radiation-sensitive parameters and camera resolution.

This paper focuses on investigating the degradation mechanism of camera resolution under a γ-ray radiation environment. Specifically, it establishes a quantitative evaluation formula for the impact of the CMOS image sensor’s Full Well Capacity (FWC) and quantum efficiency (QE) on camera resolution. This study provides a theoretical foundation for evaluating camera radiation resistance in strong nuclear radiation environments and developing radiation-resistant cameras.

2. Materials and Methods

The test camera follows a modular design and comprises three main components: an optical lens, an image sensor, and a peripheral circuit. The optical lens is connected to the CIS, while the peripheral circuit is linked to the CIS device through a flexible cable. For this test, an ON Semiconductor AR series commercial image sensor with 2.1 million pixels and a single pixel size of 3 μm × 3 μm utilizing RGB Bayer array color filters was used as the image sensor. The image sensor utilized rolling exposure mode, while a commercial automatic zoom lens was employed to capture high-quality images. Additionally, the camera features a self-designed anti-radiation circuit encompassing a power supply module, a digital signal processing module, and a network transmission module.

The irradiation test was carried out on the ⁶⁰Co-γ radiation source of the Xinjiang Institute of Physics and Chemistry, Chinese Academy of Sciences. The camera system irradiation test is shown in Figure 1. The camera system was connected to the PC outside the irradiation room via a network cable.

The irradiation test involved two parts: the camera system irradiation test and the CIS irradiation test. Firstly, the camera system was subjected to irradiation at a dose rate of 28 rad(Si)/s while in an online working state. After reaching the dose points of 70, 110, 140, 180, 210, and 280 krad(Si), the resolution of the camera was tested. When the radiation dose of the camera system exceeded 280 krad(Si), the performance index of the camera dropped significantly, and the working state became abnormal, so the irradiation test was stopped. As the TID of the camera increased, the light transmittance of the camera lens decreased. Thus, during the displacement test of a portion of the dose of the camera system, a supplementary test was conducted by replacing the unirradiated lens. In the CIS irradiation test, only the CIS was irradiated while the peripheral circuit was shielded and protected. The device worked normally during the irradiation process at a dose rate of 28 rad(Si)/s, with irradiation doses of 70, 110, 140, 180, 210, and 280 krad(Si). After reaching the corresponding dose, the key parameters of the CIS and the combined camera system were tested using the photoelectric imaging device radiation damage test system at the Xinjiang Institute of Physics and Chemistry, Chinese Academy of Sciences.

3. Results

3.1. Camera System Resolution Degradation

Camera resolution is the ability to distinguish the number of line pairs per unit length, and it serves as an important parameter to determine the clarity of camera imaging. This is crucial for the human eye to discern whether the actual image is clear and effective. MTF is a function that varies with spatial frequency, and its value ranges from 0 to 1 with constant spatial frequency. A higher MTF value indicates better imaging quality, as it represents higher restoration of the contrast between the object and the image [6]. In this study, the imaging system resolution has been evaluated by calculating the MTF value, which possesses the property of being cascadable. The formula for calculating the MTF in this experiment is given in Equation (1).

$$MTF={MTF}_{cis}\ast {MTF}_{\mathrm{P}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{p}\mathrm{h}\mathrm{e}\mathrm{r}\mathrm{a}\mathrm{l} \mathrm{C}\mathrm{i}\mathrm{r}\mathrm{c}\mathrm{u}\mathrm{i}\mathrm{t}\mathrm{s}}\ast {MTF}_{Optical Lens}=\frac{I_{max}-I_{min}}{I_{max}+I_{min}}$$

(1)

When image noise is low, the Spatial Frequency Response (SFR) test method recommended by ISO 12233 yields stable results. However, if the image noise surpasses the algorithm’s threshold value, the test outcomes will change dramatically. This discrepancy with human perception during actual use can lead to inaccurate assessment of the impact of radiation-induced noise on the camera system’s performance. To precisely assess noise’s effects on the camera’s actual performance in high-radiation environments, this study employed wedge diagrams to evaluate the camera’s resolution, along with Imatest Master to determine the value of Aliasing onset and MTF10.

MTF10 is a classical theoretical value used to describe the resolution of an optical system, while Aliasing onset is the spatial frequency at which the number of bars detected by the software is lower than the total number of wedges. The results obtained from Aliasing onset are not affected by signal processing, such as sharpening or noise reduction, making it suitable for evaluating the resolution of different types of cameras. In practical camera usage, Aliasing onset is more in line with manual subjective discrimination than the theoretical limit value of MTF10. Therefore, it can better solve the problem of evaluating camera resolution in a strong nuclear radiation environment. When pictures taken by the camera have noise, the MTF calculation formula can be deduced from Equation (1) [5].

$$MTF=\frac{{(I}_{max}+I_{\mathrm{l}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t}-\mathrm{n}\mathrm{o}\mathrm{i}\mathrm{s}\mathrm{e}})-(I_{min}+I_{\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{k}-\mathrm{n}\mathrm{o}\mathrm{i}\mathrm{s}\mathrm{e}})}{{(I}_{max}+I_{\mathrm{l}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t}-\mathrm{n}\mathrm{o}\mathrm{i}\mathrm{s}\mathrm{e}})+(I_{min}+I_{\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{k}-\mathrm{n}\mathrm{o}\mathrm{i}\mathrm{s}\mathrm{e}})}=\frac{{(I}_{max}+{\sigma }_{light}n{g}_2)-(I_{min}+{\sigma }_{dark}n{g}_1)}{{(I}_{max}+{\sigma }_{light}n{g}_2)+\left(I_{min}+{\sigma }_{dark}n{g}_1\right)}$$

(2)

The maximum and minimum grayscale values of the image target region under ideal conditions are represented by $I_{max}$ and $I_{min}$, respectively. The noise captured by the camera after irradiation can significantly impact the resolution. Thus, the ratio of the number of noise to the total number of pixels in the maximum gray value region of the image target area is ${\sigma }_{light}$, the ratio of the number of noise to the total number of pixels in the minimum gray value region of the image target area is ${\sigma }_{dark}$, the average gray value of the noise in the minimum gray value region is $g_1$, and the number of pixels is $n$. Then, the increment of the gray value of the image in the region of maximum gray value where noise exists is represented by $I_{\mathrm{l}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t}-\mathrm{n}\mathrm{o}\mathrm{i}\mathrm{s}\mathrm{e}}$, and the average gray value of the noise in this region of maximum gray value is $g_2$, while its value is ${\sigma }_{light}n{g}_2$. Similarly, the increment of the gray value of the image in the region of minimum gray value where noise exists is represented by $I_{\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{k}-\mathrm{n}\mathrm{o}\mathrm{i}\mathrm{s}\mathrm{e}}$, and its value is ${\sigma }_{dark}n{g}_1$. Finally, the ${MTF}_{camera-radiation}$ formula in the size of $I_{max}+{\sigma }_{light}n{g}_2$ and $I_{min}+{\sigma }_{dark}n{g}_1$ has been obtained, which represents the maximum and minimum gray values of the image measured.

Figure 2a shows the change in the calculated MTF value of the camera with the TID. As the irradiation dose increases, the calculated MTF value of the camera decreases. The trend of the camera system’s resolution after irradiation with the TID is presented in Figure 2b. The measured values of the camera resolution are consistent with the trend of decreasing MTF calculated values with the increase of the TID. Moreover, after the irradiation dose reaches 210 krad(Si), the rate of decrease in the camera resolution becomes faster.

For the camera system with only the irradiated CIS, the degradation of the MTF and the resolution is caused by the radiation damage of the CIS. Figure 3 illustrates the trend of camera resolution with the TID for a camera system with only the irradiated CIS. The CIS radiation damage has little effect on the resolution before reaching 210 krad(Si), but it sharply decreases after that.

3.2. The Key Parameters of CIS Degradation

Dark current, Full Well Capacity, and spectral response are key parameters that evaluate the imaging performance of the CIS after irradiation, and their degradation has a significant impact on the overall performance of the camera system. Dark current is the current generated by the pixel cell of an image sensor when it absorbs spontaneously generated electrons due to the presence of defects (interface defects and body defects) under dark conditions, and it is usually measured in e⁻/s [7]. When the CIS is exposed to ionizing radiation, the dark current signal mainly comes from the pixel cell and the peripheral circuit, with the peripheral circuit dark current being a fixed value independent of exposure time. Figure 4 presents the test results of the dark current of the CIS. The dark current of the image sensor increases with the increase of the TID, and it significantly increases at 75–100 and 175–210 krad(Si).

The FWC is the maximum number of electrons that can be stored in the pinned photodiode (PPD) in a saturated state. Figure 5 illustrates the change in the FWC of the CIS with the increase of the TID during γ-ray irradiation at different dose rates. The variation of the FWC at different doses shows some differences: before 100 krad(Si), there is no significant change in the FWC of the CIS, while after 100 krad(Si), the FWC shows a significant decreasing trend. Moreover, the decreasing rate of the FWC increases significantly after 210 krad(Si).

The spectral response of the CIS is an important parameter for evaluating its ability to convert incident photons of different wavelengths into electrical signals, and it is crucial for assessing color reproduction in color cameras. QE is used to characterize the responsiveness of the CIS to light signals of specific wavelengths. Figure 6 shows the degradation ratios of the spectral response curve of the CIS under γ-ray irradiation at 28 rad(Si)/s for incident light wavelengths of 420, 450, 516, 550, and 630 nm. The degradation of the CIS is greater in the short wavelength band, as indicated by the degradation ratios in Figure 6. The degradation ratio decreases gradually with increasing wavelength, and a larger dose is required to show significant degradation.

4. Discussion

For the CIS, the comprehensive MTF can characterize its detail resolution capability, which is composed of three types of MTF: geometric MTF, transfer MTF, and diffusion MTF. Usually, the comprehensive MTF function is obtained by multiplying these three types of MTF in the frequency domain. For the CIS in radiation cameras, because the internal pixel structure of the CIS remains unchanged, the geometric MTF remains unchanged as well. The transfer MTF refers to the charge loss generated during the charge transfer between pixels. The TID effect causes trap positive charges to be generated in the STI region near the Transfer Gate (TG), which induces the production of negative charges on the Si–SiO₂ surface of the STI due to the appearance of trap positive charges. The accumulation of these negative charges increases the regional electron density, reduces the TG channel potential barrier, and, ultimately, allows some photoelectrons in the PPD to transfer to the FD through the channel sidewalls without voltage applied to the TG [8], which leads the transfer MTF and the FWC to decrease with the increase of dose. The diffusion MTF refers to the difference in position of photogenerated carriers caused by the difference in the depth of incidence of incident light for different spectral bands after the incident light enters. The photogenerated carriers that are far away from the depletion region will diffuse freely before entering the depletion region. With increasing TID, the interface trap charge density formed at the SiO₂ layer surface due to the TID effect also increases. The energy level of interface trap charges is close to the center of the bandgap, and they can act as effective recombination centers, increasing the net recombination rate and reducing the lifetime of photogenerated carriers in this region. This directly reduces the diffusion length of carriers and, ultimately, lowers the efficiency of collecting photogenerated carriers in the depletion region. As such, the diffusion MTF also decreases with the increase of dose. Because incident lights of different wavelengths have different penetration depths, longer-wavelength light generates fewer photogenerated carriers near the interface and is less affected by the interface trap charge density [9,10]. Consequently, the degradation of the QE after irradiation is lower for longer-wavelength light.

According to Equation (2), the maximum and minimum gray values of the image before and after irradiation will have a certain impact on the MTF value, where the change in the minimum gray value is mainly affected by the CIS dark current noise, and the change in the maximum gray value is mainly affected by the FWC. The radiation-induced increase in interfacial trap charges at the Si–SiO₂ interface of the 4T pixel structured CIS, especially at the periphery of the shallow trench isolation (STI) region, the TG–PPD overlap region, and the PPD surface [11], which is the main mechanism behind the gradual increase in the CIS dark current with increasing radiation dose. During the γ-ray irradiation process, the Si–SiO₂ interface precipitated in the STI region generates broken suspension bonds and forms interface defects [12]. Unstable gaseous substances, such as silicon monoxide, generated by incomplete reactions between silicon and oxygen at the interface can be emitted from the oxide layer at high temperatures, creating dangling bonds at the interface [13]. Therefore, during ionizing radiation, the density of dangling bonds and point defects will continue to increase with the increase of the TID, becoming one of the main sources of increased dark current after irradiation [14,15]. The dark current increases more significantly at 75–100 and 175–210 krad(Si), and Figure 4 reflects the introduction of different dark current sources with the increase of irradiation dose. By substituting the corresponding gray value parameters of the entire irradiation experimentally collected image measurement area into Equation (2) to calculate the post-irradiation MTF value, the effects of the CIS dark current noise and the FWC on camera resolution are compared and analyzed. Figure 7a shows the MTF value calculated by substituting the maximum gray value measured in the target area of the image under different doses and the minimum gray value measured in the target area of the image under unirradiated conditions, while Figure 7b shows the MTF value calculated by simultaneously substituting the maximum and minimum gray values measured in the target area of the image under different doses.

From Figure 7, it can be seen that the overall trend of the MTF calculation value decreases as the dose increases, and the degree of MTF reduction gradually increases as the dose reaches a certain level. At the same time, whether the minimum gray measurement value under different doses is substituted into Equation (2) has a certain impact on the MTF calculation value, but there is no significant difference in the overall trend. Therefore, the CIS dark current noise has a certain impact on the MTF, but the degradation of the CIS FWC after irradiation has a more significant impact on the MTF.

The test card image captured by the camera is a combination of effective signal and noise, where the noise can be mainly divided into image signal noise and background noise. The image signal noise is caused by scattered photons from external incident light, while the background noise includes the CIS noise and noise from the camera’s peripheral circuit. Under light and dark fields, the RGB three-channel noise value of the CIS calculated using Imatest Master changes with the dose, as shown in Figure 8.

From Figure 8, the noise of the CIS increases with the increase of the TID, and the increase in the CIS noise under the light field condition is much larger than that under the dark field condition. This is because under sufficient light conditions, photon scatter noise is much greater than the dark current noise. In addition, the higher the dose rate, the more obvious the increase in noise. Finally, after 140 krad(Si), the growth rate of the CIS noise increases significantly.

Figure 9 shows the gray level values of the test card images captured by the test cameras with CIS combinations at different dose rates measured using Imatest Master software (imatest, 2020.2, Boulder, CO, USA). The gray level value with serial number 1 represents the maximum gray level value in the image, which is mainly affected by the degradation of the CIS FWC parameter after irradiation.

Based on Figure 8 and Figure 9, it can be seen that different gray level values show varying degrees of change after CIS irradiation. Meanwhile, the change in the CIS noise in Figure 8 results in a change value of less than 0.2 DN compared to the actual gray level value, indicating that the effect of the CIS noise on gray level values is not significant. The impact on the minimum gray level value of the image is greater than on the brightest gray level value. Therefore, combined with the previous analysis of the specific performance of the MTF calculation values in Figure 9, it can be seen that the CIS noise has a certain impact on the MTF values, but it does not change the overall trend of MTF change.

The change in the diffusion MTF is affected by the change in the number of photo-generated carriers caused by QE degradation after irradiation. Therefore, the degradation of the image brightness Y after irradiation is calculated to estimate the degradation of the diffusion MTF. In addition, the camera resolution is determined by the software based on processing of the test card image using brightness Y, which is calculated from the RGB image using Equation (3).

$$Y_0=0.299\times R_0+0.587\times G_0+0.114\times B_0$$

(3)

In Equation (3), $Y_0$ is the gray level value of the transformed image captured by the camera before irradiation, $R_0$ is the red component of the image captured by the camera before irradiation, $G_0$ is the green component of the image captured by the camera before irradiation, and $B_0$ is the blue component of the image captured by the camera before irradiation. The degradation of the R, G, and B values of the image in the camera before and after irradiation is related to the QE degradation of the CIS in the corresponding red, green, and blue light bands after irradiation. Substituting the irradiation degradation rate of the CIS in the corresponding red, green, and blue incident light bands into Equation (3) yields Equation (4).

$$Y_1=0.299\times R_0\times m+0.587\times G_0\times n+0.114\times B_0\times l$$

(4)

In Equation (4), $Y_1$ is the gray level value of the transformed image captured by the camera after irradiation, m is the QE degradation rate of the CIS in the red light band, n is the QE degradation rate of the CIS in the green light band, and $l$ is the QE degradation rate of the CIS in the blue light band. The coefficients for R, G, and B are the influence weights of the change rate of the QE in the corresponding band on the Y value.

Based on the above analysis, the change in the FWC directly affects the maximum gray level value of the image before and after irradiation; the minimum gray level value is affected by the CIS noise, but the effect of noise on the minimum gray level value does not change the overall trend of MTF change. The ratio of brightness Y before and after irradiation reflects the degree of degradation of the diffusion MTF. Considering these factors, Equation (2) is modified to obtain Equation (5), where K represents the conversion gain, which refers to the output image gray level value increase of the unit effective photo-generated electrons after system processing. According to relevant studies, the conversion gain is not a sensitive parameter for TID effects and can generally be regarded as a constant value in parameter calculations for the same device [16].

$${MTF}_{camera-1}={MTF}_{camera}\times \left[\left(\frac{{FWC}_1\times K-I_0}{{FWC}_1\times K+I_0}\right)/\left(\frac{{FWC}_0\times K-I_0}{{FWC}_0\times K+I_0}\right)\right]\times \left(Y_1/Y_0\right)$$

(5)

In Equation (5), ${MTF}_{camera}$ is the camera resolution before irradiation, ${MTF}_{camera-1}$ is the calculated camera resolution after irradiation in LW/PH, ${FWC}_0$ is the Full Well Capacity when the camera is not irradiated, ${FWC}_1$ is the Full Well Capacity after irradiation in e⁻, and K is the camera conversion gain in DN/e⁻. $I_0$ is the minimum grayscale value of the captured image when the camera is not irradiated, and the unit is DN; $Y_0$ is the converted grayscale value of the captured image before irradiation, and the unit is DN; and $Y_1$ is the converted grayscale value of the captured image after irradiation, and the unit is DN. After substituting the experimental values into Equation (5), the theoretical calculation results of the camera resolution were consistent with the actual camera irradiation measurement values. This also indicates that the defects generated by radiation have a significant impact on the CIS photodetector signals, which is an important reason for the decrease in camera resolution. In the subsequent development of radiation-resistant cameras, STI and PPD reinforcement technologies can be used to strengthen the CIS against radiation, thereby reducing the generation of radiation-induced defects, inhibiting the impact of radiation-induced defects on the photodetector signals, and reducing the influence of radiation on camera resolution.

5. Conclusions

After γ-ray radiation, the QE and FWC of the CMOS image sensor degrade with the increase of the TID. Through CMOS image sensor irradiation experiments, camera system irradiation experiments, and result analysis, combined with theoretical deduction, it was found that the degradation of the QE and FWC of the CMOS image sensor is the main cause of the decrease in camera resolution. By revealing the impact mechanism of radiation damage to the CMOS image sensor on camera resolution, a quantitative evaluation formula for the influence of the FWC and QE on camera resolution was established, laying a theoretical foundation for the assessment of camera radiation resistance and the development of radiation-resistant cameras in strong nuclear radiation environments.