Performance of Monolithic CMOS Pixel Sensors Under X-Rays

1. Introduction

Current experimental particle physics facilities require efficient tracking detectors that can operate in extremely demanding environments. Experiments at the Large Hadron Collider (LHC) and its future upgrades, such as the High Luminosity LHC (HL-LHC) [1,2], will experience significantly increased particle rates and radiation levels. As a result, detector technologies must provide excellent spatial resolution, fast signal response, and strong radiation tolerance while minimizing the material budget. These requirements have motivated the development of new types of silicon-based pixel detector capable of meeting the performance demands of future collider experiments. One promising technology in this direction is the Depleted Monolithic Active Pixel Sensor (DMAPS) [3,4,5,6,7]. In DMAPS devices, the charge produced during particle passage is collected in a fully or partially depleted silicon region and processed directly by integrated front-end electronics within each pixel. This approach enables faster signal collection through drift mechanisms, improves radiation tolerance, and allows the sensor to be fabricated as a single integrated structure. Furthermore, monolithic sensors can be produced with thinner silicon layers and without custom bump bonding, thereby reducing the material budget and improving overall detector performance. This work discusses recent advances in monolithic active pixel sensors for large-area tracking applications and their performance characteristics. The investigation focuses on the electrical behaviour and the response of the sensor threshold under X-ray irradiation. Understanding these properties is essential for optimizing detector performance and assessing the suitability of their technology for future high-energy physics experiments.

2. The CMOS Monolithic Pixel Sensor

Monolithic pixel detectors play an important role in particle physics experiments because they provide precise spatial information about particle interactions while maintaining high granularity and fast response. As the luminosity and collision rates of experiments such as the Large Hadron Collider increase, detector technologies must be able to operate efficiently at high radiation levels and high hit rates. In this context, CMOS-based monolithic pixel sensors [3,4,5,6,7] have attracted considerable attention because they combine the sensing element and the readout electronics on a single silicon substrate, enabling compact detector designs with a reduced material budget and improved scalability for large detector areas.

The monolithic active pixel sensor [8,9] is developed using a modified TowerJazz 180 nm CMOS technology. The device was developed to satisfy the demands expected in upcoming high-luminosity accelerator facilities, with emphasis on radiation-hard detectors capable of operating under intense particle fluxes. This sensor consists of an active area containing a $512\times 512$ pixel-matrix, where each pixel has a pitch of approximately $36.4\times 36.4\mu {\mathrm{m}}^2$. This fine pixel size enables precise particle tracking while maintaining high hit-detection efficiency. The sensor is capable of operating at hit rates approaching $100\mathrm{MHz}/{\mathrm{cm}}^2$, making it suitable for the demanding environment expected in upgraded collider detectors.

The second generation of this monolithic sensor [10,11] is also built in the TowerJazz 180 nm process. It features a $224\times 512$ pixel matrix over a $20.2\times 10.1{\mathrm{mm}}^2$ area, keeping the pixel size the same at $36.4\mu {\mathrm{m}}^2$. In this second-generation sensor, a shift register replaced the Ethernet-like slow control and an open-loop front-end was used for lower noise and compactness, with gain-enhancing multi-cascode transistors [12,13,14]. Its asynchronous read-out uses a 37-bit data bus and a 4322-bit control register, with read-out handled by a Kintex-7 FPGA on a KC705 board.

2.1. Description of DACs

A brief description of the DAC parameters [15] used in the second-generation sensor front-end is shown in Table 1. The sensor front-end is shown in Figure 1 [16], with the DAC parameters used. In our study, we changed $ITHR$, $IDB$ and $ICASN$ within their respective ranges, keeping $IBIAS$ constant at 43.

3. The CMOS Monolithic Pixel Sensor Setup

3.1. Electrical Characterization Setup

The electrical characterization of the second-generation CMOS monolithic pixel sensor was performed using a dedicated readout system. The setup is designed to enable precise threshold and noise measurements through controlled Digital-to-Analog Converter (DAC) scans and stable biasing conditions. The second-generation CMOS monolithic pixel sensor was wired to a custom-designed carrier board at CERN, which provides the required biasing and signal routing. The board distributes multiple independent voltage domains, including the low-voltage digital supply (LVDD), the digital core supply (DVDD), the analog supply (AVDD), and substrate-related biases such as PWELL and SUB. These voltages are critical for the operation of the front-end electronics and for tuning the sensor’s depletion and charge-collection properties. A reference voltage (DREF) is also supplied to the on-chip level shifter circuitry to ensure proper signal reset and logic-level compatibility.

The carrier board interfaces with an FPGA-based evaluation board via a high-speed FMC (FPGA Mezzanine Card) connector. The FPGA platform is responsible for sensor configuration, control, and data acquisition. Communication between the FPGA board and the host computer is established via a standard Ethernet interface, enabling remote operation and data transfer within a local network. Five independent low-voltage power supplies power the system, ensuring stable, noise-minimized operation across the different voltage domains. A high-precision digital multimeter is used to monitor and calibrate the supplied voltages and currents. Care is taken to minimize electrical noise by properly grounding and shielding the setup. Data acquisition and control are handled by the MALTA software framework, implemented in C++ and Python. The software allows for full configuration of the chip registers, execution of DAC scans, and readout of pixel responses. Threshold characterization is performed by scanning the injected charge or equivalent DAC settings and recording the pixel response at each setting. The resulting S-curves are analyzed to extract the effective threshold and noise for each pixel, which are then used in the calculation of the threshold and noise for full pixel sensors and groups.

All measurements are carried out under stable environmental conditions, with the temperature maintained at $16\pm 1^{\circ }C$. Temperature stability is ensured to minimize variations in electronic noise and threshold dispersion. The system allows reproducible measurements across different sensor samples and operating conditions.

3.2. X-Ray Irradiation Setup

CMOS monolithic pixel sensor irradiation studies were carried out using a PRECISION X-Rad160 X-ray irradiator installed in the IIT Madras High Energy Physics Laboratory, as shown in Figure 2(b). The irradiation setup is equipped with a tungsten target and can generate X-rays with tube voltages up to 160 kV, producing a broad bremsstrahlung spectrum suitable for radiation-hardness studies of semiconductor devices.

The sensor under test was placed inside the irradiation chamber at a fixed source-to-specimen distance (SSD) of 30 cm. This configuration was chosen to ensure uniform irradiation across the sensor’s active area. The device alignment within the irradiation field was carefully controlled to minimize spatial dose variations. The dose rate was calibrated to approximately 12.7 Gy/min based on the operating parameters of the X-ray tube, including the applied high voltage (up to 160 kV), the tube current (up to 18.7 mA) and the exposure time. The irradiation was performed using the Time Program mode of the X-Rad160 system, in which X-ray generation is maintained at fixed voltage and current settings and automatically terminated after a predefined exposure duration. In the present study, an exposure time of approximately 13 hours was used to achieve the total ionizing dose.

The delivered dose was determined from the calibrated dose rate and exposure time. Systematic uncertainties in dose estimation arise from irradiator calibration and possible variations in beam uniformity, and these are taken into account in the analysis. During irradiation, the sensor was kept under controlled conditions inside the chamber. The bias configuration during irradiation was maintained to be consistent with the measurement requirements, ensuring reproducibility of the radiation effects. After irradiation, the sensor was removed and subjected to electrical characterization under the same controlled temperature conditions as prior to irradiation. The time interval between irradiation and subsequent measurements was minimized to reduce the impact of annealing effects. All post-irradiation measurements were performed at a controlled temperature of $16\pm 1^{\circ }C$, consistent with the pre-irradiation characterization, allowing a direct comparison of sensor performance before and after exposure.

3.3. Measurement Strategy

To evaluate the impact of irradiation on sensor performance, a comparative analysis was performed using DAC-based threshold scans across sensors of different thicknesses. Measurements were carried out both before and after irradiation under identical operating conditions using a setup at IIT Madras, as shown in Figure 2(a). The threshold and noise distributions were extracted from the S-curve fits, enabling a systematic study of radiation-induced variations. The use of a consistent electrical setup, controlled environmental conditions, and calibrated irradiation parameters ensures the reliability and reproducibility of the results presented in this work.

4. Irradiation Results and Discussion

4.1. Threshold Behaviour

The sensor’s threshold distribution is first evaluated at a representative operating point corresponding to the DAC configuration $ITHR=20$, $IDB=80$, and $ICASN=8$. The threshold is extracted from the S-curve of individual pixels, obtained from the response as a function of the injected charge, as illustrated for a representative pixel in Figure 3(a).

Figure 3(c) shows the resulting two-dimensional threshold map of the pixel matrix. The distribution is observed to be largely uniform across the matrix, with only mild spatial variations. No significant localized structures or defective regions are visible, indicating good pixel-to-pixel uniformity at this operating point. The corresponding one-dimensional distribution of the threshold values is shown in Figure 3(b). The distribution is well described by a Gaussian with a mean threshold of approximately $367{\mathrm{e}}^{-}$ and a standard deviation of approximately $13{\mathrm{e}}^{-}$. The relatively narrow spread demonstrates good intrinsic threshold uniformity of the sensor, which is essential for stable operation and uniform detection efficiency across the pixel matrix.

4.1.1. Dependence of Threshold on DAC Parameters and Irradiation

The dependence of the threshold on the front-end tuning parameters is studied by varying $ITHR$, $IDB$, and $ICASN$. Figure 4 illustrates the dependence of the mean threshold on $ICASN$, where each point corresponds to the average over all available configurations of $ITHR$ and $IDB$.

A decreasing trend of threshold with increasing $ICASN$ is observed for both unirradiated and irradiated data. The irradiated sensor consistently exhibits lower threshold values compared to the unirradiated case across the full range of $ICASN$. The distribution spreads, represented by the error bars, remain comparable, indicating that irradiation does not significantly degrade threshold uniformity. To further investigate the combined dependence on $ITHR$ and $IDB$, two-dimensional threshold maps are constructed for a fixed value of $ICASN=4$, as shown in Figure 5. A clear monotonic increase of threshold with increasing $ITHR$ is observed, while a weaker dependence on $IDB$ is seen. A comparison between unirradiated and irradiated conditions reveals a systematic shift in threshold across the parameter space. The absolute change in threshold, shown in Figure 5 (bottom left), indicates an increase of up to $\sim 80$–$90{\mathrm{e}}^{-}$ for low $ITHR$ values. The effect is more pronounced at lower threshold settings, suggesting increased sensitivity of these configurations to irradiation. The relative variation, shown in Figure 5 (bottom right), remains below $\sim 25\%$ over most of the parameter space, with the largest deviations occurring at low $ITHR$. At higher $ITHR$ values, the relative change is significantly reduced, indicating a more stable operating regime under irradiation. These results demonstrate that, while irradiation introduces a measurable shift in the threshold, a wide range of parameter configurations remains operational, allowing optimization of detector performance by selecting appropriate DAC settings.

The dependence of the sensor threshold on $IDB$ is shown in Figure 6, with the mean threshold presented versus $IDB$ for both unirradiated and irradiated data. A relatively weak dependence on $IDB$ is observed compared to other tuning parameters. The threshold varies only mildly across the scanned range, indicating that $IDB$ plays a secondary role in threshold tuning. The two-dimensional threshold maps for a fixed value of $IDB=80$ are shown in Figure 7, where the variation with $ITHR$ and $ICASN$ is explored. A clear gradient is observed along the $ITHR$ axis, while the dependence on $ICASN$ remains moderate, consistent with the trends discussed earlier. After irradiation, a systematic shift in the sensor threshold is observed across all $IDB$ values. The absolute change in threshold is larger at lower $ITHR$ values, as seen in Figure 7 (bottom plots), while the relative variation remains below $\sim 20\%$ over most of the parameter space. This indicates that $IDB$ does not significantly affect the radiation-induced degradation, and the sensor response remains stable with respect to this parameter.

The dependence of the sensor threshold on $ITHR$ is shown in Figure 8. A strong, nearly linear increase in the threshold with increasing $ITHR$ is observed in both unirradiated and irradiated data. This demonstrates that $ITHR$ is the primary parameter controlling the sensor’s absolute threshold. The separation between the unirradiated and irradiated curves remains approximately constant across the full range of $ITHR$, indicating a systematic shift induced by irradiation. The distributions’ spreads are comparable in both cases, suggesting that threshold uniformity is preserved after irradiation. The two-dimensional threshold maps for a fixed value of $ITHR=100$ are shown in Figure 9. The dependence on $IDB$ and $ICASN$ is relatively weak compared to the dominant effect of $ITHR$, confirming that this parameter governs the global threshold scale. The absolute threshold shift, shown in Figure 9 (bottom plots), indicates variations up to $\sim 80$–$90{\mathrm{e}}^{-}$ depending on the parameter configuration. The relative change remains below $\sim 15\%$ in most regions, with slightly larger deviations observed at low $IDB$ values. These results demonstrate that $ITHR$ provides a robust handle to tune the threshold while maintaining stability under irradiation, making it the key parameter for optimizing detector performance.

Overall, the threshold behaviour of the monolithic pixel sensor is primarily governed by $ITHR$, which controls the global threshold scale, while $ICASN$ provides additional tuning of the front-end response. In contrast, $IDB$ exhibits only a weak influence on the threshold. Irradiation introduces a systematic shift in threshold across all parameter configurations, with a more pronounced effect at lower threshold settings. However, the relative variation remains within acceptable limits over a wide range of operating conditions, indicating that stable operation can be achieved through appropriate tuning of the DAC parameters.

4.2. Noise Performance

The sensor’s noise performance is evaluated using the width of the S-curve for individual pixels, which provides a measure of the equivalent noise charge (ENC). The dependence of noise on the front-end tuning parameters $ITHR$, $IDB$, and $ICASN$ is studied for both unirradiated and irradiated conditions. The Figure 10 shows the 1-dimensional noise distributions of the sensor and 2-dimensional noise distributions over all $224\times 512$ pixels, respectively.

4.2.1. Dependence of Noise on DAC Parameters

The dependence of noise on $ICASN$ is shown in Figure 11(a). The noise remains relatively stable across the scanned range of $ICASN$, with only minor variations of the order of $\sim 1$–$2{\mathrm{e}}^{-}$. This indicates that $ICASN$ has a limited impact on the sensor’s intrinsic noise performance. A comparison between unirradiated and irradiated data shows a slight increase in noise after irradiation. However, the overall variation remains small, and no significant degradation of noise performance is observed.

The variation of noise with $IDB$ is shown in Figure 11(b). A weak dependence on $IDB$ is observed, with noise values remaining nearly constant over the full range of the parameter. This confirms that $IDB$ does not significantly affect noise performance. After irradiation, a modest increase in noise is observed; however, the absolute variation remains limited to $\sim 2$–$3{\mathrm{e}}^{-}$, indicating stable behaviour under irradiation. The dependence of noise on $ITHR$ is presented in Figure 11(c). A slight increase in noise is observed with increasing $ITHR$, although the effect remains weaker than the corresponding dependence of the threshold. The separation between unirradiated and irradiated data is small across the full range, indicating that irradiation has only a limited impact on the noise performance. The spread of the distributions remains comparable, suggesting that noise uniformity across the pixel matrix is preserved.

Overall, the noise shows only a weak dependence on the tuning parameters and exhibits a modest increase after irradiation. In contrast to threshold, which is strongly affected by parameter tuning, the noise remains stable across a wide range of configurations. This demonstrates that the sensor maintains good noise performance even after radiation exposure.

4.3. Noise vs Threshold Studies

The optimization of the second-generation monolithic pixel sensor operating point is performed by studying the correlation between threshold and noise for different configurations of the front-end tuning parameters. This allows identification of regions in the parameter space that provide an optimal balance between low threshold and acceptable noise performance.

Figure 12(a) shows the correlation between threshold and noise with respect to $ICASN$, while keeping $ITHR$ and $IDB$ constant. Each point corresponds to a different $ICASN$ setting. A clear horizontal trend is observed, with the threshold varying significantly, while the noise remains nearly constant. This indicates that $ICASN$ can be used to fine-tune the threshold without introducing a substantial noise penalty. A comparison between unirradiated and irradiated data shows a consistent shift towards lower threshold values after irradiation, while the noise increases only slightly. The overall structure of the correlation remains unchanged, demonstrating that the tuning behaviour is preserved under irradiation. The dependence on $ITHR$ is shown in Figure 12(c), where a strong correlation between threshold and noise is observed. As $ITHR$ increases, the threshold shifts upward significantly while the noise increases only mildly. This confirms that $ITHR$ is the primary parameter controlling the sensor’s operating point, enabling a wide range of threshold values to be accessed. In contrast, Figure 12(b) shows the dependence on $IDB$, where only a weak variation in both threshold and noise is observed. The data points cluster in a narrow region, indicating that $IDB$ has a limited impact on the overall performance and mainly serves as a secondary tuning parameter.

The combined results demonstrate that the optimal operating region corresponds to configurations with moderate $ITHR$ values, where the threshold is sufficiently low while maintaining stable noise performance. Extremely low-threshold configurations, although achievable, tend to be more sensitive to irradiation effects and exhibit larger relative variations. Overall, the sensor provides a flexible tuning space, with $ITHR$ defining the global threshold scale, $ICASN$ enabling fine adjustment, and $IDB$ ensuring stable operation. Preservation of the threshold–noise correlation after irradiation confirms the sensor’s robustness and suitability for operation in radiation environments. These results demonstrate that the detector can be operated in a regime that simultaneously ensures low threshold, controlled noise, and stability under irradiation, fulfilling the requirements for high-efficiency tracking applications.

4.4. Radiation-Induced Threshold Shift

To further quantify the impact of irradiation, the average threshold shift is evaluated with respect to the tuning parameters. For every parameter setting, the threshold difference between irradiated and unirradiated data is calculated by averaging over all configurations of the remaining parameters.

The dependence of the threshold shift on $ITHR$ is shown in Figure 13(a). A clear decreasing trend is observed with increasing $ITHR$, indicating that configurations with lower threshold settings are more sensitive to radiation effects. The largest shifts are observed at low $ITHR$, while higher values provide a more stable response. The variation with $ICASN$, illustrated in Figure 13(b), exhibits a behaviour that is not monotonic, reaching a maximum near intermediate $ICASN$ values. This suggests that the radiation response is influenced by the front-end shaping conditions, with certain configurations being more susceptible to irradiation-induced changes. In contrast, the dependence on $IDB$, shown in Figure 13(c), is relatively weak. Only small variations in the threshold shift are observed in the full range of $IDB$, indicating that this parameter has a limited impact on radiation sensitivity.

Overall, these results demonstrate that the radiation-induced threshold shift is primarily driven by $ITHR$, while $ICASN$ introduces secondary effects and $IDB$ plays a minimal role. This reinforces the conclusion that stable operation under irradiation can be achieved by selecting appropriate $ITHR$ values, avoiding configurations with excessively low thresholds.

5. Conclusions

The performance of the second-generation CMOS monolithic pixel sensor has been systematically studied with respect to threshold and noise behaviour, with particular emphasis on the dependence on front-end tuning DAC parameters and the impact of irradiation. The threshold distribution is found to be uniform across the pixel matrix, with a narrow spread of approximately $13{\mathrm{e}}^{-}$ at representative operating conditions. A detailed scan of the tuning parameters shows that $ITHR$ is the dominant parameter controlling the absolute threshold, with a strong, nearly linear dependence. The parameter $ICASN$ provides additional fine-tuning of the threshold, while $IDB$ has only a minor influence. Irradiation results in a systematic shift in the threshold across the parameter space. Interestingly, the threshold decreases after irradiation, while the overall threshold uniformity is preserved. A dedicated study of radiation-induced threshold shift demonstrates that configurations with lower threshold settings are more sensitive to radiation effects, with the largest variations observed at low $ITHR$. In contrast, higher $ITHR$ values provide a more stable response under irradiation. The dependence on $ICASN$ is non-monotonic, while $IDB$ has a minimal impact on radiation sensitivity.

The sensor’s noise performance is observed to be stable across all tuning parameters, with only a weak dependence on $ITHR$, $ICASN$, and $IDB$. After irradiation, a modest increase in noise is observed; however, the overall variation remains small, and the noise uniformity across the matrix is preserved. The combined analysis of threshold and noise demonstrates that the pixel sensor offers a flexible tuning space, allowing optimization of the operating point. The threshold-noise correlation remains well-defined both before and after irradiation, enabling selection of configurations that achieve low thresholds while maintaining acceptable noise levels. Overall, these results confirm the sensor’s robustness under irradiation and demonstrate that stable, efficient operation can be achieved through appropriate tuning of the front-end parameters. This makes the sensor a promising candidate for use in radiation environments requiring precise tracking performance.