2.1

Sample description

HAS2 is a radiation-tolerant CMOS monochrome image sensor with 1024x1024 pixels array. The pixel is a photodiodetype with the size of 18 μm x 18 μm and 3T readout circuit. It supports rolling shutter operation with 12-bit ADC and has been described in the following references [7-10]. For the JUICE application it was decided to deposit a color filter array (CFA) on top of the photodiode array. To increase the information content of images four colors were chosen: red, green, blue, and yellow. The filters were deposited on the top of the wafers, which were then cut into single image sensors and assembled in a custom 84-pin ceramic package (see Figure 1).

Figure 1.

Color filer array (left) and assembled image sensor inside the ceramic package (right).

The image sensor has four 3.3 V power rails, namely: VDD, VPIX, VRES (reset), and VAA.

2.2

Camera elegant breadboard

The test camera set-up is an elegant breadboard developed by ALTER TECHNOLOGY and consists of necessary electronics to communicate with the image sensor, e.g. FPGA, memory, UDP, and USB to access the registers via SPI protocol. The sensor is placed in a remote socket, with necessary passive elements for decoupling (see Figure 2), making the breadboard flexible for different testing set-ups such as radiation, thermal vacuum, or thermal cycling analysis. The image sensor was powered using a high precision power supply from Keithley, 2230G-30-1, monitoring currents and voltages by means of a Data Acquisition Unit Keysight 34970A.

Figure 2.

Camera elegant breadboard (left) and remote board with the image sensor (right).

2.3

Optical setup and optical measurements

The optical setup for image sensor characterization was developed at ALTER according to the EMVA Standard 1288 - Standard for Characterization of Image Sensor and Cameras, Release 3.0, November 29, 2010. One of the provisions of the EMVA standard is that each pixel color is characterized separately. Therefore, the setup consisted of a light source with four colors (LEDs) integrating sphere (Newport 819-SF-6), power meter (Newport 2936-C) with photodetector (Newport 883-SL), and spectrometer (Thorlabs CCS200). The linearity of the light source is better than 0.995 measured from the center of the image sensor to the edges. The setup and its realization are shown in the Figure 3. The incident light properties were measured by means of the power meter and the spectrometer and reads: red – 630 nm – 1.466 μW, yellow – 590 nm – 1.688 μW, green – 535 nm – 2.454 μW, and blue – 465 nm – 4.561 μW. The full width at half maximum (FWHM) < 50 nm for each color. The exposure times used in the experiments for photon transfer method (PTM) were set up to 14.5 ms where the saturation of the flat images occurred at all colors. For the dark current (DC) measurements longer exposure times were used up to 120 ms. For spatial nonuniformities, namely photo response nonuniformity (PRNU) and dark signal nonuniformity (DSNU) the images were taken at 50% of the saturation capacity.

Figure 3.

Schematic and pictures of the setup for optical measurements.

2.4

Radiation setups

The TID test was performed at CIEMAT’s radiation facility in Madrid, Spain. It is the NAYADE installation of the water pool type and ⁶⁰Co sources placed at the bottom of the pool. The samples were placed inside hermetic vessels. The internal cavity protects the samples and permits the uniformity of the dose rates. Four samples: 2 biased and 2 unbiased were fixed to the metallic plate inside the vessel and the connections were taken outside by hermetically sealed wire tubes (about 10 m in length). The setup is shown in the Figure 4. The dose rate was about 235 rad/h resulting in a 5 krad daily dose, considering 2 hours break for the imagers characterization. The intermediate characterization was taken at 10 krad, 20 krad, 36 krad, and the final one after 57 krad. After the 48-h annealing at room temperature (RT), one week (168 h) annealing was performed at 85°C.

Figure 4.

TID setup: (left) samples on the holder and hermetic vessel, (middle) the vessel inside the pool; SEE setup (right) camera with the image sensor mounted on a holder inside the vacuum chamber.

According to the requirements of the JUICE mission the Single Event Effects (SEE) should be tested in the LET range from 1 to 100 MeV·cm²/mg. The maximum LET for heavy ions (HI) available in Europe (at UCL, Belgium) is 62.5 MeV·cm²/mg. Therefore, the tilting of the sample was necessary. The beam had about 2.5 cm of diameter and was entering the vacuum chamber where the DUT was placed. Inside the vacuum chamber were placed the PCB with the image sensor and its motherboard. The sensor was connected using 3 supply lines, VDD, VAA, and VPIX (which was connected to the VRES). The motherboard was accessed through Ethernet and a modem, the later controlled by a PC for image transfer. The motherboard was also connected via USB to the PC providing power and SPI communication. The power supply and monitoring block was used to drive the image sensor and monitor the currents and voltages during the SEE tests in order to detect the SEL. The picture of the setup is shown in the Figure 4. The board with DUT was fixed inside the vacuum chamber to the moving stage, which allows for positioning inside the beam and rotation to the desired angle to increase the LET. The flux of the HI beam can be set from tens to 10,000 ions/(cm²·s).

4.1

Experimental results

SEE test was performed on 2 samples: #5 and #6. The window was removed to allow direct access to the photodiode array of the image sensor by the heavy ions. The properties of the applied beam are shown in the Table 2.

Table 2.

Properties of the HI beam and number of events.

First, two steps were performed with the flux on device of 10,000 ions/(cm²·s) and then 1000 ions/(cm²·s). It was observed that under the higher flux the sensor had some problems with responsivity. All images were taken in a vacuum chamber under total darkness and they were acquired continuously with 5 ms exposure time. All the images were stored for further analysis (SEU events) and all the interruptions due to SEL or SEFI were counted during the test.

The SEL was monitored by means of fast external sources with reaction time in the range of 40 μs. For these tests it has been considered that a current increase of more than 25% of the nominal values would present latch-up event. The voltage limits were set to 3.45 V. There was one event in one sensor at the LET level of 62.5 MeV·cm²/mg (the 5^th step). Since the LET_th > 60 MeV·cm²/mg which is the limit for the mission, no more analysis was performed for SEL. The SEU is usually monitored in the registers, sending a pattern of registers, and monitoring them continuously until a bit flip occurs. Counting an event and repeating the task again until the desired fluence is reached. The registers of the investigated sensor are only accessible in write mode and no readout is possible (only four registers). Therefore, any change/corruption of the image meant that there was change in one of the four registers. Such changes were accounted as an SEU (see Figure 6 with example of corrupted images).

Figure 6.

Example of the SEU reflected in the images.

A SEFI is any event which makes image sensor unresponsive, but different from the SEL. It could be caused by SEU in the control register. Considering the comment about the writing mode of the registers, a SEFI was registered when the error caused in the registers (first seen as SEU) was repeated in more than one image, namely the sensor did not recover and was sending corrupted images again. In this case a restart of the camera was necessary. Based on the analysis of the images the numbers of corrupted images and the required restarts for each step are shown in the Table 2. Furthermore, a full characterization was performed before and after the test and the comparison is shown in the Table 3.

Table 3.

Comparison of PTM parameters after SEE test.

The initial values of the responsivity are even higher than 12% in respect to the ones for the TID samples. It is assumed that this must be attributed to the window characteristics allowing more photons to reach the surface of the image sensor. After the SEE test all parameters show some degradation irrespective of the pixel color. The most affected parameter is the saturation capacity and the quantum efficiency which decreased within a 14% margin.

4.2

Calculation of the SEU and SEFI rates

To calculate the SEE rates, the models implemented in the CRÈME96 software were used [12, 13, 14]. In principle, there are three main steps in the calculation of the SEE rates: (i) Measure the cross section σ using accelerator testing, where the device cross section is defined as the ratio of the number of upsets to the particle fluence. (ii) Determine the sensitive volume (SV) – usually SV is smaller than the actual device physical volume, and it is the most difficult parameter to estimate. One possible way is to use the Weibull fit to estimate the bit size. (iii) Finally, to determine the device error rate, the cross section and sensitive device volume must be integrated with the LET spectrum. The third step can be calculated by means of CRÈME96 including the calculation of the desired LET spectrum. CRÈME96 calculates the SEE rate due to direct ionization using the Rectangular Parallelepiped (RPP) model.

The Weibull function is widely used to fit direct ionization (heavy-ion) SEE cross-section data, since it provides great flexibility in fitting the “turn-on” in the cross-section and naturally levels to a plateau or limiting value. The functional form of Weibull fit reads:

where x is effective LET in MeV·cm²/mg, σ(x) is the SEE cross-section in μm²/bit; σ₀ is the limiting or plateau crosssection in μm²/bit, x₀ is the onset parameter (LET threshold) in MeV·cm²/mg, W is the width parameter in MeV·cm²/mg, and S is a dimensionless exponent. Using the number of corrupted images and the effective fluence, the cross sections were calculated for each event. After Weibull fit it was found that LET threshold is higher than 15 MeV·cm²/mg for both, SEU and SEFI. The plateau cross-section for SEFI is of about 4·10^-6 cm²/bit, whereas for SEU reads 6·10^-5 cm²/bit, which is one order of magnitude less than that reported by Beaumel [10] for the same image sensor.

According to the JUICE requirements for the SEE rates estimation following data were used:

First, the flux for GEO orbit and ions with Z numbers from 1 to 92 was calculated for the three environments: (i) Solar Minimum (GCR Maximum), (ii) Worst Week, and (iii) Peak 5 Minutes of the Solar Event. Then the transport of the input flux through the thickness of the shielding of Al with 1 g/cm² was obtained. Upon finalizing the previous task, the LET spectra for the three environments were calculated. Finally, using the LET spectra, the output of the Weibull fits of the cross section vs. LET, and assuming the depth of the sensitive volume as 6 μm, the SEU rates were obtained for GEO orbit for 1 bit. To calculate the rates for the entire device it was necessary to multiply 1-bit result by number of bits in the device. The HAS2 image sensor has 2 shift registers (Y and X) with 10 bits each, one control register with 10 bits, and 12-bit ADC resulting in 42 bits which can be flipped causing SEU. In principle, for the SEFI rates, only the control register plays significant role, but taking the worst case, also 42 bits were considered. The calculated rates for three environments are shown in the Table 4.

Table 4.

SEU and SEFI rates calculated by means of CREME96.

Obtained rates confirm the resistance of this imager to the SEE. Even in the worst conditions, namely peak 5 minutes, the probability of the SEU is less than one event per day. It is necessary to note that in space the flux of high energetic particles (i.e. heavy ions) is of about 4 particles/(cm²·s) and all the tests in the labs need to be considered as accelerated tests. The obtained rates are very small, leading to the probability of less than 1 % that the SEFI event occurs during the 10-year mission.