Detector dark currents

The HWK1910A device was tested in a thermal chamber to characterize the temperature dependence of the detector dark current with various integration times. Several temperature sensors were set inside the chamber and on the detector to monitor the detector and environmental temperatures. During the test, the detector dark current was measured between -15°C and +25°C.

Figure 3 displays the dark current profile as an ensemble average of the entire array at different temperatures. For these measurements, the detector bias was measured with the shortest integration time and subtracted for each dark frame. The dark current at 20°C is about 2.5 e^-/s/pixel, which is a factor of 8 lower than the threshold provided by the manufacturer. Based on the linear integration time dependency and the exponential temperature dependency of the detector dark current, a model was established to predict the detector dark current for each pixel, as shown in the red curve for averaged results on the detector array. According to the fitted parameters, the detector dark current doubles for every 6.4 K temperature change.

Figure 3.

Temperature dependency of the detector dark current between –15°C and +25°C. The black line represents the measurement, and the red line corresponds to the result from model estimation.

System gain and readout noise

The detector system gain usually describes to the conversion ratio from the number of electrons into Analog-Digital Units (ADUs). This parameter can be estimated from the shot noise characteristic associated with the signal intensity, namely the mean-variance method (MVM) or the photon transfer curve^10,11. The Poisson distribution of the shot noise gives the shot noise variance σ_shot² which is equal to the averaged signal intensity S in ADUs. The total noise variance σ_noise² can therefore be expressed as:

where G is the system gain in ADUs/electron and σ_readout² represents the readout noise variance which has zero mean. Accordingly, the linear fit of the total noise variance against the averaged signal intensity S gives the conversion gain G and the offset that corresponds to the readout noise variance.

For this measurement a light source was installed inside the thermal chamber. The detector temperature was cooled down below –20°C to minimize the detector dark current. The amount of photons recorded by the detector was controlled by adjusting the integration time. From the linear regression analysis the distribution of the estimated system gain under different integration times was obtained (Figure 4). This estimation indicates that the average conversion ratio is about 2 ADUs/electron and the corresponding readout noise is about 1e^-, which are close to the detector specifications provided by the manufacturer.

Figure 4.

Estimation of system gain

Another method to characterize the detector readout noise is to determine the standard deviation of various read-outs at zero integration time (Figure 5). This method does not require a light source. For the detector analyzed in this study, the median value is approximately 0.9e^- on each pixel, which fits to the results from MVM method. The corresponding root mean square (rms) is 3.9e^-, which is slightly higher than the given information.

Figure 5.

Detector readout noise

System gain and readout noise

The detector fixed pattern noise (FPN) describes a spatial variation in pixel outputs of an image under uniform illumination conditions. For CMOS detectors that use column amplifiers configuration, the FPN appears as “stripes” on the readout images. The manual calibration mode of the detector allows the correction of this pixel gain variation on the whole detector plane, as shown in Figure 2. This operation is achieved by injecting electrons into detector pixels specifically reserved for this calibration purpose. This calibration mode will change the pixel offsets slightly in each power loop, since this procedure is also bound up with the environmental conditions.

The investigation of the pixel offset variation, which is also called the detector fixed pattern, is achieved by running several power loops. The detector integration time was set to be 1ms such that the dark current is negligible for the analysis. During each loop, frames were recorded every 5 s. As illustrated in Figure 6(a), the fixed pattern in loop 1-4 changes slightly, meanwhile the result from the 5^th loop shows relative larger deviation than the first 4 loops, which is possibly associated with the environmental temperature at that time. For the instrument radiometric calibration, the detector pixel offsets need to be measured and subtracted each time when the instrument is switched on, simply by taking images at lowest integration time.

Figure 6.

Left (a): detector mean fixed pattern at shortest integration time from 5 power loops. Right (b): detector fixed pattern at certain detector region. Loop 1-4 were measured with 3 minutes in between, while the 5^th loop was measured 0.5h after the 4^th loop.

Radiation environment modelling

The ionizing effect in the space radiation environment can cause degradation and failure of electronic devices, especially when using optoelectronic detectors. Basically the radiation effects consist of two parts: the single event effect (SEE) and the total ionizing dose (TID). The SEE induced by high energy particles will result in device logic failure as an instantaneous failure mechanism, whereas the TID will make cumulative long term ionizing damage on the detectors for low earth orbit (LEO) mission due to protons and electrons exposure. These particles and rays create shifts in threshold voltages and leakage currents, therefore causing permanent radiation damage to Silicon devices.

The increase of dark current is one of the most critical effects on the detector degradation in space radiation environments. The growth of the CMOS dark current under radiation is mainly caused by the generated interface traps and bulk defects, producing additional surface leakage current and bulk current. The ionizing energy transfer generally leads to an increase of the detector dark currents, whereas the non-ionizing mechanism results in the displacement damage dose (DDD) effect, increasing the dark signal non uniformity (DSNU)^12,13.

The commercial and industrial components generally do not provide radiation tolerance performance, therefore those devices need to be tested under expected space environmental radiation. The TID accumulated during satellite missions depends on the orbit, the mission duration and the shielding. TID effects can be reduced using radiation hardened devices and shielding.

To estimate the reliability of the COTS parts used in our experiment and to determine the required thickness of a radiation shield, the TID was calculated depending on the shielding thickness and different mission scenarios using the SPENVIS tool provided by ESA¹⁴. For this estimation, a satellite flying in a sun-synchronous orbit at an altitude of 1000 km is assumed. Mission durations are 3 years and 5 years, respectively. As a worst case scenario, models for the trapped proton and electron fluxes were the Proton model AP-8 at solar minimum and the Electron model AE-8 at solar maximum¹⁵.

Figure 7 illustrates the SPENVIS output for the accumulated radiation dose for a mission lifetime of 5 years with different Aluminum layer thicknesses. The SHIELDOSE-2 is selected for the ionizing dose model, and the shielding is considered as center of Al spheres¹⁴. The trapped electrons, as the dominate source of the TID, are largely reduced by increasing the thickness of the medium, since the electrons have a low penetration depth and are therefore easy to shield. Meanwhile, the decrease of trapped protons are not pronounced.

Figure 7.

Effective dose vs. aluminum shielding thickness for a 5 years mission from SPENVIS. A sun-synchronous orbit at an altitude of 1000km is assumed.

The TID reaching the electronics becomes inefficiently reduced for aluminum thickness greater than 4 mm. The TID for 3 year and 5 year missions with 3mm shielding are 11.4 krads and 19.4 krads, these values reduce to 2.4 krads and 4.1 krads when the semi-infinite aluminum medium model is defined for the shielding. Therefore, to guarantee a nominal lifetime of 3 years the instrument should be able to survive under 11.4 krads radiation without considering the shielding of the satellite.

Radiation test

During the radiation test, the readout electronics in the instrument were classified in modules, which were irradiated in separate steps with different total doses. Radiation tests were made to assure the functionality of the electronics and the detector for the designed mission lifetime. These tests were performed at the Fraunhofer Institute for Technological Trend Analysis (INT), where the samples were irradiated with Co-60 gamma radiation. The actual received radiation was calculated based on the distance to the radiation source. In this paper we mainly focus on the radiation effect on the detector.

Figure 8 shows a photograph taken for the experiment of the radiation test on the CMOS detector. To guarantee the full functionality of the electronics during the detector test, lead blocks were used to protect the remaining parts of the printed circuit board (PCB). The detector was irradiated up to 30 krads under uncalibrated mode, the manual calibration mode was activated from then on. The standalone test of the detector indicates that the detector component worked normally until accumulating 85.25 krads radiation, then all detector pixels became saturated.

Figure 8.

Radiation test on the detector board, the lead blocks were used to shield other electronics.

The detector was set to record images continuously during the radiation test. Dark current measurements were taken when the radiation source was switched off. Figure 9 illustrates the dark images measured at two different radiation doses. Since the manual calibration mode was switched on after 30 krads irradiation on the detector, the dark current at 30 krads is already quite remarkable. Nevertheless, the detector dark current increases with the increasing radiation dose. In contrast, the “hot pixels” were not significant since the Co-60 gamma radiation generally causes less DDD effect. Figure 10 shows two dark current histograms of the detector for the radiation test, where the long tail of the distribution after radiation is clearly shown. The dark signal after radiation shows a remarkable increase, where the corresponding rms changes from 1538 counts to 3247 counts.

Figure 9.

3-D intensity plots of pixel dark current measured at two different radiation doses for 50 x 50 pixels, detector operated under manual calibration mode with integration time 100ms. Left: radiation dose for 30 krads, right: radiation dose for 60 krads.

Figure 10.

Dark current histograms of the detector measured under manual calibration mode, an image area of 600 rows and 600 columns is chosen.