Mean dark current increase

The mean dark current is the number of electrons generated per second when the detector is in darkness, and averaged over a large number of pixels.

The mean dark current increase is due to both the ionizing dose and the non-ionizing dose. For the considered missions and shielding configurations, ionizing dose is mainly generated by the protons interaction with the detector. Secondary emission of electrons by surrounding materials can also contribute to this effect.

In order to compare degradations observed on various detectors, we need to use a common scale. We decide to express the mean dark current increase as a function of the ionizing dose received, even if non-ionizing dose has an impact as well. Furthermore, we express the dark current as a function of the pixel active surface. To calculate the dark current in pA/cm2, the total surface of storage areas has been taken into account, including MOS storage capacity oxides as well as LOCOS oxides.

Moreover the mean dark current is expressed for a temperature of 20°C, using well known rule for Silicon (on satellites, detector temperature is regulated at a constant value).

The mean dark current increase will be then expressed in pA / cm² / kRad(Si) @ 20°C.

Figure 5 shows mean dark current increase for different satellites:

Fig 5:

Cumulative mean dark current measured in orbit, vs ionizing dose

We clearly see that for all detectors (on different satellites) mean dark signal increases quite linearly with the Total Ionizing Dose. Nevertheless the slope is very different from one detector to the other, reflecting the fact that mean dark current increase depends on the detector technology, especially pixel design.

Worst slope, for SPOT 5 PAN detector (linear array of photodiodes) is 2,3 nA/cm2/kRad(Si) @ 20°C

Another example could be given for Corot satellite, with astronomical purposes (astero-sismology and search for exoplanets). The CCD is backthinned, in MPP mode, and at low temperature, giving a very low dark signal. Indeed, in an inverted mode the Si/SiO2 interface can be flooded with holes during integration. As a consequence, the surface dark current is suppressed and the bulk dark current will dominate the total dark signal.

The pixel pitch is 13.5μm and the integration time was set to 32 s (exoplanet channel). The ionizing dose deposited per year is about 700 Rad.

Corot CCD dark signal increase is given Figure 6, which exhibit real measurements, same points recomputed for -35°C, and predicted lines [13]

Fig 6:.

Corot CCD Dark signal increase

Effects of mean dark signal increase

Dark signal increase has several consequences for the detector, and thus the mission, performances:

The impact depends on the type of missions

For Earth Observation, mean dark signal is quite low and does not degrade the overall performances. Figure 6b is shows the SNR (Signal to noise ratio) over SPOT 5 life time: no degradation is present

Fig 6

bis: Evolution of SPOT 5SNR over time

For astronomy mission, the main impact is the noise increase leading to a degradation of the capability of detecting low signal variations.

The noise increase will lead to a certain degradation of the capability of detecting faint signals. This is illustrated by the ratio between real noise and photon noise. Figure 7 gives the drift of this ratio over time

Fig 7:

Evolution of Corot detection ratio (Credit P. Bordé)

This figure exhibits a slight drift of this ratio but very low (0.12 per year), with very few impact on overall mission performance.

Same impact is present for star trackers, reducing the number of useful stars

Dark signal non uniformity (DSNU)

The DSNU is the non uniformity of the dark signal over the detector: it can be expressed in % (related to he mean dark signal) or in physical units. It can be given as peak-to-peak or RMS value, but in certain cases, we give the number of “bright” (or “hot”) pixels (with a dark signal higher than a certain limit)

DSNU increase is meanly due to displacement damage in the silicon created by energetic particles. These defects act as generation centres lying within the depleted region, and therefore raise dark signal in the volume [4-6]. Each displacement damage generation centre is localized and implies a sharp dark signal increase in the considered pixel. As a consequence of the statistical nature of displacement damage, the main effect of the non-ionizing energy transfer is an increase in DSNU, and especially the apparition of a tail in the dark current distribution.

Examples For Earth Observation satellites [12]

Figure 8 gives an histogram of the dark signal over all pixels of the detector of PARASOL Satellite after 800 days in orbit (PARASOL is a CNES microsatellite that observes aerosol distribution in the atmosphere) the corresponding Displacement damage: equivalent fluency (30MeV protons) is 3.10⁹ proton/cm²: We clearly see the tail in the distribution

Fig 8:

Dark current histograms of one of the detectors just after launch and after 800 days in orbit

One interesting parameters is the amplitude of dark current jumps. We call “jump” a sudden (but permanent) increase in a pixel dark signal after proton interaction. Figure 9 shows the dark current of 1 pixel over time

Fig 9:

PARASOL pixel jump observed in orbit

In order to compare the different detectors behaviours, jump amplitudes are expressed in electrons generated per millisecond at a temperature of 20°C

Figure 10 shows the histogram of the jump values for several detectors Spot 5 payload, Parasol payload, Parasol star tracker (in orbit duration: 3 years for Parasol, 5 years for Spot)

Fig 10:

histogram of jump values

It clearly appears that the values are quite similar, leading to the question: why this similarity?

First element to explain that is that a jump can present an annealing [11], so it can be either transient (with rapid and strong annealing),or semi-permanent (with slow or delayed annealing), or permanent (without any annealing).

Figure 11 shows both jump values and DSNU for Parasol satellite. It is obvious that after event 2, DSNU decreases quickly to its former value, showing that the jump does not exist anymore: it is clearly as sign of cure of the pixel

Fig 11:

DSNU and jump vs time

Another way to analyse this phenomenon is given by Corot data. Figure 12 shows the number of “bright” pixels vs time:

Fig 12:

number of bright pixels vs time for Corot

it clearly shows a saturation of the number of bright pixels : this is a sign that a kind of equilibrium exists between the degradation rate, and the cure rate, explaining why the long term distribution of bright pixels is the same for different satellites (cf fig 9).

This theory has nevertheless to be confirmed by further in orbit satellites.

Impact of DSNU evolution on mission performances

for Earth Observation satellites, DSNU leads to increase the noise of the most impacted pixels: searching to limit this effect will lead either to reduce focal plan temperature, or temperature drift, during one image sequence) or to increase the rate of darkness calibration : the first one will induce constraints on satellite design (area of dissipative surfaces) or on operational constraints (rate of darkness calibration).

For Astronomy satellites, DSNU increase will reduce the number of pixels (and thus the area) really useful: if too much pixels are impacted, so the useful area may become too small.

RTS noise

RTS noise is due to bi or multi stable state of darkness current of a pixel (after displacement damage).

RTS noise is clearly present in orbit Figure 13 shows a real signal acquired on Parasol satellite detector.

Fig 13:

RTS signal on Parasol detector

Obviously, all missions are impacted by such phenomenon, which is random, and so quite impossible to calibrate.

For pushbroom Earth Observation satellites, this leads to “dotted” lines in images (cf fig 14).

Fig 14:

RTS effects on images (Credit CNES - Distribution Spot image)

This kind of phenomenon is clearly to be reduced.

Charge transfer inefficiency

Displacement damage could lead to an increase in charge transfer inefficiency (CTI).

This kind of phenomenon is clearly visible on images from Corot satellites (fig 15 and 16):

Fig 15:

Image showing Vertical transfer inefficiency

Fig 16:

Image showing CCD register inefficiency

Impact of CTI degradation on performance mission

For Corot satellites, which purpose is to follow the radiance of faint stars, images are de-focused: so even CTI is present, it has no consequence on mission performance, because it is possible to gather all pixels by ground software processing.

For Star Trackers, the question is more tricky, because process is in flight and in real time : Even if star images are de-focused, if CTI becomes too large, the level of signal measured on a star image will decrease leading to a more difficult detection, and even a false detection if the star apparent magnitude changes.

This phenomenon is particularly awkward for satellites at medium altitude, where proton environment is hard: that is the case for JASON satellite (1355 km)

For Earth Observation satellites, CTI degradation could lead to a smoothing of sharp black to white edge, which is characterized by the MTF (Modulation Transfer Frequency) degradation.

Up to known we have not noticed any significant degradation for in orbit MTF: Figure 17 shows SPOT 5 MTF evolution during 9 years in orbit:

Fig 17:

SPOT 5 MTF evolution