2.1

A digital data interface

There are several ways to provide for a digital interface, such as:

These solutions do solve the digital interface issue, but merely turn an analogue Sunsensor into an analogue Sunsensor with a digital interface. As a result, the sensors will still be hampered by significant measurement errors caused by albedo signals and reflections by spacecraft parts. In addition, also the cost associated with the interface electronics will in many cases supersede the cost of the bare sensor if a discrete solution is chosen.

One of the main advantages of using analogue Sunsensors is that the core sensors can be extremely robust. This is an important property as the sensors are always located on the outside of the satellite and as such directly exposed to the harsh space environment.

Depending on how well-designed the interface circuit is, the radiation tolerance can be quite high. When the interface electronics are located inside the spacecraft, the radiation tolerance will be increased by the shielding provided by spacecraft components. Integrating the interface electronics in the sensor itself, in the form of an ASIC, therefore requires careful consideration of the radiation environment and the lack of shielding by the satellite structure. If the sensor is not carefully optimised for radiation tolerance, this will lead to Single Event Upsets (SEU’s) or even Single Event Latch-ups (SEL’s), which are not only detrimental to the accuracy and availability of the sensors, but can even lead to destruction of the sensors.

In case of small sensors with low generated current levels there may however be no other option than to integrate a pre-amplifier or a more complicated form of data processing in the sensor itself. No level of processing on an analogue Sunsensor however will be able to avoid albedo errors, which brings us to the second reason to develop a true digital Sunsensor.

2.2

Albedo insensitivity with true Digital Sunsensors

Analogue Sunsensors use either a single photodiode or four-quadrant photodiodes to determine the direction of the Sun. With a single diode, Coarse Sunsensors, this is done by looking at the cosine behaviour of the generated current as a function of the incidence angle. With the four-quadrant diodes, Fine Sunsensors, the ratio of the four generated currents is used. The main issue with both types of sensors however is the fact that they integrate all incoming light into a single output current per detector element. This leads to significant albedo induced errors, especially for low Earth orbiting satellites.

The intensity of the Sunlight is much higher than the intensity of light reflected by the Earth, but this reflected light will enter the sensor over a much wider opening angle. Where the opening angle of the Sun is approximately half a degree, the Earth can be visible over as much as 120 degrees when in low Earth orbit. Integrated over the full opening angle this leads to a significant albedo current. As this current cannot be separated from the current generated by the Sunlight, analogue Sunsensors will always be affected by Earth albedo signals. Similarly, reflections by spacecraft parts (be it specular or diffuse) will also lead to albedo errors.

Contrary to Star trackers, Sunsensors have no problems with albedo signals generated by the Moon. This is because the opening angle of the moon (0.6°) is almost the same as the opening angle of the Sun, but the reflection from the moon is semi-diffuse and the average albedo of the Moon is only 0.14. Therefore, the Moon is actually relatively black and not generating signals that are of significance to Sunsensors.

Earth albedo signals can easily lead to errors in the order of 10° or even in excess of 20° for analogue Sunsensors, depending on opening angle orientation and orbit height. Even though these albedo signals only exist temporarily, they do lead to massive disturbances in the attitude control loop and significant software effort is required to mitigate the effects of these disturbances. Under rare circumstances, the use of wide FOV analogue Sunsensors has even led to a phenomenon that is called Earth lock, where the satellites attitude control system started to use the reflection from the Earth to control the orientation of the satellite rather than the position of the Sun. This can specifically happen when extensive ice cloud cover leads to a high average albedo signal.

Due to the large disturbances caused in the attitude control loops and the risk of Earth locking there is a strong preference to use true digital Sunsensors. A true digital Sunsensor uses a 2D array of sensing elements and a relatively small pinhole. As a result, the sensor divides the entire field of view in many smaller fields of view, each covering a smaller angular extend. By looking at the intensity within this smaller FOV, the sensor can discriminate between direct Sun illumination and illumination via a reflection on either a diffuse or reflective surface. Diffuse reflection will lead to a much lower signal because the light energy is spread over a much larger angular extend. Specular reflections can lead to signals in the same order of magnitude as those caused by direct Sun illumination but will only exist temporarily as the satellite is continuously moving around the Earth. Specular reflection will therefore only cause a “flash like” signal at detector level.

Splitting up the entire FOV in many smaller fields of view and by using thresholding the pixels illuminated by reflections can be simply distinguished from the sun illuminated pixels, leading to a highly reliable Sun detection and albedo insensitivity.

Even though various companies have tried to develop albedo insensitive Sunsensors, none of them have been able to develop a commercially successful product for high reliability applications. The available sensors were either too large and too expensive to be commercially successful (for instance the Galileo Avionica S3 or the JenaOptronik FSS) or not radiation tolerant enough (Newspace SS411 or Cube space Cubesense S).

Well documented attempts to make a single chip version by both Galileo Avionica and TNO have not led to a commercially available solution, and both attempts have been terminated to our knowledge.

4.1

Low material and component costs

The first criterion might be the least important one as the overall sensor costs will only be marginally influenced by the actual costs associated with the materials of the housing and standard components unless a number of active space qualified components are needed to build the sensor. As the number of sensors to be produced are expected to be limited, recoup of development costs and non-recurring engineering costs are expected to have a major influence on the overall cost of the sensors. As most of the functionality of the sensor is expected to be integrated in a single chip and the use of only passive support components is envisaged, the developments hold the promise to significantly reduce the overall costs associated with Sun sensing.

4.2

Low mass solution

Even though the actual component costs are expected to be limited, it is still expensive to launch items into space and mass budgets are still critical. Depending on the required level of radiation shielding, a significant wall thickness of the sensor will be required to meet the 15 or 18 years in GEO orbit radiation tolerance requirement.

Electric orbit raising is increasing in popularity but also has a significant drawback. During the electric orbit raising the Van Allen Belts are crossed on a regular basis, leading to a significant increase in collected radiation. As a result, a significantly higher level of shielding will be required for missions that use this orbit transfer method. The size and number of components in the cavity will determine the required cavity size. This cavity, plus the required level of shielding, will largely determine the total mass of the system.

It should be realized that radiation shielding is related to the specific mass of the applied material and that using materials providing higher levels of radiation shielding will decrease the wall thickness but will not decrease the mass due to the higher density of the material. Consequently, the only way to keep the mass low is to keep the sensors small.

4.3

Low power absorption and dissipation

In addition to the above, it should be realized that a larger sensor will lead to a larger Sun-exposed area. For ease of operation, it is strongly preferred if the use of the sensors doesn’t require the application of any multi-layer isolation material to keep the sensor from being heated by the incident Sun light. As the surface area quadratically increases with both the length and the width of the cavity, the absorbed energy quickly increases with increasing sensor size. As the units are intended to be conductively controlled to keep the temperature within operational limits, the added absorption means increased temperatures and temperature differences as well as potentially the need for increased thermal conductance of the mechanical interface.

Larger sensors can easily absorb several Watts. This should be easily recognized as a strong rational to keep the sensors as small as possible to avoid thermal issues. The majority of the absorbed heat together with the power dissipated by the electronics will have to be rejected through the mounting feet of the sensors in case the unit is to be conductively controlled. Therefore, a small surface and low power dissipation are definitively high on the priorities list for the new development.

4.4

Wide Field Of View

While looking at Figure 4, it can be seen that one of the critical measures in a highly integrated digital Sunsensor is the height of the membrane above the sensing surface. This height (together with the physical size of the detectors sensor part) will determine the sensors FOV. By reducing this height, the same sensor will exhibit a larger FOV, but there are a number of effects which put a limit on this. With al FOV larger than 75° in diagonal even full spherical FOV coverage with less sensors can be achieved. Therefore, a larger FOV is preferred, as long as the albedo sensitivity is not negatively affected. Consequently, an investigation was started into the FOV limiting factors.

4.4.2

Chip size

The digital Sunsensor is implemented as a single chip device. Preliminary designs show that the chip size is mainly determined by the size of the sensor array. Presuming that the number of pixels in the array remain constant, the size of the processing electronics will be constant, but at a constant height of the membrane, the chip will have to be significantly larger as shown in below Figure 5.

Figure 5

chip size for 60° and 70° FoV design

The 70° design has approximately twice the size of the 60° design, leading to a larger overall sensor, less chips per wafer and a lower production yield. Therefore, from a cost, size and yield point of view, the 60° design is preferred.

4.4.3

Negative effects and unwanted variations on Signal level

One of the main concerns while designing a digital Sunsensor is the expected signal level. Basically, a digital Sunsensor determines the Sun by looking at the pixels for which a certain intensity threshold is exceeded. This will automatically discriminate between direct Sun inputs and inputs created by reflections (Earth Albedo) and leads to the so much desired albedo insensitivity.

The Cosine Effect

Increasing the FOV however has a number of negative effects on the intensity variations experienced by the pixels. The first prominent intensity variation is related to the cosine of the angle of incidence. This effect is barely noticeable around Zenith, but already reduces the intensity to half of the Zenith intensity at the end of the field of view as can be seen in Figure 6.

Figure 6

cosine

Where a reliable Sun detection will require an intensity compensation of a factor of 2 for angles up to 60°, extending to 70° FOV already requires a factor of 3 compensation. Whereas the FOV in diagonal will extend significantly beyond the FOV on axis, a 75° FOV in diagonal leads to a FOV less than 70° on axis but will require an intensity compensation by already a factor 4. Such a wide intensity variation is far more difficult to compensate than the factor of 2 associated with a 60° FOV.

The cosine effect on the intensity however is not the only effect relevant for the intensity of the generated signal. The spot size variation as a result of the Sun’s opening angle is another issue of concern for very large FOV’s.

The Twilight Effect

The Sunlight is generally considered to be near parallel. However, due to the size and distance of the Sun, the light has an opening angle of approximately half a degree though. This leads to certain pixels being in what is dubbed “the twilight zone”.

In order to explain this phenomenon, one has to look at Figure 7. The light from the Sun is entering the sensor through a membrane (on the left) that has a translucent pinhole with a diameter D. This will project a spot with an exact diameter of D on the chip (as depicted on the right) if the beam would be 100% parallel. But, while the sun has an opening angle of 0.5° (ω), this will lead to a bigger and partly overlapping area of the sensor receiving light with different intensity (ΔD) This area will grow with increasing input angles, whereas the core size of the image that receives full illumination remains the same. It should be realized however that the core intensity already decreases with the cosine of the input angle as previously discussed and that the twilight pixels will have a decreased intensity the further they move from the side of the image spot.

Figure 7

twilight zone pixels

The current design of the sensor is based on a configuration where the pinhole has a nominal size of ten pixels and each pixel represents a 0.5° opening angle. Under these conditions, a fairly straight forward calculation can be made to show how many pixels are influenced by the opening angle of the incoming radiation. The result of this calculation is shown in Figure 8.

Figure 8

Twilight zone pixels

As can be seen, the number of pixels is relatively constant up to some 50° but then starts to increase tangentially. The increase in number of twilight zone pixels is such that for angles larger than 65° there are no core pixels left that are illuminated with the full beam intensity. This would further complicate a reliable detection leading to the need to increase the size of the pinhole (in turn leading to a larger pixel array) or a more complicated intensity compensation.

Polarization effects

Another phenomenon that needs to be taken into account when calculating the expected signal levels is the polarization dependent behaviour. As the light passes from vacuum into the silicon detector a significant polarization effect exists as a function of the angle of incidence and the polarisation state of the light, as shown in Figure 9.

Figure 9

polarization dependent reflection

Where the reflection at normal incidence is calculated to be 4% (determined by the difference in refraction coefficient), the reflectivity for S and P polarized light is strongly dependent on the polarization state of the light.

If we look at the signal loss due to the reflection on the vacuum to silica interface it becomes obvious that this has a major impact for angles larger than 60 degrees. The graph in Figure 10 not only shows the reflection losses as a function of input angle, but also tabulates the extra signal loss over nominal incidence signal loss.

Figure 10

calculated extra polarization loss

The 4% nominal signal loss can be contributed for in the calculations and even the 9% at the 60° input angle could be accounted for (leading to a ±5% intensity variation). The 25% signal loss for input angles of 75° however are deemed problematic, especially because the exact configuration of the lower layers is not known and the reflection on the lower SiO2 to Si interface will also show the polarization dependent reflection, thus reflecting even more of the S polarized light. This transition will lead to another 15.5% reflection and signal loss at nominal incidence, but this signal loss is as before taken into account during the definition of the quantum efficiency. Nevertheless, it is strongly advised to try and stay below the Brewster angle of the vacuum to SiO2 interface as much as possible (which is 56°) as above this angle the reflection will strongly increase.

Track Shadowing

Despite the fact that the tracks of the integrated circuit are located very close to the actual junction, for small pixels (as used in the digital Sunsensor), the shadowing effect can play an important role if not properly considered. The basis of this phenomenon is shown in Figure 11. This figure shows the metallization stack as defined by the supplier of the CMOS chips.

Figure 11

track shadowing

The separating material between the tracks (forming the optical window) is Silicon dioxide. The thickness of these separating layers is given in Figure 11, but are close to 1 μm for each of the layers. This means that using a four-layer metal stack, the total stack thickness (excluding the covering layer) can be calculated to be around 6 μm). Due to the higher refractive index of the silica, the optical angle will be reduced from 60° to some 37°, but this still means that the spot will hit the actual silicon not closer than 3μm from the edge of the upper metal layer. As the current design anticipates to use 10 μm pixels it is obvious that this will have a detrimental effect on the apparent quantum efficiency and a significant amount of signal will be lost unless specific design measures are taken. As the effect grows with the tangent of the angle, it will worsen drastically with increasing input angles.

This is the final reason identified why it seems ill advised to try and make a digital Sunsensor with a FOV larger than 60° in diagonal.

Adding things up

There is no ease, in adding ten degrees

The main reason to increase the Field Of View would be to allow covering a full spherical field of view with 8 instead of 10 sensors while providing single sensor failure redundancy. Covering the full spherical FOV with six sensors requires a FOV of 60° in diagonal. This can be calculated to be equivalent to 54° on axis. Such a sensor can also be used to provide single failure redundant coverage with 10 sensors.

As a smaller FOV significantly reduces signal variations and implementation difficulties, it is proposed to reduce the FOV to slightly over 60° in diagonal so as to have some overlap but refrain from increasing the field of view.