Preliminary Design of a Micro-Satellite for Earth

Observation
Sara Morales, Marc Alomar
22nd of February, 2015

1
Contents
1 Orbitography and Satellite Visibility 3
1.1 Visibility analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Coverage analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Accuracy of the orbit calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Consequences on the ground antenna pointing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Radio Communication 6
2.1 Link budget - Downlink . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Analysis with SATORB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Effects of antenna depointing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.4 Reception capacity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.5 Uplink sizing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3 Thermal Subsystem 11
3.1 Thermal fluxes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Thermal analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

4 Energy Subsystem 14
4.1 System specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.2 Dimensioning the solar panels and battery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.3 Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

5 Attitude Control Subsystem 20

5.1 Performance requirements for the ACS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
5.2 Environmental disturbing torques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5.3 Assessment for different possible architectures for the ACS . . . . . . . . . . . . . . . . . . . . . . . . 23
5.4 Dimensioning of the actuators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

2
1 Orbitography and Satellite Visibility
We begin the preliminary design of the satellite by performing an orbitography analysis. The study is done for 4
different altitudes (800 km, 1000 km, 1200 km and 1400 km), with a sun-synchronous orbit. Let’s recall that this
orbit maintains always the same angle with respect to the Sun, precessing about 1º per day. This orbit was chosen
because it provides constant illumination during the year, which is very convenient for taking pictures of the Earth.
The satellite has access to three different ground stations, and the minimum elevation angle needed is 5º. These
stations are located in Cayenne, Toulouse and La Réunion. All these parameters are set in SATORB in order to
perform the visibility and coverage analyses. SATORB generates visibility reports, that is, it propagates the orbit
and indicates when the ground stations are visible.

1.1 Visibility analysis

In this section we study the visibility periods for each of the ground stations. This is a very important parameter
because it determines the maximum number of pictures that can be downlinked per day.
Table 1 lists the mean daily visibility of each station at different altitudes. The mean value is calculated over a
period of 5 consecutive days. As we can see, the visibility period increases with the altitude. Among the different
ground stations, the one in Toulouse has the higher visibility period, having about a 30% higher visibility than the
other ground stations. We have to note that these calculations were done with a Keplerian propagator.

1.2 Coverage analysis

After a few days, the footprint of the satellite will have covered all the Earth. However, it won’t cover all the
areas with the same frequency: the satellite will overfly the polar latitudes more often than the equator, due to the
curvature of the Earth. With regards to the altitude, if we set a constant minimum elevation angle, the footprint
increases with the altitude. That means that we expect a higher coverage when the satellite is at higher altitudes.
In order to study the coverage we select a zone of the Earth that is 30º large in longitude from the equator
to the pole. The simulation was performed for a period of 3 days. Figure1 shows the coverage percentage for the
4 different altitudes. Regarding to these figures, we can see that the coverage increases with the latitude and the
altitude.
This coverage analysis explains why the Toulouse ground station has the highest visibility. Toulouse is at a
much higher latitude than the other ground stations, which explains the higher visibility periods (see Figure 1,
which shows the coverage over the three ground stations).
To confirm that the visibility periods calculated before are consistent with the coverage analysis, we calculate
the coverage percentage from the mean daily visibility. Table 2 compares the results obtained from the mean daily
visibility to those read directly from Figure 2. We can conclude that the visibility analysis is consistent with the
results from the coverage analysis.

1.3 Accuracy of the orbit calculation

So far, we have used a Keplerian propagator to perform all the calculations. This propagator considers only the
gravitational attraction from a spherical Earth, and neglects all the perturbations that come from the gravitational
attraction of other celestial bodies and the oblateness of the Earth. To quantify the magnitude of the error, we
perform the simulation at 800 km using the SGP4 (NORAD) propagator. The Simplified General Perturbations

Mean daily visibility duration (in minutes)

Altitude (km) Cayenne (4.9º N) La Reunion (21.1º S) Toulouse (43.6º N)
800 36.35 37.9 53.01
1000 46.27 49.91 67.41
1200 57.78 63.3 84.49
1400 66.39 70.84 99.77

Table 1: Mean daily visibility.

3
 Figure 1: Coverage percentage.

Figure 2: Coverage percentage including the 3 stations in Cayenne, La Réunion and Toulouse.

4
 Cayenne (4.9º N) La Réunion (21.1º S) Toulouse (43.6º N)
Altitude (km) Data (%) Graph (%) Data (%) Graph (%) Data (%) Graph (%)
800 2.5 2.5 2.6 2.5 3.7 3.8
1000 3.2 3.1 3.4 3.6 4.7 4.4
1200 4.0 4.3 4.4 4.7 5.9 5.9
1400 4.6 4.7 4.9 5 6.9 7

Table 2: Coverage analysis.

Figure 3: Error on the x axis for the 800 km orbit.

(SGP4) propagator considers secular and periodic variations due to Earth oblateness, solar and lunar gravitational
effects, gravitational resonance effects, and orbital decay using a drag model. Therefore, the simulation is much
more accurate than using the Keplerian propagator.
Figure 3 shows the error on the x axis between both propagators. The error follows a sinusoidal pattern, with
a maximum error of 51 km after one orbit. However, the error diverges rapidly: after one day, the error increases
up to 400 km, which is half the altitude of the satellite.

1.4 Consequences on the ground antenna pointing

Let’s consider that the ground station in Cayenne has a temporal tracking error of 4 seconds. To calculate the
tracking error in terms of depointing angle, we use the following relation,
q
2
θerror = 4E 2 + (4A cos E)

where ∆E is the error on the elevation angle and ∆A the error on the azimuth. Using this equation we have found
a maximum error of 2.05 º. In the next section we study in more detail the consequences of a pointing error.

5
2 Radio Communication
In this section we analyze the radio communication link between the satellite and the ground station. This analysis
is very important to size the communications subsystem: the characteristics of the ground stations and the specifi-
cations of the digital link determine the needed power emission on the satellite segment. In the following, we study
the downlink and uplink budget, and analyze the consequences of depointing errors.

2.1 Link budget - Downlink

The specifications of the downlink are given in table 3. In order to determine the needed power emission from
the satellite, we have to start with the link specifications. The maximum bit error rate, 10−6 , and the modulation
scheme, BPSK, determine the minimum energy per bit. In this case, the required value is Eb /N = 10.5 dB. This is
a theoretical value, but we have to consider a minimum margin of 5 dB and a demodulator degradation of 1.5 dB,
which leads to Eb /N = 17 dB. To obtain the required carrier to noise ratio at the ground station, we multiply the
previous result by the data rate, R,
C Eb
= + 10 log R = 71 dB Hz
N0 N0
This is the required carrier to noise ratio to ensure a correct decoding of the radio link, with a margin of 5
dB. Using the link budget equation, we can deduce the required power emission on the satellite. The link budget
equation combines the characteristics of the satellite transmitter (EIRP), the orbit (path losses), and the ground
receiving antenna (figure of merit, i.e. G/T ),
   
C 1 G 1
= (PT GT )SL · · ·
N0 L T GS κB

The figure of merit, G/T , characterizes the receiving antenna. The maximum gain of the antenna can be
2
calculated from the values given in table 3, GR,max = η (πD/λ) = 29.6 dB. Taking this value and the system noise
temperature, we obtain the figure of merit,
G
= G − 10 log T = 6.6 dB
T
The path loss accounts for the attenuation between the satellite and the ground station. This term has many
contributions, the main one being the free space loss. To calculate the free space loss, we have to determine the
maximum length between a ground station and the satellite. If we consider that the minimum elevation angle is
E = 5º, the maximum distance is (see fig. 4)
sin α
R = RE = 2783 km
sin θ
To find this value, we need the angle θ, which is the maximum depointing between the satellite and the ground
station (figure 4). We can find this value using some simple geometrical arguments. Taking r = RE + h, with
h = 800 km,

Link Satellite Ground Station

Frequency: 2.270 GHz Antenna providing a 3 dB coverage Antenna diameter: 1.80 m
Bit Error Rate: 10−6 for a min. elevation of 5º Antenna efficiency: 0.5
Modulation: BPSK (No encoding) Max. depointing: 5º Noise temperature: 200 K
Rate: 250 kbps (25% for TM) Antenna efficiency: 0.6
Demodulator degradation: 1.5 dB Internal losses: 1 dB
Atmospheric and rain losses: 0.2 dB Power amplifier ratio: 20%
Margin: 5 dB

Table 3: Downlink specifications.

6
 θ = sin−1 (RE cos (E) /r) = 62.3º

One of the requirements says that the ground stations on the 5º contour have a -3 dB gain. This requirement
determines the beamwidth of the satellite, θ3dB = 2 (θ + ∆θ) = 134.5º, where ∆θ = 5º is a margin for depointing
error.
The free space loss is calculated in the worst case
scenario, i.e. ground station at the - 3 dB contour.
The maximum length, R, is calculated using figure 4, Satellite
R = RE sin α/ sin θ = 2783 km. This leads to a free
space loss of
2 θ
R

4πR
Lf s = = 168.4 dB
λ
r
The total path loss is the sum of the free space loss E GS
and the atmospheric and rain losses, Lp = Lf s +Latm =
168.6 dB. The EIRP of the satellite transmitter can be
now calculated,
  α RE
C T
EIRPSL = Lp κB = 4.43 dBW
N0 G GS

The EIRP is the product of the satellite antenna Figure 4: Geometry relations to derive the maximum dis-
gain and the emission power, which is the value we are tance between the satellite and the ground stations.
looking for. The gain of the satellite antenna is deter-
mined by the θ3dB beamwidth,
 
70π
Gmax = η = 2.05 dB
θ3dB

In the contour zone, the gain of the antenna is G3dB = Gmax − 3 dB = −0.95 dB. Finally, the power transmitted
by the antenna shall be

PSL = LF T x EIRP/G = 6.4 dBW

where we have considered the line feeder losses, LF T x . During a downlink, the power that the satellite has to supply
to the communication subsystem is

PCOM = PSL /η = 21.7 W

Table 4 summarizes the power emission requirements for different altitudes. The needed power emission increases
with altitude, because of the higher path loss. At the same time, the gain of the antenna increases with altitude, in
order to reduce the beamwidth. This is a consequence of the 3 dB coverage requirement for a minimum elevation
of 5º.

2.2 Analysis with SATORB

We have analyzed the link budget between the satellite and the ground station of Cayenne during one of the orbital
passes. The objective of the analysis was to verify that the link budget, given as Eb /N0 , was above the worst case
figure calculated before. During one pass, the link budget values were comprised between 15.8 dB and 28.6 dB,
which is above the worst case value of 17 dB.

7
 PT x [dBW] PCOM [W] θ3dB [º] GT x, max [dB]
800 km 6.4 21.7 134.5 2.05
1000 km 7.2 26.2 128.9 2.42
1200 km 7.8 30.3 123.9 2.76
1400 km 8.3 34.1 119.5 3.08

Table 4: Summary of the communication subsystem requirements for different altitudes.

Figure 5: Link margin with depointing.

2.3 Effects of antenna depointing

Up to this point, we have assumed that the ground station doesn’t have any depointing errors. In practice, this is
not the case: the ground station uses an orbit propagation algorithm to know the trajectory of the satellite, and
over the time the estimation will have some errors, which lead to a depointing error. This effect is particularly
negative for high gain antennas, which have a narrow beamwidth.
In this section we study the consequences of a depointing error on the ground station. We consider a tracking
error of 4 seconds, and study the consequences on the link budget. When there is a depointing error, the gain of
the ground station is reduced by
 2
∆θ
LGS = 12
θ3dB GS

where ∆θ is the depointing error. As a consequence, the link margin is reduced. Figure 5 shows the link margin1
in the case of depointing, compared to a perfect pointing scenario. The calculations for 4s are obtained from the
first section analysis; bigger depointing errors are linearly propagated from the 4s results. As we can see, the link
margin is greatly reduced when the satellite is at the zenith. When the tracking delay is of 10 s, the link margin
turns negative at zenith. Therefore, the maximum temporal error allowed to have a positive margin is 10 seconds.

2.4 Reception capacity

In this section we estimate the number of pictures that can be downlinked during one day. Recall that the transmitter
only operates during sunlight periods, and that there are three available ground stations. We have analyzed the
visibility passes on the 8/01/15, and we have found a total of 45 minutes downlink in sunlight. Each image has a
size of 2000 x 2000 pixels, and each pixel needs 12 bits. This gives a total size of 48 Mbits per image. If we consider
that the data rate is of 250 kbps and that only a 75% is used for the payload, the information rate is of 187.5 kbps.
The total number of pictures per day is
1 To calculate the link margin, we compare the actual value of E /N to 17 dB, which accounts for 10.5 dB of theoretical minimum
b 0
value, 1.5 dB of demodulator losses, and a minimum margin of 5 dB to establish the connection.

8
 Link Satellite Ground Station
Frequency: 2.092 GHz Patch antenna Antenna diameter: 1.80 m
Bit Error Rate: 10−6 Gain: 3 dB Antenna efficiency: 0.5
Modulation: FSK (no encoding) Noise temperature: 400 K
Rate: 10 kbps Losses: 2 dB
Atmospheric and rain losses: 0.2 dB
Margin: 5 dB

Table 5: Uplink specifications.

Ri Tsun
N. P ictures = = 9 @ 800 km
Image size
The table below summarizes the number of pictures collected as a function of the altitude. As we can see, the
higher the altitude, the larger the number of pictures downloaded per day. This is because the visibility period
increases with the altitude, as we could see in the first section. Notice that this calculation was made for one single
day in the spring equinox, when the period of sunlight is approximately the same as the night period. These values
will vary significantly during summer and winter, because of the latitude of the ground station in Toulouse. During
summer, the visibility periods will be much higher than during winter.

Pictures / day Tsunlight [min]

800 km 9 39.9
1000 km 18 79.0
1200 km 26 112.1
1400 km 33 142.7

2.5 Uplink sizing

Up to now, we have only considered the downlink between the satellite and the ground stations. Even if it is the
most important with regards to the scientific mission, it is crucial to have a good uplink between the ground stations
and the satellite, to be able to command the satellite. The specifications for the uplink are given on table 5. Notice
that the data rate is now much lower, 10 kbps, compared to 250 kbps for the downlink.

To calculate the power of the ground station transmitter, we follow the same procedure used for the downlink.
In this case, the modulation scheme is FSK, which leads to a minimum energy per bit ratio of 14.5 dB. If we account
for a 5 dB margin, we obtain
C Eb
= + 10 log R = 59.5 dB Hz
N0 N0
The free space loss is slightly smaller, because of the longer wavelength, λ = 14.3 cm. In this case, the free space
loss is Lf s = 167.7 dB, and the total path loss, accounting for the atmospheric losses, is Lp = 167.9 dB. The figure
of merit of the satellite antenna is
G
= 3 dB − 3 dB − 2 dB − 10 log 400 = −28 dB
T
where we consider the satellite losses of 2 dB and the worst case of a ground station at the 3 dB zone. Notice that
this value is much lower than for the ground station, because of the dimensions of the antenna. The EIRP of the
ground station shall be
 
C T
EIRPSL = Lp κB = 26.9 dBW
N0 G GS

9
 800 km 1000 km 1200 km 1400 km
PGS [W] 0.63 0.82 1.03 1.25

Table 6: Required power emission of the ground station, uplink.

The gain of the ground station uplink antenna is smaller than in the downlink case, because of the longer
wavelength. In this case, Gmax = 28.9 dB. If we consider perfect pointing from the ground segment, the needed
power emission is

PGS = EIRP/Gmax = −2.04 dBW

10
3 Thermal Subsystem
The analysis of the thermal model of a satellite is of great importance in the design of a mission. Depending
on the orbital parameters, the spacecraft will be subject to very different thermal regimes. In a LEO orbit, the
satellite is subject to the solar radiation, the albedo reflection, and the infrared emission from the Earth. One of
the characteristics of this orbit is the high number of eclipses per day, which is in the order of 14. This leads to
high variations of temperature during the orbital period.
In this section we calculate the temperature of the different parts of the satellite, and try to find the optimum
thermal covers. We begin by analyzing the different contributions to the thermal flux, and then calculate the
temperature of the six faces and the internal node. All the analyses are performed over 5 orbits during spring
equinox, with RAAN = 0º. This choice is of particular interest, since it is the most stringent in terms of solar flux:
at the equator, the +z face is directed towards the sun (Fig. 6).

3.1 Thermal fluxes

The solar flux is the most important source of radia-
tion. It acts on the satellite during the sunlight period
of the orbit, and it follows a symmetrical pattern over -y
the different faces of the satellite (see Fig. 7). The max- Earth
imum solar flux is 1357 W/m2 , reached on the ±x faces Sun +z -x -z
γ
(velocity and anti-velocity) and on the +z face. Figure +y
8 shows the solar flux incident on the solar panels. As
we can see, there is a symmetrical pattern between the
±x faces: just after eclipse, the sun starts radiating the Figure 6: Satellite at the equator during the spring
+x face. The angle between the Sun and face deter- equinox. The Space face (+z) faces directly the Sun.
mines the total solar flux, P = Pmax cos θ. In fact, we
can clearly see on the figure the sinusoidal shape of the
solar flux, which is a consequence of the rotation of the satellite.
The albedo flux is the solar radiation reflected by the
atmosphere. This term is only important at Low Earth
Orbit, and decreases with the altitude. An important
characteristic of the albedo flux is that the radiation
is diffuse. As a consequence, faces that would never
see the Sun directly will be affected by the albedo flux.
Figure 8 shows the albedo flux over the solar panels.
We can see that it has a maximum magnitude of 110
W/m2 . The face that will receive the higher albedo flux
is the Earth face, where the CCD is located. Over that
face, the albedo reaches about 300 W/m2 .
Finally, the infrared flux is the black-body radiation Figure 7: Solar flux over one orbit.
emitted by the Earth. Notice that this radiation is very
different from the solar radiation: the spectrum is on the infrared region, due to the low temperature of the Earth.
The infrared emission is constant along the orbit, with a magnitude of 86 W/m2 .

3.2 Thermal analysis

Now, we use SIMUSAT to calculate the temperature of
the satellite during the orbit. The temperature of the α ε
faces is a result of the thermal budget: the radiation White paint 0.2 0.9
absorbed by the satellite, coming from the solar flux, Black paint 0.95 0.9
infrared emission, and albedo, is equal to the radiation MLI 0.25 0.05
emitted by the satellite to deep space. The radiation Solar panels 0.8 0.9
emitted/absorbed by a gray body is defined by φ =
ασT 4 . The term α depends on the spectrum: if it is Table 7: Thermal characteristics of the different covers.

11
 Figure 8: Solar flux (left) and albedo (right) over the solar panels.

Figure 9: Temperature of the different faces and the interior of the satellite.

12
solar radiation, α is the solar absorptance, and if it is
infrared emission, α = ε is the infrared emissivity. The key point is that different materials have different absorption
coefficients depending on the spectrum. For instance, white paint is a very good reflector of solar radiation, but it
is a good absorber of infrared radiation. By choosing different covers, we can change the thermal characteristics of
the satellite. The requirements are the following: the temperature of the solar panels shall be below 60 ºC, and the
temperature of the interior shall be at least 0 ºC, to keep the batteries operative.
The satellite is covered by solar panels on the ±x and ±y faces. Therefore, we can only play with the covers for
the Earth (-z) and Space (+z) faces. Table 7 lists the different materials that are available. We first try white paint
on the +z face, to reject the solar radiation, and black paint on the -z face, to absorb the albedo flux on the Earth
side. The results of the simulation over 6 orbits are shown on Figure 9. We can see that the temperature variations
of the ±x solar panels are extreme: they go from 70 ºC to -40 ºC in less than one hour. The temperature of all the
other faces and of the interior is more stable, but is quite cold. For instance, the interior has a temperature of - 20
ºC, which is incompatible with the requirements. For the next simulation, we choose MLI on both sides, because we
want to increase the temperature of the interior. The use of MLI over non-active surfaces is very common, because
it insulates the satellite from radiative coupling. The temperature of the interior is much higher now, in the order
of -5 ºC. However, the temperature of the solar panels is still excessive.
To reduce the temperature of the solar panels, we have to increase the conductive coupling between the panels
and the central node. The objective is to bring more heat from the solar panels to the central body, which has a
high specific heat of 50000 J/º. By this way, we reduce the temperature of the solar panels and increase it a little
bit on the central node. The nominal value was of 4 W/m2 º, and we will try a higher coupling of 10 W/m2 º. The
simulation with white and black paint on the external faces is quite interesting (see Fig. 9, second row). Now,
the temperature variations of the solar panels are reduced to + 45 ºC, - 30 ºC. However, the interior is still too
cold, due to the high specific heat. We will try MLI covers with high coupling. Looking at the simulation, we can
conclude that this configuration has the best thermal characteristics. The temperature of the solar panels is well
below the limit of 60 ºC, and the interior is close to 0 ºC. If this temperature is too cold for certain components,
like the batteries, a solution would be the use of heaters.

13
4 Energy Subsystem
The Power Supply Subsystem produces the electrical power needed by the payload and the platform. This subsystem
has different functions: it converts the energy from the primary source, the solar radiation, into electrical energy,
stores the energy into batteries, and regulates, distributes, and controls the power through the satellite.
In this section we dimension the batteries and the solar arrays for our satellite. Some simulations have been
performed using SIMUSAT, to check that the system behaves as expected.

4.1 System specifications

The specifications of the power subsystem are the following,
• Non regulated architecture (24 V - 38 V)
• Nickel-Cadmium batteries (3A.h.), Max DOD = 30 %
• Maximum charging current of the battery = Cbat /2
• Power during eclipse = 30 W, Power during sunlight = 35 W
• Solar cells: AsGa (4 cm x 4cm), on the 4 lateral sides

4.2 Dimensioning the solar panels and battery

4.2.1 Battery
The batteries are made of electrochemical cells. These cells are devices made of two electrodes separated by an
electrolyte and are able to store energy in an electrochemical from. Normally, in space it is used three major kinds of
batteries: NiCd, NiH2 and Li-ion. All of them have different characteristics but all of them are used as a secondary
energy source. Their purpose is to replace the solar array during eclipse or when the solar array does not produce
enough power.

Calculate the number of cells in series taking into account the tension values
The batteries can be connected in series or in parallel. The number of batteries connected in series is determined by
the required bus voltage, while the number of strings connected in parallel is determined by the minimum capacity.
For this satellite we are using a non regulated architecture, therefore the voltage can vary from 24 V to 38 V. To
calculate the number of batteries in series, we have to take into account the maximum voltage that we can have
(38V). Each cell can give a maximum voltage of 1.55 V, giving a total number of
Vbus−max 38V
Nbatteries = V bat−max = 1.55V = 24.52 → 24

We need 24 cells in series. Now, we have to check the ranges. The minimum voltage of the battery is Vmin =
N ·Vmin = 24 V, where Vmin = 1 V is the minimum voltage provided by each cell. The maximum voltage is
Vmax = N ·Vmax = 37.2 V, where Vmax = 1.55 V. Therefore, the battery voltage will be within the specified bus
voltage.

Calculate maximum eclipse duration

The maximum eclipse duration is an important parameter to dimension the batteries. During this time, no solar
power is available, and the batteries should provide enough power to all the subsystems of the satellite.
To calculate the eclipse duration we first have to obtain the angle in which the satellite will be shadowed by the
Earth (Fig. 10). The satellite will be in eclipse if

90 + α < θ < 270 − α

 
α = arccos RR E
E +h
= 27.3 º, h = 800 km

14
 Therefore the angular eclipse duration will be 180 −
2α = 125.4º. The duration is

360 ·T
180−2α
Teclipse = = 35.07 min
q
(RE +h)3
where T = 2π µ = 6040.36 s and µ = 3.99.1014 .

Calculate the value of the whole charging Qd of

the battery during an eclipse
The duration of the orbit at this altitude is approxi-
mately of 100 min, having an eclipse of 35min we have
a sunlight period of 65 min. As it is specified in the re-
quirements, the power required during eclipse is 30 W,
and 35 W during the sunlight periods.
Therefore the value of the whole charging Qd will
be:
Figure 10: Duration of the eclipse.
35 min
Qd = ·1A = 0.584 Ah
60 min/h

Deduce the minimum capacity

To calculate the minimum capacity we have to take into
account that the maximum DOD is 30%. Using the value of Qd obtained before:
Qd 0.584
DOD = C < 30% → Cmin = 0.3 = 1.95 Ah

Calculate the number of cells in parallel and the predicted max DOD
In order to meet the value of Cmin obtained before, we calculate the number of cells in parallel taking into account
that the capacity of a string of batteries in series is the same as the capacity of one single cell. If the batteries are
in parallel, the capacity of the string is multiplied by the number of strings in parallel. The capacity of one cell is 3
A.h. which is higher than the minimum capacity, 1.94Ah. Therefore, we need only one string of 24 cells in series.

Calculate the maximum power which is needed to charge the battery

To calculate the maximum power needed to charge the battery, we have to calculate the power of the batteries at
maximum voltage and current. The charge current is C/2 at LEO, and the maximum voltage is 1.55 V for each
cell. Therefore, V max = 24·1.55 = 37.2 V, I max = C2 = 1.5 A, and the maximum power is

P max = Imax ·Vmax = 37.2·1.5 = 55.8 W

The charging cycle of the battery is shown in Figure 11.

4.2.2 Solar panels

The solar panels are the primary source of energy. They power all the subsystems of the satellite and also recharge
the batteries after an eclipse. The solar panels are made of solar cells that are basic units of a solar generator. They
are made of a semiconductor junction which converts solar energy (photons) into electrical energy (electrons). The
solar cells are characterized by four parameters: Isc, Imp, Voc and Vmp (Figure 12).
The solar cells can be assembled in the solar array in series and in parallel. The maximum temperature that
they can withstand is 60 ºC.

15
 Figure 11: Battery charging cycle.

Figure 12: Electrical characteristics of a solar cell.

16
 GaAs GaAs
Isc (bol) at 28º 0.53A dIsc
dT (eol) 3.10−4 A/º Imp(eol)
0.88 Imp (eol) at 60º 0.446A
Imp(bol)
Imp (bol) at 28º 0.5A dImp
dT (eol) 2.5.10−4 A/º V mp(eol)
0.93 V mp (eol) at 60º 0.736V
Voc (bol) at 28º 1V dV oc
dT (eol) −2.10−3 V /º V mp(bol)
P mp(eol)
Vmp (bol) at 28º 0.86V dV mp
(eol) −2.10−3 V /º P mp(bol) 0.88
dT

Table 8: Characteristics of the GaAs solar cells at the beginning of life (left). Electrical characteristics of the solar
cells at 60 ºC (right).

Calculate the minimum number of cells in series of a chain

The solar cells in a string (series) are dimensioned in order to get an overall voltage compatible with the bus voltage.
To calculate the minimum number of solar cells connected in series, we have to study the electrical characteristics
of the solar cell at the maximum temperature of 60 ºC. The characteristics of the AsGa solar cells are summarized
in Table 8. The parameters are given at beginning of life (bol) and room temperature, 28 ºC, but we have to
extrapolate them to end of life (eol) and maximum temperature, 60 ºC.
To calculate the values at end of life, we have to take into account the degradation due to space radiations (total
dose of 4.14 electrons of 1MeV at end of life). The diagram of the solar cells degradation due to radiation shows
that the Vmp is reduced by 7%, and the Imp by 12%. Therefore the electrical characteristics of the cells at end of
life and 60 ºC are shown in Table 8
The voltage of one string will be equal to the voltage of one cell multiplied by the number of cells, we can
calculate the number of cells:

Vbus Vbus 37.2

V cell = < Vmp → Ns > = = 50.54
Ns Vmp 0.736

Therefore, the minimum number of cells in series will be 51.

Calculate the maximum number of cells chains which can be put on each lateral face of the satellite
Each face of the satellite is a square of 80 cm. If the cell dimensions are 4 cm x 4 cm, and adding a gap of 1mm
between the cells, in each face of the satellite we can put:

80 = 0.1 + (4 + 0.1)·Ncells

79.9
N cells = = 19.48 → 19 cells/row
4.1
The total number of cells in one face will be Ncells = 19·19 = 361. Since the number of cells in series is 51, we
can have up to
361
51 = 7.08 → 7 strings

Calculate the cells filling ratio on the lateral sides of the satellite
Each face will have 7 strings of 51 cells. Therefore, the total number of solar cells in one face is Nf ace = 7·51 = 357
cells.

4.3 Simulation
4.3.1 Simulation specifications
The simulation done in SIMUSAT has the following parameters:

17
 Figure 13: Simulation of the power subsystem during sunlight (left) and eclipse (right).

• Initial date: 21/03/2003, date of the launch of the satellite

• Maximum duration eclipse: the Sun is in the orbital plane (21st of March), therefore the right ascension of
the ascending node is equal to zero.
• Lifetime: 3 years, therefore the end of life will be on 21/03/2006
• ROL (ψ = 0, φ = 0, θ = 0):

– +X = Velocity vector
– Y perpendicular to the orbital plane
– Z axis towards the Earth

• Sunlight power will be equal to the power during eclipse plus the emitted power (mean value during the
sunlight duration).

4.3.2 Working plan

Simulate the energy subsystem behavior with the previous defined conditions
The simulation shows how the solar arrays provide energy during the sunlight periods and the batteries during
eclipse (Figure 13). We can see also that during the sunlight periods the satellite receives 35 W of power and
the batteries are being charged. We can also see that during eclipse the batteries deliver power to the satellite,
providing 30 W.

Identify the 3 battery charging modes and verify the validity of the previous calculations
Figure 14 shows a simulation of the battery current and voltage during one orbit. We can identify 3 modes of
charge, where the battery is at constant current, constant voltage and zero current (trickle). We can appreciate
that when the voltage is constant the intensity decreases. During the trickle mode the battery is being charged,
but at low intensity. After the trickle part we can observe that the batteries provide energy to the satellite, even
during a sunlight period. During this period of the year, the+Z face is directly pointing the Sun (Figure 13) and
the satellite does not have solar panels on this face. The batteries, therefore, have to power the satellite during this
period.
In order to compare the validity of the previous calculations we can take a look to the values of the graphs
(Figure 14) and see how the maximum current of the batteries corresponds to 1.5 A, while the minimum is −1.13
A as it was defined previously. With regards to the voltage, we can see that the minimum voltage is 29.5 V and the
maximum is 36.94 V, which are within the operational range of the bus. All the simulations confirm the accuracy
of the previous calculations.

18
 Figure 14: Battery voltage (left) and current (right).

Figure 15: Simulation on the 21st of June (left). Power characteristics of the batteries, solar panels and available
power (right).

Calculate the maximum available power (without charging the battery) taking into account the mean
power which is used by the emission system: the energy bilan must be balanced on several following
orbits (the battery must be totally charged at least 5 min before the eclipse).
The maximum available power after one orbit is 123.1 W. We can see also that the batteries are providing 33.3 W
during the eclipse and they receive up to 55.4 W during the sunlight in order to be charged.

Calculate the influence of the season on the energy balance. (By defining an orbit at the date of
21st of June, when the Earth-Sun distance is maximum)
It can be observed that for this period of the year there is no change in power. The battery power, the Solar Arrays
power an the available power are constant. This is due to the fact that in this day we are in the Solstice of Summer,
the Sun has the bigger distance with the Earth. As we can observe in Figure 15, the +Y face is always pointing the
Sun during this period of the year. Therefore, there is no need to use the batteries during sunlight periods. The
power produced by the solar panels is 114 W, but the available power is 103 W, due to dissipation.

19
5 Attitude Control Subsystem
The goal of the Attitude Control Subsystem is to keep the satellite in a good orientation, pointing to the Earth,
regardless of any disturbing torque. We first obtain the performance requirements according to the mission objec-
tives, and then calculate the disturbing torques that affect our satellite. Once we have quantified the maximum
disturbing torques, we can dimension the actuators, and select an architecture.

5.1 Performance requirements for the ACS

The requirements of the Attitude Control Subsystem are driven by the mission, which consists of taking pictures of
the Earth. The quality of these pictures will depend on the capacity of the satellite to maintain a stable attitude,
and to point the camera accurately. In this section, we deduce the pointing requirements and design an Attitude
Control Subsystem which copes with them. In order to make the computation easier, the performance requirements
are evaluated at the center of the instrumental field of view.

5.1.1 Pointing requirements

The instrumental field of view is 2000 km wide, with a 10 km margin allocated to pointing errors (i.e. 0.5%). This
means that, at the center of the field of view, the line of sight must be orientated within ±5 km from the target,
in the direction that is perpendicular to the satellite track. This relation can be seen on Figure 16. If we consider
α = θ/2, the angular error that we can admit due to pointing requirements in roll axis is equal to

α = arctan ∆y 5
h = arctan 800 = ±0.36º = ±22
0

The position of the images along the satellite track must be controlled as well. The main error sources are
the accuracy of the prediction of the satellite orbital position, the accuracy of the date of imaging, and finally
the pointing accuracy around pitch axis. If we have an error in the allocated position of ±5 km, the pointing
requirements around the pitch axis will be of ±0.36º, as we have found for the roll axis.

5.1.2 Angular speed requirements

The satellite takes pictures using the push-broom technique: the CCD detector, which is a row of square detectors,
takes pictures at a certain sampling rate. The sampling period is chosen in such a way that the sub-satellite points
move one pixel during the sampling period. Therefore, the image is formed row by row.
The sampling rate can be calculated using simple geometrical relations. Figure 16 shows the satellite at two
samplingqpoints, A and B. We just have to calculate how long it takes to move from A to B. The orbital frequency
is ω 0 = (R µ+h)3 = 1.038·10−3 rad/s, and the angle subtended by A and B is θS = xs /RE . Then, the sampling
E

period TS is the time needed by the satellite to move an angle θs ,

θs 1km
Ts = ω0 = RT ω0 = 0.151 s

It is crucial that the pitch and roll angular rates are low between sampling periods. Otherwise, the spacing
between pixels will be unequal, leading to a deformation of the image. Let’s consider that the size of the pixels
shall be within 10%. This means that, between two sampling periods, the maximum displacement induced by the
angular rate is 0.1 km. To calculate the maximum angular rate, we calculate the maximum angular displacement
θ, (Fig. 16)
0.1 km
θ = arctan 800 km = 7.15·10
−3
º = 0.430

This is the maximum angle that the satellite can rotate between two pictures. Therefore, maximum angular
rate on pitch and roll is
θ
θ̇ = = 0.05 º/s
TS
With regards to the yaw axis, we can consider the same requirement. The only difference is that for pitch and
roll axis the distortion will be longitudinal, whereas for the yaw axis the distortion will occur in the corners due to

20
 xs= 1 km

A B

RE
θs

Figure 16: Error in roll axis (left), sampling distance (center), and regularity of the pixels size (right).

Figure 17: Total disturbing torque (red) and magnetic torque (blue) over six orbits. The solar, aerodynamic, and
gravity gradient torques are negligible compared to the magnetic torque.

21
the angular movement. Thereafter, and as a simplifying hypothesis, we will take into account the same requirements
on the yaw axis than on the pitch and roll axes.
Pointing Accuracy Angular Stability
3 axes ± 0.36 º ± 0.05 º/s

5.2 Environmental disturbing torques

As in all the altitudes chosen we are in low Earth orbit, we will have to take into account how the environmental
disturbing torques will affect our satellite and evaluate their order of magnitude and their evolution with respect
to the time.

5.2.1 Magnetic torque

At Low Earth Orbit, the magnetic field of the Earth has a magnitude of about 2·10−5 T. The satellite has a
remanent magnetic moment of M = 10 Am2 , which interacts with the magnetic field and produces a torque,
T mag = M̄ ∧ B
The maximum magnitude of the magnetic torque will be about T mag = 2·10−4 N m. The magnetic torque is not
constant, and varies over the different axes. This is because the Earth’s magnetic field is not a constant vector: in
a first approximation, we could think of it as a dipole. This torque is periodic.

5.2.2 Aerodynamic torque

Drag is an important effect at Low Earth Orbit. The torque that it exerts over the satellite can be calculated using
1 2
T aero =
ρV SCla
2
being ρ the atmospheric density, S the wet surface, C the aerodynamic coefficient, V the linear speed of the satellite
and la the longitude of the lever arm (i.e. distance between the center of pressure and the center of gravity). In
this case, we consider that there is a 1 cm shift between the center of pressure and the center of gravity, along the
Earth pointing axis, towards the Earth. The aerodynamic coefficient is about 2.5 and, for an orbit at 800 km, the
density is approximately 10−15 g/cm3 for high solar activity (hsa) and 10−17 g/cm3 for low solar activity (lsa). The
linear velocity can be calculated as
2πR 2π(800+6378)km
V = TE = 6052s = 7.45km/s
The magnitude of the aerodynamic torque is

T aerohsa = 4.4·10−9 Nm

T aerolsa = 4.4·10−7 Nm
This disturbance is reduced at higher altitudes. It also depends on the solar activity: when the Sun is more
active, the density of the atmosphere is higher, thus increasing the torque.

5.2.3 Solar torque

The solar radiation exerts a pressure on the satellite. If the center of gravity is different from the center of pressure,
a torque is applied on the satellite,

T sun = Psun Sla

The solar pressure radiation has a magnitude of about P sun = 5.10−6 N/m2 . Thus, the magnitude of the solar
torque will be about

T sun = 3.2·10−8 Nm
This torque is periodic along the orbit. In fact, during eclipse, the Sun doesn’t exert any pressure on the satellite.

22
 Magnetic Aerodynamic Solar Gravity Gradient
Theory [N·m] 2·10−4 4.4·10−7 3.2·10−8 8.1·10−6
Simulation [N·m] 6·10−4 7·10−7 8·10−8 8.1 · 10−6

Table 9: Maximum values of the disturbing torques, from the simulation and back-of-the-envelope calculations.

5.2.4 Gravity gradient torque

The Earth’s gravitational field decreases with the square of the distance. As a consequence, the satellite doesn’t
feel the same gravitational attraction all over the body: the parts that are closer to the Earth are more attracted.
This asymmetry produces a torque,
→
−
C = 3ω02 ·→
−
u ∧ Isat/sat →
−
u

where →−u stands for the unit vector from the Earth to the satellite, and I sat is the inertia matrix of the satellite
with respect to its center of gravity.
For the calculation we have to take into account that we have a non-diagonal term of the inertia matrix for the
x and z axis, Ixz = 2.5 kg m2 , whereas all the other non-diagonal terms are zero. With → −
u = (0, 0, 1) in the local
orbital frame, we have
      
Ixx 0 Ixz 0 Ixz  i j k 
→
−
C = 3ω02 ·→
− 0   0  = 3ω02 ·→−
u ∧  0  = 3ω02 ·  0 0 1  = 3ω02 Ixz j =

u ∧  0 Iyy
Ixz 0 Izz 1 Izz  Ixz 0 Izz 
−6 →
−
8.1·10 j N m

Therefore, the gravity gradient torque is constant and directed along the y axis.

5.2.5 Comparison of the different torques

The previous calculations are summarized on Table 9. As we can see, the most important contribution is the
magnetic torque, which is two orders of magnitude higher than the gravity gradient. Therefore, all the disturbing
torques are negligible compared to the effects of the magnetic torque. Figure 17 shows a simulation of the disturbing
torques over 5 orbits. We can see that the total disturbing torque is essentially equal to the magnetic torque. With
regards to the values obtained with the simulations (Table 9), we can see that our calculations are close to the real
values.

5.3 Assessment for different possible architectures for the ACS

5.3.1 Spin Stabilization
The spin stabilization technique consists of giving a constant angular velocity to the satellite around the pitch axis,
to gain gyroscopic stiffness. However, this technique is not at all adequate for a satellite that takes pictures of the
Earth. As we have seen in the previous section, we require a high angular speed stability on all the axes, which is
incompatible with a spin stabilization technique.

5.3.2 Stabilization with an on-board kinetic momentum

A momentum wheel provides gyroscopic stiffness to the satellite. It is essentially the same concept as spin stabiliza-
tion, however, it is only a wheel inside the satellite that spins, not the whole satellite. The wheel spins at constant
angular speed, and provides a momentum equal to Hw = Iω, where ω is the angular speed of the wheel.
For a nadir pointing satellite, the momentum wheel must be oriented along the pitch axis. Let’s recall that a
nadir pointing satellite must rotate 360º about the pitch axis in one orbit. If we placed the momentum wheel along
the yaw or roll axis, it would act against a rotation about the pitch axis. Placing a momentum wheel along the
pitch axis provides stiffness against disturbing torques on the roll and yaw axis.

23
 -Roll

Yaw -Yaw
ΔH ΔH

Roll
Figure 18: Roll - yaw coupling. Every 1/4 orbit, ∆H changes from the roll to the yaw axis.

The coupling between yaw and roll axes is of particular importance. Let’s consider that the satellite has a residual
angular momentum ∆H on the x-z plane, which induces an error on the attitude. By momentum conservation, the
vector ∆H is constant in the inertial frame, i.e. it doesn’t rotate along the orbit. However, the local orbital frame
rotates along the orbit. Figure 18 shows the situation. If the vector ∆H is along the roll axis at a given point, it
will be along the yaw axis 1/4 orbit later.
Now, let’s calculate the required wheel kinetic momentum to satisfy the pointing requirements. The performance
around the yaw axis is
Tz
ψ<
ω0 Hw
where Tz is the disturbing torque around the yaw axis, and Hw is the wheel kinetic momentum. According to the
pointing requirements, ψ < ψmax = 0.36º, and we can find the required Hw ,
Tz 6·10−4
Hw = = = 92 N·m·s
ω0 ψmax 1.04·10−3 ·0.36π/180
This value is far beyond the capability of momentum wheels for small satellites. From the available actuators
(Annex 2), the wheel with the maximum angular momentum has Hw = 2.2 Nms, one order of magnitude below
the required momentum. Therefore, it is not possible to use a momentum wheel at Low Earth Orbit for precise
pointing.

5.3.3 Gravity gradient stabilization

The gravity gradient stabilization technique consists in adding a mast along the z axis, with a mass at the end. The
mast increases the inertia around the x and y axes, and ensures that the satellite will be nadir pointing thanks to
the gravity gradient. Now, we calculate the gravity gradient torque when there is a small pointing error (ψ, θ, ϕ),
i.e. yaw - pitch - roll. We consider also that the non-diagonal terms are negligible. The gravity gradient torque is
→
−
given by C = 3ω02 ·→−u ∧ Isat/sat →
−u . The vector →−
u is now →
−
u = (− sin θ, cos θ sin ϕ, cos θ cos ϕ) w (−θ, ϕ, 1), where
we have used the small angles approximation. The gravity gradient torque is
     
−θ −Ix θ (Iz − Iy ) ϕ
~ = 3ω02 ~u ∧ I →
C −
u = 3ω02  ϕ  ∧  Iy ϕ  w 3ω02  (Iz − Ix ) θ 
1 Iz 0

As we can see, there are only components about the pitch and roll axis, which means that gravity gradient stabi-
lization doesn’t control the yaw axis. With regards to the pitch and roll axes, we need Cx , Cy < 0 when ϕ, θ > 0,
to counteract the pointing error. This requires Ix , Iy > Iz , and this is what we obtain when we put a mast along
the z axis: the inertias along the x and y axis are significantly increased.

24
 Now, we calculate the characteristics of a mast that would stabilize our satellite. We suppose that when the
satellite has a depointing error equal to the pointing requirement, ϕmax = 0.36º, the torque produced by the gravity
gradient is equal to the maximum disturbing torque, Tmax = 6·10−4 Nm. When we add a mast, the inertia about
the x and y axis is Ix,y
0
= Ix,y + Imast , where Ix,y is the inertia of the satellite without mast, and Imast = M D2 ,
where M is the mass of the mast and D the longitude of the boom. The gravity gradient torque about the roll or
pitch axis will be

C = 3ω02 (Iz − Ix0 ) θ ' −3ω02 Imast θ

The inertia of the mast shall be

Cmax
Imast = ' 29400 kg·m2
3ω02 θmax

It is not possible to use a mast with such a big inertia on a small satellite, since it would be too heavy and too
long. Just as an example, the mast used for UoSat had a mass of 3 kg and a length of 6 m, Imast = 108 kg·m2 , two
orders of magnitude below the required inertia. Therefore, we cannot use gravity gradient stabilization to meet the
pointing requirements.

5.3.4 Three axes stabilization

We have seen that a momentum wheel and a mast are not enough to meet the pointing requirements of this mission.
Therefore, it is necessary to use three axes stabilization. We will use three reaction wheels, along the three axes.
However, it is necessary to use also magnetorquers, to dump the angular momentum built up by the reaction wheels.

5.4 Dimensioning of the actuators

In the previous section we have concluded that it is necessary to use three axis stabilization. Now, we will dimension
the reaction wheels. There are two parameters that have to be taken into account: the maximum torque authority,
and the maximum angular momentum.
The torque authority is the torque that the reaction wheel can generate at a given time. Let’s recall that a
reaction wheel exerts a torque by increasing the angular speed of a kinetic wheel, C = Iw dω/dt. In order to keep
the satellite within the pointing requirements, the maximum torque authority shall be well above the maximum
disturbing torque. As a rule of thumb, we consider Cmax = 10Cdist . Then, the required torque authority is about
Cmax ' 5 mN·m.
The maximum angular momentum of the reaction wheels is determined by the maximum angular speed. The
angular momentum of the wheels builds up along the orbit: when a disturbance torque affects the satellite, the
angular speed of the reaction wheels increases. The momentum build up during one orbit is approximately Hw =
C̄·Torbit ' 4·10−4 ·6042 = 2.4 N·m·s. Therefore, the magnetorquers should be able to dump 2.4 N·m·s per orbit. An
important parameter is the frequency at which the reaction wheels must be unloaded. Let’s say that the maximum
angular momentum is 2.4 Nms. In this case, it is enough to unload the wheels once per orbit. However, if the
maximum angular momentum is 0.6 Nms, the wheels must be unloaded four times per orbit. As a rule of thumb,
let’s consider that the wheels shall be unloaded once per orbit.
With regards to the magnetotorquers, the unloading capacity is 0.03·M Nms during one orbital period, where
M is the magnetic momentum of the magnetotorquer. If we want to unload 2.4 Nms per orbit, the magnetotorquer
shall have at least a magnetic momentum of 80 Am2 .
To summarize, the ideal actuators would be a set of three reaction wheels, each of them with a torque authority
of 5 mN·m and a maximum angular momentum of 2.4 N·m·s, and a set of three magnetotorquers with a magnetic
momentum of 80 Am2 .
If we take a look to the list of available actuators, we can see that only the heaviest ones satisfy the requirements.
However, they are probably going to be out of the power and mass budgets. A good strategy would be to try to
reduce the magnetic moment of the satellite, since it produces a high disturbance torque. This requirement could
be achieved during a second iteration of the design.

25