Accepted Manuscript

Thermal analysis of the SMOG-1 PocketQube satellite

Róbert Kovács, Viktor Józsa

PII: S1359-4311(17)35091-3
DOI: https://doi.org/10.1016/j.applthermaleng.2018.05.020
Reference: ATE 12159

To appear in: Applied Thermal Engineering

Received Date: 8 August 2017

Accepted Date: 7 May 2018

Please cite this article as: R. Kovács, V. Józsa, Thermal analysis of the SMOG-1 PocketQube satellite, Applied
Thermal Engineering (2018), doi: https://doi.org/10.1016/j.applthermaleng.2018.05.020

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
 Thermal analysis of the SMOG-1 PocketQube satellite
Róbert Kovácsa,b , Viktor Józsaa,∗
a
Department of Energy Engineering, Faculty of Mechanical Engineering, Budapest
University of Technology and Economics, H-1111, Budapest, Műegyetem rkp. 3.,
Hungary
b
Department of Theoretical Physics, Wigner Research Centre for Physics, H-1121,
Budapest, Konkoly-Thege Miklós út 29-33., Hungary

Abstract

CubeSats have revolutionized the space industry in the past two decades. Its
successor, the PocketQube class seems to be a lower size limit for a satellite
which can operate continuously and can be received by radio amateur equip-
ment. The present paper discusses the simulation of the thermal environment
of the SMOG-1 PocketQube satellite at low Earth orbit by both thermal net-
work and finite element models. The major findings of the analyses are the
following. Even a single node per printed circuit board model can provide ad-
equate information about the thermal behavior without tuning the physical
parameters. By applying a finite element model with few magnitudes more
nodes, the predicted inner temperature increased as the losses were reduced
in the radiation-dominant environment compared to the thermal network
model. Therefore, this latter method provides a more conservative tempera-
ture estimation. The most sensitive component of small-sized satellites is the
battery which remains in the desired positive temperature regime even in this

∗
Corresponding author
Email addresses: [email protected] (Róbert Kovács),
[email protected] (Viktor Józsa)

Preprint submitted to Applied Thermal Engineering March 25, 2018

 satellite class according to the finite element model. However, the thermal
network model predicted a restricted battery charging protocol to ∼50% of
the lit duration. Nevertheless, this condition still results in a positive energy
balance by a factor of 1.5.
Keywords: picosatellite, satellite, finite element method, thermal network,
thermal analysis, low Earth orbit

1 1. Introduction

2 A new era has started in space industry with the introduction of CubeSats
3 in 1999 [1]. The 10x10x10 cm satellite design (1U) brought outstanding at-
4 tention since it is characterized by low development time and launch cost [2],
5 also reviewed by NASA [3]. Such a small satellite is an excellent platform for
6 low-cost experiments or measurements. An even smaller variant, a 5x5x5 cm
7 (1P) satellite size was proposed in 2009 by Bob Twiggs, called PocketQube,
8 to cut further the costs and development time [4]. However, after the great
9 success of the CubeSats, only a few groups wanted to adopt the new size. To
10 date, there are more than fifteen known PocketQubes are under development,
11 according to the topical web pages and the PocketQube Workshop, organized
12 by Delfi Space. Up to now, only four PocketQubes were launched from which
13 only the WREN satellite was 1P in size; the others were 1.5P and 2.5P [5].
14 These satellites were released by the UniSat-5 mother satellite. Due to the
15 low interest, the PocketQube specification was removed from the internet. A
16 reason was that it was impossible to substitute eight 1P units with a 1U in
17 a CubeSat launch pod which became a widespread standard over the years.
18 However, the PocketQube seems to be the smallest satellite class which can

2
19 operate continuously and be detected by low-cost radio amateur equipment.
20 The SMOG-1 PocketQube-class satellite, shown in Fig. 1, is currently
21 being developed at the Budapest University of Technology and Economics
22 by students and professors. This spacecraft is planned to be the successor
23 of the MaSat-1, the first Hungarian satellite. Its principal mission is the
24 measurement of electrosmog emitted by the land-based digital TV stations
25 in the 430-860 MHz band [6, 7, 8], being the name-giver of the satellite.
26 The secondary payload is a total dosimeter [9]. A Sun Synchronous Orbit
27 (SSO) at not higher than 600 km altitude and inclination of 98◦ is desired to
28 ensure an acceptably low orbit life since common commercial electronic parts
29 are used principally to keep the project expenses low and avoid being space
30 debris for a long time. Note that the satellite has doubled circuits which can
31 handle a single point of failure while it runs on only one battery. The lowest
32 acceptable altitude for the project is 400 km which ensures at least three
33 months lifetime before re-entry to perform all the planned measurements.
34 Currently, the estimated launch time is Q4 2019 at a 550 km SSO by the
35 Falcon 9 launch platform. The satellite is scheduled to have a double launch
36 since it will be carried by the UniSat-7, developed by GAUSS Srl. However,
37 all of the satellite components and software are developed in the university,
38 the present paper is confined to the thermal analysis. All the documentation
39 are planned to be publicly available upon successful launch, in accordance
40 with the pursuit of Scholz and Juang [10].
41 Predicting the thermal behavior of small-sized satellites is crucial since
42 there is insufficient room for their thermal control which is often desired in
43 the space industry [11, 12, 13]. The goal of the present paper is similar

3
 Figure 1: The assembled satellite model for qualification tests.

44 to the ones of Bulut and Sozbir [4], Corpino et al. [11], and Anh et al.
45 [14], namely, performing thermal analysis of a small-sized satellite. How-
46 ever, the mentioned papers evaluate the thermal characteristics of CubeSats
47 which have considerably larger heat capacity and internal volume than that
48 of PocketQubes. Therefore, 1P units require a different conceptual design
49 [5, 15, 16, 17, 18, 19]. For further differences originated from the satellite
50 size, see [20, 21].
51 The novelty of the present paper is to enrich the thin literature of the
52 thermal design of small-sized satellites. Since a detailed thermal analysis is
53 absent in the 1P size, this work might give an insight to fellow small-size satel-
54 lite developers. A thermal network (TN) model in MATLAB
R
Simulink
R

55 and a Finite Element Method (FEM) analysis in ANSYS

R
Workbench envi-
56 ronment were established, and their results were compared. After a successful
57 launch, the thermal behavior will be published along with the validation of
58 the below-discussed models.

4
59 2. Materials and methods

60 The reason behind performing the calculations by TN and FEM is that

61 TN is usually not computationally intensive, allowing long-term simulations
62 to run on a single PC. The FEM model helps in setting the appropriate ther-
63 mal resistances of the TN model besides revealing the detailed temperature
64 field. The present section begins with the introduction of the satellite struc-
65 ture which is followed by a summary of the thermal environment. Finally,
66 the details of the TN and FEM models are discussed.

67 2.1. Structure

68 The satellite has five internal Printed Circuit Boards (PCBs) and six side
69 PCBs equipped with solar cells. Table 1 summarizes the mass (m) and the
70 averaged specific heat (c) of each panel, including all electric components.
71 The regulated voltage of the system is 3.3 V. Since selected electronic com-
72 ponents were used in the PCBs, the efficiency of them was maximized to
73 achieve minimal losses. The dominant power consumer is the communica-
74 tion and spectrum analyzer integrated circuit which dissipates 30 mA during
75 measurement and 100 mA during transmission. A normal operating cycle
76 consists of 9 s measurement followed by 1 s status report. This transmission
77 begins with a Morse code for radio amateurs which is followed by a binary
78 code of basic information about the condition of the satellite. If it passes
79 over a trusted ground station, it might be commanded to transmit all mea-
80 surement data back to Earth to clear its internal memory. Such an operation
81 can extend up to 6 minutes, depending on the orbit. The concept here was
82 to sacrifice transmission bandwidth to maximize the signal-to-noise ratio as

5
83 there is very low available power on board.

Table 1: Estimated mass and average heat capacities of the main parts.
Panel m [g] c̄ [J/kgK] Role
Top 14.4 869
Bottom 19.0 877
Left 13.3 862 Solar cells with Maximum
Right 13.3 862 Power Point Tracking
Front 14.4 850
Back 13.4 861
Capacitator bank and
CAP 14.8 798
connection of side panels
OBC 8.4 882 Onboard computer
Communication and
COM 34.6 799
signal analysis
PCU 10.5 839 Power control unit
Electrical power system
EPS 11.8 827
and dosimeter
BAT 17.2 1200 Li-Ion battery
TH 7.9 450 Threaded rod and nuts

84 The mass of the assembled satellite is 200 g, and its overall heat capacity
85 is 171.5 J/K. Due to the low available volume (1/8 liter), there is no room
86 for a separate frame which hosts the PCBs. Therefore, these panels must
87 serve as load carriers. However, the satellite size forces such a design choice,
88 the structure can survive the launch procedure, proven by the desired shock

6
89 tests included in the Falcon 9 User’s Guide.

90 2.2. Thermal environment

91 The mean lit surface of the orbiting satellite is governed by the angu-
92 lar velocity. It is convenient to handle the angular velocity in a spherical
93 coordinate system, allowing the separation of the direction vector from the
94 magnitude. The latter affects the Maximum Power Point Tracking (MPPT)
95 logic of the side PCBs and does not affect the mean lit surface. By neglecting
96 the magnitude, it is assumed that the power generation by the solar cells is
97 working at 100% efficiency. However, at high angular velocities (above 100
98 rotation per minute), the MPPT cannot work, making the solar cells inef-
99 fective. Therefore, all of the incoming solar flux is absorbed by the satellite,
100 slightly increasing the temperature of all components.
101 The mean value of the direction vector of the angular velocity tells the
√
102 average lit surface area between the minimum (1) and the maximum ( 3) of
103 a perfect cube with unit length. It is calculated as follows:

16 π/4 π/4
Z Z
[sin(φ) + cos(φ)] cos(θ) + sin(θ)dφdθ = 1.519, (1)
π2 0 0
104 where the φ angle is the axis of rotation which is normal to the top side
105 of the cube. θ angle is its tilt from the Earth-Sun line. The problem is
106 self-symmetrical by 90◦ , therefore, it is enough to calculate the integral from
107 zero to 45◦ for both angles. 1.519 means 72◦ tilt angle which is used in
108 the subsequent calculations. However, rotation at 1 rotation per minute is
109 employed in the TN model, the average incoming heat was added in the
110 FEM model as a boundary condition for the side panels to avoid excessive
111 run times.

7
112 Table 2 contains the key parameters of the thermal environment based
113 on refs. [21, 22] and the satellite properties. The most power is dissipated
114 by the communication and spectrum analyzer integrated circuit, located in
115 the COM panel. It is followed by the battery, however, the heat generation
116 drops as its internal resistance decreases with the increasing temperature
117 which means a favorable self-controlling behavior. The radiation of Earth
118 at 255 K falls in the infrared (IR) regime according to Wien’s displacement
119 law. Consequently, this radiation source cannot be used for energy generation
120 since the solar cells absorb electromagnetic rays in the micrometer wavelength
121 regime with very poor efficiency. Hence, all the incident IR radiation on
122 the solar cells is converted into heat which was taken into account during
123 boundary condition setup.
124 As for the other components, they maintain > 95% efficiency, meaning
125 few mW heat generation distributed among all the other panels. Therefore,
126 this thermal power was neglected which brings one to the safe side. The
127 concentrated thermal powers were considered as the average of the satellite
128 operation during a full orbit. In the present case, Spectrolab Ultra Triple
129 Junction Solar Cells are used. These are principally made for CubeSats,
130 therefore, they were halved to fit the quarter-sized sides. After the breaking
131 process, the full functionality was conserved with a slight asymmetry in the
132 maximum output power. However, this cruel solution causes no problem
133 since each side has a dedicated MPPT circuit. The size of the solar cells
134 is ∼1500 mm2 . The infrared radiation of Earth can be substituted by an
135 equivalent heat flux to the satellite, see, e.g., [11]. Since it varies with the
136 view factor, it is theoretically more correct to calculate the radiative heat

8
 Table 2: Thermal boundary conditions of the TN model.
Space temperature 2.7 K
Solar heat flux 1367 W/m2
View factor of Earth 0.305
Avg. temperature of Earth 255 K
Avg. albedo 21%
Orbit height 550 km
Orbit period 5669 s
Lit/shadow ratio 64%/36%
Heat generation in the COM panel 203 mW
Heat generation inside the battery 10 mW
Efficiency of the solar cells 27%
Absorptance of the PCBs 0.85

137 transfer with Earth, like in ref. [14] at 255 K average temperature [21].
138 The guaranteed operating temperature range of the used commercial elec-
139 tric components is at least T = −40 – +80 ◦ C. The Li-ion battery is the most
140 sensitive to its operating temperature which is the bottleneck of the satellite,
141 emphasized by other researchers as well [11, 14]. It can withstand T = −10
142 – +60 ◦ C during use and T = 0 – +45 ◦ C while charging. In order to reduce
143 the amplitude of temperature fluctuations, it is placed in the center of the
144 frame with insulation. Note that using more than double-layer insulation has
145 a negligible gain on the battery temperature since conduction becomes the
146 governing heat transfer process via the connecting electric wires.

9
147 2.3. The TN model

148 In the thermal network model, all PCBs were assumed as a single node.
149 Due to its limited operating temperature range, the battery is handled sep-
150 arately. Hence, the TN model consists of twelve nodes: six PCB sides, five
151 inner PCBs, and the battery. The thermal characteristics of the threaded
152 rods were split between the other components, considering their connection
153 with the rods. The temperature variation was determined by Eq. 2 as the
154 effect of convection is negligible in space:

n n
dT X X
c·m· = Q̇rad + Q̇cond + Q̇int , (2)
dt i=1 i=1

155 where t is the time, T is the temperature, Q̇ is the thermal power, and
156 rad and cond refer to radiation and conduction, respectively. Q̇int notes
157 the internal heat generation which is considered only in the battery and the
158 COM panel. All the other electric components are highly efficient, and the
159 panels showed negligible temperature difference during operation, verified by
160 a thermal camera with 60 mK sensitivity per pixel. The absorptivity and
161 emissivity of the surfaces were uniformly α = ε = 0.85, as noted in Table 2.
162 The view factors of the surfaces are calculated by the position of each panel
163 without taking the effect of the mounted electric components into account.
164 Thermal conduction was modeled between the OBC and the battery as the
165 battery cap flanges are to be soldered to the back of the OBC panel. However,
166 the battery is connected to the PCU with two wires which are responsible
167 for the majority of the heat transfer. The third conduction contact group
168 consists of the connected PCBs. All the thermal resistances used here were
169 synchronized with those of the FEM simulation. Therefore, biased results

10
170 are avoided, often caused by the human factor as there are some constants
171 to tweak manually to get visually better results. Regarding boundary con-
172 ditions, the two thermal models are strictly similar here. Nevertheless, the
173 number of nodes of the FEM model is two magnitudes higher which leads to
174 a more detailed solution, discussed in Section 3.

175 2.4. The FEM model

176 By applying a FEM model, there is an opportunity to obtain local results
177 which depend upon the mesh size. It is essential to utilize a simplified geom-
178 etry to reduce the mesh complexity which results in less number of equations
179 to solve. Hence, the computational time decreases significantly, especially,
180 when radiation is present which scales with T 4 .
181 In order to simplify the geometry, the particular contacts between the
182 PCBs, i.e., wires, bus connectors, were substituted by rectangles. The es-
183 timated thermal resistances were given between the contact and target sur-
184 faces. Furthermore, all the electronic parts were removed from the model,
185 but their heat capacity was considered in their thermophysical properties. A
186 mesh dependency analysis was performed first at three different face sizes,
187 summarized in Table 3.

Table 3: Settings of the mesh dependency analysis.

Mesh quality Face size [mm] # of nodes # of elements
coarse 30 6606 2421
mediocre 10 9321 3008
fine 2 55780 15646

188 Time advancing was set to automatic with 10 s as initial and 50 s as

11
189 maximum step sizes with duration of five full orbits. The effect of initial
190 condition vanished after four orbits. Calculations by using the fine mesh
191 resulted in <0.65 ◦ C deviance in terms of minimum, maximum, and average
192 temperatures of each part compared to the use of the mediocre mesh. For
193 further simulations, the mediocre mesh was chosen as it reveals the tem-
194 perature distribution better than the coarse mesh which resulted in ∼1 ◦ C
195 deviance from the temperature field of the fine mesh.
196 Figure 2 shows the mediocre mesh and the simplified internal structure
197 with named parts. The size of the base plate is 58 x 64 mm which will be
198 inserted into the rail of the launch pod. Besides the PCBs, the battery was
199 included since it is more sensitive to the operating temperature than all the
200 other electric components. The threaded rods were also considered since they
201 establish a notable thermal bridge between the inner PCBs and the top and
202 the bottom sides.

Figure 2: The FEM model with mediocre mesh and simplified geometry.

12
203 2.5. Additional boundary conditions of the FEM model

204 In the FEM model, the local domains were distinguished from each other
205 like the solar cell and the domain around it, shown in Fig. 3. Hence, by
206 considering the heat flux boundary condition, its value locally varies even in
207 one side since the solar cell transforms a portion of the incoming solar flux
208 to electricity. Figure 4 shows the square wave-like incident solar flux to a
209 solar cell on orbit. The heat flux was averaged for each section to simplify
210 the effect of rotation and significantly reduce the computational time. Table
211 4 summarizes all the averaged heat fluxes defined for the side panels of the
212 satellite.

Figure 3: Separately handled heat flux boundary condition on the middle side solar cells,
highlighted in blue.

213 Ambient-to-surface type radiation was considered between the outer sides
214 and space, summarized in Table 2. As for the inner surfaces, surface-to-
215 surface radiation was defined. The view factors were automatically com-

13
 Figure 4: Periodic heat flux boundary condition on the middle side solar cells, shown in
Fig. 3.

Table 4: Thermal radiation boundary conditions in the FEM model.

Part Heat flux source Magnitude
Top panel Incident solar 357 W/m2
Top solar cell Incident solar 282 W/m2
Bottom panel Albedo + IR 101 W/m2
Bottom solar cell Albedo + IR 74 W/m2
Middle side panels All three 473.8 W/m2
Middle side solar cells All three 395.7 W/m2

216 puted. Regarding other parameters and the solver, solely the default settings
217 were used.

218 3. Results and discussion

219 As two different methods were applied to estimate the thermal balance
220 of the SMOG-1 satellite, the present section starts with the TN due to its
221 simplicity over the FEM model. Secondly, the results of the latter method
222 are discussed in the light of the outcome of the former analysis. Note that

14
223 the results are corresponding to a 550 km SSO, but a generalization of the
224 findings was performed where it was reasonable.

225 3.1. TN results

226 The temperature variation of the battery is shown in Fig. 5 while Figs. 6
227 and 7 contain that of the side panels of the first five orbits, starting from a
228 uniform +10 ◦ C initial temperature. Since the solar flux varies by the orbit
229 of the Earth around the Sun, minimum and maximum values correspond to
230 1322 and 1414 W/m2 , respectively.

Figure 5: Temperature variation of the battery during five orbits at minimum and maxi-
mum solar flux.

231 The battery temperature violates the 0 – +45 ◦ C regime, shown in Fig. 5.
232 Nevertheless, the discharge conditions are met in the simulated case by the
233 TN model. Therefore, charging should be intermittent as the battery tem-
234 perature is expected to fall below 0 ◦ C. The problem is that the low heat
235 capacity results in a rapid heat loss in the shadow of Earth. Hence, when the
236 satellite comes to light again, the battery temperature still falls due to the
237 cooler panels. Based on the results at 1322 W/m2 solar flux, only the second
238 half of the orbit is suitable for storing the generated power by the solar cells.

15
239 Therefore, an onboard logic must be implemented which detaches the battery
240 from the power bus when the conditional minimum temperature is met. By
241 assuming an estimated 203 mW average power consumption of the SMOG-1
242 and if only the half of the lit period can be used for storing power, the over-
243 all energy balance is still positive by a factor of at least 1.5. Consequently,
244 the operation of the battery is safe from both thermal and energy point of
245 views. This conclusion indicates that the battery technology set the current
246 size limitation of the satellite which is able to operate continuously with full
247 functionality in orbit.
248 The temperature variation of the top, the bottom, and a side panel is
249 shown in Figs. 6 and 7 at the minimum and maximum solar flux, respec-
250 tively. The bottom side sees only the Earth, therefore, only the albedo and
251 the infrared radiation increases its temperature which results in fluctuations
252 between -14 ◦ C and -34 ◦ C. The side panel receives fluctuating solar flux
253 due to its rotation which causes roughly 1 ◦ C variation during a full turn
254 at 1 rotation per minute. This PCB is characterized by the highest peak
255 temperature of the satellite which is +25 ◦ C.

Figure 6: Temperature variation of three sides during five orbits at 1322 W/m2 solar flux.

16
 Figure 7: Temperature variation of three sides during five orbits at 1414 W/m2 solar flux.

256 Figure 8 shows the characteristics of three inner PCBs which are not
257 in connection with the battery in the same time frame as Fig. 6 at 1322
258 W/m2 solar flux. CAP is subjected to the largest temperature fluctuations
259 while EPS fluctuates with reduced amplitude and its temperature remains
260 below of that of the other two PCBs. The temperature variation of the
261 inner PCBs is closely in phase with the sides. Consequently, it can be stated
262 based on the TN model that the electric components except the battery will
263 operate within their allowed temperature range during the mission. However,
264 the situation of the battery requires a more complicated programming and
265 electrical layout, it fulfills all the requirements of safe and reliable operation.
266 Based on the results of the TN model, it can be stated that the system is
267 able to operate safely in the described thermal environment with a reasonable
268 safety factor. Therefore, e.g., the loss of a single solar cell is expected not to
269 harm the mission. Note that all these conditions are pessimistic since any
270 penalty or damage to the power generation system results in more absorbed
271 heat, hence, higher temperatures. Consequently, the increasing battery tem-
272 perature leads to wider charging regimes.

17
 Figure 8: Temperature variation of three of the inner PCBs at 1322 W/m2 solar flux.

273 3.2. FEM results

274 In the previous section, it was shown that there is no significant difference
275 between the temperature values in the case of minimum and maximum solar
276 fluxes. Therefore, the present analysis proceeds with the time-averaged solar
277 flux applied, i.e., 1367 W/m2 .
278 Figure 9 shows the global temperature extremes over time, and Fig. 10
279 shows the global temperature distribution after six hours (roughly three and
280 a half orbit) simulation time. It corresponds to the quasi-steady behavior;
281 the transient effects due to the initial conditions already vanished. Note that
282 there is a global temperature peak at this time instant. According to the
283 simulations, a stable periodic temperature oscillation appears by orbiting as
284 it was also concluded by Gaite et al. [23, 24]. In the followings, the results
285 of the FEM and TN models will be presented and discussed.
286 Among the electronic parts, the battery has a critical role as it was em-
287 phasized earlier; its temperature variation is shown in Fig. 11. The internal
288 thermal resistance of the battery can be neglected since there is no significant
289 temperature difference within its physical structure. Consequently, modeling

18
 Figure 9: The global temperature history according to the FEM model at 1367 W/m2
solar flux.

Figure 10: The global temperature distribution after six hours simulation time at
1367 W/m2 solar flux.

290 a battery with a single node might be a reasonable simplification for small-
291 scale satellites. Comparing the results to the TN model, the FEM predicts
292 higher temperatures under the same boundary conditions which confirm the
293 safe operation, including the stricter conditions of charging the battery. Fur-
294 thermore, a noticeable phase lag appears between these time histories. It is

19
295 originated from the simplicity of TN model; the overall thermal resistances
296 are smaller due to the one PCB – one node simplification.

Figure 11: The battery temperature history at 1367 W/m2 solar flux.

297 Regarding the sides of the cube, there are three characteristically distin-
298 guishable parts. The first is the top side, which receives only the solar flux,
299 shown in Fig. 12. Figure 13 presents the bottom side which is affected by the
300 albedo and the infrared radiation of Earth. The third one discussed here is
301 the front side which is thermally the same as the remaining three sides. Fig-
302 ure 14 shows the temperature history of this face, influenced by all radiative
303 effects. Paying attention again to the phase lag between the FEM solution
304 and the TN model, shown in Fig. 13, it appears in the opposite direction
305 compared to the battery temperature history of Fig. 11. However, this lag
306 disappears in case of the rotating sides in Fig. 14.
307 The spatial temperature distribution of the inner surface of the sides after
308 six hours simulation time is shown in Figs. 15, 16, and 17, calculated by the
309 FEM model. The blue regime in the middle of Fig. 15 shows that the inner
310 PCBs, hence, the CAP, are cooler and draw heat from the top panel via
311 radiation. The outer regions are characterized by higher temperature since

20
 Figure 12: The top side temperature history at 1367 W/m2 solar flux.

Figure 13: The bottom side temperature history at 1367 W/m2 solar flux.

Figure 14: The front side temperature history at 1367 W/m2 solar flux.

312 the side panels are warmer. This phenomenon peaks below the slot which
313 allows wire data connection and battery charging. It also hosts the Remove

21
314 before flight pin which switches off the system while inserted.

Figure 15: Spatial temperature distribution of the top side at 1367 W/m2 solar flux.

315 Since the bottom panel, shown in Fig. 16, receives mostly the IR radiation
316 of Earth in addition to albedo, it is cooler than the top panel. Here, the
317 thermal bridge effect of the threaded rods is clearly visible. This can also be
318 found on the top side, however, its magnitude almost fades since that side is
319 warmer.
320 Figure 17 shows the temperature distribution of a side panel. The hori-
321 zontal asymmetry is due to the diagonally placed threaded rods which trans-
322 port heat to the cooler parts. Since the side panels receive the most heat
323 flux, shown previously in Table 4, the central region is characterized by higher
324 temperatures. Note that the global minimum and maximum temperatures
325 at 21600 s simulation time were 15 and -26 ◦ , respectively. The temperature
326 extremes at this time instant were located on the outer surface of the sides.
327 As for the summary of the comparison of the two methods, it can be
328 stated that the FEM model estimates generally higher temperatures. Since

22
 Figure 16: Spatial temperature distribution of the bottom side at 1367 W/m2 solar flux.

Figure 17: Spatial temperature distribution of the front side at 1367 W/m2 solar flux.

329 the temperature of a PCB in the case of the TN model is concentrated at a

330 single point, it leads to increased heat losses through radiation to the ambient
331 as it scales with the fourth power of the absolute temperature. However, the
332 TN model was created in MATLAB
R
Simulink
R
environment, it contains

23
333 no special functions which allow the construction of similar solvers in any
334 open source environment. Moreover, a single run takes ∼10 seconds while
335 the FEM setup at mediocre mesh requires 30 minutes on a PC, being an-
336 other advantage of the TN model. The underprediction of the temperature
337 compared to measurements is shown by [17]. Note that advanced numer-
338 ical approaches like genetic algorithm [15], differential evolution, or other
339 metaheuristic methods require reliable measurement data which are often
340 not available in the conceptual design phase.
341 If one wishes to further improve the accuracy of a TN model in a radiation-
342 dominated environment like space, the number of divisions should be in-
343 creased which provides similar results to that of a FEM simulation [11]. If
344 both methods are available at will, then a wise introduction of virtual ther-
345 mal resistances can be performed to get an accurate but less computation-
346 intensive TN model. It was concluded that even a tiny satellite is thermally
347 safe in the vicinity of Earth. Bulut and Sozbir [4] have shown that the equi-
348 librium temperature of the satellite is expected to decrease by a small extent
349 up to a 2000 km orbit. Therefore, the results of the present paper might be
350 used at other low-Earth orbit profiles as well.

351 4. Conclusions

352 Thermal analysis of the SMOG-1 PocketQube (or picosatellite) was pre-
353 sented by comparing the results of the thermal network (TN) and finite ele-
354 ment methods (FEM). This investigation assumed a 550 km Sun Synchronous
355 Orbit at 98◦ inclination, but the methods shown here are generally applica-
356 ble to similar problems. Based on the results, the following conclusions were

24
357 derived:

358 • The FEM model predicts higher temperatures than the one panel-one
359 node TN model. The reason behind this is that the outer panels con-
360 sisting of numerous nodes which allow lower temperature losses due to
361 radiation as the thermal resistance between them is considered, unlike
362 in the TN model. As a side effect, the FEM model requires significantly
363 more time to solve (30 minutes against 10 seconds in the present case).

364 • Despite the simplicity of the TN model regarding the concentrated

365 thermal capacities and resistances, it is adequate for temperature pre-
366 diction over extended time and real-time analysis. Compared to the
367 FEM model, the deviations are explained by the different mathemati-
368 cal approaches and the number of nodes.

369 • Even though the small-sized PocketQube satellites have very low over-
370 all heat capacity, they can perform continuous operation since a Li-ion
371 battery can be kept within its operating temperature range at a low-
372 Earth orbit. Nevertheless, the more conservative TN model predicted
373 at a minimum solar flux of 1322 W/m2 that its temperature might fall
374 below 0 ◦ C when discharge is unrestricted, but charging is not advised.
375 However, the reduced time for charging still allows a continuous oper-
376 ation since 1.5 times the energy demand of a single orbit can be stored
377 during the restricted time frame.

25
378 Acknowledgments

379 This project was funded by numerous institutes, corporations, and prin-
380 cipally by Budapest University of Technology and Economics. For the full list
381 of supporters and their contribution, visit http://gnd.bme.hu/smog1/tamogatok.php.
382 The authors gratefully acknowledge the support of NVIDIA Corporation with
383 the donation of the Quadro P6000 used for this research. In addition, the
384 support of Ágnes Welsz and Gergely Kun is also acknowledged in developing
385 the thermal network models.

386 References

387 [1] S. Lee, A. Hutputanasin, A. Toorian, W. Lan, R. Munakata, J. Carna-

388 han, D. Pignatelli, A. Mehrparvar, CubeSat Design Specification. rev 13
389 (2014).

390 [2] J. Bouwmeester, J. Guo, Survey of worldwide pico- and nanosatellite

391 missions, distributions and subsystem technology, Acta Astronautica
392 67 (7-8) (2010) 854–862. doi:10.1016/j.actaastro.2010.06.004.

393 [3] R. Shimmin, Small Spacecraft Technology State of the Art

394 (NASA/TP2015216648/REV1), Tech. rep., NASA, Mission De-
395 sign Division Ames Research Center, Moffett Field, California (2016).
396 URL https://www.nasa.gov/centers/ames/engineering/state-of-the-art

397 [4] M. Bulut, N. Sozbir, Analytical investigation of a nanosatel-

398 lite panel surface temperatures for different altitudes and panel
399 combinations, Applied Thermal Engineering 75 (2015) 1076–1083.
400 doi:10.1016/j.applthermaleng.2014.10.059.

26
401 [5] S. Speretta, T. Pérez Soriano, J. Bouwmeester, J. Carvajal Godı́nez,
402 A. Menicucci, T. G. Watts, P. P. Sundaramoorthy, J. Guo, E. K. A. Gill,
403 Cubesats to pocketqubes: Opportunities and challenges, in: Proceedings
404 of the 67th International Astronautical Congress (IAC), IAF, 2017, pp.
405 1–5.

406 [6] L. Dudás, L. Szűcs, A. Gschwindt, The spectrum monitoring system by

407 Smog-1 satellite, in: Conference on Microwave Techniques (COMITE),
408 2015, IEEE, 2015, pp. 1–4. doi:10.1109/COMITE.2015.7120316.

409 [7] L. Dudás, A. Gschwindt, The communication and spectrum monitoring

410 system of Smog-1 PocketQube class satellite, in: 21st International Con-
411 ference on Microwave, Radar and Wireless Communications (MIKON),
412 2016, IEEE, 2016, pp. 1–4. doi:10.1109/MIKON.2016.7491999.

413 [8] B. D. Olaszi, V. Józsa, J. Ladányi, Energy management of a PocketQube

414 satellite, in: 5th International Youth Conference on Energy (IYCE),
415 2015, IEEE, 2015, pp. 1–5.

416 [9] G. Géczy, Results from SMOG-1 PocketQube, in: PocketQube Work-
417 shop, Delft, 2017, p. 31.

418 [10] A. Scholz, J. N. Juang, Toward open source CubeSat design, Acta As-
419 tronautica 115 (2015) 384–392. doi:10.1016/j.actaastro.2015.06.005.

420 [11] S. Corpino, M. Caldera, F. Nichele, M. Masoero, N. Viola, Thermal de-

421 sign and analysis of a nanosatellite in low earth orbit, Acta Astronautica
422 115 (2015) 247–261. doi:10.1016/j.actaastro.2015.05.012.

27
423 [12] A. Torres, D. Mishkinis, T. Kaya, Mathematical modeling of a new satel-
424 lite thermal architecture system connecting the east and west radiator
425 panels and flight performance prediction, Applied Thermal Engineering
426 65 (1-2) (2014) 623–632. doi:10.1016/j.applthermaleng.2013.11.040.

427 [13] P. Fortescue, G. Swinerd, J. P. W. Stark (Eds.), Spacecraft Sys-

428 tems Engineering, Fourth Edition, Wiley Online Library, 2003.
429 doi:10.1002/9781119971009.ch2.

430 [14] N. D. Anh, N. N. Hieu, P. N. Chung, N. T. Anh, Thermal radiation

431 analysis for small satellites with single-node model using techniques of
432 equivalent linearization, Applied Thermal Engineering 94 (2016) 607–
433 614. doi:10.1016/j.applthermaleng.2015.10.139.

434 [15] E. Escobar, M. Diaz, J. C. Zagal, Evolutionary design of a

435 satellite thermal control system: Real experiments for a Cube-
436 Sat mission, Applied Thermal Engineering 105 (2016) 490–500.
437 doi:10.1016/j.applthermaleng.2016.03.024.

438 [16] J. Rotteveel, A. Bonnema, Thermal control issues for nano- and pi-
439 cosatellites, in: 57th International Astronautical Congress, International
440 Astronautical Congress, International Astronautical Federation, 2006,
441 pp. 1–8. doi:10.2514/6.IAC-06-B5.6.07.

442 [17] J.-R. Tsai, Overview of satellite thermal analytical model, Journal of
443 Spacecraft and Rockets 41 (1) (2004) 120–125. doi:10.2514/1.9273.

444 [18] M. F. Diaz-Aguado, J. Greenbaum, W. T. Fowler, E. G. Lightsey, Small

28
445 satellite thermal design, test and analysis, in: Proceedings of the SPIE,
446 2006, p. 12. doi:10.1117/12.666177.

447 [19] C.-J. Fong, A. Lin, A. Shie, M. Yeh, W.-C. Chiou, M.-H. Tsai, P.-Y.
448 Ho, C.-W. Liu, M.-S. Chang, H.-P. Pan, S. Tsai, C. Hsiao, Lessons
449 learned of NSPOs picosatellite mission: YamSat-1A, 1B & 1C, in:
450 Proceedings of the 16th Annual AIAA/USU Conference on Small
451 Satellites, SSC02-X-6, 2002, p. 12.
452 URL http://digitalcommons.usu.edu/smallsat/2002/all2002/60/

453 [20] D. Barnhart, M. N. Sweeting, Right-sizing small satellites, in: Proceed-

454 ings of the 28th Annual AIAA/USU Conference on Small Satellites,
455 2014, 2014, pp. 1–8.

456 [21] D. G. Gilmore, Spacecraft thermal control handbook. Vol. 1. Funda-

457 mental Technologies, El Segundo, California, Aerospace Press, 2002.
458 doi:10.2514/4.989117.

459 [22] J. H. Henninger, Solar Absorptance and Thermal Emittance of Some

460 Common Spacecraft Thermal-Control Coatings (No. NASA-RP240400-
461 1121), Tech. rep., National Aeronautics and Space Administration, God-
462 dard Space Flight Center, Greenbelt, Maryland (1984).

463 [23] J. Gaite, A. Sanz-Andrés, I. Pérez-Grande, Nonlinear analysis of a sim-

464 ple model of temperature evolution in a satellite, Nonlinear Dynamics
465 58 (1) (2009) 405. doi:10.1007/s11071-009-9488-x.

466 [24] J. Gaite, Nonlinear analysis of spacecraft thermal models, Nonlinear

467 Dynamics 65 (3) (2011) 283–300. doi:10.1007/s11071-010-9890-4.

29
Highlights

Thermal analysis of the SMOG-1 picosatellite is presented.

Results of a simple thermal network and finite element methods were compared.
The sensitive battery just fulfills all the requirements for continuous operation.
The thermal network model underpredicts the temperature due to its simplicity.
A simple thermal network is able to predict the temperature variations appropriately.