Dimensional Analysis for the Design of Satellites in...
Transcript of Dimensional Analysis for the Design of Satellites in...
Lehrstuhl fur Raumfahrttechnik
Prof. Dr. rer. nat.
Ulrich Walter
Diplomarbeit
RT-DA 2010/14
Dimensional Analysis
for the Design of Satellites
in LEO
Author: Tanja Nemetzade
Primary Advisor: Dipl.-Ing. Andreas Hein
Lehrstuhl fur Raumfahrttechnik / Institute of Aeronautics
Technische Universitat Munchen
Local Advisors: Prof. Olivier de Weck and Dr. Afreen Siddiqi
Engineering Systems Division
Massachusetts Institute of Technology
Bestatigung der eigenstandigen Arbeit
Ich erklare hiermit, dass ich diese Arbeit ohne fremde Hilfe angefertigt und nur die in dem
Literaturverzeichnis angefuhrten Quellen und Hilfsmittel benutzt habe.
Garching, den
Name: Tanja Nemetzade
Matrikelnummer 2723592
Page III
Danksagung
Ich danke meinen Professoren und Betreuern, die mir wahrend meines kompletten Studiums
ermoglicht haben, den Weg zu dieser Arbeit zu ebnen. Insbesondere danke ich Herrn Professor
Ulrich Walter und Herrn Dipl.-Ing. Andreas Hein von der TU Munchen sowie Herrn Professor
Olivier de Weck und Frau Dr. Afreen Siddiqi vom Massachusetts Institute of Technology
fur die Moglichkeit am MIT forschen zu durfen wie auch fur die fachliche und personliche
Unterstutzung.
Ein ganz besonderer Dank gilt meiner Familie und meinen Freunden fur ihre bedingungslose
Unterstutzung durch mein Studium hindurch und daruber hinaus. Vor allem danke ich meiner
Mutter von ganzem Herzen. Diese Arbeit sei ihr gewidmet.
Page IV
Zusammenfassung
In dieser Arbeit wird die Dimensionsanalyse als leistungsstarke Methode fur den Entwurf
von Satelliten vorgestellt. Diese Anwendung basiert auf der Annahme, dass mit Hilfe der
Ahnlichkeitstheorie, die sich dimensionsloser Kennzahlen aus der Dimensionsanalyse bedient,
Satelliten schneller und damit kostengunstiger entworfen werden konnen als es die gebrauch-
lichen Methoden erlauben. Ein weiterer Vorteil der Konzeption von Satelliten mit dimensions-
losen Großen ist die mogliche Uberprufung, ob ein Design sich in den richtigen Großenordnun-
gen bewegt. Außerdem wird gezeigt, dass die Dimensionsanalyse auch verwendet werden
kann, um Satellitenstandards in Ihrer Missionsperformance zu untersuchen und damit auch
zu uberdenken indem in einem neuartigen Ansatz physikalische Charakeristiken des Satelliten
mit der erreichbaren Missionsperformance des Raumflugkorpers in Beziehung gesetzt werden.
Ein besonderer Schwerpunkt der Arbeit liegt auf der Anwendung der theoretischen Ergebnisse
auf den CubeSat-Standard da diese Satellitenklasse an Bedeutung gewinnt, jedoch nur wenige
systemspezifische Ergebnisse bis jetzt vorhanden sind.
Nach einer Einfuhrung in die theoretischen Grundbegriffe dieser Arbeit, die CubeSat-spezifische
Definitionen wie auch Begriffsdefinitionen aus der Dimensionsanalyse aufgreift, folgt die Vorstel-
lung des Buckingham-Π-Theorems. Diese vor allem in der Aerodynamik verwendete Methode
aus der Dimensionsanalyse wird in dieser Arbeit genutzt um den Satelliten mit dimensionslosen
Kennzahlen zu modellieren. Die Anwendung der theoretischen Grundlagen auf den Satellite-
nentwurf erfolgt mit zwei verschiedenen Ansatzen. Der Top-Down-Ansatz bedient sich als
Ausgangspunkt nutzlastspezifischer Parameter, wohingegen der Bottom-Up-Ansatz Charak-
teristiken des Satelliten als Eingangsgroßen fur den Satellitenentwurf verwendet. Als Einstieg
in die Dimensionsanalyse werden zunachst im Satellitenentwurf gebrauchliche Verhaltnisse an-
hand von Daten von nicht-geosynchronen Kommunikationssatelliten validiert. Anschließend
werden dimensionslose Kennzahlen fur einen Top-Level Ansatz und den Entwurf von drei ex-
emplarischen Subsystemen und einer optischen Nutzlast hergeleitet. Die Analysen munden
in der Erstellung eines sogenannten Missionsperformanceindex, welcher die Quantifizierung
des Missionserfolgs eines gegebenen Satelliten ermoglicht. Eine Anwendung der theoretis-
chen Ergebnisse anhand von drei verschiedenen hypothetischen Satelliten veranschaulicht ab-
schließend das praktische Vorgehen mit dimensionslosen Kennzahlen. Insbesondere der dritte
Beispielfall, ein 2U CubeSat, verdeutlicht das Potential der Ergebnisse dieser Arbeit, namlich
die Moglichkeit Satellitenstandards auch in Ihrer Missionsleistung zu untersuchen und sie da-
raufhin gegebenenfalls zu andern.
Dass die Dimensionsanalyse vielseitig einsetzbar ist, zeigt sich anschließend in einer kurzen
Einfuhrung in die potentielle Anwendung der Ergebnisse auf Satellitencluster und der damit
verbundenen Moglichkeit anhand der Dimensionsanalyse die Missionsperformance von Clus-
Page V
Zusammenfassung
terarchitekturen mit jenen von einzelnen Satelliten zu vergleichen. Der Appendix mit einer
Detaillierung der Berechnung der Eklipsenzeit veranschaulicht schließlich beispielhaft, dass die
Modellierung des Satelliten in dieser Arbeit nicht die vollstandige Komplexitat der Wirkungszu-
sammenhange der einzelnen Parameter in der Realitat darstellt und folglich weiterentwickelt
werden muss. Eine kurze Beschreibung einer im Rahmen dieser Arbeit erstellten Datenbank,
die die momentan auf dem Markt erhaltlichen CubeSat-spezifischen Standardkomponenten
zusammenfasst, schließt die Diplomarbeit ab.
Page VI
Abstract
In this work the powerful technique of dimensional analysis is used to benefit from the simi-
larity of spacecrafts and missions to facilitate and accelerate the design of newly developed
satellites. By means of the Buckingham-Π-Theorem, a widely employed method in dimen-
sional analysis, ratios and non-dimensional similarity parameters have been identified in this
work which formalize and facilitate the comparison between the characteristics of satellites of
different sizes. As a consequence new satellite designs can be based on results from former
missions and existing designs can be verified. Furthermore it will be shown that dimensional
analysis can be used in a newly developed approach to examine the mission performance of a
satellite and therefore provide a method for the verification of satellite standards by relating
physical characteristics of the spacecraft to the achievable mission performance of the satellite.
A special emphasis during the work is put on the application of the theoretical results on the
CubeSat standard as this satellite class is gaining more and more significance but only few
systems engineering guidelines are developed for them.
After the presentation of the theoretical basis of this work, including an introduction to the
CubeSat standard, a presentation of the most common notions in dimensional analysis and
an overview of the history and applications of dimensional analysis is given. Subsequently, an
introduction to the Buckingham-Π-Theorem with an exemplary application in aerodynamics
is presented. The theoretical basis will be applied to two different design approaches: in the
Top-Down approach payload specific quantities will be the input parameters for the satellite
design whereas the Bottom-Up approach uses spacecraft specific variables as point of depar-
ture for the satellite design. The analysis firstly consists of the validation of ratios common in
satellite design by means of data from non-geosynchronous communication satellites. After-
wards, non-dimensional parameters for a top-level approach, three exemplary subsystems and
an optical payload are derived. These results are then used to create the Mission Performance
Index which enables the quantification of the mission accomplishment of a satellite. The the-
oretical results are finally applied on three different hypothetical satellites. Especially the third
example, a 2U CubeSat, shows the potential of the results of this work, namely the possibility
to quantify the mission performance of satellite standards in order to optimize them.
A short introduction to the application of dimensional analysis on cluster architectures is given
afterwards, proving that dimensional analysis can be used to develop a possibility to compare
single satellite missions with cluster architectures. Finally, the appendix deals with an accu-
rate calculation of the eclipse time, showing indirectly the complexity of the interdependencies
between the various quantities in the system and the need for further detailed development
of the theoretical results. The work finishes with a brief presentation of a CubeSat specific
COTS component database, which was also created during this work.
Page VII
Contents
1. Introduction 1
2. Theoretical Basis 4
2.1. CubeSat-Specific Definitions and Standards . . . . . . . . . . . . . . . . . . . 4
2.2. International Vocabulary in Metrology . . . . . . . . . . . . . . . . . . . . . . 6
2.3. Standardized Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.4. Dimensional Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4.1. Buckingham-Π-Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4.2. Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4.3. Advantage of the Use of Dimensional Analysis during the Design Phase
of Satellites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4.4. Limits of the Buckingham-Π-Theorem and Dimensional Analysis . . . . 17
3. Dimensional Analysis of a Single Satellite 19
3.1. Design Approaches: Top-Down and Bottom-Up . . . . . . . . . . . . . . . . . 19
3.2. Proceeding of the Validation of the Non-Dimensional Parameters . . . . . . . 21
3.3. Ratios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.3.1. Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.3.2. Volume: Packing Factor . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.3.3. Power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.3.4. Mass and Power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.4. Non-Dimensional Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.4.1. A Top-Level Approach . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.4.2. Subsystem Power: Battery and Solar Array . . . . . . . . . . . . . . . 63
3.4.3. Subsystem AOCS: Reaction Wheel . . . . . . . . . . . . . . . . . . . 70
3.4.4. Subsystem Communication . . . . . . . . . . . . . . . . . . . . . . . 75
3.4.5. Payload . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.5. Mission Performance Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.6. Application of the Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
3.6.1. The CubeSat - a Bottom-Up Approach . . . . . . . . . . . . . . . . . 90
3.6.2. The NGSO-Satellite - a Top-Down Approach . . . . . . . . . . . . . . 96
3.6.3. A New CubeSat Standard? - a Bottom-Up Approach . . . . . . . . . . 103
4. Dimensional Analysis of Clusters 112
4.1. The F6-Project and the Concept of Fractionated Spacecrafts . . . . . . . . . . 112
Page IX
Contents
4.2. Ratios and Non-Dimensional Parameters . . . . . . . . . . . . . . . . . . . . 113
4.3. Mission Performance Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
5. Conclusion and Future Prospects 117
A. Calculation of the Eclipse Time tEclipse 118
B. CubeSat-Specific Commercial off-the-Shelf Component Database 121
Page X
List of Figures
1.1. Thesis roadmap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1. Artistic view of a 1U CubeSat in orbit . . . . . . . . . . . . . . . . . . . . . . 4
2.2. 3U P-POD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.1. Top-Down design approach . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2. Bottom-Up design approach . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3. Mass ratios of NGSO-satellites . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.4. Mass ratios of NGSO-satellites plotted against mS/Cdry. . . . . . . . . . . . . 26
3.5. Mass ratios of NGSO-satellites plotted against mS/Cdryexcluding
mS/Cwet
mS/Cdry
. . . 27
3.6. Mass ratios of NGSO-satellites with mS/Cdryabove 1000 kg . . . . . . . . . . 27
3.7. Mass ratios of NGSO-satellites with mS/Cdryunder 300 kg . . . . . . . . . . . 28
3.8. Histogram formS/Cwet
mS/Cdry
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.9. Histogram formP/L
mS/Cdry
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.10. Histogram for mAOCS
mS/Cdry
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.11. Histogram for mStructure
mS/Cdry
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.12. Histogram for mPower
mS/Cdry
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.13. Histogram for mThermal
mS/Cdry
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.14. Histogram for mC&DH+TT&C
mS/Cdry
. . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.15. Histogram formPropulsion
mS/Cdry
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.16. 1U CubeSat MOVE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.17. Exploded view of the CubeSat MOVE . . . . . . . . . . . . . . . . . . . . . . 32
3.18. mStructure
mS/Cdry
for hypothetical CubeSats, calculated based on data from COTS-
components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.19. mStructure
mS/Cdry
for hypothetical CubeSats, calculated based on Aluminium wall as-
sumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.20. VStructure
VS/Cfor hypothetical CubeSats, calculated based on Aluminium wall as-
sumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.21. Histogram formS/Cwet
mP/L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.22. Histogram formS/Cdry
mP/L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.23. Histogram for mAOCS
mP/L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.24. Histogram for mStructure
mP/L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.25. Histogram for mPower
mP/L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Page XI
List of Figures
3.26. Histogram for mThermal
mP/L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.27. Histogram for mC&DH+TT&C
mP/L. . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.28. Histogram formPropulsion
mP/L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.29.PP/L
PSAeol
and PBus
PSAeol
plotted against PSAeol. . . . . . . . . . . . . . . . . . . . . 43
3.30.PP/L
PSAeol
and PBus
PSAeol
plotted against PSAeolwithout outlier . . . . . . . . . . . . . 43
3.31.PP/L
PSAeol
and PBus
PSAeol
plotted against mS/Cdry. . . . . . . . . . . . . . . . . . . . 44
3.32.PSAeol
PP/Land PBus
PP/Lplotted against mS/Cdry
without outliers . . . . . . . . . . . . 44
3.33. Histogram forPP/L
PSAeol
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.34. Histogram for PBus
PSAeol
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.35. Histogram forPSAeol
PP/L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.36. Histogram for PBus
PP/L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.37.mS/Cdry
mP/L· PSAeol
PP/Lplotted against mS/Cdry
. . . . . . . . . . . . . . . . . . . . . 48
3.38.mP/L
mS/Cdry
· PP/L
PSAeol
plotted against mS/Cdry. . . . . . . . . . . . . . . . . . . . . 49
3.39.mS/Cdry
mP/L· PP/L
PSAeol
plotted against mS/Cdry. . . . . . . . . . . . . . . . . . . . . 49
3.40.mP/L
mS/Cdry
· PSAeol
PP/Lplotted against mS/Cdry
. . . . . . . . . . . . . . . . . . . . . 49
3.41. Histogram formS/Cdry
mP/L· PSAeol
PP/L. . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.42. Histogram formP/L
mS/Cdry
· PP/L
PSAeol
. . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.43. Histogram formS/Cdry
mP/L· PP/L
PSAeol
. . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.44. Histogram formP/L
mS/Cdry
· PSAeol
PP/L. . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.45.mS/Cwet
mP/L· PS/Ceol
PP/Lplotted against mS/Cwet . . . . . . . . . . . . . . . . . . . . 51
3.46.mP/L
mS/Cwet
· PP/L
PS/Ceol
plotted against mS/Cdry. . . . . . . . . . . . . . . . . . . . . 51
3.47.mS/Cwet
mP/L· PP/L
PS/Ceol
plotted against mS/Cwet . . . . . . . . . . . . . . . . . . . . 51
3.48.mP/L
mS/Cwet
· PS/Ceol
PP/Lplotted against mS/Cwet . . . . . . . . . . . . . . . . . . . . 52
3.49. Histogram formS/Cwet
mP/L· PS/Ceol
PP/L. . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.50. Histogram formP/L
mS/Cwet
· PP/L
PS/Ceol
. . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.51. Histogram formS/Cwet
mP/L· PP/L
PS/Ceol
. . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.52. Histogram formP/L
mS/Cwet
· PS/Ceol
PP/L. . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.53. mS/Cwet plotted against tOrbit . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.54. PSAeolplotted against tOrbit . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.55. ρS/C plotted against tOrbit . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.56. Histogram for Π1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.57. Histogram for Π1 for the NGSO-satellites in LEO . . . . . . . . . . . . . . . . 58
3.58. Π∗3 plotted against mS/Cwet . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.59. Zoom of Π∗3 plotted against mS/Cwet . . . . . . . . . . . . . . . . . . . . . . 59
Page XII
List of Figures
3.60. Π∗3 plotted against tOrbit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.61. Zoom of Π∗3 plotted against tOrbit . . . . . . . . . . . . . . . . . . . . . . . . 59
3.62. Histogram for Π∗3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.63. Histogram for Π∗3 for the NGSO-satellites in LEO . . . . . . . . . . . . . . . . 60
3.64. Π∗4 plotted against mS/Cwet . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.65. Zoom of Π∗4 plotted against mS/Cwet . . . . . . . . . . . . . . . . . . . . . . 62
3.66. Π∗5 plotted against mS/Cwet . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.67. Histogram for Π∗4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.68. Histogram for Π∗5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.69. Top-Down network flow diagram for an Earth Observation mission . . . . . . . 110
3.70. Bottom-Up network flow diagram for an Earth Observation mission . . . . . . 111
B.1. Screenshot of a part of the AOCS-sheet of the CubeSat COTS component
database . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Page XIII
List of Tables
2.1. The characteristics of the most common CubeSats . . . . . . . . . . . . . . . 5
2.2. SI base quantities, units and dimensions . . . . . . . . . . . . . . . . . . . . . 9
3.1. xav, s and sxav
for Bottom-Up mass ratios - Part 1 . . . . . . . . . . . . . . . 25
3.2. xav, s and sxav
for Bottom-Up mass ratios - Part 2 . . . . . . . . . . . . . . . 26
3.3. xav for mass ratios for smaller NGSO-satellites, a hypothetical 1U CubeSat and
real flown/planned CubeSats - Part 1 . . . . . . . . . . . . . . . . . . . . . . 30
3.4. xav for mass ratios for smaller NGSO-satellites, a hypothetical 1U CubeSat and
real flown/planned CubeSats - Part 2 . . . . . . . . . . . . . . . . . . . . . . 31
3.5. xav, s and sxav
for Bottom-Up mass ratios with mS/Cwet - Part 1 . . . . . . . . 35
3.6. xav, s and sxav
for Bottom-Up mass ratios with mS/Cwet - Part 2 . . . . . . . . 35
3.7. xav, s and sxav
for Top-Down mass ratios - Part 1 . . . . . . . . . . . . . . . 36
3.8. xav, s and sxav
for Top-Down mass ratios - Part 2 . . . . . . . . . . . . . . . 36
3.9. xeq for standard CubeSats . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.10. xav, s and sxav
for power ratios . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.11. xav, s and sxav
for power ratios excluding outliers . . . . . . . . . . . . . . . . 42
3.12. xav, s and sxav
for rMass Power 1 to rMass Power 4 with mS/Cdry. . . . . . . . . 47
3.13. xav, s and sxav
for rMass Power 1 to rMass Power 4 with mS/Cdrywithout outlier . 47
3.14. xav, s and sxav
for rMass Power 1 to rMass Power 4 with mS/Cwet . . . . . . . . . 47
3.15. xav, s and sxav
for rMass Power 1 to rMass Power 4 with mS/Cwet without outlier . 47
3.16. xav, s and sxav
for Π1 for all NGSO-satellites and those in LEO . . . . . . . . 57
3.17. xav, s and sxav
forPS/C
VS/C23mS/Cwet
for all NGSO satellites and those in LEO . . . 57
3.18. xav, s and sxav
for Π∗4, Π∗5 and Π∗6 . . . . . . . . . . . . . . . . . . . . . . . . 62
3.19. Top-level Mission Performance Parameters . . . . . . . . . . . . . . . . . . . 82
3.20. Ranges for the subsystem exponents α, β, χ, . . . for single satellites . . . . . . 88
3.21. Mass ratios for the hypothetical 1U CubeSat - Part 1 . . . . . . . . . . . . . 91
3.22. Mass ratios for the hypothetical 1U CubeSat - Part 2 . . . . . . . . . . . . . 91
3.23. Mass ratios for the hypothetical NGSO-satellite - Part 1 . . . . . . . . . . . . 97
3.24. Mass ratios for the hypothetical NGSO-satellite - Part 2 . . . . . . . . . . . . 97
3.25. Mass ratios for the hypothetical 2U CubeSat - Part 1 . . . . . . . . . . . . . 104
3.26. Mass ratios for the hypothetical 2U CubeSat - Part 2 . . . . . . . . . . . . . 104
4.1. Ranges for the subsystem exponents α, β, χ, . . . for cluster architectures, in-
cluding tendency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
Page XV
Symbols
AS/C m2 spacecraft surface area
b bitspixel
number of bits to encode each pixel
BER − bit error rate
c ms
velocity of sound
CBattery Ws capacity of the battery
cSolarWm2 solar constant
d 1yr
degradation per year
D m aperture diameter of payload instrument
DAntenna m antenna diameter
DTransmitter m transmitter antenna diameter
DReceiver m receiver antenna diameter
e ◦ pointing loss
ECA ◦ Earth Central Angle
f 1s
frequency of communication signal
GAntenna − antenna gain
GParaboleAntenna − gain of parabole antenna
hOrbit m orbital altitude
HRWkgm2
sangular momentum of reaction wheel
i − number of non-dimensional products in a system
i ◦ orbit inclination
IRW kgm2 reaction wheel moment of inertia
IS/C kgm2 spacecraft moment of inertia
k − number of quantities in the k-set of a system
ki variable part of a non-dimensional parameter
l m distance between two points
Page XVII
Symbols
Li − communication losses
Ma − Mach number
mi kg mass
n − number of quantities in a system
p % packing factor
Pi W power consumption or production
PTransmitter W transmitter output power
Qi variable quantity
r − ratio
R bitss
data rate
REarth m radius of the Earth
Re − Reynolds number
s − standard deviation
SCom m communication path length
SSolarWm2 product of cSolar and cos(θ)
SW m swath width
tOrbit s orbital period
tSatellite s satellite time
tEclipse s eclipse time
tDaylight s daylight time
TAero Nm aerodynamic torque
TSolar Nm solar radiation torque
TGravity Nm gravity torque
TOperating K operating temperature
Vi m3 volume
vims
velocity
X mpixel
cross-track ground pixel resolution
x mm dimension of a CubeSat along the x-axis
xav variable arithmetical average
xdeployed m equivalent edge length of spacecraft with deployed devices
xeq m equivalent edge length
Y mpixel
along-track ground pixel resolution
y mm dimension of a CubeSat along the y-axisPage XVIII
Symbols
z mm dimension of a CubeSat along the z-axis
βs◦ incident sun angle on the orbital plane
ε ◦ elevation angle
η − antenna efficiency
λi m wavelength
µ m3
s2Earth’s gravity constant
ωRW1s
angular velocity of reaction wheel
Φ ◦ angular of the eclipse from the orbit
Πi - non-dimensional parameter
Ψi variable component of the Mission Performance Index
ρ ◦ angular radius of the Earth
ρikgm3 mass density
θ ◦ average incident sun angle on the spacecraft surface
Page XIX
Abbreviations
AOCS Attitude and Orbit Control System
BER Bit Error Rate
BOL Begin of Life
C&DH Command and Data Handling
Cal Poly California Politechnic State University
CDS CubeSat Design Specification
CGPM Conference General de Poids et Mesures (French for General Confer-
ence on Weights and Measures
Com Communication
COTS Commercial Off-The-Shelf
DARPA Defense Advanced Research Projects Agency
DOD Depth of Discharge
ECA Earth Central Angle
EOL End of Life
EOM Earth Observation mission
ESA European Space Agency
Fig Figure
GEO Geostationary Orbit
IEC International Electrotechnical Commission
ISO International Organisation for Standardization
ISQ International Systems of Quantities
LEO Low Earth Orbit
LRT Lehrstuhl fur Raumfahrttechnik
MIT Massachusetts Institute of Technology
MPI Mission Performance Index
Page XXI
Abbreviations
MPP Mission Performance Parameter
NASA National Aeronautics and Space Administration
NGSO Non-Geosynchronous Orbit
P/L Payload
POD Picosatellite Orbital Deployer
Prop Propulsion
SA Solar Array
S/C Spacecraft
SI Systeme International d’unites (French for International System of
Units)
SSM Space Science mission
SSDL Space Systems Development Laboratory
Tab Table
TDM Technology Demonstration mission
TT&C Telemetry, Tracking and Command
TUM Technische Universitat Munchen
VIM Vocabulaire international des termes fondamentaux et generaux de
metrologie (French for International vocabulary of basic and general
terms in metrology)
U Unit
Page XXII
1. Introduction
In this work the powerful technique of dimensional analysis is used to benefit from the sim-
ilarity of spacecrafts and missions to facilitate and accelerate the design of newly developed
satellites. By means of the Buckingham-Π-Theorem, a widely employed method in dimen-
sional analysis, ratios and non-dimensional similarity parameters have been identified in this
work which formalize and facilitate the comparison between the characteristics of satellites of
different sizes. As a consequence new satellite designs can be based on results from former
missions and existing designs can be verified. It will be shown that dimensional analysis can
be used to examine the mission performance of a satellite and therefore provide a method
for the verification of satellite standards by relating physical characteristics of the spacecraft,
like its size, to the achievable mission performance of the satellite - this being a connection
which have been not implemented yet in current design approaches. Two approaches will be
in the focus of this work: the results of the payload-centric Top-Down approach will enable
the engineer to predict the design characteristics of a satellite correspondingly to a required
Mission Performance Index. In the spacecraft-centric Bottom-Up approach the mission ac-
complishment of a satellite with set design characteristics will be predicted by means of an
achievable Mission Performance Index. Consequently, the Mission Performance Index will also
be an assessment of the efficiency of satellite classes.
A special emphasis during the work is put on the application of the theoretical results on the
CubeSat standard as this satellite class is gaining more and more significance but only few
systems engineering guidelines are developed for them. Especially the efficiency of the CubeSat
standard will be in the focus of this work as the efficiency of this satellite concept is yet not
investigated.
Within the last 20 years CubeSats have established themselves in the educational program
of many universities worldwide, in order to provide its students hands-on-experience and to
offer universities and industries a low-cost and quickly designed solution to test their new
technologies and hardware components under real conditions. Also companies are showing
increasing interest in picosatellites, trying to exploit their enormous potential (e.g. QuakeSat
of Quakefinder launched in 2003, AeroCube 2 of The Aerospace Corporation, CSTB from
Boeing and MAST of Tethers Unlimited Inc., all last three launched in 2007). Conferences
and workshops around the world are hold to debate over the use of CubeSats and small satel-
lites (in 2010 i.a.: the 7th Annual CubeSat Developers’ Workshop by Cal Poly within the
24th Small Satellite Conference in Utah, USA, the 3rd European CubeSat Workshop initiated
by ESA in Noordwijk, Netherlands) and even the market is reacting to the increasing need
of CubeSat specified commercial off-the-shelf components, accelerating the development of
CubeSats even more by creating new businesses which offer only CubeSat related products.
Page 1
Currently 38 CubeSats were successfully launched of which 19 are still operational (two of them
since 2003) and two other launches are upcoming: Taurus XL is planned to be launched in
February 2011, bringing three more CubeSats in position and ESA’s new launch rocket VEGA
is supposed to carry eleven CubeSats into orbit on its maiden flight mid 2011. Furthermore,
NASA recently announced the second round of its CubeSat launch initiative, enabling more
CubeSat missions in 2011 and 2012 [NASA, 2010]. Another approximately 70-100 missions
are planned throughout the institutions of the world in the next years, above all the campaign
QB50 by the Von Karman Institute, ESA and NASA [Institute, 2010].
However, the mission efficiency of the CubeSat standard was not yet sufficiently questioned.
Consequently, this work tries to analyze the efficiency of the CubeSat standard with the help
of the Mission Performance Index and gives guidelines to a possible new CubeSat standard
which combines all the advantages of the current CubeSats as their cost-, size- and mass-
effectiveness but also a higher Mission Performance Index.
Furthermore, the work can be also used to create a design tool based on dimensional anayl-
sis which accelerates the design of CubeSats. Some related work has been already done by
[Aas et al., 2009a], [Aas et al., 2009b], however, not by means of dimensional analysis and not
providing a possibility for the quantification of the mission performance of a satellite either.
A method for the comparison of single satellite missions and cluster architectures in terms of
their mission performance is also believed to be not yet developed. First answers to this open
research questions are given in this work.
After the introduction, chapter 2 provides the theoretical basis of this work, including an
introduction to the CubeSat standard in 2.1, a presentation of the most common notions
in dimensional analysis in 2.2 and 2.3 and an overview of the history and applications of
dimensional analysis in 2.4. An introduction to the Buckingham-Π-Theorem in 2.4.1 with
an exemplary application in aerodynamics is given in 2.4.2, as well as an assessment of the
theorem’s assets and drawbacks in satellite design in 2.4.3 and 2.4.4. Chapter 3 presents
the results of the dimensional analysis for satellite design by means of ratios in 3.3, non-
dimensional parameters for the spacecraft and its subsystems in 3.4 as well as the Mission
Performance Indices in 3.5. Afterwards, the theoretical results are applied on three different
hypothetical satellites in 3.6. A short introduction to the application of dimensional analysis
on cluster architectures is given in chapter 4, proving the power of dimensional anaylsis based
on its various application areas. A conclusion of the work, based especially on the results in
chapter 3, is given in chapter 5. Finally, Appendix A deals with an accurate calculation of the
eclipse time, showing indirectly the complexity of the interdependencies between the various
quantities in a system. The work finishes with Appendix B which provides a brief presentation
of a CubeSat specific COTS component database, which was also created during this work.
Page 2
Introduction
Figure 1.1.: Thesis roadmap
Page 3
2. Theoretical Basis
2.1. CubeSat-Specific Definitions and Standards
The CubeSat standard originates from the efforts of the California Polytechnic State Univer-
sity, San Luis Obispo (Cal Poly) and the Space Systems Development Laboratory (SSDL) at
Stanford University in 1999 within the CubeSat program in order to develop a new class of
satellites [Heidt et al., 2001], [Toorian et al., 2005].
In its standard configuration the satellite measures 100.00 mm x 100.00 mm x 113.50 mm
with a weight of up to 1.33 kg and notably uses commercial off-the-shelf (COTS) electrical
components. This configuration is known as 1 unit CubeSat or 1U CubeSat. Other built
and launched standardized configurations are 2U (100.00 mm x 100.00 mm x 227.00 mm,
mass up to 2.66 kg) and 3U (100.00 mm x 100.00 mm x 340.50 mm, mass up to 4.00 kg)
CubeSats. Smaller formats like 0.5U CubeSats and larger ones (4U, 5U and 6U) are also
possible. The CubeSat Design Specification (CDS) [CalPoly, 2009] defines all these standards
by stating the nominal dimensions of the standard 1U CubeSat, dimension tolerances, the
reference coordinate system, acceptable materials and other information.
Please note that the presented density of the CubeSats in table 2.1 represents the theoretical
upper bound density. It is calculated assuming homogeneous mass distribution in the CubeSat,
thus
ρCubeSat =mCubeSat
VCubeSat(2.1)
with
VCubeSat = x · y · z (2.2)
where x, y and z represent the dimensions of the CubeSat.
Figure 2.1.: Artistic view of a 1U CubeSat in orbit [NASA, 2010]
Page 4
Theoretical Basis
Table 2.1.: The characteristics of the most common CubeSats
System Dimensions Mass Max. Density
mm mm mm kg kgm3
1U 100.00 100.00 113.50 1.33 1171.81
2U 100.00 100.00 227.00 2.66 1171.81
3U 100.00 100.00 340.50 4.00 1171.81
In practice, this density is never achieved and the packing factor p is of more significance for
assembly and integration of the CubeSat. A more detailed discussion about the differentiation
between ρS/C and p is given in section 3.3.2.
Figure 2.2.: 3U P-POD [Cal Poly, 2010]
A unique feature of the CubeSat program is the use of a standard deployment system: the
CubeSats are launched in standardized Picosatellite Orbital Deployers (PODs) as piggybacks
of large spacecrafts and hence achieve their orbit at greatly reduced costs. The POD is
the interface between the launch vehicle and the CubeSats. Its most important objective is
to guarantee the protection of the launch vehicle and other payloads from any mechanical,
electrical or electromagnetic interference from the CubeSats in the event of a catastrophic
CubeSat failure. Beyond that, the POD enables a simple implementation of the CubeSats
with most launch vehicles. The most popular version is the P-POD, the Poly-POD, designed
and built by Cal Poly [CalPoly, 2007]. It is a rectangular container of anodized aluminium.
After a signal is sent from the launch vehicle to the release mechanism, a spring-loaded door
opens and the CubeSats are deployed by a spring and glide along smooth flat rails into their
orbits. The most common configuration is to launch three 1U CubeSats in one P-POD.
However, also CubeSats of different lengths can be accommodated in the same P-POD (e.g.
one 1U and one 2U CubeSat) which gives the launch provider some flexibility.
Page 5
International Vocabulary in Metrology
2.2. International Vocabulary in Metrology
Before explaining the theoretical foundation of this work, namely the Buckingham-Π-Theorem,
it is important for the later comprehension to clarify some notions and expressions which will be
used in this work. Most of them are essential and widely used without further thoughts about
their definitions but are mentioned here for the sake of completeness. All citations in this sec-
tion are extracted from the International Vocabulary in Metrology (VIM) [JCGM/WG2, 2008],
the standard work for terms in metrology. One of the basic terms frequently used in this work
is quantity. According to the VIM, it is the ”property of a phenomenon, body, or substance,
to which a magnitude can be assigned”. Hence, quantities are e.g. the mass and the length of
a body. In this context a so called system of quantities is a ”set of quantities together with
a set of non-contradictory equations relating those quantities”. A sizable amount of different
systems is possible as many different combinations of quantities are thinkable. Therefore it is
always important to indicate the system of quantity which is used as it can differ from case to
case and especially from scientist to scientist. An important example for a system of quantities
is the International System of Quantities (ISQ), which will be explained in the next section.
So called base and derived quantities are part of a system of quantities. The definition of a
base quantity states that it is ”chosen by convention, used in a system of quantities to define
other quantities”. Hence, the characteristic of a base quantity is the fact that it cannot be
derived from other quantities with the help of an equation. The base quantities are therefore
independent by definition. We will see that seven independent base quantities are used in the
today’s most common unit system, the International System of Units [BIPM, 2006], from
which all other quantities can be derived with the help of the equations within the respec-
tive system of quantities, namely the ISQ. The choice of the independent base quantities is
made more or less arbitrarily with regard to the actual scientific theories in order to choose
the quantities which are the most convenient. In the end, it is up to every scientist to fix
his/her base quantities as it is possible to establish a sizable number of system of quanti-
ties. Consequently to the definition of a base quantity, a derived quantity is a ”quantity, in a
system of quantities, defined as a function of base quantities”. Velocity is an example for a de-
rived quantity in the ISQ, as it is composed of length and time, both base quantities in the ISQ.
Physical quantities are organized in a system of dimensions. A dimension is the ”dependence
of a given quantity on the base quantities of a system of quantities, represented by the product
of powers of factors corresponding to the base quantities”. The symbolic representation of
the dimension of a base quantity (in the following referred to as base dimension) is a single
upper case letter in roman (upright) sans-serif type and its denotation equates the name of
the respective base quantity (e.g. M for mass or L for length). The symbolic representation
of the dimension of a derived quantity (in the following referred to as derived dimension) is
the product of powers of the dimensions of the base quantities using the equations that relate
Page 6
Theoretical Basis
the derived quantities to the base quantities. Consequently, the dimension of any quantity Q
can be derived from the base dimensions in the following form
dimQ = D1αD2
βD3γD4
δ . . .Dnν (2.3)
with D1, . . . ,Dn as the n base dimensions in the chosen system of quantities and the expo-
nents α, β, γ, δ, . . . , ν as the n dimensional exponents. The dimensional symbols and expo-
nents are manipulated using the usual rules of algebra. This notation and the fact that every
dimension can be expressed in terms of base dimensions is a fundamental requirement for the
Buckingham-Π-Theorem. It is important to notice that quantities having the same dimension
are not necessarily quantities of the same kind. An example is given by the quantities torque
and work/heat in the ISQ where all three quantities have the same dimension ML2T−2 with
the length L, the mass M and the time T as base dimensions. As we will see later on, so called
quantities of dimension one or dimensionless quantities play a very essential role in the
Buckingham-Π-Theorem. Therefore we quote its definition here: a dimensionless quantity is
a ”quantity for which all the exponents of the factors corresponding to the base quantities in
the representation of its dimension are zero”. However, the term ’quantity of dimension one’
reflects the convention that the symbolic representation of the dimension for such quantities
is the symbol 1. Examples of dimensionless quantities are the Reynolds number Re and the
Mach number Ma, both fundamental quantities in aerodynamics which can be derived by
using the Buckingham-Π-Theorem. The derivation of the former one will be shown in section
2.4.2. In everyday life the wordings ’unit’ and ’dimension’ can be misunderstood as synonyms.
Buckingham himself did not separate the terms sharply in his statements to dimensional anal-
ysis [Buckingham, 1914], [Buckingham, 1915a], [Buckingham, 1915b]. In order to clarify the
difference between them, the definition of the term unit shall be given: a unit is a ”scalar
quantity, defined and adopted by convention, with which other quantities of the same kind are
compared in order to express their magnitudes”. Hence when talking about masses, for exam-
ple, we indirectly compare the mass of the considered object to the world’s standard kilogram
in Paris. Several units can be chosen for a quantity with a dimension (e.g. ft, inch, mile, mm,
cm, m, and km for the dimension length), but only one dimension can clearly refer to a unit.
So, when building dimensionless expressions, non-dimensionality can be also achieved when
one quantity in the equation is given in inch and the other in cm. Only a conversion factor
has to be considered in that case. However, for the sake of coherence, it is paid attention
in this work to always use the same units in numerator and denominator before cancelling
down. Another important point to notice is that conformable to the fact that the symbol 1
denotes the dimension of a dimensionless quantity, the units of quantities of dimension one
are simply numbers, e.g. radian and steradian. Pursuant to the existence of base dimensions
for base quantities, there also exist base units which are defined to be ”conventionally and
uniquely adopted for a base quantity in a given system of quantities”. We will see that all the
base quantities in the ISQ have corresponding base units which are designated in a well-known
system of units, namely the International System of Units (SI). Conformable to the definition
of a base unit, a derived unit is a ”unit for a derived quantity”. Thus the meter per second
Page 7
Standardized Systems
is a derived unit of velocity in the SI as meter and second are base units in the SI for the base
quantities length and time in the ISQ.
As we already mentioned the International System of Units as a system of units, it may be
useful to state its definition: it is a ”conventionally selected set of base units and derived units,
and also their multiples and submultiples, together with a set of rules for their use”. Multiples
and submultiples of units are simply units ”formed from a given unit by multiplying/dividing
by an integer greater than one” e.g. the kilometer is a multiple and the millimeter a submultiple
of meter.
The following definitions shall complete the list of definitions: A coherent derived unit is
a ”derived unit that, for a given system of quantities and for a chosen set of base units, is
a product of powers of base units with the proportionality factor one”. A coherent system
of units is a ”system of units, based on a given system of quantities, in which the unit for
each derived quantity is a coherent derived unit”. The value of a quantity or the value is
the ”magnitude of a quantity represented by a number and a reference”. A quantity value
can be expressed inter alia as a product of a number and a unit, or a number for a quantity
of dimension one (the unit one is generally not written out). A quantity equation is an
”equation relating quantities”. Correspondingly, an unit equation is an ”equation relating
units” and a numerical value equation or a numerical quantity value equation is an
”equation relating numerical quantity values”. In order to convert values of quantities of the
same kind but with different units conversion factors between units are necessary which
are defined as ”ratio of two units for quantities of the same kind”, e.g. ft/cm = 30,48 and
thus 1 ft = 30,48 cm. Please note finally that in literature the term ”base” is often used
synonymously with ”fundamental”. Thus fundamental units/dimensions/quantities are base
units/dimensions/quantities.
2.3. Standardized Systems
One of the fundamental requirements of our everyday communication is the fact that we express
our scientific knowledge in standardized systems of base and derived quantities, units and
dimensions. Without a standardization, not only our everyday life would be more complicated
(conversions between the European and Anglo-American unit system still produce confusion
and errors), but also the scientific communication would be harder (e.g. imagine when symbols
for quantities would not be the same at all in the different countries).
Today’s most common standardized quantity system is The International System of Quan-
tities (ISQ), the ”system of quantities, together with the equations relating the quanti-
ties, on which the SI is based”. Since November 2009 the ISQ is published in the In-
ternational Standard ISO 80000 or IEC 80000 - formerly ISO 31 and partially IEC 60027
- [ISO-TC12 and IEC-TC25, 2009], depending on which of the two international standards
Page 8
Theoretical Basis
bodies, either the International Organization for Standardization or the International Elec-
trotechnical Commission, is responsible for each respective part. The base quantities used
in the ISQ are length, mass, time, electric current, thermodynamic temperature, amount of
substance, and luminous intensity.
Today’s most common unit system is The International System of Units (or SI from its
French name Systeme International d’unites) which was already mentioned in the definition of
the ISQ above. It is a ”coherent system of units based on the ISQ, their names and symbols,
and a series of prefixes and their names and symbols, together with rules for their use, adopted
by the General Conference on Weights and Measures (CGPM)”. The system is not static but
evolves in terms of technological advances in scientific measurement [BIPM, 2006]. Therefore
additions and changes are done regularly. With regard to the base quantities in the ISQ, the
CGPM chose the base units of the SI to be the meter, the kilogram, the second, the ampere,
the kelvin, the mole, and the candela. Finally, each of the seven base quantities chosen for the
ISQ has its own dimension. The symbols used to signify the dimensions are L, M, T, I, θ,
N and J. Table (2.2) summarizes the seven base quantities with their respective base units
and base dimensions [BIPM, 2006].
Table 2.2.: The seven base quantities as defined in the SI [BIPM, 2006] with their respective
base units and base dimensions
Base Quantity Base Unit Base Dimension
Name Symbol Name Symbol Name Symbol
length l meter m length L
mass m kilogram kg mass M
time t second s time T
electric current I, i ampere A electric current I
thermodyn. temp. T kelvin K thermodyn. temp. θ
amount of substance n mole mol amount of substances N
luminous intensity IV candela cd luminous intensity J
Please note that although the seven base quantities are by convention seen as independent,
their respective base units are in a number of instances interdependent. Thus the definition
of the meter incorporates the second; the definition of the ampere incorporates the meter,
kilogram, and second; the definition of the mole incorporates the kilogram; and the definition
of the candela incorporates the meter, kilogram, and second. In the following the analyses
will be done essentially within the ISQ and the SI. However, it is also possible to carry out the
investigations in this work with a different system of quantities and units. Correspondingly the
choice of base units is also dependent on each engineer or scientist: it is also true to choose
Page 9
Dimensional Analysis
e.g. gram instead of kilogram as a base unit for the base quantity mass. Thus, whenever
another system as the SI and ISQ is used in this work, it will be explicitly mentioned.
2.4. Dimensional Analysis
The basic ideas of the dimensional analysis, as they are used today, date back to the 19th
century: the French mathematician Joseph Fourier spread the idea that physical equations
should be dimensional homogeneous equations in their different units of measurement in his
work Theorie Analytique de la Chaleur of 1822 [Buckingham, 1914], [Macagno, 1971]. Hence,
the dimensional analysis is still one important and very simple method to check the plausibility
of physical equations: the two terms of an equation have to have the same dimension to be
valid in terms of a first plausibility check. Or, as expressed in [Langhaar, 1951] ”an equation
will be said to be dimensionally homogeneous if the form of the equation does not depend on
the fundamental units of measurement”.
However, Fourier failed to see the connection between dimensional analysis and similarity. It
was up to the English physicist and mathematician Lord Rayleigh to develop a method to
the principle of Fourier in his work Theory of Sound from 1877/8 [Strutt and Rayleigh, 1894].
Even dimensionless groups were found by Rayleigh, like the Reynolds number Re in its inverse
form.
In continental Europe, the French physicist and mathematician Aime Vaschy achieved a more
general formulation of the theorem for the method of dimensions and came closer to a justifica-
tion of the theorem in mathematical form [Vaschy, 1892]. The Russian Dimitri Riabouchinsky
is said to have used dimensional analysis independently from the results of Vaschy and Rayleigh
[Riabouchinsky, 1911]. He also initiated a discussion about the temperature as fundamental
unit with Lord Rayleigh and Edgar Buckingham after the publication of an article by Lord
Rayleigh which was intended to stimulate the use of his method of dimensions among engi-
neers [Buckingham, 1915b], [Rayleigh, 1915].
Buckingham himself is said to have adopted the idea of the method of dimensional analysis
from his antecessors, so the theorem should not only be cited after him. Correspondingly, the
theorem is often cited as Vaschy-Buckingham-Theorem in France. However, Buckingham is
the only one, who very clearly pointed out that dimensional analysis has two further interesting
results besides the homogeneity of equations which are explained in the following.
A very important application of dimensional analysis can be often found in scientific situations
where characteristics of nearly unknown phenomena are sought. Here dimensional analysis
helps to state hypotheses about the influence of quantities of a complex system on the at-
titude and performance of the system, without knowing the exact relation between these
quantities. On top of that, even exact equations, namely scaling laws, can be found by di-
mensional analysis with some contribution of experimental results or physical reasoning.
One of the most famous examples of dimensional reasoning was in real not due to the method
of dimensional analysis. Geoffrey Taylor is to be said to have used dimensional analysis in
1950 to easily determine the discharged energy by stating a formula relating the radius r
Page 10
Theoretical Basis
of a spherical blast wave produced by the release of a quantity of energy E, at a point
in the air of density ρAir 0 and polytropic index γ to the two-fifths power of the time t
[Taylor, 1950a], [Taylor, 1950b]. In reality, fastidious calculations with the equations of motion
and the Rankine-Hugoniot equations as starting point guided him to the following equation:
r = t25
(E
ρAir 0
) 15
f(γ) (2.4)
Nevertheless, this equation can easily be reproduced by the method of dimensional anal-
ysis.Genuine examples of dimensional analysis include the most common ones in aerody-
namics/fluid dynamics [Riabouchinsky, 1911], [Anderson, 2007] and heat and mass trans-
fer [Rayleigh, 1915], [Richardson, 1919], electrotechnology [Vaschy, 1892], chemical process-
ing [Zlokarnik, 2006], mechanics [Vaschy, 1892], [Sedov, 1993] and biology [Stahl, 1961a],
[Stahl, 1961b], [Tennekes, 1996], [McMahon and Bonner, 1983]. But also not so evident sci-
entific areas as economy [de Jong, 1967] and psychology [Lehman and Craig, 1963] use di-
mensional analysis.
References for literature that deals with dimensional analysis in general are [Bridgman, 1948],
[Langhaar, 1951], [de St. Q. Isaacson and de St. Q. Isaacson, 1975], [Taylor, 1974] and also
[Sedov, 1993]. They give various examples from biology over cosmology to the classical fields
of fluid mechanics and heat transfer. The latter reference [Bridgman, 1948] is often referred
to as the first work which outlines the Π-theorem in a book with various examples.
Finally, Buckingham also stated clearly in his explanations the significant role of dimensional
analysis in the similitude theory. It is in fact especially frequently used in aerodynamics and
thermodynamics: instead of taking measurements of real sized objects, the similitude theory
states that results of a real sized object and a smaller corresponding model are the same under
the condition that the so called similarity parameters of the investigated system are the same
in reality and in the modeled situation. Similarity parameters, in turn, can be found with
dimensional analysis. Hence, this approach makes wind tunnels for real sized airplanes, for
example, nearly dispensable for most of the investigated problems.
In this context a very useful side effect of the similarity parameters and the dimensional anal-
ysis has to be mentioned which is especially useful when series of experiments are done: the
number of the influential quantities in the investigated system is reduced so that time and
effort for the series as well as the amount of data are significantly reduced since the number
of quantities which have to vary during the series, is reduced.
Another purpose of dimensional analysis which was not mentioned yet, is to quickly check the
order of magnitude of quantities and thus to see if a design is plausible or not. This is mainly
done by comparing the numerical value of non-dimensional parameters of similar designs. The
values are supposed to be in the same range for plausibility. The theory of the non-dimensional
parameters will be presented in detail in the next section 2.4.1.
Page 11
Dimensional Analysis
2.4.1. Buckingham-Π-Theorem
The Buckingham-Π-Theorem is a fundamental theorem in dimensional analysis named after
the American physicist Edgar Buckingham. It states that an equation written in the general
form
f(Q1, Q2, . . . , Qn) = 0 (2.5)
where the n symbols Q1, . . . , Qn denote physical quantities of n different kinds with k inde-
pendent dimensions can be rewritten as
F (Π1,Π2, . . . ,Πi) = 0 (2.6)
where the Πs represent i = n− k independent dimensionless products of the form
Πi = Qa11 Q
a22 . . . Qan
n (2.7)
The use of Π as a dimensionless product was introduced by Edgar Buckingham in his original
paper on the subject in 1914 [Buckingham, 1914] from which the theorem draws its name.
However, in order to do justice to the historical facts/origins, the theorem should instead be
called Vaschy- or Rayleigh-Theorem than Buckingham-Theorem as already seen in Chapter
2.4.
Because of the principle of dimensional homogeneity, equation (2.7) can be rewritten as
[Πi] = [Qa11 Q
a22 . . . Qan
n ] = [1] (2.8)
Since only k out of n dimensions are independent these equations can be further developed to
[Π1] = [Qα11 Q
β12 . . . Qκ1
k P1] = [1]
[Π2] = [Qα21 Q
β22 . . . Qκ2
k P2] = [1]
.................................................
[Πi] = [Qαi1 Q
βi2 . . . Qκi
k Pi] = [1]
(2.9)
with i = n − k in which the P ’s represent Qk+1, . . . , Qn, thus the quantities which are
”derived” from the base quantities Q1, . . . , Qk, also called the k-set. It is important to notice
that the base quantities with their respective base units and dimensions, as defined in the
SI-system, are not forcingly to be used as the fundamental units for the theorem. Buckingham
himself states in his explanations [Buckingham, 1914] that the dimension F, Newton as the
unit of force, could be used as a base dimension.
Hence the n− k dimensional products Π can be written as
Π1 = f1(Q1, Q2, . . . , Qk, Pk+1)
Π2 = f2(Q1, Q2, . . . , Qk, Pk+2)
................................................
Πi = fi(Q1, Q2, . . . , Qk, Pn)
(2.10)
Page 12
Theoretical Basis
thus reducing the variables of the problem from n to i. It is very important to choose the k
repeating quantities, the k-set, in a way that the quantities include all the k base dimensions
used in the problem [Anderson, 2007], [Buckingham, 1914]. The derived quantity P , however,
appears only in one Π-product.
A ”scaling law” results when the function F of equation (2.6) is determined in its nature. A
usual approach for that is a product approach following
(Π1)a1(Π2)a2 . . . (Πi)ai = c (2.11)
with a1, . . . , ai being the exponents in the function and c being a constant.
As it can occur that quantities of the same kind play a role in the system, briefly shall it be
discussed how they are to be treated during the analysis. Two different masses can for example
be considered to be significant to model a system. As Buckingham states, however, that
Q1, . . . , Qn in equation (2.5) are physical quantities of different kinds, he proposes to include
further quantities of a kind as ratios to the one which is included in the list of Q1, . . . , Qn
[Buckingham, 1914, p.345]. Equation (2.5) is thus only a reduction of the more general
equation
f ∗(Q1, Q2, . . . , Qn, r′, r′′, . . . ) = 0 (2.12)
In this equation the r’s represent ”all the independent ratios of quantities of the same kind”
[Buckingham, 1915a, p. 291] and each Π can be determined as discussed above. Ratios are
strictly speaking not considered as non-dimensional products in terms of the Buckingham-
Π-Theorem. Nevertheless, they can have an influence on the system and thus, have to be
considered as further parameters of the system, and so as further arguments in the function
F of equation (2.6). Therefore equation (2.6) can be rewritten as
F ∗(Π1,Π2, . . . ,Πn−k, r′, r′′, . . . ) = 0 (2.13)
Only when the ratios are considered to be constants of the investigated systems, can they
be neglected during the determination of the scaling laws. In Buckingham’s own expla-
nations the ratios are usually geometrical ones and considered to be of fixed value dur-
ing the analysis (e.g. dimensional analysis of the resistance of immersed bodies at moder-
ate speeds [Buckingham, 1915a, p. 271 ff.]). But as they can potentially play an essen-
tial role, they should be taken into consideration when scaling laws have to be established
[Buckingham, 1915a, p.291]. Thus equation (2.11) can be rewritten to
(Π1)a1(Π2)a2 . . . (Πi)ai(r′)b1(r′′)b2 · · · = c (2.14)
with a1, . . . , ai and b1, b2, . . . being the exponents in the function and c being again a constant.
Please note as an anticipation, that we will use another method in this work to show the
influence of the ratios on the system as we will not be able to determine scaling laws in this
work because of the lack of data: the non-dimensional parameters Πi will be directly expanded
Page 13
Dimensional Analysis
with the system influencing ratios. Examples of that proceeding will be shown in sections 3.4.2
to 3.4.5.
To finish the theoretical part of this work, it can be important to know that equation (2.6)
can be rewritten as
Π1 = G(Π2, . . . ,Πi). (2.15)
thus allowing to express the derived quantity P1 in Π1 as function of Π2, . . . ,Πi and the
residual quantities in Π1.
2.4.2. Example
One of the most well known applications of the Buckingham-Π-Theorem can be found in
the aerodynamics of a body (e.g. an airfoil) [Anderson, 2007]. Dimensional analysis is used
in order to determine the so called similarity parameters to compare the performances of the
model in a wind tunnel with the original object under real conditions. We consider a body with
a given shape and a given angle of attack and try to find out what physical quantities influence
the resulting aerodynamic force R being applied on the body. Thus, in a first step, we are
looking for the quantities Q1, . . . , Qn which are driving the performance of the investigated
system. It is up to the physical understanding about the system to identify these quantities.
We suppose R to be dependent on
• the freestream velocity V∞ [ms
]
• the freestream density ρ∞ [ kgm3 ]
• the viscosity of the fluid, represented by the dynamic viscosity coefficient µ∞ [ kgm s
]
• the size of the body, represented by the chord length c [m]
• the compressibility of the fluid, represented by the freestream speed of sound a∞ [ms
]
Thus the number of the independent quantities is n = 6 and we can write
R = g(ρ∞, V∞, c, µ∞, a∞) (2.16)
Equation (2.16) can be rewritten as follows
f(R, ρ∞, V∞, c, µ∞, a∞) = 0 (2.17)
Page 14
Theoretical Basis
which is equal to equation (2.5). In a next step, we consider the dimensions of the quantities.
Being in the SI-system, the fundamental dimensions of the problem are to be chosen as defined
in the standardized system:[R] = MLT−2
[ρ∞] = ML−3
[V∞] = LT−1
[c] = L
[µ∞] = ML−1T−1
[a∞] = LT−1
(2.18)
therefore the fundamental dimensions of this problem are mass M, length L and time T, thus
k = 3. As seen in section 2.4.1 the number of dimensionless products Πi can be determined
as i = n − k = 6 − 3 = 3. We choose the repeating quantities Q1, Q2 and Q3 to be ρ∞,
V∞ and c. Another choice of parameters is possible as long as the rules as discussed in 2.4.1
are followed. Note that the resulting aerodynamic force R is not a repeating quantity as we
defined it to be one of the dependent variables for which we want to find a relation. Hence,
resulting from equation (2.10) we can write
ΠEx 1 = f1(ρ∞, V∞, c, R)
ΠEx 2 = f2(ρ∞, V∞, c, µ∞)
ΠEx 3 = f3(ρ∞, V∞, c, a∞)
(2.19)
Then, with (2.18)
ΠEx 1 = ρ∞α1V∞
β1cγ1Rδ1
ΠEx 2 = ρ∞α2V∞
β2cγ2µ∞δ2
ΠEx 3 = ρ∞α3V∞
β3cγ3a∞δ3
(2.20)
or in dimensional terms
[ΠEx 1] = [ρ∞]α1 [V∞]β1 [c]γ1 [R]δ1 = [1]
[ΠEx 2] = [ρ∞]α2 [V∞]β2 [c]γ2 [µ∞]δ2 = [1]
[ΠEx 3] = [ρ∞]α3 [V∞]β3 [c]γ3 [a∞]δ3 = [1]
(2.21)
Thus[ΠEx 1] = (ML−3)α1(LT−1)β1(L)γ1(MLT−2)δ1 = 1
[ΠEx 2] = (ML−3)α2(LT−1)β2(L)γ2(ML−1T−1)δ2 = 1
[ΠEx 3] = (ML−3)α3(LT−1)β3(L)γ3(LT−1)δ3 = 1
(2.22)
For demonstration purposes we concentrate on the determination of Π1. In order to fulfill
the above equations, the exponents of each dimension must be cancelled. Thus we get k
equations by equating coefficients, one for each fundamental dimension we use:
For M: α1 + δ1 = 0 (2.23)
For L: −3α1 + β1 + γ1 + δ1 = 0 (2.24)
For T: −β1 − 2δ1 = 0 (2.25)
Page 15
Dimensional Analysis
So we find α1 = −1, β1 = −2, γ1 = −2 and δ1 = 1. Thus equation (2.22) can be rewritten
as follows
ΠEx 1 =R
ρ∞V∞2c2
(2.26)
where c2 has the dimension of an area. So it can be replaced by any reference area (e.g. the
planform area S of a wing) and ΠEx 1 will still be dimensionless. We can also multiply Π1
with a pure number and still, it will remain dimensionless [Buckingham, 1914]. Thus we can
rewrite equation (2.26) and get
ΠEx 1 =R
12ρ∞V∞
2S(2.27)
We know the term 12ρ∞V∞
2 as dynamic pressure q∞, so we can rewrite equation (2.27) to
ΠEx 1 =R
12ρ∞V∞
2S=
R
q∞S(2.28)
and can identify ΠEx 1 to be the force coefficient CR. The same analysis as for ΠEx 1 is to
be done for ΠEx 2 and ΠEx 3 and we get
ΠEx 2 =ρ∞V∞c
µ∞
ΠEx 3 =V∞a∞
(2.29)
Hence we can identify ΠEx 2 as the freestream Reynolds number Re and ΠEx 3 as the
freestream Mach number Ma. As a result of our analysis we can rewrite equation (2.17)
F
(R
12ρ∞V∞
2S,ρ∞V∞c
µ∞,V∞a∞
)= 0 (2.30)
or
F (CR, Re,Ma) = 0 (2.31)
Thus
CR = G(Re,Ma) (2.32)
So, one clear advantage of the Buckingham-Π-Theorem is the reduction of the number of
independent variables from which the resulting aerodynamic force R is dependent, from five
to two. This is especially important for investigative tests in a wind tunnel. Instead of varying
five different parameters, we only have to vary two, namely Re and Ma to get a formulation
for R. But the most important power of our results lies in the fact that we can use these
parameters to define dynamic similarity of two different flows: we are allowed to compare
the results of a wind tunnel test with the original conditions for a object when Re and Ma
are the same in both cases and when geometric similarity (i.e. the bodies and any other solid
boundaries are geometrically similar for both cases) is given. Re and Ma are called similarity
parameters.
Page 16
Theoretical Basis
2.4.3. Advantage of the Use of Dimensional Analysis during the
Design Phase of Satellites
One main purpose of this work is to find non-dimensional similarity parameters, similar to Re
and Ma in section 2.4.2, to formalize and facilitate the comparison between the performances
and characteristics of satellites of different sizes. In comparison to the aerodynamic example,
the CubeSat will be ideally what is the the model in the wind tunnel, and the bigger satellite
will be what is the real object like the plane. In addition, comparisons of satellites within whole
satellite classes will be possible as it is expected that the non-dimensional products will take at
least constant numerical values within a satellite class and hopefully also over a wider range.
Thus a very important advantage of using dimensional analysis during the spacecraft design
compared to usual calculation formulas presented in literature (i.e. [Larson and Wertz, 1999]),
is the fact that comparisons of the new design to already existing satellites in different missions
are possible. This enables the engineer ideally to avoid the whole set of calculations which is
normally necessary during the early development of a new satellite and just fall back on the
design of a ”similar” spacecraft, with the same non-dimensional parameters and hence adjust
its design right in the beginning, if necessary. Furthermore, existing designs can be checked
for their plausibility by investigating if their non-dimensional parameters can be found in the
usual value range of similar spacecrafts.
As this work will culminate in the creation of Mission Performance Indices, scalars which will
combine several non-dimensional parameters and ratios - thus characterizing a mission as a
whole - the engineer will quickly be able to get an idea of the performance that the designed
spacecraft is able to achieve. Thus, the user of the results of this work will be able to forecast
the ability of the so designed spacecraft already in the beginning of the design process and will
be, when required, easily and quickly redirected to other designs with the required performance.
Concludingly dimensional analysis helps to enable and formalize the potential hidden in the
similarity of spacecrafts and missions in order to facilitate and accelerate the design phase of
a satellite.
2.4.4. Limits of the Buckingham-Π-Theorem and Dimensional
Analysis
The quality of the results of the Buckingham-Π-Theorem is highly dependent on the choice
of the quantities Q1, . . . , Qn which drive the performance of the investigated system as they
are not fixed by the theorem itself. It is up to the physical understanding of the user of the
theorem to identify them. So, when a chosen quantity does not appear in a later dimensionless
product Π, it could mean that this quantity is redundant to the other chosen quantities of the
system, but it could also give the hint that another quantity is still missing in the modell of
the system.
Dimensional analysis can help to find non-dimensional parameters describing a problem but
does not determine the exact relation between them as it became clear in 2.4.2: equation (2.32)
Page 17
Dimensional Analysis
only states a dependence between CR, Re and Ma but does not give more information about
the function G. Physical comprehension and an extensive database of already flown satellites
are necessary to exactly determine the relation between the non-dimensional parameters and
the ratios and thus to determine the exponents a1, . . . , ai and b1, b2, . . . in equation (2.13).
Furthermore, the non-dimensional parameters and the Mission Performance Indices will provide
only a first estimation for the spacecraft design. The use of the parameters cannot substitute
detailed calculations that are definitely indispensable for a more sophisticated design.
Finally, as with every usual calculation formula, the non-dimensional parameters and also the
Mission Performance Indices will only be a model of reality and thus a simplification of all the
interdependencies which determine the characteristics of a system in the real world.
Page 18
3. Dimensional Analysis of a Single
Satellite
The aim of this chapter is to provide the engineer with theoretical scaling laws based on di-
mensional analysis which facilitate and accelerate the design phase of a satellite in general
and especially the design of a CubeSat by using the potential lying in the concept of simi-
larity. Based on the Buckingham-Π-Theorem presented in 2.4.1, non-dimensional parameters
will be presented in this chapter which will interrelate the characteristics of a satellite and its
subsystems to each other. This will establish correlations between the design parameters of
the spacecraft such as the power consumed by the spacecraft, the orbital period or simply
the volume and the size of the satellite. Based on these investigations, Mission Performance
Indices which characterize the performance of a mission as a whole by connecting the design
parameters, will be presented for three different mission types, namely Earth Observation,
Space Science and Technology Demonstration, these being analyzed as representing the most
frequent ones in recent CubeSat missions [Thomsen, 2010].
Additionally, using data from communication satellites in non-geosynchronous orbits (NGSO)
[Springmann and de Weck, 2004], some of the results of the dimensional analysis are empiri-
cally validated. Data from previously developed CubeSats was intended to be used instead but
the current available information is not extensive enough to be exclusively used for the valida-
tion of the products. Some estimations, however, are given based on this information and that
on the CubeSat-specific commercial off-the-shelf components (see Appendix B). Ideally, the
scaling laws would thus be a combination of both physics based as well as empirically derived
parameterized relationships.
3.1. Design Approaches: Top-Down and Bottom-Up
In general, two different problem assignments are possible for the design of a satellite: either
the parameters of the mission are fixed (i.e. the type of the mission, the payload to be used
on the spacecraft) and the ideal spacecraft fulfilling the requirements of the mission has to be
found - thus the bottom limits of the performance of the spacecraft are set; or the parameters
of a possible spacecraft are fixed (i.e. the size of the spacecraft, the power which can be
provided; thus the upper performance limits of the spacecraft are fixed) as well as the orbital
parameters and a mission fitting to the spacecraft and orbit has to be determined. In this
work, the first payload-centric approach is called ”Top-Down” design approach while the
Page 19
Design Approaches: Top-Down and Bottom-Up
latter spacecraft-centric approach is called ”Bottom-Up” design approach. These notations
will be used equally in this work. Figures 3.1 and 3.2 display the two approaches graphically.
Figure 3.1.: Top-Down design approach
Figure 3.2.: Bottom-Up design approach
Most of the usual industrial and scientific missions are done in a Top-Down design approach
as weight and space issues of the spacecraft play often only a secondary role as long as the
main function of the mission (i.e. scientific research with special payload) is accomplished.
However, this is one important difference to the CubeSat-missions: as most of the CubeSats
do obey a standard, size and mass limitations are the consequence and hence limitations in
power and reduced room for payloads are the result. Furthermore, as CubeSats are mostly
launched in a piggyback launch, the orbit of the CubeSat cannot be chosen independently but
is fixed by the primary payload on-board of the launcher. Thus CubeSats are definitely more
Page 20
Dimensional Analysis of a Single Satellite
often build in a Bottom-Up approach than in a Top-Down approach. However, in order to
keep the results of this chapter as general as possible, both approaches will be investigated
and are distinguished since according to the chosen one, the results are going to be different
in most of the analyses.
In order to have applicable results, the non-dimensional products Π as result of the dimensional
analysis are composed of quantities in a way, that they include ideally only one parameter
which is unknown in the chosen design approach and which can be determined by employing
the equation for the non-dimensional product with its numerical value. In the Top-Down
approach the known quantities will be characteristics of the payload (i.e. size, mass, power
consumption) whereas the unknown parameter is a characteristic of the spacecraft (i.e. size,
mass, power production). In the Bottom-Up approach it will be exactly vice-versa with known
spacecraft characteristics and unknown payload parameters. In coherence to the Buckingham-
Π-Theorem the unknown parameter will be the P and the known parameters will be equal to
the Qk’s in equation (2.10).
Please note as a final remark that in practice the orbital altitude, an approximate satellite
lifetime and the mission type are usually also given as input parameters in the Bottom-Up
approach.
3.2. Proceeding of the Validation of the
Non-Dimensional Parameters
The results of this chapter, the non-dimensional parameters, will be as general as possible and
thus will be applicable to any class of satellite. Only the empirical validation of them will
lead to different results depending on the various satellite classes as the numerical values of
the non-dimensional products Π can be different for each satellite class. Hence, some of the
theoretical results of this chapter will be confirmed by use of data from non-geosynchronous
satellites [Springmann and de Weck, 2004] as far as the data is available. Data for the flown
CubeSats was tried to be gathered together but the current available information is not exten-
sive enough to be used for the validation of the non-dimensional parameters and the Mission
Performance Indices. Some numerical estimations for the expected mass ratios of the picosatel-
lites, however, are given based on some real flown satellite data and the characteristics of the
commercial-off-the-shelf components which are gathered together in a component database
that was created during this work (see Appendix B). Thus, the main emphasis during the em-
pirical validation is put on the condition in similarity (see section 2.4) if the non-dimensional
products Π do have a fixed value for a given satellite class and thus can be applied as design
tool, and, in a first order approximation, not which value they will take.
In order to quantify the quality of the results of the application of the non-dimensional pa-
rameters on the NGSO-data, different statistical parameters were used. In a first step, the
Page 21
Proceeding of the Validation of the Non-Dimensional Parameters
arithmetical average of the results over all satellites was built. Secondly, the standard deviation
was calculated based on the following equation:
s =
√√√√ 1
N − 1
N∑i=1
(xi − xav)2 (3.1)
with N as the total number of the samples, xi the result for the satellite i and xav the arith-
metical average of the results over all satellites.
And finally, in order to be able to compare the results, the percentage of the deviation from
the arithmetical average, the coefficient of variation sxav
was calculated. The smaller this
characteristical value, the higher is in general the quality of the non-dimensional parameter,
as the deviation from the arithmetical average is as small as possible.
The validation of the non-dimensional parameters for the subsystems is more complex than for
the ratios as not only more quantities play a role but also complete ratios based on equation
(2.13). Their influence on the subsystem and correspondingly on the numerical value of the
non-dimensional parameter is expressed by exponents a, b, c, . . . that have to be determined
during the validation process. The influence of the non-ratio part nr of a non-dimensional
parameter is expressed by an exponent 1. Thus every non-dimensional parameter can be
rewritten to
1 = k · nr1r1ar2
br3c . . . (3.2)
k is a constant and the reciprocal of the latter numerical value of the non-dimensional param-
eter Π. a, b, c, . . . are the exponents of the ratios ri, expressing the level of influence of the
different ratios within the system which is described by the non-dimensional parameter. Before
computing k, nr and a, b, c, . . . , however, data of the investigated satellite class is required so
that the numerical values for nr and ri can be calculated. After that, the exponents a, b, c, . . .
and the constant k have to be determined in a way that they take ideally the same value for
every satellite in the given database, so that equation (3.2) is satisfied for every element of the
database, thus for every satellite in it. As this is very difficult and in general all but impossible
the larger the number of samples in the database gets, the left term of the equation (3.2) has
to be allowed to be slightly variable, for example variations of +/- 0.1 has to be possible in
order to get a solution. The solution which allows the slightest variations in the left term shall
be then the solution (3.2) which gives us a scaling law for the investigated satellite class.
In our case, however, this validation needs more detailed information on the satellites than it
is available in the NGSO-database. Therefore this explanations shall only be theoretical and
provide the user with the necessary information for the practical validation. An example of the
notions and terms will be given in section 3.4.2 by means of ΠBat 3 and ΠBat 4.
Theoretically, the results, especially the non-dimensional parameters, have to be checked in
their physical sense. This is formalized in an approach by [Bhaskar and Nigam, 1990] and
Page 22
Dimensional Analysis of a Single Satellite
is called intraregime-approach. Basically a regime is a physical aspect of a system. Thus a
non-dimensional parameter Π describes a regime. The intraregime approach consists of the
creation of partial differentials out of the quantities generating the non-dimensional parameters
in order to see how the quantities within the regime, thus the Π, are related to each other. In
the end this is a formalized expression of what will be described later for every non-dimensional
parameter. We will check for each of them if they are ”physically” consistent, thus we will
investigate, for example, if a raise in the quantities of the denominator which leads to an
increase in the quantities of the numerator for a constant Π, is logical in terms of the usual
relationships between the quantities. For example a raise in the mass of the spacecraft usually
entails also a raise in the power consumption of the spacecraft. Thus we will find the termPS/C
mS/C
in the top-level approach. Correspondingly, the intraregime approach indicates us a differential∂PS/C
∂mS/C> 0. For further interest in that topic, please refer to [Bhaskar and Nigam, 1990].
3.3. Ratios
3.3.1. Mass
A commonly used first approach in the sizing of satellites is to estimate the mass of the subsys-
tems of a satellite as percentage of the dry mass mS/Cdryor wet mass mS/Cwet of the spacecraft
[Larson and Wertz, 1999], [Pritchard, 1984], [Kiesling, 197172], [Saleh et al., 2002]. This point
of departure is definitely not in the classical sense of the dimensional analysis and the Buckingham-
Π-Theorem, as ratios of quantities are not seen as non-dimensional parameters. However, as
approximate mass ratios can be often found in literature, especially for communication satel-
lites or satellites in general, they appear to offer a quick and reliable possibility to get a first
idea of the preliminary mass budget of a satellite. Therefore, mass ratios were considered to
be worth further investigation.
In the Top-Down approach, the mass of the payload mP/L is considered to be given, thus
all the subsystems’ masses are divided by it. The Bottom-Up approach requires a mass of
the spacecraft, either the dry mass mS/Cdryor the wet mass mS/Cwet , to be fixed. Hence the
following ratios were investigated with the data of non-geosynchronous satellites as already
mentioned in section 3.2:
Page 23
Ratios
For the Top-Down approach we will investigate:
rMass TD1 =mS/C
mP/L
rMass TD2 =mAOCS
mP/L
rMass TD3 =mPower
mP/L
rMass TD4 =mStructure
mP/L
rMass TD5 =mThermal
mP/L
rMass TD6 =mC&DH+TT&C
mP/L
rMass TD7 =mPropulsion
mP/L
(3.3)
and for the Bottom-Up approach we will have a closer look at:
rMass BU1 =mP/L
mS/C
rMass BU2 =mAOCS
mS/C
rMass BU3 =mPower
mS/C
rMass BU4 =mStructure
mS/C
rMass BU5 =mThermal
mS/C
rMass BU6 =mC&DH+TT&C
mS/C
rMass BU7 =mPropulsion
mS/C
(3.4)
with mS/C being either mS/Cdryor mS/Cwet .
The data for the C&DH and TT&C subsystem was not differentiated in the given database of
the NGSO-satellites so in the following investigations they will be considered together as one
subsystem. Generally speaking, but especially in case of the CubeSats, it is recommended to
consider these subsystems separately.
Propulsion is normally not provided on CubeSats as it is a very heavy subsystem. Accord-
ingly the propulsion ratio in both approaches is not of importance for CubeSats. However, in
order to demonstrate a non-dimensional approach for satellites of all different classes which
do include propulsion, the mass of the propulsion system is considered in this investigation.
Furthermore, intensified research is ongoing in this area, also at MIT in the Space Propulsion
Laboratory under Professor Lozano who tries to find a new technology to provide propulsion
especially for small satellites [Courtney and Lozano, 2010]. Thus, this issue is assumed to be
one of the main picosatellite research topics in the near future and this work can help to model
Page 24
Dimensional Analysis of a Single Satellite
the resource allocation, including the propulsion subsystem.
In the following, the demonstration of the analysis is focused on the Bottom-Up approach with
the use of the dry spacecraft mass mS/Cdry, as this will be the principal case for designing the
CubeSats. The investigation of the Bottom-Up approach with use of the wet spacecraft mass
mS/Cwet and the Top-Down approach were done in the same way. Results of these two cases
will be discussed shortly afterwards.
Figure 3.3 shows the values of the ratios of equation (3.4) with the different satellites repre-
sented on the abscissa. Tables 3.1 and 3.2 represent the arithmetical average xav as well as
the standard deviation s and the coefficient of variation sxav
for the available data points.
At first glance the ratios shown in figure 3.3, seem to be similar for the different satellites
with some outliers. Tables 3.1 and 3.2 confirm this impression by showing percentages of
deviation of only 30% for mPL
mS/Cdry
, 39% for mPower
mS/Cdry
and 43% for mThermal
mS/Cdry
. Considering the
large number of satellites in the database of different masses, sizes and orbits, these deviations
were expected and are quite small.
Figure 3.3.: Mass ratios of NGSO-satellites plotted against the various satellites, represented
by a number
Table 3.1.: xav, s and sxav
of the mass ratiosmS/Cwet
mS/Cdry
,mP/L
mS/Cdry
and of the subsystems AOCS,
Power and Structure over mS/Cdry
mS/Cwet
mS/Cdry
mP/L
mS/Cdry
mAOCS
mS/Cdry
mPower
mS/Cdry
mStructure
mS/Cdry
Arithm. Average xav 1.217 0.363 0.072 0.297 0.195
Std. Deviation s 0.255 0.110 0.049 0.115 0.103
Coeff. of Variation sxav
[%] 21.00 30.43 68.73 38.74 52.55
Page 25
Ratios
Table 3.2.: xav, s and sxav
for the mass ratios of the subsystems Thermal, C&DH+TT&C and
Propulsion over mS/Cdry
mThermal
mS/Cdry
mC&DH+TT&C
mS/Cdry
mPropulsion
mS/Cdry
Arithm. Average xav 0.071 0.060 0.090
Std. Deviation s 0.030 0.060 0.070
Coeff. of Variation sxav
[%] 42.76 99.91 77.77
The high deviations occur mainly in the subsystems where the subsystem contributes to a very
small amount of the total mass of the satellite, namely in C&DH, AOCS and Propulsion (the
Structure subsystem is an exception as it does contribute to almost 20% of the total mass
but is still very dispersed). Small changes in the design of the subsystem can easily lead to
high percentage changes in the mass contribution. Those systems are hence more sensitive to
changes in their design, which, in turn, leads to higher coefficients of variation.
As a next step, it was assumed that the mass ratios would show more constancy in different
mass ranges. Figure 3.4 shows this to be a reasonable assumption. In the figure the mass ratios
are plotted against the dry mass mS/Cdryand here areas of constant ratios can be observed,
with the outliers inmS/Cwet
mS/Cdry
becoming more important with rising mS/Cdry:
Figure 3.4.: Mass ratios of NGSO-satellites plotted against mS/Cdry
Figure 3.4 also shows thatmS/Cwet
mS/Cdry
ranges between 1 and 1.5 for nearly all of the satellites with
an increase in wet mass for heavier satellites. This increase, however, could not be explained.
It is assumed that it can be due to the further propulsion mass heavier satellites require. In
case of most of the CubeSats this value will be equal to 1 anyway, as no propulsion is used on
CubeSats and thus mS/Cwet = mS/Cdry, the graphical presentation of
mS/Cwet
mS/Cdry
was eliminated
from the next figure to focus on the other mass ratios (see figure 3.5).
Page 26
Dimensional Analysis of a Single Satellite
Figure 3.5.: Mass ratios of NGSO-satellites plotted against mS/Cdryexcluding
mS/Cwet
mS/Cdry
A slight difference in constant values is visible between the satellites of masses above 1000
kg and under 300 kg. So, the data was divided into these two areas and the satellites were
investigated in their different mass classes.
In the dry mass range of 1000 kg and more (see figure 3.6), it becomes clear at first glance
that the payload and the Power subsystem are responsible for a good portion of the com-
plete satellite mass as it is also modelled in the known rules of thumb for first mass budget
estimations [Larson and Wertz, 1999, p.316]. Their values are concentrated around 40% formP/L
mS/Cdry
and around 30% for mPower
mS/Cdry
. The very few available ratios for the subsystems AOCS,
Thermal, C&DH + TT&C and Propulsion are around 10% and the ratio for the Structure
subsystem is around 15%.
Figure 3.6.: Mass ratios of NGSO-satellites with mS/Cdryabove 1000 kg
In the range of dry masses under 300 kg (see figure 3.7) some more data is available, thus the
conclusion of these results are based on a slightly wider foundation as for satellites over 1000
kg. The unfortunately few payload ratios are gathered around 25%, thus a small difference
Page 27
Ratios
to the satellite class with mS/Cdryabove 1000 kg can be observed. A correlation between the
two mass ranges can be detected for the power ratio which is again around 30%, as well as
for the subsystems AOCS, Thermal, C&DH + TT&C and Propulsion which are again in the
range of 10%. Only the subsystem Structure seems to be relatively heavier for the smaller
satellites as its ratio can be found between 10 to 35% with a rise to smaller satellite masses.
Figure 3.7.: Mass ratios of NGSO-satellites with mS/Cdryunder 300 kg
So, with the exception of the payload and structure ratio where small differences between
the mass ranges can be detected, all the other ratios were roughly in the same area for a
representative number of all 38 satellites.
The numerical results read from the figures are confirmed by the use of histograms (see figures
3.8 to 3.15) which plot the frequency of the different ratios against different value ranges, so
called classes. Important peaks can be seen for nearly every ratio except for the Structure
subsystem where the values range almost equally from 10 to 40%. (Please note that the values
are rounded up to the nearest class. Thus, all the ratios from 0.31 till 0.4 are allocated to 0.4
for example).
To summarize, it can be stated that the results of the analysis for a representative number of
satellites which ranges from 9 to up to 33, clearly show that the mass ratios can be mostly
found in a fixed value range. Deviations from the arithmetical average are of 30% for the
payload ratio up to nearly 100% for the C&DH+TT&C subsystem. Outliers do exist and
are included in the deviations, since the database is composed of satellites of very different
sizes and masses, thus with very different densities (between 4.44 and 698.6 kgm3 ). But as
the CubeSats all have the same idealized maximum density of 1171.81 kgm3 (see table 2.1),
one can assume that the mass ratios for this class of satellites will be much less dispersed
than it is the case for the NGSO-satellites. However, there is also a likely tendency that
the value for the payload mass ratio will decrease relative to the structure ratio as this trend
to higher structure ratios and lower payload ratios for smaller satellites was already seen for
the small NGSO-satellites. And in fact, a rough estimation based on the characteristics of
Page 28
Dimensional Analysis of a Single Satellite
Figure 3.8.: Histogram formS/Cwet
mS/Cdry
Figure 3.9.: Histogram formP/L
mS/Cdry
Figure 3.10.: Histogram for mAOCS
mS/Cdry
Figure 3.11.: Histogram for mStructure
mS/Cdry
the CubeSat-specific COTS components and some data on flown CubeSats retrieved only by
personal communication, confirms this assumption (see tables 3.3 and 3.4).
Page 29
Ratios
Figure 3.12.: Histogram for mPower
mS/Cdry
Figure 3.13.: Histogram for mThermal
mS/Cdry
Figure 3.14.: Histogram for mC&DH+TT&C
mS/Cdry
Figure 3.15.: Histogram formPropulsion
mS/Cdry
Table 3.3.: xav [%] for the Bottom-Up approach of the smaller NGSO-satellites in comparison
to an estimation for 1U CubeSat based on COTS components and real flown/-
planned CubeSats for the mass ratiosmP/L
mS/Cdry
and of the subsystems AOCS, Power
and Structure over mS/Cdry
mP/L
mS/Cdry
mAOCS
mS/Cdry
mPower
mS/Cdry
mStructure
mS/Cdry
NGSO 25 7.2 29.7 27
1U CubeSat 0-20 10-25 30-40 30-40
CP3 29.4 11.8 11.8 20
CP4 40.0 10.0 10.0 17
Cute1.7+APD 12.0 7.0 8 57
Exoplanet 13.5 30.8 28.6 18.2
SMAD 15-50 - - 15-25
Page 30
Dimensional Analysis of a Single Satellite
Table 3.4.: xav [%] for the Bottom-Up approach of the smaller NGSO-satellites in comparison
to an estimation for 1U CubeSat based on COTS components and real flown/-
planned CubeSats for the mass ratios of the subsystems Thermal, C&DH+TT&C,
C&DH and Communication over mS/Cdry
mThermal
mS/Cdry
mC&DH+TT&C
mS/Cdry
mC&DH
mS/Cdry
mCom
mS/Cdry
NGSO 7.1 6.0 - -
1U CubeSat 1-5 - 5-10 15-20
CP3 0 - 4 23.5
CP4 0 - 3 20.0
Cute1.7+APD 0 - 3 13
Exoplanet 0 - 2.8 6.1
SMAD 2-5 - - -
In this context, it should be noticed that in general the mass of the AOCS system highly
depends on the chosen components. The mass ratio for the 1U CubeSat is expected to around
10% when solar panels are used which already have sun sensors, magnetorquers, gyroscopes
and thermal sensors embedded. Only reaction wheels and star trackers are then potentially
added. 15 to 20% mass ratio are expected when the solar panels do not include AOCS
components and no star tracker is used. About 25% mass ratio is assumed when additionally
a star tracker is included in the design. This assumption can also be confirmed by the data on
the planned and flown CubeSats. 10% AOCS mass seems to be usual except for Exoplanet, a
3U CubeSat from MIT students. The high AOCS mass is due to the use of reaction wheels.
The assumed high power ratios for a hypothetical 1U CubeSat are due to the COTS-batteries
on-board (over 60 g for each battery with a capacity of 1200 mAh) and especially the solar
panels. The flown satellites possess lower ratios but it could not be confirmed why. We assume
that smaller or fewer batteries have been used.
Figure 3.16.: 1U CubeSat MOVE [MOVE, 2010a]
Page 31
Ratios
Figure 3.17.: Exploded view of the CubeSat MOVE [MOVE, 2010b]
The structure ratio is assumed to be higher for CubeSats than for the NGSO-satellites, taking
not only the outer walls of the spacecraft into account but also two Printed Circuit Boards
with a weight of 150g where especially the AOCS sensors are mounted on. However, two
1U CubeSats, CP3 and CP4 of Cal Poly, show smaller structure ratios than even the smaller
NGSO satellites. It is assumed that this is due to a special own build light weight structure.
Cute1.7+APD, in turn, is flying with a very high mass ratio of 57%, although the satellite is
Page 32
Dimensional Analysis of a Single Satellite
in the class of a 3U CubeSat. It seems that the structure ratio for CubeSats is as dispersed as
for the NGSO-satellites. One clear trend, however, can be seen by means of the rule of thumb
out of [Larson and Wertz, 1999, Tab.10-10]: larger satellites are assumed to have smaller
structure ratios of 15 to 25% of the dry spacecraft mass than nano- and picosatellites. The
tendency that the structure ratio gets larger for smaller CubeSats can be also confirmed when
plotting the mass structure ratio for several CubeSats, calculated with the potential CubeSat-
COTS components (see figure 3.18). Figures 3.19 and 3.20 show the mass and volume ratio
evolution when assuming a cube with edge length xeq and a outer structure consisting of a
5mm 5052-H32 Aluminum wall with a mass density of 2.68 gcm3 . This metal was chosen as it is
also used on the CubeSatKit from Pumpkin Inc. that builds the structure COTS components
on which the ratios from figure 3.18 are based.
Figure 3.18.: mStructure
mS/Cdry
for hypothetical CubeSats, calculated based on data from COTS-
components
Figure 3.19.: mStructure
mS/Cdry
for hypothetical CubeSats, calculated based on Aluminium wall
assumption
Page 33
Ratios
Figure 3.20.: VStructure
VS/Cfor hypothetical CubeSats, calculated based on Aluminium wall
assumption
The differences in the structure ratios in figure 3.18 and 3.19 are due to the different assump-
tions. In the calculation which is based on the COTS-components (please note the difference
when calculating with the skeletonized or solid walls provided by Pumpkin Inc.), the CubeSat
is assumed with its standard values as defined in table 2.1, thus with dimensions of a quarder.
In case of the calculations with the aluminium wall, the CubeSat is a cube with an edge length
which equals the so called equivalent edge length, a quantity which will be explained in detail
in the next section 3.3.2. For the moment it is enough to know that we assume cubes. The
main result is the fact that the mass structure ratio decreases with increasing dimensions of
the CubeSat. This can be advantageous for the payload mass.
Back to tables 3.3 and 3.4 we can see that the estimated CubeSat’s Thermal mass ratio is
smaller than for the NGSO-satellites as usually only thermal sensors are integrated in a CubeSat
and no complicated thermal control system. The real flown CubeSats even declared a thermal
mass ratio of zero percent. The C&DH subsystem of a CubeSat consists principally of a
motherboard including the microprocessor and the required memory storage causing relatively
low mass contribution. The hypothetical assumption is not far from the real flown CubeSat
data.
The communication subsystem is basically, depending on the chosen design, composed of
transceivers or transmitters and receivers and antennas. According to the chosen redundancy
a greater or fewer number of components are integrated into the satellite. The numerical
value for the theoretical 1U CubeSat in table 3.4 is based on a redundancy which necessitates
two transceivers or two transmitter/receiver pairs for the CubeSat. The real flown CubeSat
data confirms this assumed range with a tendency to smaller Communication ratios for bigger
satellites as it can be seen for Exoplanet and Cute1.7+APD. This tendency can not be as easily
developed with further evaluations as for the structure mass ratio. However, it is assumed
that the communication mass ratio decreases further for bigger satellites which would be again
Page 34
Dimensional Analysis of a Single Satellite
advantageous for the payload mass.
So, in the end, the residual payload mass is approximated to be between 10 and a maximum
of 20%. Surprisingly, the two bigger CubeSats Exoplanet and Cute1.7+APD show smaller
payload ratios than CP3 and CP4, both 1U CubeSats. This was not expected and is due to
the heavy subsystem of the satellites, the structure subsytem for Cute1.7+APD and Power
and AOCS for Exoplanet. CP4, on the other hand is the forth developed CubeSat by CalPoly,
thus clearly showing the gain in experience in design of CubeSats by the high payload ratio.
However, still a lot of flown CubeSats are platforms for Technology Demonstration missions.
In this respect, the payload is part of the subsystems, for example as in the case of the
CubeSat BeeSat of the Technische Universitat Berlin, Germany, which tested newly developed
reaction wheels, and is not considered as payload anymore. By that the mass restrictions for
the payload due to the other subsystems are avoided.
Despite the very rough estimations of the numerical values of the mass ratios for CubeSats,
the most important point from this section, is to see that the mass of the subsystems of a
satellite can definitely be estimated as percentage of the spacecraft’s dry mass for a whole
satellite class as a starting point in the design. This result was proven for the NGSO-satellites
and is also very likely to be valid for the CubeSats.
For the sake of completeness, the NGSO-results for the Bottom-Up approach with considera-
tion of the wet spacecraft mass mS/Cwet can be found in tables 3.5 and 3.6.
Table 3.5.: xav, s and sxav
for the mass ratiosmS/Cdry
mS/Cwet
,mP/L
mS/Cwet
and of the subsystems AOCS,
Power and Structure over mS/Cwet
mS/Cdry
mS/Cwet
mP/L
mS/Cwet
mAOCS
mS/Cwet
mPower
mS/Cwet
mStructure
mS/Cwet
Arithm. Average xav 0.848 0.296 0.066 0.262 0.179
Std. Deviation s 0.133 0.087 0.048 0.105 0.106
Coeff. of Variation sxav
[%] 15.62 29.44 72.59 40.04 59.11
Table 3.6.: xav, s and sxav
for the mass ratios of the subsystems Thermal, C&DH+TT&C and
Propulsion over mS/Cwet
mThermal
mS/Cwet
mC&DH+TT&C
mS/Cwet
mPropulsion
mS/Cwet
Arithm. Average xav 0.063 0.057 0.075
Std. Deviation s 0.028 0.062 0.063
Coeff. of Variation sxav
[%] 44.26 109.22 83.87
Page 35
Ratios
In comparison to the results of the ratios which used mS/Cdryas main spacecraft mass (see ta-
ble 3.1 and 3.2) the percentage of the standard deviation to the arithmetical average is higher
here for the Bottom-Up design approach with mS/Cwet in all ratios except for the one which
relates the spacecraft’s dry mass and wet mass with each other and the payload ratio. But as
the first is not important in case of the CubeSats as they do not have a propulsion subsystem,
and as the latter only varies with marginal 0.99% difference and the other ratios do show
better values in the range of an delta of 1.3% (Power) to up to 9.31% (C&DH+TT&C), it
can be clearly concluded that the Bottom-Up approach with the spacecraft’s dry mass mS/Cdry
is more promising for the CubeSats.
To complete the analysis, the discussion of the results of the Top-Down design approach
follows. Tables 3.7 and 3.8 present the results of the investigation and one can clearly see that
the values for the Top-Down ratios are more dispersed than for the Bottom-Up approaches:
the values sxav
have a delta of plus 2.8% (Power) to 20.8% (Structure) for the subsystems’
ratios of the Top-Down approach in comparison to the Bottom-Up wet mass approach and
a delta plus of 4.1% (Power) to 38.47% (C&DH+TT&C) for the subsystems’ ratios of the
Bottom-Up approach with dry mass.
Table 3.7.: xav, s and sxav
for the mass ratios of wet mass over payload mass, dry mass over
payload mass and of the subsystems AOCS, Power and Structure over payload
mass
mS/Cwet
mP/L
mS/Cdry
mP/L
mAOCS
mP/L
mPower
mP/L
mStructure
mP/L
Arithm. Average xav 3.719 3.049 0.262 0.906 0.619
Std. Deviation s 1.259 1.080 0.233 0.388 0.494
Coeff. of Variation sxav
[%] 33.84 35.41 88.60 42.84 79.81
Table 3.8.: xav, s and sxav
for the mass ratios of the subsystems Thermal, C&DH+TT&C and
Propulsion over the payload mass
mThermal
mP/L
mC&DH+TT&C
mP/L
mPropulsion
mP/L
Arithm. Average xav 0.225 0.147 0.284
Std. Deviation s 0.107 0.201 0.261
Coeff. of Variation sxav
[%] 47.65 137.38 91.84
The histograms 3.21 to 3.28 confirm graphically the higher distribution of the ratios. All
of them represent more values with lower frequencies than the histograms 3.8 to 3.15 of the
Page 36
Dimensional Analysis of a Single Satellite
Bottom-Up approach. One clear example is the power ratio: figure 3.12 shows only four values
with a high peak in 0.3 whereas figure 3.25 shows a distribution of ten values from 0.3 to 1.5
without a clear peak.
The reason for the higher distribution here is definitely the smaller divisor (mP/L is only 44%
of mS/Cdryand even 34% of mS/Cwet). However, a clearer picture was expected as the payload
mass coefficient of variation sxav
is only 67.9% in comparison to 80.3% for mS/Cdryand 86.0%
for mS/Cwet . Hence, the reason for the higher distribution seems to be a lack in a continuous
interdependency between the payload mass and the other subsystems’ masses for the NGSO
satellites. However, it is again expected for the CubeSats that they will show a higher coherence
between the payload mass and the subsystem masses as they are more standardized in their
mass distributions because of the mass limitations due to the standard they obey.
But in the end, a non-dimensional first mass estimation will be best when using the Bottom-Up
dry mass approach which is fortunately also the most important design approach in terms of
the CubeSats.
Figure 3.21.: Histogram formS/Cwet
mP/LFigure 3.22.: Histogram for
mS/Cdry
mP/L
Figure 3.23.: Histogram for mAOCS
mP/LFigure 3.24.: Histogram for mStructure
mP/L
Page 37
Ratios
Figure 3.25.: Histogram for mPower
mP/LFigure 3.26.: Histogram for mThermal
mP/L
Figure 3.27.: Histogram for mC&DH+TT&C
mP/LFigure 3.28.: Histogram for
mPropulsion
mP/L
3.3.2. Volume: Packing Factor
In order to get design information about the dimensions of the considered spacecraft, the next
step in the development of the non-dimensional parameters is to put the volume and thus
the edge length of the spacecraft into consideration. Especially important for CubeSats is a
packing factor as CubeSats are subject not only to mass but also to size restrictions. One
possible definition of this parameter is
p :=VusedVS/C
(3.5)
with
Vused = VP/L + VBus = ΣVi = Σmi
ρi(3.6)
Page 38
Dimensional Analysis of a Single Satellite
i being the representative for every different subsystem, thus AOCS, Power, Structure, Ther-
mal, C&DH, Communication and Propulsion so that we can write
p =VBusVS/C
+VP/LVS/C
(3.7)
Thus this parameter is again a ratio and not a non-dimensional product in terms of the
Buckingham-Π-Theorem. In almost the same manner as the packing factor p was defined
in equation (3.5) and which is especially of importance in the Bottom-Up design approach,
other ratios are possible to be created and investigated. Ratios which can be of great interest
for the Top-Down approach are VBus
VP/Land
VS/CVP/L
, both giving the engineer a first estimation of
how much space is occupied by a payload in comparison to the spacecraft bus or the whole
spacecraft.
Since no data was available for the volume of the payload or the subsystems, the usefulness
of those ratios can not be investigated with NGSO-data. However, as the volume ratios are
supposed to give information about the dimensions of the satellite which is very important
with regard to the volume restrictions of CubeSats, the alluded ratios are assumed to be of
high interest during the design of picosatellites.
Generally speaking, the higher the packing factor, the better is the conception and design
of the spacecraft. However, the value of the above limit is constrained and can practically
never be 100%, not only because of unadept ways of putting the spacecraft together but also
because of the dimensions of the different components which can be unpropitious, too. After
discussion with some student teams involved in CubeSat projects (CP3 and CP4 from CalPoly,
Exoplanet at MIT and MOVE at TUM), typical values for CubeSats are assumed to be in the
range of 60% and ideally up to 80% for the smaller ones. Thus an increase is expected for
smaller CubeSat units and not a constant value over the whole range of CubeSats which is
important to notice in terms of normally desired constancy over a whole satellite class for the
non-dimensional parameters.
Please note that it is very important to differentiate between the density ρi of the whole space-
craft or a subsystem and the packing factor p which are physically and also from a practical
point of view not the same. The definition of the density assumes that the mass is homoge-
neously distributed over the complete volume of the spacecraft, a fact that is not realized in
praxis. Thus the definition of the packing factor which takes account of this fact. Finally, the
packing factor is of higher and more practical importance for the design of the spacecraft than
its density.
As it was already mentioned in table 2.1, it shall be kept in mind that CubeSats in their
standard configuration all have the same maximum density of ρCubeSat = 1171.81 kgm3 . This
allows to simplify the mass and volume relationship
mS/C = ρS/CVS/C (3.8)
Page 39
Ratios
to
mS/C ∼ VS/C (3.9)
To get a relation between the mass and the edge length of the spacecraft, it is necessary to
establish a relationship between the volume and the edge length of the spacecraft. In order to
facilitate the further investigations, an equivalent edge length xeq is introduced in this work
which is defined as
xeq := VS/C13 (3.10)
so that equation (3.9) can be rewritten as
mS/C ∼ xeq3 (3.11)
For the standard CubeSats the values of the equivalent edge length are presented in table 3.9.
Table 3.9.: Equivalent Edge Length xeq for standard CubeSats
System Volume Equivalent Edge Length xeq
units mm3 mm
1U 1.135 · 106 104.31
2U 2.270 · 106 131.42
3U 3.405 · 106 150.44
Please notice that it is assumed that the concept of the equivalent edge length is still ”phys-
ical” for the whole spacecraft but not for every single subsystem as they consist of different
components and are often not integrated closely together.
3.3.3. Power
In the same way as the mass of the spacecraft is influenced by the mass of the payload, the
power consumption or requirement of the spacecraft is driven by the power requirement of the
payload. Thus there is an interdependency between payload and spacecraft power and thus
an interest rises in power ratios just in the same manner as the interest in the mass ratios.
The equations which are going to be investigated in this section are as follows:
For the Top-Down approach:
rPower TD1 =PS/CPP/L
rPower TD2 =PBusPP/L
(3.12)
Page 40
Dimensional Analysis of a Single Satellite
and for the Bottom-Up approach:
rPower BU1 =PP/LPS/C
rPower BU2 =PBusPS/C
(3.13)
with PS/C being the power consumption of the spacecraft and PP/L and PBus being the
normal (and not the peak) power consumption of the payload and the spacecraft bus. It is
also desirable to analyze the power consumption of every single subsystem in comparison to
the payload or spacecraft power consumption as in the case of the mass ratios. However, as
those values are not presented in the data base, only the bus power consumption was used
here. Nevertheless, further investigations should definitely take the subsystems separately into
consideration as demonstrated with the mass ratios.
Please note that the total spacecraft consumption PS/C was not given in the NGSO-database.
Instead, we use PSAeol. Both values should, however, be in the same order of magnitude and
differ from each other only by a margin and the influence of the power need during eclipse
PEclipse so that PSAeol is slightly higher than PS/C . This is typically determined during the
design process for the solar arrays where a closer investigation of the different power quantities
is done. Results of the interdependencies can be find in section 3.3.3. Thus, in the following
investigations PSAeol is used instead of PS/C .
Table 3.10 shows the results of the investigation for equations (3.12) and (3.13).
Table 3.10.: xav, s and sxav
forPSAeol
PP/L, PBus
PP/L,
PP/L
PSAeol
and PBus
PSAeol
PSAeol
PP/L
PBus
PP/L
PP/L
PSAeol
PBus
PSAeol
Arithm. Average xav 4.351 0.488 0.715 0.201
Std. Deviation s 14.792 0.900 0.239 0.186
Coeff. of Variation sxav
[%] 339.95 184.49 33.42 92.78
It is easy to see directly the high percentage of deviation forPSAeol
PP/Land PBus
PP/L. A closer look
at the data base shows the responsible outlier to be the satellites of Pentriad, a constellation
with very low normal payload and bus power consumption of 100 W and 448 W but an end
of life power of 7684 W and a begin of life power of 10247 W. Those high differences can
only be possible when the power budget was not well calculated or, more likely, the peak
power consumptions are very high. Unfortunately, the data base does not contain the peak
power consumptions of the Pentriad satellites. Thus, these explanations are only specula-
tions. Excluding Pentriad of the investigations gives the results shown in table 3.11, thus an
improvement in every ratio, from 1.89% for PBus
PSAeol
to up to 308.68% forPSAeol
PP/L.
Page 41
Ratios
Table 3.11.: xav, s and sxav
forPSAeol
PP/L, PBus
PP/L,
PP/L
PSAeol
and PBus
PSAeol
excluding Pentriad
PSAeol
PP/L
PBus
PP/L
PP/L
PSAeol
PBus
PSAeol
Arithm. Average xav 1.451 0.335 0.743 0.207
Std. Deviation s 0.454 0.426 0.195 0.188
Coeff. of Variation sxav
[%] 31.27 127.24 26.27 90.87
Especially the results for the ratiosPSAeol
PP/Land
PP/L
PSAeol
are promising. Furthermore, they do con-
firm the results of [Springmann and de Weck, 2004] where a correlation between the spacecraft
power consumption PSAeoland the payload power consumption PP/L was discovered.
The results for PBus
PP/Land PBus
PSAeol
are less promising. On the other hand, a look at their dis-
tributions plotted against the corresponding divisor (see 3.29 and 3.30, showing a clearer
distribution of the ratios than when plotted against the dry mass mS/Cdryin figures 3.31 and
3.32) states that the ratios are not much varying for higher abscissa values, thus especially for
end of life powers and payload power consumption bigger than 2500 W. The high deviations
in table 3.11 are thus especially due to the satellites with low payload power consumption
and/or low end of life power. A glance at the corresponding histograms 3.34 and 3.36 fortifies
this statement by showing clear peaks in the distribution for these ratios which are even more
developed as forPSAeol
PP/Land
PP/L
PSAeol
. (Please note that the outlier Pentriad is still present in the
histograms and can be detected as the single ratio 0.1 in figure 3.33, as the single ratio 77 in
figure 3.35 and as the single ratio 5 in figure 3.36. It is not visible in figure 3.34 as its value
is of 0.05 and counts to the 0.1 bar).
Thus a conclusion only out of the arithmetical average, standard deviation and the coefficient
of variation is not sufficient here and one needs to be careful with those analyses.
What becomes also obvious in figures 3.29 and 3.30 is the fact, that the values for the bus
power ratios are much less distributed graphically than those of the payload or the spacecraft.
The reason for the high percentage of deviation for the bus ratios in table 3.11 becomes here
clear, too: the values for the bus power ratios are so dispersed that they partially vary with
a factor of 20 (see values at 0.1 and then one value at 2 in figure 3.30). The factor for the
payload and spacecraft power values is far less important and can be analyzed to be about 3
in figure 3.29.
In the end, for the purpose of this section to detect constant ratios and/or constant non-
dimensional parameters, all four ratios show potential to be similarity parameters for the
design of the spacecraft and mission but are not as satisfying as expected. Furthermore as
the outliers are more present in the areas of smaller payload or spacecraft power as the divisor
becomes smaller and thus the ratios more sensitive to deviations, high deviations are also
Page 42
Dimensional Analysis of a Single Satellite
expected for the CubeSats as they in general have power consumptions of only several watts.
Nevertheless, constant values for the power ratios of CubeSats are assumed to be possible as
they obey a standard in mass and size which could possibly also lead to a restricted range in
power requirement and power production.
Figure 3.29.: Power ratiosPP/L
PSAeol
and PBus
PSAeol
plotted against PSAeol
Figure 3.30.: Power ratiosPSAeol
PP/Land PBus
PP/Lplotted against PP/L without outliers
Page 43
Ratios
Figure 3.31.: Power ratiosPP/L
PSAeol
and PBus
PSAeol
plotted against mS/Cdry
Figure 3.32.: Power ratiosPSAeol
PP/Land PBus
PP/Lplotted against mS/Cdry
without outliers
Figure 3.33.: Histogram forPP/L
PSAeol
Figure 3.34.: Histogram for PBus
PSAeol
Page 44
Dimensional Analysis of a Single Satellite
Figure 3.35.: Histogram forPSAeol
PP/LFigure 3.36.: Histogram for PBus
PP/L
Finally, which parameter is best to describe the power interactions, cannot be definitely decided.
Hence the idea for the next step arose in which it was tried to find an interdependency between
mass and power.
3.3.4. Mass and Power
In the classical Top-Down approach, the dry mass of a satellite mS/Cdryis primarily influenced
by the mass of its payload mP/L which in turn is driven by the payload power PP/L. On the
other hand the spacecrafts’ power consumption PS/C is influenced by the power requirement
of the payload, too. These are only two examples of the interactions between mass and
power which play an important role during the spacecraft design and which can be also found
in literature [Larson and Wertz, 1999, p.334],[Springmann and de Weck, 2004]. Because of
these interdependencies it was assumed that the combination of power and mass ratios would
lead to more satisfying, thus less dispersed results than the mass and the power ratios alone.
That is why it was tried in a next step to connect power and mass in ratios which are still
not non-dimensional products in terms of the Buckingham-Π-Theorem but considered to be
worth the investigation. As the payload and the complete spacecraft are the most important
players in the design process, the further investigation will be focused on their characteristics.
For the Top-Down approach it is again important to chose the characteristics of the payload
to be given, thus the payload power consumption PP/L and the payload mass mP/L. For
the Bottom-Up approach it is again vice versa: the mass of the spacecraft (either mS/Cdryor
mS/Cwet) is supposed to be known as well as the power consumption of the spacecraft PS/C .
As the NGSO-data set is the most complete for PSAeol, its value was used for the investigation
instead of PS/C as in the case of the power ratios.
Page 45
Ratios
Four different combinations of the four quantities are possible for the Top-Down and the
Bottom-Up approach likewise and shown in the following:
rMass Power 1 =mS/C
mP/L
·PS/CPP/L
(3.14)
its reciprocal
rMass Power 2 =mP/L
mS/C
·PP/LPS/C
(3.15)
then
rMass Power 3 =mS/C
mP/L
·PP/LPS/C
(3.16)
and its reciprocal
rMass Power 4 =mP/L
mS/C
·PS/CPP/L
(3.17)
Tables 3.12 to 3.15 show the arithmetical average, the standard deviation and the coefficient
of variation for all the four above combinations but with the small difference whether the
dry mass mS/Cdryor the wet mass mS/Cwet of the satellite is used in the ratios and, further,
whether or not the satellites of the Pentriad constellation are taken into account. Generally
speaking, the results are again better without Pentriad. So, the further analysis is based
on the values built without Pentriad which are shown in tables 3.13 and 3.15. First of all,
the percentage of deviation for all the four combination is very similar and also relatively
low considering again all the different satellites in the data base. Thus, it can be generally
concluded that this approach is definitely not on the wrong track to useful results. There are
also no significant differences between the ratios when using the spacecraft’s dry or wet mass.
However, a remarkable difference is the fact that the deviation is smaller for rMass Power 1 and
rMass Power 2 when using mS/Cwet instead of mS/Cdryand, in contrast, for rMass Power 3 and
rMass Power 4 smaller deviations are achieved when using mS/Cdryinstead of mS/Cwet . The
first observation can be explained by the higher deviation in the mass ratios formS/C
mP/Land
mP/L
mS/Cwhen using the dry spacecraft mass mS/Cdry
in comparison to the values when using the
spacecraft’s wet mass mS/Cwet .
Surprising, however, are the coefficients of variation for rMass Power 3 and rMass Power 4. These
are smaller when using the spacecrafts dry mass, whereasmS/C
mP/Land
mP/L
mS/Cshowed higher
deviations in section 3.3.1 when using the spacecraft’s dry mass mS/Cdry. This cannot be
easily explained and is surely associated with the interdependency between mass and power:
the power ratios seem to compensate the deviation in the mass ratios. This effect is apparently
so complimentary that it even produces the lowest percentage of deviation of 33.30% for
rMass Power 3 with use of dry spacecraft mass in comparison to all other combinations. The
lowest deviation of 36.62% for rMass Power 2 by using mS/Cwet was, however, expected as this
ratio combines the ratios with the lowest single deviation, namely 29.3% for the mass ratio
and 26.27% for the power ratio.
Page 46
Dimensional Analysis of a Single Satellite
Table 3.12.: xav, s and sxav
for rMass Power 1 to rMass Power 4 with mS/Cdry
mS/C
mP/L· PSAeol
PP/L
mP/L
mS/C· PP/L
PSAeol
mS/C
mP/L· PP/L
PSAeol
mP/L
mS/C· PSAeol
PP/L
Arithm. Average xav 15.017 0.253 2.013 2.238
Std. Deviation s 43.471 0.131 0.847 7.247
Coeff. of Variation sxav
[%] 289.48 51.93 42.09 323.82
Table 3.13.: xav, s and sxav
for rMass Power 1 to rMass Power 4 with mS/Cdrywithout Pentriad
mS/C
mP/L· PSAeol
PP/L
mP/L
mS/C· PP/L
PSAeol
mS/C
mP/L· PP/L
PSAeol
mP/L
mS/C· PSAeol
PP/L
Arithm. Average xav 4.791 0.268 2.130 0.530
Std. Deviation s 2.826 0.120 0.709 0.214
Coeff. of Variation sxav
[%] 58.99 44.66 33.30 40.26
Table 3.14.: xav, s and sxav
for rMass Power 1 to rMass Power 4 with mS/Cwet
mS/C
mP/L· PSAeol
PP/L
mP/L
mS/C· PP/L
PSAeol
mS/C
mP/L· PP/L
PS/C
mP/L
mS/C· PSAeol
PP/L
Arithm. Average xav 19.224 0.202 2.481 1.492
Std. Deviation s 60.886 0.089 1.258 4.658
Coeff. of Variation sxav
[%] 316.71 44.04 50.69 312.32
Table 3.15.: xav, s and sxav
for rMass Power 1 to rMass Power 4 with mS/Cwet without Pentriad
mS/C
mP/L· PSAeol
PP/L
mP/L
mS/C· PP/L
PSAeol
mS/C
mP/L· PP/L
PSAeol
mP/L
mS/C· PSAeol
PP/L
Arithm. Average xav 5.624 0.212 2.609 0.451
Std. Deviation s 2.823 0.078 1.150 0.1932
Coeff. of Variation sxav
[%] 50.20 36.62 44.09 42.87
So, in the end, what becomes clear from this short analysis is that single ratios are not neces-
sarily showing smaller deviations than combinations of ratios. In the case ofmS/Cdry
mP/L· PP/L
PS/Ceol
smaller deviations are possible than taking only its mass ratio alone (35.41%). Hence, for
an exhaustive analysis all possible combinations rMass Power 1 to rMass Power 4 should be in-
vestigated. On the other hand, the clear advantage of combining ratios and the desired
Page 47
Ratios
compensating effects for a less dispersed distribution, could not be confirmed for the NGSO-
satellite database for the whole range of possible parameters. Further validations are therefore
recommended in order to decide if the single ratios or the combinations are best to describe
the relations between mass and power for a given database.
The numerical results are confirmed by the graphical presentation of the ratios, firstly for the
ratios created with the the dry spacecraft mass mS/Cdry. Figure 3.37 and the corresponding
histogram 3.41 show how dispersed the results for the ratio rMass Power 1 are, ranging from 2
to 12, although a small peak in frequency can be spotted around 3 and 4, reflecting the ratios
behavior to be concentrated more densely around this values for increasing mS/Cdry. But this
trend is not interesting in terms of the CubeSats whose mS/Cdrydecreases in comparison to the
NGSO-satellites. The reciprocal of rMass Power 1, rMass Power 2, reflects the slightly smaller
deviation of 44.66% in figures 3.38 and 3.42 where the results are distributed over a smaller
range, namely from 0.1 to about 0.6. The graphical representations of rMass Power 4 are very
similar to those of rMass Power 2. However, the smaller occupied range became also obvious,
as most of the data points can be found between 0.4 and 0.6. The smaller percentage of
deviation is provided by rMass Power 3 which can be especially seen in its histogram 3.43 where
the most significant frequency peak of all the four investigated ratios with dry spacecraft mass
can be found. Thus, keeping in mind that the main CubeSat design approach is a Bottom-Up
approach with use of the dry spacecraft mass, ratio rMass Power 3 seems to be the most useful
for that purpose.
But one should also note that all four ratios had most of their outliers in the small dry mass
range, thus the potential mass range of the picosatellites. Intensive investigations of the ratios
(3.14) to (3.17) for the application on CubeSats is thus highly recommended.
Figure 3.37.:mS/Cdry
mP/L· PSAeol
PP/Lplotted against mS/Cdry
Page 48
Dimensional Analysis of a Single Satellite
Figure 3.38.:mP/L
mS/Cdry
· PP/L
PSAeol
plotted against mS/Cdry
Figure 3.39.:mS/Cdry
mP/L· PP/L
PSAeol
plotted against mS/Cdry
Figure 3.40.:mP/L
mS/Cdry
· PSAeol
PP/Lplotted against mS/Cdry
Page 49
Ratios
Figure 3.41.: Histogram formS/Cdry
mP/L
PSAeol
PP/LFigure 3.42.: Histogram for
mP/L
mS/Cdry
PP/L
PSAeol
Figure 3.43.: Histogram formS/Cdry
mP/L
PP/L
PS/Ceol
Figure 3.44.: Histogram formP/L
mS/Cdry
PSAeol
PP/L
For the sake of completeness, a brief analysis of the results of the ratios (3.14) to (3.17)
follows which focuses on the comparison of the results when using the mS/Cwet compared to
those using mS/Cdry. A comparison of the corresponding figures and histograms shows that
the value ranges for the ratios (3.14) to (3.17) using the wet mass are smaller for rMass Power 1
and rMass Power 2 (see figures 3.45, 3.46, 3.49 and 3.50) than when using the spacecraft’s dry
mass. That is why the coefficient of variation for rMass Power 1 and rMass Power 2 is smaller
when using mS/Cwet than when applying mS/Cdry. And similarly, the figures and histograms
for rMass Power 3 and rMass Power 4 (see figures 3.47, 3.48, 3.51 and 3.52) show bigger value
ranges for the corresponding ratios when using mS/Cwet in comparison to the results when
using the dry mass mS/Cdry, and hence the bigger the coefficient of variation.
Page 50
Dimensional Analysis of a Single Satellite
Figure 3.45.:mS/Cwet
mP/L· PS/Ceol
PP/Lplotted against mS/Cwet
Figure 3.46.:mP/L
mS/Cwet
· PP/L
PS/Ceol
plotted against mS/Cdry
Figure 3.47.:mS/Cwet
mP/L· PP/L
PS/Ceol
plotted against mS/Cwet
Page 51
Ratios
Figure 3.48.:mP/L
mS/Cwet
· PS/Ceol
PP/Lplotted against mS/Cwet
Figure 3.49.: Histogram formS/Cwet
mP/L
PS/Ceol
PP/LFigure 3.50.: Histogram for
mP/L
mS/Cwet
PP/L
PS/Ceol
Figure 3.51.: Histogram formS/Cwet
mP/L
PP/L
PS/Ceol
Figure 3.52.: Histogram formP/L
mS/Cwet
PS/Ceol
PP/L
Page 52
Dimensional Analysis of a Single Satellite
3.4. Non-Dimensional Parameters
3.4.1. A Top-Level Approach
In the previous sections, interdependencies between different masses, powers and the volume
and edge length became obvious. Apparently, these characteristics are highly interconnected
as the following examples summarizes the results of the previous sections and also give further
relations:
• The spacecraft mass is influenced by the power requirement because according to the
power requirement the corresponding components are heavier or lighter.
• The spacecraft mass is limited by the volume of the spacecraft as no infinite density is
possible.
• The volume is driven by the mass (as no infinite density is possible: usually, the higher
the mass, the bigger the volume) and power (i.e. the higher the power requirement, the
more complex is the power requirement, thus the bigger and numerous the components).
• Volume and mass are both driven by the orbital period (or, even more precisely, the
eclipse time) and power requirement: the longer the eclipse time or the power require-
ment, the bigger and heavier are the batteries and thus the bigger are the solar panels.
Thus, the orbital period is supposed to interplay in the system, too, as it can also be seen
in the figures 3.53 and 3.54 for the NGSO-satellites: the higher the orbit, the bigger is the
orbital period, thus in general the higher is the spacecraft mass and hence also the spacecraft
power requirement.
Figure 3.53.: mS/Cwet plotted against tOrbit Figure 3.54.: PSAeolplotted against tOrbit
Page 53
Non-Dimensional Parameters
Figure 3.55.: ρS/C plotted against tOrbit
Because of the high interdependencies between those four parameters mass, power, vol-
ume/length and orbital period, it was assumed that a non-dimensional parameter build with
these four quantities could be very useful for design purposes. Hence, taking these four main
parameters of a spacecraft into consideration, one can write:
0 = f(mS/C , PS/C , VS/C , tOrbit) (3.18)
With the number of system influencing quantities n to be four and with L, M and T being
the base dimensions of the quantities, thus k = 3, one non-dimensional product has to be
determined. The k-set is chosen to contain mS/C although its importance is minimal here
as only one non-dimensional parameter is expected. Dimensional analysis according to the
Buckingham-Π-Theorem described in section 2.4.1 directs us to the following non-dimensional
product
Π1 =PS/C · t3OrbitVS/C
23 ·mS/C
(3.19)
As the edge length of a spacecraft is more practical for the design of a spacecraft than its
volume, the non-dimensional product can also be expressed with the equivalent edge length
xeq by applying equation (3.10):
Π2 =PS/C · t3Orbitx2eq ·mS/C
(3.20)
In terms of the intraregime approach [Bhaskar and Nigam, 1990], these two parameters can
also be physically confirmed. A higher spacecraft mass mS/C leads in general to higher power
consumption of the spacecraft PS/C . The same argumentation is valid for bigger satellites since
a raise in the dimensions of a satellite, here presented by the equivalent edge length x2eq, usually
leads to a raise in the mass of the spacecraft and thus again in the power consumption. The
position of the orbital period tOrbit in the parameter can be explained by the fact that usually
heavier and bigger satellites can be found in higher orbital altitudes, thus with higher orbital
Page 54
Dimensional Analysis of a Single Satellite
period. And those heavier and bigger satellites, in turn, have again higher power requirements.
Thus all these relationships are perfectly expressed by Π1 and Π2.
In the following, we will also deal with dimensional parameters derived from the non-dimensional
parameters Π1 and Π2 in order to reduce the number of influencing quantities in the similarity
parameters. These dimensional parameters are assumed to be useful for the design of satellites
with similar quantities. For example, as mS/C scales with x3eq (see equation (3.11)) Π2 can be
rewritten to
Π∗1 =PS/C · t3Orbit
x5eq
(3.21)
for satellites with equal mass densities. This replacement changes the numerical value of
Π2 by the reciprocal of the spacecraft’s density ρS/C and makes Π2 dimensional. In order to
distinguish between non-dimensional and dimensional quantities the dimensional Πs are marked
with a ”*”. The approximation of constant mass densities is especially true for CubeSats as
they have constant maximum densities over the whole CubeSat range. This assumption is
unfortunately not true for the NGSO-satellites as it can be seen in figure 3.55. The lack
in density constancy of the NGSO-satellites will be one of the main reasons for the high
distribution of the parameters when applied for NGSO-satellites later on in this section.
Supposing equal orbital periods for the investigated satellites, the orbital period tOrbit can be
integrated in Π∗1, leading to
Π∗2 = Π∗1 · t3Orbit =PS/Cx5eq
(3.22)
This equation is especially interesting as it reflects a result of [Springmann and de Weck, 2004]:
with increasing edge length the power of the spacecraft has also to increase in order to main-
tain a constant non-dimensional parameter in terms of the similitude theory. For example,
doubling the edge length of the satellite necessitates to take the power times 25. Thus an
increase in the specific power Peol
xeqwith increasing equivalent edge length xeq can be observed,
or in other words, some economies of scale can be detected.
Integrating tOrbit in Π1 which is again valid for satellites with same orbit periods, leads us to
Π∗3 = Π1 · t3Orbit =PS/C
VS/C23 ·mS/C
(3.23)
Some remarks shall follow these equations before the analysis of the results by the NGSO-data
is presented:
• The above equations can be applied to each different subsystems by using the corre-
sponding quantities. As the volume and also the equivalent edge length of a subsystem
is difficult to be determined, it should to be reasonable to substitute the volume in equa-
tion (3.19) by the mass of the subsystem which is more easily to be identified. With
(3.9) Π1 can be rewritten to
Π∗4 = Π1 · ρS/C =PS/C · t3Orbit
VS/C53
(3.24)
Page 55
Non-Dimensional Parameters
• In case of the subsystem Power, one possibility is to replace PS/C by PBattery and to
substitute tOrbit by the eclipse time tEclipse. mPower includes, among other things, the
mass of the batteries, the solar panels and also the mass of the power distribution unit.
Thus equation (3.19) becomes
Π∗Power =PBattery · t3Eclipse
mPower53
(3.25)
Other replacements with tchargingtime instead of tEclipse or PSolarPanel instead of PBatteryare also reasonable and should be investigated in future work.
• As the payload is in the focus during the design of the spacecraft, its consideration in
above equations should not be ignored. Possible replacements for the quantities in Π1
are simply mP/L, PP/L, VP/L and Tdutycycle.
ΠP/L 1 =PS/C · t3dutycycleVP/L
23 ·mP/L
(3.26)
A ”hybrid” approach in terms of relating spacecraft specific quantities and payload
characteristics simultaneously in equation (3.19) may also be possible. One example
could be
ΠP/L 2 =PP/L · t3dutycycleVP/L
23 ·mP/L
(3.27)
• Discussing the time choice, it was considered to use the energy over an orbit or the
whole lifetime instead of the power requirement and the orbital period, in order to get
a more general equation. However, as most of the satellites do have similar lifetimes
and orbits in their own class (i.e. communication satellites are normally in GEO with a
lifetime of about 15 years; CubeSats are in LEO with a design lifetime of about a year)
this idea was considered to account for nothing more than the original equation with the
orbital period and the power requirement. Furthermore, as the same orbital period can
lead to different eclipse and sunlight times, it is supposed that it is more useful to take
the eclipse time tEclipse instead of tOrbit as time quantity. The calculation of the eclipse
time is presented in Appendix A. Especially for the minimal eclipse time and also for
sun-synchronous missions in general the knowledge of the inclination i of the orbit plane
is essential. For the sun-synchronous orbits the Right Ascension of Ascending Node also
needs to be known for the computation of the eclipse time. In order to keep the design
problem as simple as possible, however, the consideration of the orbital period tOrbit is
assumed to be sufficient enough for our estimations at this point.
The non-dimensional and dimensional parameters of this section were investigated for the
available data of the NGSO-satellites. Tables 3.16 and 3.17 summarize the results for Π1 =PS/Ct
3Orbit
VS/C23mS/Cwet
and Π∗3 =PS/C
VS/C23mS/Cwet
:
Page 56
Dimensional Analysis of a Single Satellite
Table 3.16.: xav, s and sxav
for Π1 for all NGSO-satellites and those in LEO
Π1 Π1 for satellites in LEO
Arithm. Average xav 5.74082E+12 4.44645E+11
Std. Deviation s 1.61752E+13 6.51342E+11
Coeff. of Variation sxav
[%] 281.76 146.49
Table 3.17.: xav, s and sxav
forPS/C
VS/C23mS/Cwet
for all NGSO satellites and those in LEO
Π∗3 Π∗3 for satellites in LEO
Arithm. Average xav 1.360 1.696
Std. Deviation s 2.074 2.663
Coeff. of Variation sxav
[%] 152.50 157.02
First of all, it shall be mentioned that the available data for those products was extremely rare,
so that only 10 data points out of 38 possible satellites were available for the investigation
of Π1. What becomes obvious, however, is the relatively large deviation of nearly 282%.
Consequently, the next step was again to eliminate the orbital period tOrbit and investigate
Π∗3. The results are definitely better then, showing a coefficient of variation of only 152.5%
which, in comparison to the result for Π1, is an improvement of nearly 130%.
To confirm the assumption that the orbital period is not a suitable time choice, the numerical
values for Π1 and Π∗3 were computed only for those satellites in LEO, all with a similar orbital
period of about 100 minutes, of which data points were available for both parameters Π1
and Π∗3. But, surprisingly, the percentage of deviation is smaller for the LEO satellites when
taking the orbital period into account, namely 146.5% in comparison to 157.0%. One can
assume that this is due to computational mistakes in MS Excel since the numerical values for
Π1 get very high because of the contribution of the orbital period, but also when introducing
tOrbit in minutes and not seconds the coefficient of variation becomes even better for Π1,
namely 142%, thus an improvement of 10% in comparison to the parameter Π∗3. Thus, for
the CubeSats which are all suited in orbits with relatively similar parameters (mostly LEO and
sun-synchronous), it seems reasonable to use Π1 and take account of the orbital period tOrbit.
When the orbital parameters are more scattered for the class of the investigated satellites, Π∗3will be better suited. However, given the very rare datapoints, these are only speculations and
need to be verified with more empirical data.
As the available data points for Π1 are so random, only the histograms of the results are
presented (see figures 3.56 and 3.57), showing the dispersed characteristic of the results but
also the improvement when eliminating the non-LEO satellites.
Page 57
Non-Dimensional Parameters
Figure 3.56.: Histogram for Π1 Figure 3.57.: Histogram for Π1 for the NGSO-
satellites in LEO
Figure 3.58 shows the data points of Π∗3 plotted against mS/Cwet . The outliers in those figures
are detected to be the satellites of the constellation Ellipso and Orbcomm, both having very
small volumes: 0.6 m3 for Orbcomm and 0.09 m3 for Ellipso. Pentriad is not an outlier
in that case, although it has a very large end of life power production, since its volume of
approximately 5 m3 together with its mass of about 2000 kg compensate the high power.
A closer examination of the graph 3.58 leads to the figure 3.59 that highlights the range of
most of the data points to be between 0.2 and 1.5.
A presentation of the results plotted against tOrbit (see figure 3.60 and 3.61) does not help
much further at first glance as the number of available data points is reduced. What becomes
clear, however, is that the outliers are LEO satellites and not satellites with complete different
orbital parameters. This means that their deviation from the other satellites is not due to
a difference in the orbital periods but really in the spacecraft’s characteristics. Thus the
neglection of tOrbit in Π∗3 does not cause any outliers. Besides, the only two data points
which are not representing satellites in LEO are located in the range between 0.2 and 1.5, as
well, thus reflecting the fact that the consideration of the orbital period with power 3 in the
non-dimensional product would definitely let disperse the data points.
A look at the histograms 3.62 and 3.63 possibly reveals why the percentage of deviation for
Π∗3 is slightly worse when taking only the satellites in LEO into consideration whose values are
available for both Π1 and Π∗3: on the one hand, this eliminates three data points on the bar
with value 2 (these being unfortunately the data points where no information on the orbital
parameters was available), on the other hand, this also deletes three satellites on the bar with
value 1 which reflects definitely the majority of the data points and the significant outliers on
4 and 9 are not touched at all by this selection.
It can be summarized so far that the inclusion of the orbital period tOrbit reduces the disper-
sion for satellites with similar orbital parameters. Thus the conclusion from the theoretical
presentation of the findings that for satellites with similar orbital period Π∗3 instead of Π1 is
Page 58
Dimensional Analysis of a Single Satellite
suited, is not generally true and proves the importance of numerical validation of the theoreti-
cal findings. On the other hand, when considering satellites with different orbits, taking tOrbitinto account leads only to more dispersion. This result is not satisfying and has to be further
investigated with a more extensive database than given, hoping that the high dispersion is due
to the lack datapoints.
Figure 3.58.: Π∗3 plotted against mS/Cwet Figure 3.59.: Zoom of Π∗3 plotted against
mS/Cwet
Figure 3.60.: Π∗3 plotted against tOrbit Figure 3.61.: Zoom of Π∗3 plotted against tOrbit
Page 59
Non-Dimensional Parameters
Figure 3.62.: Histogram for Π∗3 Figure 3.63.: Histogram for Π∗3 for the NGSO-
satellites in LEO
In a further step, the substitution of the mass by the volume (instead of the equivalent edge
length) was investigated which leads us to
Π∗4 = Π∗3 · ρS/C =PS/C
V53
(3.28)
as well as the elimination of the orbital period tOrbit and the mass of the spacecraft mS/C
from Π1 supposing similar masses and orbits
Π∗5 = Π∗3 ·mS/C =PS/C
V23
(3.29)
Also of interest is to eliminate the spacecraft’s volume VS/C and the orbital period tOrbit.
Π∗6 = Π∗3 · V23 =
PS/CmS/C
(3.30)
with mS/C = mS/Cwet .
The numerical results for these investigations can be found in table 3.18. Π∗4 and Π∗5 were
especially investigated to see how significant the influence of the diverse densities of the
satellites is. First of all, the figures 3.64, 3.65 and 3.67 for Π∗4 prove the fact that for the
given NGSO-database mass and volume of the spacecrafts do not relate all with the same
proportionality factor, which is the density, since the results are more scattered for Π∗4 than for
Π∗3. Thus the substitution of the mass by the volume is not possible. The higher dispersion
not only becomes clear in the higher coefficient of variation of 272.5% in comparison to
152.5% for Π∗3 but also in the histogram 3.67: here, in comparison to figure 3.62 a third,
very significant outlier can be determined on 6000. Furthermore the peak on the first class is
not as significant as in case of Π∗3, the second class, however, increases in frequency. For Π∗4,
Ellipso is the significant outlier at 6000 again because of its very small volume which is now
even more emphasized because of its higher exponent 53
in equationPS/C
VS/C53
. The outlier on 900
Page 60
Dimensional Analysis of a Single Satellite
is Orbcomm, again because of its small volume, and the new outlier on 500 is Pentriad where
the relatively small volume alone is appareantly no longer able to compensate the immense
power production at the end of life.
Thus, the mass and volume of the presented spacecrafts cannot be substituted by each other
without changing the numerical value of the non-dimensional product. In the NGSO-case,
Π∗4 showed significant differences and even higher dispersion than the results for Π∗3. For the
CubeSats, however, no important difference in the dispersion is expected as they are expected
to show a constant interdependency between mS/C and VS/C with the proportionality factor
ρS/C .
Additionally, investigating Π∗5 confirms that mass and volume are definitely not playing hand
in hand for the satellites in the given database, as the coefficient of variation gets even smaller
when ignoring the mass at all in the non-dimensional product. This can not only be seen
by a smaller coefficient of variation (120.9% in comparison to 152.5% when taking also the
mass into consideration in Π∗4) but also in the histogramm 3.68 where the only bars can be
found on 1000, 2000 and 3000 in comparison to 3.62 where four different bars were visible
and the first bar’s frequency was smaller than here. One can argument that the value range
is now multiplied by 103 but this is only a constant which can be eliminated by inserting the
numerical value of PS/C in kW and not in W. The presentation in figure 3.66 lets assume
the same concern but a closer look at the y-axis confirms the smaller value range (from 0 to
2600, or clearer, from 0 to 2.6 · 103) in comparison to figure 3.66 where the numerical values
ranged from 0 to over 8. Outliers, thus satellites with values above 1000, in figure 3.66 are
Pentriad, Virgo, Orblink and Teledisc (enumerated with anticlimatic deviation from 1000), all
satelittes with very high powers and/or small volumes. Comparatively, even better results can
be obtained with Π∗6 , thus completely eliminating the volume of the spacecraft.
However, the non-dimensional product and its derivatives are expected to show better results
when validated for CubeSat use. Π∗3 and Π∗4 are expected to provide promising results, as well
as the parameters Π∗1 and Π∗2 with the equivalent edge length xeq as quantity. Π∗5 and Π∗6were only investigated to show the discrepancy of volume and mass of the NGSO-satellites,
but could possibly be useful for other satellite classes again. Π1 and Π2, however, have to be
further investigated especially because of the influence of the orbital period tOrbit.
Nevertheless, especially the two parameters Π1 and Π2 are assumed to be very useful for quick
design estimations as they combine four main quantities of a satellite design: mass, power,
length/volume and orbital period. Three parameters of them, mass, power and length/volume,
are both for the Top-Down and the Bottom-Up approach input quantities for the design
process, for the Top-Down approach with payload specific parameters and for the Bottom-Up
approach with spacecraft specific quantities. We will see later on in figures 3.69 and 3.70
that Π1 and Π2 are not directly integrated in the flow of the design calculations based on
dimensional analysis. However, they are especially advantageous when one of the four design
quantities is missing as input parameters. It can then easily be calculated with the knowledge
of the numerical value of the non-dimensional parameter and three given quantities.
Page 61
Non-Dimensional Parameters
Table 3.18.: xav, s and sxav
for Π∗4, Π∗5 and Π∗6
Π∗4 Π∗5 Π∗6
Arithm. Average xav 542.072 619.360 2.365
Std. Deviation s 1477.130 748.720 1.638
Coeff. of Variation sxav
[%] 272.50 120.89 69.24
Figure 3.64.: Π∗4 plotted against mS/Cwet Figure 3.65.: Zoom of Π∗4 plotted against
mS/Cwet
Figure 3.66.: Π∗5 plotted against mS/Cwet
Page 62
Dimensional Analysis of a Single Satellite
Figure 3.67.: Histogram for Π∗4 Figure 3.68.: Histogram for Π∗5
3.4.2. Subsystem Power: Battery and Solar Array
One of the most important subsystems of a satellite is its power system as its enables the
satellite to be operational. The payload, the on-board computer, the communication system
none of these could work without power. On the other hand, the power subsystem is also
one of the most significant mass contributors. Thus, because of its high importance for the
success of the mission and its significance in the design, the two most typical design solutions
shall be discussed in the following.
BATTERY
The basic power source used on satellites is a battery. If no other power source is present
on-board, it provides the satellite with the required power over its whole lifetime. More
sophisticated solutions use solar arrays in addition to batteries for storing and providing energy
when the satellite is in eclipse and during the start phase. Thus a closer look on the design of
these devices in this work is considered to be significant.
The main characteristic performance parameter of a battery is its capacity CBattery[Ws]. On
one hand, the capacity of a battery is driven by the power required of the satellite during
eclipse PEclipse and the eclipse duration tEclipse. On the other hand, the satellite lifetime
tSatellite determines the number of charges/discharges, thus the number of duty cycles of the
battery and thus the Depth of Discharge (DOD). The DOD, in turn, contributes to the choice
of the battery’s capacity: the higher the Depth of Discharge, the lower is usually the required
capacity for a fixed power requirement. Hence four parameters play a role in the battery
system:
0 = f(CBattery, PEclipse, tEclipse, tSatellite) (3.31)
with
• the capacity of the battery CBattery [Ws] = [kgm2
s2]
Page 63
Non-Dimensional Parameters
• the power required of the spacecraft during eclipse PEclipse [W ] = [kgm2
s3]
• the eclipse duration tEclipse [s]
• and the satellite lifetime tSatellite [s]
So, with the definition of a combined non-fundamental SI-dimension [ML2], it is n = 4 and
k = 2, thus the number of non-dimensional parameters is i = 2. With a k− set consisting of
PEclipse and CBattery, equation (3.31) can be rewritten to
ΠBat1 = f1(PEclipse, CBattery, tEclipse)
ΠBat2 = f2(PEclipse, CBattery, tSatellite)
}(3.32)
and the following non-dimensional parameters can be found:
ΠBat1 =CBattery
PEclipse · tEclipse
ΠBat2 =CBattery
PEclipse · tSatellite
(3.33)
The two non-dimensional parameters can also be confirmed physically in terms of the intra-
regime approach [Bhaskar and Nigam, 1990]: for both of them, an increase in the power
requirement of the spacecraft PEclipse during eclipse goes along with an increase in the nec-
essary capacity of the battery CBattery for a constant Π. Furthermore, supposing ΠBat1 to be
constant over the satellite class, an increase in the eclipse time tEclipse leads to an increase
in the necessary capacity of the battery CBattery for constant PEclipse. Considering ΠBat2 , it
is known that an increase in the lifetime of the satellite causes an increase in the number
of charges and discharges of the battery, thus the cycle life of the battery. This, in turn,
reduces the depth of discharge (see [Larson and Wertz, 1999, p.421]) which then necessitates
an increase in the capacity of the battery for a constant energy requirement PEclipse · tEclipseof the spacecraft during eclipse. These interdependencies are hence appropriately expressed
by ΠBat2 . The non-dimensional products ΠBat1 and ΠBat2 could be now further developed to
a scaling law according to (2.11). As we will not be able to determine scaling laws in this
work because of the lack of data, another method of expansion is chosen to directly show the
influence of ratios on subsystems: the non-dimensional parameters are directly expanded with
the subsystem influencing ratios. This proceeding will be carried out for this and the following
subsystems.
A further expansion of the non-dimensional products ΠBat1 and ΠBat2 is possible with the
mass and power consumption of the spacecraft since these quantities are also the main input
during the Bottom-Up design approach. We will focus on ΠBat1 to show the expansion. Hence
we can write
ΠBat3 =CBattery
PEclipse · tEclipse·(
PS/CPEclipse
)a·(mS/C
mPower
)b(3.34)
This is also physically consistent and a constant Π can be achieved as an increase in the
power requirement of the spacecraft PS/C normally leads to a higher power requirement in
Page 64
Dimensional Analysis of a Single Satellite
eclipse PEclipse (it can be assumed that PS/C is equal to PEclipse although this is not generally
true as the power consumption is usually held lower in eclipse by turning off the payload
for example) which in turn necessitates a higher battery capacity CBattery. A higher battery
capacity usually causes a higher power mass mPower - easy to understand when imagining that
the requirement for higher capacity is satisfied by an additional battery - which, in turn, entails
a higher spacecraft mass mS/C .
From the payload-centric point of view which necessitates the use of the payload mass and
power consumption, ΠBat3 can be re-expressed with PP/L instead of PS/C as a change in the
payload power involves the same reasoning and changes as a change in the spacecraft power
requirement. In terms of the payload mass, it is assumed that a higher payload mass implies a
higher payload power requirement [Springmann and de Weck, 2004] and thus a higher battery
capacity which in turn causes a higher power mass. All these interdependencies are implied in
the following representation of ΠBat4 :
ΠBat4 =CBattery
PEclipse · tEclipse·(
PP/LPEclipse
)a·(mP/L
mPower
)b(3.35)
Interdependencies with the other subsystems are certainly possible as well since the power
subsystem is highly interconnected with every subsystem of the satellite that consumes power.
Thus it would be physically consistent to express ΠBat3 with for example PCom and mCom in-
stead of PS/C and mS/C . However, those interdependencies are already expressed by the power
and mass ratios in sections 3.3.1 and 3.3.3 and are thus implied when using the spacecraft
mass and the spacecraft power.
To complete this section, it shall be mentioned that is was attempted to express the eclipse
time tEclipse by the orbital period tOrbit in order to standardize the quantities in the non-
dimensional products used in accordance to the use of tOrbit in section 3.4.1. However, as
shown in the Appendix A of this work, tEclipse is not simply proportional to the orbital period
tOrbit but a complex function of the orbital altitude hOrbit. In case of the CubeSats which are
mostly launched in sun-synchronous orbits, tEclipse is additionally dependent on the inclination
i of the orbital plane and the Right Ascension of the Ascending Node RAAN . In order to
keep the level of complexity in this work adequate to a first order model, the approach with
tEclipse as a quantity of the battery system is considered to be sufficient.
Another remark has to be done regarding the efficiencies of the battery. Efficiencies are not
directly considered in the non-dimensional products above but can be retrieved in the numerical
value of the non-dimensional products. For a satellite class and the same battery type (i.e.
Lithium-Ion), the non-dimensional products describing the battery should have the same value
for every kind of satellite. The last remark in this section refers to the number of batteries
used on the satellite. The amount of batteries is not explicitly listed but can be retrieved in
CBattery which shall be the representative for the accumulated capacity of all batteries used
on the spacecraft.
As a conclusion, it can be stated that the design of the battery is influenced by several
quantities. However, in a first order model, the capacity of the battery CBattery, the power
requirement during eclipse PEclipse, the eclipse time tEclipse, the satellite lifetime tSatellite, and,
Page 65
Non-Dimensional Parameters
depending on the design approach, the mass of the spacecraft mS/C and the power requirement
of the spacecraft PS/C or the mass of the payload mP/L and the power requirement of the
payload PP/L are the most significant quantities for the battery-system.
To finish this section, some notions and expressions of 3.2 shall be exemplary shown on the
non-dimensional parameters of the battery in terms of the practical validation. The same
explanations will be valid for the following sections. Equations (3.34) and (3.35) show the
nature of (3.2). Thus nr can be identified to beCBattery
PEclipse·tEclipsefor both parameters.
PS/C
PEclipse,
mS/C
mPower,
PP/L
PEclipseand
mP/L
mPowerare ratios ri in terms of equation (3.2) and k is equal to 1
ΠBat3
for equation (3.34) and equal to 1ΠBat4
for equation (3.35). Hence the validation of the
non-dimensional parameter would include the determination of the numerical value of every
quantity as described in 3.2. As the currently available information is not sufficient for this
kind of detailed validation, the results in this and the following sections will remain theoretical.
A first order numerical application will be given in section 3.6.
SOLAR ARRAY
Solar arrays are used on spacecrafts in order to provide the satellite with power over a longer
lifetime than it is possible with primary batteries. They not only provide the satellite with power
during daylight but also recharge the secondary batteries on-board which are used during an
eclipse. Solar arrays have been widely adopted, ranging from the International Space Station
to CubeSats. Its investigation in terms of the Buckingham-Π-Theorem shall be done in this
section.
In a first order approximation, the power amount produced by a solar array is mainly dependent
on the type of the cells, the solar array area, the operating temperature, the sun angle and
the solar constant:
0 = f(PSAbol, type of cells, ASolarArray, θ, Toperating, cSolar) (3.36)
with
• the begin of life power production by the solar arrays PSAbol[kgm
2
s3]
• the type of the solar array cells (i.e. Silicium, Gallium Arsenide)
• the surface of the solar array ASolarArray [m2]
• the average sun angle over the whole mission θ [deg]
• the operating temperature Toperating [K]
• the solar constant cSolar which is equal to 1367 kgs3
The characteristic ”type of the cells” is ideally an efficiency which is neglected in the following
analysis as it will emerge again in the numerical value of the non-dimensional product. Typ-
ically, the non-dimensional product should take a constant value for solar arrays of the same
type over a satellite range.
Page 66
Dimensional Analysis of a Single Satellite
The influence of the orbit on the performance of the solar array is represented by θ, the aver-
age incident sun angle over the whole mission, and the solar constant cSolar, which normally
changes in conjunction with the season but is supposed to be constant in our investigations.
Both quantities are considered together in the quantity SSolar which equals cosθ · cSolar. A
remark in terms of θ should be stated before continuing the analysis: as in case of the batteries
and the eclipse time, the average solar angle is also dependent on the orientation of the orbital
plane in the ecliptic as well as the satellite itself (in case of body mounted solar panels) or
the orientation of the deployable panels. As this work is a first attempt to analyze satellites
with dimensional analysis, dependencies between θ and the orbital parameters as well as the
orientation of the satellite are not taken into consideration in order to reduce the complexity
of the analysis. Please note, however, that θ equals the solar angle βs (see Appendix A) for
satellites which are launched in sunsynchronous orbits and whose orientation to the Sun is
supposed to be constant.
Continuing now on the analysis, it can be stated that the operating temperature Toperating of
the solar array affects the performance of the array and is driven by the orbital parameters. The
highest power is produced when the satellites comes out of eclipse, thus when the temperature
of the array is as low as possible. But since its influence is also complex, it was neglected in
the following first order analysis.
In general, the cells do not cover the whole outside area of the spacecraft as some space is
also required for the Communication subsystem (antenna) and for the payload (i.e. camera).
However, for the sake of simplification, it is assumed that the whole outside area of the satellite
is used for the solar cells, thus ASolarArray ∼ x2eq (an equalization is intentionally not made
as the amount of illuminated area changes during the orbit. In the case of a cube, though, a
maximum of three sides is simultaneously illuminated). In the case of deployable solar arrays,
another equivalent length quantity than xeq must be used, which represents not only the
body-mounted solar cells but also the deployable solar panels. Furthermore, a more detailed
approach should also consider areas of the satellite’s surface which are not covered by solar
cells or obscured by the deployables. In this work, however, we will continue with the body
mounted solar panels since this is usually adopted for CubeSats. Taking into account these
remarks, equation (3.36) can be finally simplified to
0 = f(PSAbol, xeq, SSolar) (3.37)
The base dimensions of the quantities describing the system are the length L and the combined
dimension watt W. Thus with the number of quantities describing the system n to be three
and the number of the base dimensions to be two, one non-dimensional product can be found:
ΠSA1 =PSAbol
x2eq · SSolar
(3.38)
In terms of the intraregime approach [Bhaskar and Nigam, 1990] ΠSA1 perfectly expresses
that the power produced by the solar array PSAbolincreases with increasing solar array area,
represented by x2eq, and/or increasing quantity SSolar which increases with higher average sun
angle θ assuming cSolar to take the constant average value of 1367 Wm2 . A more sophisticated
Page 67
Non-Dimensional Parameters
approach would take into account that cSolar actually changes during the seasons (i.e. maximal
in winter).
In order to get information about the design of the spacecraft over the whole mission length, it
is necessary to incorporate also the lifetime degradation of the solar cells into the considerations
which can be expressed by Ld = (1− d)tSatellite [Larson and Wertz, 1999, p.417] with
• the yearly degradation d = degradationyear
[ 1yr
] which takes a typical value for every kind of
solar cell (i.e. satellites in LEO with cells out of Gallium Arsenide are subject to a yearly
degradation of 2.75% [Larson and Wertz, 1999, p.417])
• and the satellite lifetime tSatellite [yr]
Thus equation (3.37) can be expanded to
0 = f(PSAeol, xeq, SSolar, tSatellite, d) (3.39)
and hence
ΠSA2 =PSAeol
· d · tSatellitex2eq · SSolar
(3.40)
which can be also confirmed physically in terms of the intraregime approach again: assuming
ΠSA2 to be constant over the satellite class, an increase in the yearly degradation d or in the
satellite lifetime tSatellite will cause a decrease in PSAeolfor a constant PSAbol
.
On the other hand, an expression with PSAbolis also possible:
ΠSA3 =PSAbol
x2eq · SSolar · d · tSatellite
(3.41)
thus an increase in the yearly degradation d or in the satellite lifetime tSatellite will lead to a
necessary increase in PSAbolin order to achieve a constant PSAeol
.
Taking both these approaches into consideration, we can write
ΠSA4 =PSAbol
x2eq · SSolar · d · tSatellite
·(PSAbol
PSAeol
)a(3.42)
which expresses the same interdependencies as the non-dimensional products ΠSA2 and ΠSA3
before but in a combined equation.
As the overall aim of this work is to determine key design characteristics of a spacecraft
and/or a mission with few known characteristics of the payload (Top-Down approach) or of
the spacecraft (Bottom-Up approach), it is of special interest now to include not only the
equivalent edge length of the satellite xeq into the non-dimensional products but also its mass
and power. In terms of the mass of the solar arrays, this weight is already included in mPower
and its relation to mS/C and mP/L was investigated in section 3.3.1. As for the power, the
power productions PSAboland PSAeol
are a function of PS/C , tEclipse and tOrbit (representative
for tDaylight) since
PSolarArray · tDaylight = PS/C · tDaylight + PEclipse · tEclipse
PSolarArray = PS/C + PEclipse ·tEclipsetDaylight
(3.43)
Page 68
Dimensional Analysis of a Single Satellite
with
tDaylight = tOrbit − tEclipse (3.44)
Furthermore, in a simplified model, PEclipse can be supposed to be equal to PS/C (which is
normally not true as the operations on board are reduced in eclipse). Hence, equation (3.43)
can be rewritten to
PSolarArray = PS/C ·1
1− tEclipse
tOrbit
(3.45)
and finally
ΠSA5 =PSAbol
x2eq · SSolar · d · tSatellite
·(PSAbol
PSAeol
)a·(mPower
mS/C
)b·(PSAbol
PS/C
)c·(tOrbittEclipse
)d(3.46)
Again, this can be physically confirmed: ΠSA5 appropriately expresses that for a satellite class
an increase in the satellite mass mS/C usually entails an increase in the spacecraft’s power
requirement PS/C , and thus an increase in PSAboland also in the mass of the power system
mPower. Furthermore, equation (3.45) states that with increasing eclipse time tEclipse the
power produced by the solar array has also to be increased for a constant power requirement
PS/C . This interdependency is perfectly expressed by ΠSA5 as well.
While ΠSA5 is again a spacecraft-centric approach, it can be rewritten with mP/L and PP/Linstead of mS/C and PS/C as in the case of the battery and we get
ΠSA6 =PSAbol
x2eq · SSolar · d · tSatellite
·(PSAbol
PSAeol
)a·(mPower
mP/L
)b·(PSAbol
PP/L
)c·(tOrbittEclipse
)d(3.47)
The analysis for the Top-Down approach will again be the same as for the Bottom-Up approach.
Interdependencies to other subsystems are again possible as in the case of the batteries.
Please notice as a final remark of this section that the non-dimensional product ΠSA2 can be
expanded in the same way as ΠSA3 leading to
ΠSA7 =PSAeol
· d · tSatellitex2eq · SSolar
·
·(PSAeol
PSAbol
)a·(mPower
mS/C
)b·(PSAeol
PS/C
)c·(tOrbittEclipse
)d (3.48)
for the Bottom-Up approach and for the Top-Down approach:
ΠSA8 =PSAeol
· d · tSatellitex2eq · SSolar
·
·(PSAeol
PSAbol
)a·(mPower
mP/L
)b·(PSAeol
PP/L
)c·(tOrbittEclipse
)d (3.49)
In conclusion it can be stated that the solar array system is at least as complex as the battery
system since again a significant number of quantities play an important role in their design,
beginning with the powers PSAboland PSAeol
produced by the solar array, over spacecraft and
payload parameters such as mS/C , PP/L or tSatellite up to orbital quantities such as the eclipse
time PEclipse.
Page 69
Non-Dimensional Parameters
3.4.3. Subsystem AOCS: Reaction Wheel
The Attitude and Orbit Control System (AOCS) of a spacecraft keeps the spacecraft on the
planned orbit and enables the accuracy for the pointing of instruments and/or the antenna(s)
on-board. In order to fulfill these functions, propulsion systems are often used which allow
significant attitude corrections. A propulsion system also increases the lifetime of the satellite
as it is able to increase the orbital altitude of the satellite which decreases in time due to
atmospheric drag and other disturbances. However, propulsion systems are not yet mature
enough for use on CubeSats (see Appendix B). Another common approach to the AOCS of
spacecrafts, however, is to use reaction wheels, mechanical devices which provide the satellite
with angular momentum in order to compensate for disturbance torques by aerodynamic drag,
gravity and solar radiation. The first two disturbance torques are especially significant in Low
Earth Orbits where the CubeSat are deployed. It should be noted that a reaction wheel can not
compensate for altitude loss due to these disturbances, it will only control the orientation of
the spacecraft. Propulsion systems, in turn, are a possibility for restoring the orbital altitude.
Reaction wheels are relatively heavy devices and hence not very often used on CubeSats be-
cause of their weight limitations. However, specially designed reactions wheels for nano- and
picosatellite applications have risen in interest in the industry. For example the CubeSat BeeSat
from the Technical University of Berlin, launched in September 2009, tested successfully reac-
tions wheels from the company Astro- und Feinwerktechnik Adlershof GmbH for picosatellite
use [TUBerlin, 2010]. Further designs from other companies are also expected. A brief ap-
plication of the Buckingham-Π-Theorem for dimensioning reactions wheels shall therefore be
presented in this section.
As already mentioned, a reaction wheel is supposed to compensate the disturbance torques
which act on a spacecraft. So its most important performance characteristics are its angular
momentum HRW [kgm2
s] and its inertial momentum IRW [kgm2]. The disturbance torques, on
the other hand, are dependent on a number of factors [Larson and Wertz, 1999, p.366]:
• Aerodynamic Torque: TAero = f(ρAir, vS/C , AS/C)
• Gravity Gradient: TGravity = f(IS/C , hOrbit)
• Solar Radiation Torque: TSolar = f(AS/C , surface properties)
with
• the air density ρAir [ kgm3 ]
• the velocity of the spacecraft vS/C [ms
]
• the surface of the spacecraft AS/C [m2] normal to the spacecraft velocity vector
• the inertial momentum of the spacecraft IS/C [kgm2]
• and the orbital altitude hOrbit [m]
Page 70
Dimensional Analysis of a Single Satellite
In section 3.3.2 the concept of the equivalent edge length was introduced, making the simpli-
fication that the spacecraft is designed as a cube. Furthermore, supposing that the mass is
homogeneously distributed inside the spacecraft, the principal moment of inertia of the space-
craft around its principal axes can be easily expressed with the mass of the spacecraft mS/C
and its equivalent edge length xeq [Gross et al., 2010]:
IS/C =1
12·mS/C · (x2
eq + x2eq)
=1
6·mS/C · x2
eq
(3.50)
This representation of the spacecraft does not take into account deployables by using xeqsince this quantity represents the dimensions of the stowed spacecraft. However, the inertial
momentum of a body is always a function of its mass and representative lengths. So, in case of
deployables, another length representing the complete deployed dimensions of the spacecraft
has to be added to the list of parameters which will cause a supplementary ratio in the non-
dimensional representation, namely for example xeqxdeployed
. A detailed explanation of how to
calculate the inertial momentum of bodies with extensions can be found in [Gross et al., 2010].
In the following, however, the representation without deployables will be considered since
CubeSats do have deployables only in a minority of cases. Thus the surface of the spacecraft
AS/C on which the disturbances act, can also be modelled with the equivalent edge length
xeq:
AS/C ∼ x2eq (3.51)
Here again, deployables are not taken into consideration. For a more sophisticated approach, a
representative length should be taken into consideration which takes account of the spacecraft
extensions.
Assuming circular orbits around the Earth, the velocity of the spacecraft vS/C can be repre-
sented by the orbital altitude hOrbit as v =√
µREarth+hOrbit
. The density of the air ρAir is also a
function of the orbital altitude hOrbit, decreasing with increasing orbital altitude [Hedin, 1991].
Developing this further, the orbital period is also a function of the orbital altitude with
tOrbit = 2π
√(REarth + hOrbit)
3
µ(3.52)
Instead of expressing the torques as functions of orbital altitude hOrbit, they can be presented
as functions of the orbital period tOrbit:
• Aerodynamic Torque: TAero = f(tOrbit, xeq)
• Gravity Gradient: TGravity = f(mS/C , xeq, tOrbit)
• Solar Radiation Torque: TSolar = f(xeq, surfaceproperties)
Thus the system of the reaction wheel can be described as following
0 = f(HRW , IRW , tOrbit, xeq,mS/C) (3.53)
Page 71
Non-Dimensional Parameters
with tOrbit, xeq and mS/C being the parameters of the disturbance torques. Please note that
the surface properties of the spacecraft which have an influence on the solar radiation torque
are neglected in the following for the sake of simplification. Furthermore the accuracy range
is assumed to be 100% (this is to say we assume that the disturbance torque is completely
compensated by the torque produced by the reaction wheels) which is also a simplification. In
practice an accuracy of 100% is not feasible and precision near 100% is very expensive in terms
of mass, power and complexity. An accuracy less than 100% would imply another parameter
in equation (3.53) which can be an angle representing the accuracy range. The numerical
value of the non-dimensional parameter Π would change according to that range but finally
not the relations between the above parameters describing the system.
For the sake of simplification the reaction time of the satellite to a change in the angular
momentum by the reaction wheel is not considered, either. However, a more detailed model
of the system should take its influence into account as well.
It is also important to mention that in practice disturbances of different magnitudes act on
different satellite axes. In addition, the satellites are mostly not designed as cubes and have
extensions which also influences the degree of disturbance. Thus different sized reaction wheels
can be needed for the satellite. Hence, in this approach the non-dimensional parameter will
only provide a first approximation of the dimensions of the reaction wheels.
Given these caveats, the non-dimensional analysis can now be discussed. In accordance to
relation (3.53), five parameters describe this subsystem with three base dimensions namely
mass M, length L and time T. Thus with n = 5 and k = 3, i will be equal to two. The k-set
is chosen to contain mS/C , xeq and tOrbit. For ΠRW1 we can rewrite
0 = f1(mS/C , xeq, tOrbit, HRW ) (3.54)
and for ΠRW2
0 = f2(mS/C , xeq, tOrbit, IRW ) (3.55)
After the usual proceeding as described in section 2.4.1, it can be found that
ΠRW1 =HRW · tOrbitmS/C · x2
eq
ΠRW2 =IRW
mS/C · x2eq
(3.56)
which can be confirmed in terms of the intraregime approach [Bhaskar and Nigam, 1990]:
supposing ΠRW1 to be constant over the satellite class, an increase in spacecraft mass or
spacecraft dimensions increases the exposure to the disturbance torques. That is why more
angular momentum HRW is required. However, increasing the orbital period assuming circular
orbits means increasing the orbital altitude. As aerodynamic torque and disturbances due to
gravity decrease with increasing orbital altitude, less angular momentum has to be provided
by the reactions wheels as expressed by ΠRW1 .
Considering ΠRW2 , an increase in spacecraft mass or the equivalent edge length of the space-
craft causes an increased requirement of the reaction wheel’s inertial momentum IRW . An
Page 72
Dimensional Analysis of a Single Satellite
increase in mS/C and/or xeq also causes an increase in HRW as seen in ΠRW1 . Thus in terms
of the inter-regime approach, the partial derivative ∂HRW
∂IRWis positive which is not surprising as
it is also known that
HRW = IRW · ωRW (3.57)
thus an increase or decrease in IRW leads to an increase or decrease in HRW with constant
angular velocity of the wheel ωRW and vice versa.
In order to get clearer understanding of the dimensions of the reaction wheel, it was decided
to represent the principal moment of inertia of the reaction wheel IRW with the mass of the
reaction wheel mRW and its radius rRW . It is known according to [Gross et al., 2010] that
IRW =1
2·mRW · r2
RW (3.58)
Thus the system can be described with the following parameters
0 = f(HRW ,mRW , rRW ,mS/C , xeq, tOrbit) (3.59)
with n = 6 and k = 3. Thus i = 3, however, taking into account that two of those assumed
non-dimensional parameters are simply ratios. Hence after applying the process of dimensional
analysis, we get for the spacecraft-centric Bottom-Up approach
ΠRW3 =HRW · tOrbitmS/C · x2
eq
·(mRW
mS/C
)a·(rRWxeq
)b·(mAOCS
mS/C
)c(3.60)
including an additional expansion by(mAOCS
mS/C
)c.
As the power consumption and the power production of the spacecraft also play an important
role in the design of the spacecraft, the dimensioning of the reaction wheel shall also be
considered from this point of view as its operation also consumes power. The system will then
be best described by
0 = f(HRW ,mRW , rRW , tOrbit, PS/C ,mS/C , xeq) (3.61)
Thus with the spacecraft-centric chosen k-set PS/C ,mS/C , xeq, the following results can be
found
ΠRW4 =H3RW
PS/C ·m2S/C · x4
eq
ΠRW5 =mS/C · x2
eq
PS/C · t3OrbitΠRW6 =
mRW
mS/C
ΠRW7 =rRWxeq
(3.62)
ΠRW4 expresses in terms of the intraregime-approach [Bhaskar and Nigam, 1990] that an in-
crease in power PS/C leads to an increase in the performance of the reaction wheel namely the
Page 73
Non-Dimensional Parameters
angular momentum HRW . This is true as with more power, more operation of the reaction
wheel is possible. The second parameter ΠRW5 is already well-known and was first derived in
section 3.4.1. Here, however, the orbital period tOrbit represents the orbital altitude and thus
the magnitude of disturbances which act on the spacecraft. With increasing orbital period
tOrbit the orbital altitude hOrbit also rises which means that the disturbances get less signif-
icant. This, on the other hand, means that the need for compensation of the disturbance
torques by the reactions wheels is reduced. Consequently, the operation time of the reaction
wheels is lower and hence the power consumption by the wheels is less which is perfectly
expressed by ΠRW5 . On the other hand, an increase in spacecraft mass or edge length means
a more significant requirement for disturbance torque compensation. The operation time of
the reaction wheels is higher and hence also the power consumption by them, which is also
represented by ΠRW5 .
ΠRW6 expresses that an increase in spacecraft mass leads to an increase in the mass of the
reaction wheels which is also comprehensible as higher spacecraft masses raise the requirement
for disturbance torque compensation which can be realized by heavier reaction wheels. The
same reasoning is valid for ΠRW7 as bigger spacecraft dimensions raise the disturbance torques
which act on the spacecraft and thus cause a higher requirement in compensating angular
momentum HRW which can be fulfilled by bigger wheels.
It shall mentioned that the inclusion of the angular velocity ωRW was attempted in the analysis
but only limited satisfying results were found. However, one approach shall be presented here
which includes ωRW as parameter. The system is described with
0 = f(HRW , ωRW , PS/C) (3.63)
With n = 3 and k = 2 (assuming the combination of dimensions ML2 to be one base
dimension) i equals one. Through dimensional analysis one gets
ΠRW8 =HRW · ω2
RW
PS/C(3.64)
This expresses that a rise in the power of the spacecraft leads to an increase in HRWω2RW .
The influence of PS/C is intentionally not considered in terms of the dividend separately, as
a rise in ωRW implies a rise in HRW and vice versa. So one term of the dividend can not be
held constant while the other changes. That would mean the inertial momentum of the wheel
IRW decreases at the same time which is not the intention.
In conclusion, it can be stated that the dimensioning of the reaction wheel is not only dependent
on the orbital parameters but also on the characteristics of the spacecraft namely its mass
mS/C , its equivalent edge length xeq (and another length representing the completely deployed
spacecraft) and the power provided by the spacecraft PS/C . The above presented approaches
are completely spacecraft-centric (Bottom-Up) which is also the most important design case for
CubeSats. In order to have a payload-centric (Top-Down) design one can include the accuracy
range of the payload into the list of system describing quantities. Other than that, the same
parameters as for the spacecraft-centric approach can be used as still spacecraft mass and its
Page 74
Dimensional Analysis of a Single Satellite
external dimensions are the most important quantities determining the disturbance torques.
The mass and the power consumption of the payload do play an indirect role in designing the
reaction wheels as they are related to the spacecraft mass and spacecraft power consumption.
However, this is enough to rewrite the equations with payload-centric ratios, as for example
ΠRW3 . So we get
ΠRW4 =HRW · tOrbitmS/C · x2
eq
·(mRW
mP/L
)a·(
rRWxeq P/L
)b·(mAOCS
mP/L
)c(3.65)
The same porcedure can be done for the remaining parameters.
3.4.4. Subsystem Communication
The communication subsystem is one of the most significant subsystems as it enables the
communication between the ground station and the satellite. Without it, no data transmission
and health monitoring would be possible.
The path of the signal from the satellite to the ground station is characterized by a number
of influences. The satellite itself carries a communication subsystem which consists of at least
an antenna and a transmitter and receiver or a transceiver, a device combining the functions
of the latter two. The performance of the transmitter is characterized by its output power
PTransmitter. The performance of the antenna is characterized by its peak gain GAntenna which
is in general a function of the dimensions of the antenna and the wavelength λCom of the
transmitted signal. As an example, the gain for a parabolic antenna can be calculated by
Gparabolic antenna = η ·(π ·DAntenna
λCom
)2
(3.66)
The gain of the antenna is therefore a function of the antenna diameter DAntenna and the
signal wavelength λCom. η describes the efficiency of the antenna and takes a value between 0
and 1. In the case of the parabolic antenna, η is equal to 0.55. In the following we will assume
this type of antenna but other antenna types are also possible. The exact calculation of their
gain can be found in common literature, for example [Larson and Wertz, 1999, p. 571].
The important aspect for the further calculation is, however, that the gain of an antenna can
be calculated based on its dimensions and the signal wavelength.
It is needless to say that not only the satellite has to provide an antenna for the communication
but also the ground station. The calculation of its gain is performed using the same equation
as for the satellite antenna.
A number of losses occurs in the transmission. The space loss Ls is by far the most important
loss. It is calculated by(
λCom
4ΠSCom
)2
, with SCom being the communication path length, thus
the distance between the spacecraft and the ground station, and λCom expressing the signal
wavelength. So the space loss becomes more important with higher distance SCom.
The transmission path loss La is another loss, taking into account atmospheric and rain
absorption. It is often far less significant than the space loss, but still recommended to be
Page 75
Non-Dimensional Parameters
always considered in the calculations. It is highly dependent on the chosen frequency and on
the elevation angle ε [Larson and Wertz, 1999, p.564] and can become significant when ε and
λCom are not chosen with caution.
The transmitter line loss Ll occurs between the antenna and the transmitter. It is a loss due
to the hardware and not the environment or the orbital parameters as Ls and La.
The antenna pointing losses Lp Transmitter and Lp Receiver are based on a imperfect orientation
of the on-board antenna towards the one on the ground station and vice versa. They can be
expressed as a function of the pointing error e and the half-power beamwidth θ which in turn
is a function of the signal wavelength λCom and the geometrical dimensions of the antenna.
We can sum up the losses and their interdependencies to
• space loss Ls = f(λCom, SCom) = g(λCom, hOrbit, ε)
• transmission path loss La = f(λCom)
• line loss Ll
• and antenna pointing loss Lp = f(e, θ)
As the line loss Ll is only dependent on the quality of the hardware, it is a characteristic of
its own, without any interrelations with other parameters.
The most important quantity of the system, however, is the data rate R [ bits
], the amount of
information that is transmitted by the communication subsystem.
As important as the communication subsystem is, as heavy and power intensive is it very often.
The high mass is usually due to the mass of the antenna mAntennaT which in turn is driven by
the antenna’s dimensions (see [Saleh et al., 2002]). Thus it is recommended to take account
of the antenna mass when listing the quantities influencing the communication subsystem.
The high power requirement of the subsystem is due to the transmitter or transceiver power
requirement which in turn is a high contributor to the power budget of the spacecraft in
general.
Taking all the above discussions into account, the communication subsystem can be expressed
by
0 = f(PTransmitter, PS/C ,mAntennaT , DTransmitter, DReceiver, λCom, SCom, R, e) (3.67)
with
• the power output of the transmitter PTransmitter [kgm2
s3]
• the power consumption of the spacecraft PS/C [kgm2
s3]
• the mass of the transmitter antenna mAntennaT [kg]
• the diameter of the transmitter antenna DTransmitter [m]
• the diameter of the receiver antenna DReceiver [m]
• the signal wavelength λCom [m]
• the communication path length SCom [m]
Page 76
Dimensional Analysis of a Single Satellite
• the data bit rate R [ bits
]
• and the pointing error e [deg]
Please note that SCom is a function of the orbital altitude hOrbit, the position of the ground
station to the subsatellite point and also of the elevation angle ε (see [Larson and Wertz, 1999,
p. 113]). Assuming the best possible orientation of satellite and ground station, namely the
satellite in the zenith of the ground station, SCom is then equal to hOrbit. The worst position
is given when satellite and ground station are just in optical range (assuming that the minimal
ε required for communication is 0◦ which is practically not true as a minimum angle of 5◦ is
usually needed). This is when the ground station is in the true outer horizon of the satellite.
SCom can then be calculated with
SCom =
√(REarth + hOrbit)2 −REarth
2 (3.68)
In both extreme cases which are important when designing a spacecraft, S is only a function of
the orbital altitude hOrbit. Furthermore, no pointing losses will be considered in the following,
assuming perfect alignment between the antenna of the spacecraft and the ground station.
Thus the number of the system’s quantities can be reduced and changed to
0 = f(PTransmitter, PS/C ,mAntennaT , DTransmitter, DReceiver, λCom, hOrbit, R) (3.69)
After dimensional analysis and in consideration of the ratios, equation (3.69) can be rewritten
as
ΠCom1 =mAntennaT · h2
Orbit ·R3
PTransmitter·(
λComDTransmitter
)a·(
λComDReceiver
)b·(
PS/CPTransmitter
)c(3.70)
Please note that the bit error rate (BER) is included in the numerical value of ΠCom1 as BER
is already a non-dimensional quantity. However, in order to display the influence of the bit
error rate more clearly, it can be simply added to equation (3.70) and we get
ΠCom2 =mAntennaT · h2
Orbit ·R3
PTransmitter·
·(
λComDTransmitter
)a·(
λComDReceiver
)b·(
PS/CPTransmitter
)c·(mCom
mS/C
)d·BERe
(3.71)
The signal modulation is taken into consideration indirectly by adding the bit error rate BER
into the list of the influencing quantities. It can be confirmed that this non-dimensional
product is also physically consistent: a higher transmitted data rate R generally requires a
higher transmitter output power PTransmitter; an increase in hOrbit leads to an increase in the
communication path length S. This in turn implies an increased space loss which necessitates
higher transmitter output power in order to transmit the fixed data rate R; a larger antenna
diameter, be it the one of the transmitter DTransmitter or of the receiver DReceiver, leads to a
decrease in the transmitter output power PTransmitter for a constant data rate R; an increase
in the antenna mass usually comes along with an increase in its diameter and thus leads
Page 77
Non-Dimensional Parameters
to a decrease in required transmitter output power PTransmitter for a constant data rate R;
and finally, a raise in the transmitter output power PTransmitter is usually possible by heavier
devices, thus increasing the mass of the communication subsystem mCom which leads to an
increase in the spacecraft mass mS/C . All these interdependencies are appropriately expressed
by ΠCom2 .
To conclude it can be stated that the dimensioning of the communication subsystem is de-
pendent on a number of factors, taking not only spacecraft specific parameters into account
but also the characteristics of the ground station(s) and the orbital parameters as well as the
environment.
This approach is nearly completely spacecraft-centric (Bottom-Up design) except for the data
rate R which is not only influenced by the number of ground stations the satellite can have
contact with to transmit the data but also the performance of the payload. Depending on
the data rate and data volume produced by the payload, the data rate R to be transmitted
to Earth is determined. It is expected that there will be a direct interdependency between the
communication subsystem and the payload, expressed by a common quantity of the systems,
namely the data rate R. This assumption will indeed be confirmed in the next section.
Other than that, a payload-centric Top-Down approach uses nearly the same quantities as a
spacecraft-centric Bottom-Up approach, since the antenna diameters are payload-independent
and the mass and the power consumption of the payload do only play an indirect role in
designing the communication subsystem. But as in case of the battery and the solar arrays
the mass and power ratios can be adapted to the Top-Down approach so that we can re-express
equation (3.71) to
ΠCom3 =mAntennaT · h2
Orbit ·R3
PTransmitter·
·(
λComDTransmitter
)a·(
λComDReceiver
)b·(
PP/LPTransmitter
)c·(mCom
mP/L
)d·BERe
(3.72)
.
3.4.5. Payload
The payload is the most important part of the satellite as it gives the satellite its significance
and value. An immense range of payloads is possible as there are many possible mission pro-
files. Three kinds of missions will be in focus in this work, namely Earth Observation, Space
Science and Technology Demonstration. The payload will be different, depending on the type
of the mission.
Page 78
Dimensional Analysis of a Single Satellite
When taking a closer look at the payload of the satellite fulfilling an Earth Observation mission,
the payload will most likely be a sensor measuring the incoming radiation from the Earth. Based
on the characteristic wavelength of the radiation which will be measured during the mission,
the type of the sensor is then chosen. However, as the type of the design of the optical
sensors is independent from the radiation wavelength their performance is described by the
same characteristics.
First of all, there is the ground pixel resolution of the instrument (often indicated at nadir)
X [ mpixel
] and Y [ mpixel
], the first one representing the cross-track ground pixel resolution, the
latter one the along-track pixel resolution. The higher the resolution is, the smaller is X
and/or Y and the better is the quality of the image taken. The ground resolutions have direct
interdependencies with the instrument on-board, especially the diameter of the aperture of the
sensor D [m], and the orbital altitude hOrbit. Their interrelations can be expressed by
X = Y = 2.44 ·hOrbit · λP/L
D(3.73)
from [Larson and Wertz, 1999, p.264].
Thus with increasing orbital altitude hOrbit which implies an increase in tEclipse the resolution
gets worse, thus bigger, but with a bigger aperture D the resolution gets better. The data
rate R, which is produced by the instrument, depends in first order on the ground resolution
and the coverage area of the instrument: with the swath width SW of the instrument and
the cross-track ground pixel resolution X, SW
Xpixels per swath line has to be taken by the
instrument. vGround
Yswath lines, in turn, are scanned in a second as vGround expresses the
spacecraft ground velocity and Y the along-track ground pixel resolution. Hence the data rate
R produced by the instrument can be calculated with
R =SWX· vGround
Y· b (3.74)
[Larson and Wertz, 1999, p.287] with
• the swath width SW [m]
• the cross-track ground pixel resolution at nadir X [ mpixel
]
• the ground-track velocity of the spacecraft vGround [ms
]
• the along-track ground pixel resolution at nadir Y [ mpixel
]
• and the number of bits used to encode each pixel b [ bitspixel
].
The term SW
X· vGround
Ycan be thus understood as the total number of pixels taken by the instru-
ment during one second. As the swath width SW represents the coverage of the instrument,
it is assumed that it is a function of the orbital altitude hOrbit, so
SW = 2 · ECAmax (3.75)
with the Earth Central Angle ECA whose absolute maximum can be calculated by means of
equation (A.3) since ECA = 90◦ − ρ. In practice the finally chosen ECA for the design of
Page 79
Non-Dimensional Parameters
the payload is less than the theoretical maximum in order to prevent distortions in the pixels
on the limits of the coverage area due to the curvature of the Earth and the position of the
instrument to the scanned area. Thus, as in case of the communication path length SCom in
section 3.4.4, ECA is a function of the orbital altitude hOrbit and the elevation angle ε. For the
sake of simplification, however, it is assumed in the following, that the width of the coverage
area is only a function of the orbital altitude (or, also possible, a function of the orbital period
tOrbit as with tOrbit = 2π√
(REarth+hOrbit)3
µ) like in the case of ECAmaxabs).
Furthermore, the ground velocity of the spacecraft is also a function of the orbital period tOrbitas
vground =2 · π ·REarth
tOrbit(3.76)
Thus, taking all these explanations into consideration, the payload for a Earth Observation
mission can be described by
0 = f(R, tOrbit, b, λP/L, D) (3.77)
and it is
ΠP/LEarth Observation 1 =R · tOrbit
b·(λP/LD
)a(3.78)
With the help of equations (3.73) and (3.74), this non-dimensional product can also be con-
firmed physically in terms of the intraregime approach [Bhaskar and Nigam, 1990]: supposing
a constant ΠP/LEOover a satellite class, an increase in b leads to an increase in the data rate
R. An increase in the aperture length D causes a decrease in the ground resolutions X and
Y , thus causes a better resolution and hence also a higher data rate R. An increase in tOrbitthus in hOrbit causes higher ground track resolutions, thus worse resolution, and decreases the
coverage area by decreasing SW and the ground velocity vground, both effects leading to lower
data rate R.
As mass and power of payload and spacecraft play an important role in the dimensioning of
the payload, these quantities shall also be taken into account:
ΠP/LEarthObservation 2 =R · tOrbit
b·(λP/LD
)a·(mS/C
mP/L
)b·(PS/CPP/L
)c(3.79)
ΠP/LEarthObservation 2 can be equally used for the Top-Down as well as for the Bottom-Up design
approach.
Space Science missions are often missions where the payload measures special wavelengths
emitted by a certain object. So the design of an space science instrument is in general not
very different to an Earth Observation instrument: an object with distance l to the satellite
emits a signal with a special wavelength λP/L which, in turn, is detected by an instrument
on-board of the satellite with a lens diameter D. The data rate R is consequently produced.
So based on ΠP/LEO 3 the following non-dimensional product can be derived
ΠP/LSpaceScience 1 =R · l
c
b·(λP/LD
)a·(mS/C
mP/L
)b·(PS/CPP/L
)c(3.80)
Page 80
Dimensional Analysis of a Single Satellite
with l as the distance to the measured object and c the velocity of sound and thus lc
is the time
the signal needs to get to the satellite. Another difference between the Earth Observation and
Space Science mission will be seen in the following section 3.5 where the Mission Performance
Index will be presented.
In terms of the Technology Demonstration, the payload can be any kind of component from
any kind of subsystem thus no further non-dimensional parameter is required to express it. It
is perfectly described by the ratios and already created parameters for the subsystems and the
spacecraft. The only aspect added is pointed out in section 3.5, where the mission-driving
parameters are presented. Those will be different from mission to mission and thus those
of a Technology Demonstration mission are different from those of the Earth Observation
mission and Space Science mission. To already take it upfront: important parameters for
the Technology Demonstration mission are the mass of the payload mP/L, the volume of
the payload VP/L, the satellite lifetime tSatellite and also the data rate R which results from
observing the performance of the tested component from the ground station.
3.5. Mission Performance Index
An all-embracing scaling law shall be found in this section which expresses the Mission Perfor-
mance Index, including implicitly all the relevant non-dimensional products and ratios which
are assumed to be significant for influencing the performance of a mission. The MPI shall be
designed to take values in a given range, for example between 0 and 100, in order to be able to
compare different missions simply by their numerical values of their MPI. The higher the MPI,
the ”better performant” is the mission. In order to get a relation between the MPI and the
non-dimensional parameters and ratios of section 3.3 and 3.4, it is not possible to simply insert
the non-dimensional parameters in the MPI as they are as they do not express compellingly a
better performance with higher numerical value. The non-dimensional parameters are build to
save their constant value within a satellite class when the quantities generating them change.
First of all, it shall be defined what a ”performant mission” is. In order to do so, it is
important to define the Mission Performance Parameter (MPP) for every type of mission,
thus the quantities whose magnitude shall be as large as possible for a successful mission. We
differentiate between primary and secondary top-level MPPs. The primary MPPs express the
most important quantities of the mission, the latter describing quantities which are significant
for the mission accomplishment, too, but which are not considered to be the main parameters
of the mission. This differentiation is used for the order of maximization: a primary MPP
has to be maximized for a ”performant” mission. The simultaneous maximization of the
secondary parameters, however, is desirable, but not critical. ”Top-Level” means that only the
overall characteristics are considered. As we will see later on, a satellite can be also ”of high
performance” for a maximum or minimum of subsystems’ characteristics.
With regard to the already completed, planned and current CubeSat missions [Thomsen, 2010],
we identified three types of missions on which we will focus on, namely Earth Observation mis-
Page 81
Mission Performance Index
sions (EOM), Space Science missions (SSM) and Technology Demonstration missions (TDM).
Their Mission Performance Parameters can be found in table 3.19.
Table 3.19.: Top-level Mission Performance Parameters
Mission type Primary MPPs Secondary MPPs
EOM R mP/L, VP/L, tSatellite
SSM R mP/L, VP/L, tSatellite
TDM R, mP/L, VP/L, tSatellite -
For an Earth Observation mission, for example, a maximization of the data rate R is de-
sirable: the higher the data rate, the more performant is the mission. A good performant
mission would make a simultaneous maximization of the secondary parameters, the mass of
the payload mP/L, its volume VP/L and the satellite lifetime tSatellite, also possible - these
maximizations, however, are considered to be secondary. The numerical value for the MPI
of an Earth Observation mission should therefore express the achievable maximum datarate
ideally under the condition of maximum possible payload mass and volume as well as satellite
lifetime. As already mentioned above, a satellite can be also considered to be performant in
terms of its subsystems characteristics. It is, for example performant when as much power PSAas possible is produced by the solar cells of the satellite. The ratios and non-dimensional pa-
rameters presented in sections 3.3 and 3.4 were thus investigated for their Mission Performance
Parameters. In terms of subsystems’ characteristics, we define a satellite to be performant
when CBattery is minimized (this is not valid for payloads which are only active in eclipse; here a
performant mission would have a maximized CBattery) and PSAbol/eoland HRW are maximized.
The definitions are based on the argumentation that it is possible to influence the choice of
an orbit in LEO for a mission so that no eclipse occurs. Hence a mission that needs as less
battery power as possible is a performant mission. However, disturbations cannot be avoided
for LEO satellites. Therefore, a mission that is prepared to compensate as much disturbances
as possible, is a performant mission, thus HRW needs to be maximized.
Please note for the understanding of further explanations, that the consequence of a maximum
mP/L is that for the Top-Down approach
{mS/C
mP/L
,mAOCS
mP/L
,mPower
mP/L
,mStructure
mP/L
,mThermal
mP/L
,mC&DH
mP/L
,mCom
mP/L
,mProp
mP/L
}are minimized
(3.81)
and for the Bottom-Up approach{mAOCS
mS/C
,mPower
mS/C
,mStructure
mS/C
,mThermal
mS/C
,mC&DH
mS/C
,mCom
mS/C
,mProp
mS/C
}are minimized (3.82)
Page 82
Dimensional Analysis of a Single Satellite
andmP/L
mS/C
are maximized. (3.83)
mCom
mS/C, for example, shall be minimized, expressing that as much mass as possible shall be
assigned to the payload. The remaining subsystems shall be as light weighted as possible.
Therefore the interest in a low mCom. This remark is especially important for the arrangement
of the ratios in the non-dimensional parameters and thus in the MPI later on.
In order to fall back on the results expressed in the non-dimensional parameters, we rearrange
the quantities in them for use in the MPI. To understand the following explanations better,
the ideas and resulting proceeding are shown exemplarily by means of the Communication
subsystem and its non-dimensional parameter ΠCom 2. The key driving parameter for the
mission success which can be found in ΠCom 2 is the data rate R that shall be maximized. For
the sake of simplification ΠCom 2 is re-expressed to
ΠCom 2 =R3
kCom(3.84)
with
kCom =PTransmitter
mAntennaT · h2Orbit
·(DTransmitter
λCom
)a·
·(DReceiver
λCom
)b·(PTransmitter
PS/C
)c·(mS/C
mCom
)d·(
1
BER
)e (3.85)
R, in turn, is maximal, for a maximal kCom - for ΠCom 2 being constant which is the condition
for applying the similitude theory within a satellite class. Please note that kCom is arranged
in a way that the conditions of the mass ratios explained above are satisfied within the kCom.mCom
mS/C, for example, shall be minimized. This is equivalent to the condition that its reciprocal
mS/C
mComshall be maximized. That is the reason why the latter ratio is presented in kCom as kCom
can be only maximized when its single factors are maximized.
Concludingly, the MPI will include a part (R3 ·kCom)χ which expresses a good performant mis-
sion when maximized. The same proceeding can be done with the remaining non-dimensional
parameters so that the MPI can be defined for both the Bottom-Up and the Top-Down ap-
proach asMPI = (ΨP/L
α) · (ΨAOCS)β · (ΨPower)χ · (ΨStructure)
δ·· (ΨThermal)
ε · (ΨC&DH)φ · (ΨCom)ϕ · (ΨProp)γ
(3.86)
with ΨCom = R3 ·k. The exponents α, β, χ, . . . express a weighting in terms of the importance
of every subsystem to the performance of the mission. Thus, the higher the exponent will be,
the higher is the significance of the corresponding subsystem. A possible value range of them
for the three chosen mission types is presented in table 3.20.
Please notice that Ψi is a dimensional quantity. However, as long as one is aware of this fact
and uses the same dimensions for the quantities for the MPIs of the missions one wants to
Page 83
Mission Performance Index
compare, only the numerical values of the Φs are important for the numerical value of the
MPI. This idea is leaned from aeronautics where the performance indices are also dimensional.
For example the specific fuel consumption (SFC) is defined as
SFC =mFuel
F(3.87)
thus being the mass of fuel mFuel needed to provide the net thrust F for a given period and
thus often having the unit gkN ·s .
However, a normalization of the MPI avoids the dimensionality of MPI. We will see at the end
of this section which different kinds of normalization are possible.
The other dimensional Ψi based on the results in section 3.4 for the Bottom-Up approach are
for the payload
ΨP/L BU = R · kP/L = R · b
tOrbit·(
D
λP/L
)a·(mP/L
mS/C
)b·(PP/LPS/C
)c(3.88)
For the AOCS we can write
ΨAOCS BU = HRW · kRW
= HRW ·mS/C · x2
eq
tOrbit·(mS/C
mRW
)a·(xeqrRW
)b·(mS/C
mAOCS
)c (3.89)
The Power subsystem is described by batteries and solar arrays. Therefore ΨPower BU consists
of a part related to the batteries and one referring to the solar arrays:
ΨPower BU =1
ΨBattery BU
·ΨSolar Array BU (3.90)
with
ΨBattery BU = CBattery · kBattery
= CBattery · PEclipse · tEclipse ·(
PS/CPEclipse
)a·(mS/C
mPower
)b (3.91)
andΨSolar Array BU = P 1+a+c
SolarArraybol· kSolar Array
= P 1+a+cSolarArraybol
· x2eq · SSolar · d · tSatellite·
· P aSolarArrayeol
·(mS/C
mPower
)b· P c
S/C ·(tEclipsetOrbit
)d (3.92)
Please note that CBattery shall be minimized for a performant mission, thus CBattery · kBatteryshall be minimized. This is equivalent to a maximization of 1
CBattery ·kBatterywhich is demanded
for a maximum MPI. That is why the reciprocal of ΨBattery is taken into account in ΨPower.
The subsystems C&DH, Structure, Thermal and Propulsion where not further investigated in
the sections 3.3 and 3.4. For the latter two this is especially because of their lack of importance
Page 84
Dimensional Analysis of a Single Satellite
for the CubeSats. The Structure and C&DH subsystem where not further investigated because
of their simplicity in CubeSats. They are most easily described by a simple mass ratio. Further
development of the ratios and non-dimensional parameters is surely possible and desirable
but as this work is a first order modeling of a satellite by dimensional analysis, the following
definitions of the Ψs are considered to be sufficient for our purposes. For the Structure
subsystem
ΨStructure BU =mS/C
mStructure
(3.93)
and for the C&DH subsystem we can write
ΨC&DH BU =mS/C
mC&DH
(3.94)
The Thermal subsystem can be represented by
ΨThermal BU =mS/C
mThermal
(3.95)
and the Propulsion subsystem by
ΨPropulsion BU =mS/C
mPropulsion
(3.96)
The above explanations refer to the Bottom-Up approach. As the ratios and non-dimensional
parameters change for the Top-Down approach, the Ψs shall be presented for that approach
in the following. It is for the Earth Observation payload
ΨP/L TD = R · kP/L = R · b
tOrbit·(
D
λP/L
)a·(mP/L
mS/C
)b·(PP/LPS/C
)c(3.97)
and for the AOCS subsystem
ΨAOCS TD = HRW · kRW = HRW ·mS/C · x2
eq
tOrbit·(mP/L
mRW
)a·(xeq/P/LrRW
)b·(mP/L
mAOCS
)c(3.98)
The Power subsystem is as in the case of the Bottom-Up approach represented by a battery
related and a solar array related part, leading to
ΨPower TD =1
ΨBattery TD
·ΨSolar Array TD (3.99)
with
ΨBattery TD = CBattery · kBattery
= CBattery · PEclipse · tEclipse ·(
PP/LPEclipse
)a·(mP/L
mPower
)b (3.100)
Page 85
Mission Performance Index
andΨSolar Array TD = P 1+a+c
SolarArraybol· kSolar Array
= P 1+a+cSolarArraybol
· x2eq · SSolar · d · tSatellite·
· P aSolarArrayeol
·(mP/L
mPower
)b· P c
P/L ·(tEclipsetOrbit
)d (3.101)
A simple mass ratio represents as in the case of the Bottom-Up approach the subsystems
Structure, C&DH, Thermal and Propulsion. So
ΨStructure TD =mP/L
mStructure
(3.102)
ΨC&DH TD =mP/L
mC&DH
(3.103)
ΨThermal TD =mP/L
mThermal
(3.104)
and
ΨPropulsion TD =mP/L
mPropulsion
(3.105)
For the Communication subsystem we can write
ΨCom TD = R3 · kCom
=PTransmitter
mAntennaThOrbit2 ·(DTransmitter
λCom
)a·
·(DReceiver
λCom
)b·(PTransmitter
PP/L
)c·(mP/L
mCom
)d· 1
BER
e
(3.106)
Please note that equations (3.88) to (3.105) include a choice of the non-dimensional param-
eters and ratios presented in sections 3.3 and 3.4. Not every parameter of sections 3.3 and
3.4 is integrated into the MPI but the parameters which are taking the most quantities into
consideration. Furthermore the parameters of section 3.4.1 provide a top-level approach to
the design. They can thus, as mentioned in that section, be used for quick design estimations
or provide a possibility to compute a missing input quantity for the Top-Down and Bottom-
Up design approach, but are not integrated into the MPI. However, the MPI can be further
developed as it was the case for the non-dimensional parameters, or also newly composed.
Nonetheless, a standardized definition of the MPI is considered to be essential and desirable
in order to be able to compare and exchange results through the satellite world.
A further important remark is that equation (3.86) is valid for any kind of mission or satellite.
The numerical values of the Φi, ki and the exponents change, however, depending on the mis-
sion and the satellite. Table 3.20 gives an overview of the estimated ranges for the exponents
α, β, χ, . . . for the three mission types we identify to be the most frequent and important
ones. As already mentioned, the higher the exponent, the higher is the functional importance
Page 86
Dimensional Analysis of a Single Satellite
of the subsystem for the mission accomplishment. The Earth Observation and the Space Sci-
ence missions can be considered together, as the only difference lies in the exponents for the
Thermal subsystem. This will be explained later on. The exponent ranges for the Technology
Demonstration mission are bigger as various payloads are possible so that the subsystems are
differently engaged.
The payload as most important part of the spacecraft has the highest exponent. The Power
and Communication subsystems have the second biggest exponent, as without power the
satellite and most importantly the payload cannot work. Without a working Communication
subsystem a satellite is operational but the value of the mission, the collected data, cannot
be transmitted to the Earth. In case of operational problems the mission is often lost because
of the lack of command and health monitoring links. The C&DH subsystem is responsible
for the data processing and distribution. Its significance for the mission success is almost
as important as the Communication subsystem. Once in orbit, the AOCS and Propulsion
subsystem basically carry out the same function, namely the compensation of disturbances.
Their significance is determined by the required payload accuracy. Here, among others con-
sidering CubeSat missions where a minimum of accuracy is required, a medium importance
is chosen for the AOCS and Propulsion subsystem to take CubeSats and more sophisticated
missions into consideration. A low exponent range is assigned to the Structural subsystem
assuming that the design principally gets on without mechanisms. Every mechanism, however,
adds complexity and thus the exponents should be adjusted for systems where critical parts
needs deployment, for example an antenna. The Thermal subsystem concludes the list of
subsystems, being generally the less important subsystem in most of the CubeSat missions.
However, in special cases, it is especially this subsystem which can get really critical and decide
on the mission success. It is partly taking account of this fact by widening the range for the
Space Science missions to higher exponents and thus to higher importance of the Thermal
subsystem. As satellites of Space Science missions can be placed literally everywhere in the
universe, thermal aspects can get significant when being far in the outer solar system or in
the near of the Sun. The Thermal subsystem can also be of high importance for sensitive
components and the payload on-board. However, in the case of CubeSats, Space Science and
Earth Observation missions are both executed in Earth orbits where most satellites need a
passive thermal subsystem at the most.
Page 87
Mission Performance Index
Table 3.20.: Ranges for the subsystem exponents α, β, χ, . . . for single satellites
Subsystems EOM SSM TDM
Payload 0.8-1.0 0.8-1.0 0.8-1.0
Power 0.7-0.9 0.7-0.9 0.6-0.9
Communication 0.7-0.9 0.7-0.9 0.6-0.9
Structure 0.2-0.5 0.2-0.5 0.2-0.5
AOCS 0.5-0.7 0.5-0.7 0.4-0.8
Thermal 0.1-0.3 0.1-0.6 0.1-0.6
C&DH 0.6-0.8 0.6-0.8 0.5-0.8
Propulsion 0.5-0.7 0.5-0.7 0.4-0.8
The product approach for the composition of the MPI entails the fact that a non-dimensional
parameter which takes the value zero, leads to a Mission Performance Index of zero. In order
to avoid this result, non-dimensional parameters which are equal to zero shall be considered
during the calculation of the MPI with a very low value for Ψ close to zero, for example 10−20,
assuming that a complete lack in a subsystem is not conducive for the mission performance.
This can for example be the case for the Propulsion subsystem on CubeSats. This assumption
makes it still possible to compare satellites which are not equally equipped with subsystems.
Otherwise, only ignoring the subsystem in the MPI could potentially lead to better results for
satellites with a lacking subsystem.
As the numerical values of the non-dimensional parameters are theoretically not limited, so it
is with the numerical value of the MPI. In order to avoid unwieldy large values to compare the
MPIs of different satellites more easily, it is therefore recommended to normalize the numerical
values of the MPI so that it can only take values in a range, for example 0 < MPI ≤ 100. The
normalization can be classically done by means of a reference mission. As a reference mission is
always difficult to determine, another possibility shall be presented here on which we will focus.
It is the normalization by means of a numerical value based on the highest MPI value within a
satellite group which shall be investigated. It can for example simply be a hundredth of the MPI
of the mission with the highest MPI. Another possibility is to take for example 110% of this
value as normalization figure, leaving still some space to the upper MPI border in case further
satellites shall be included into the group with higher MPIs than the actual highest performant
one. The normalization can also be already done for the single Ψs, thus for each subsystem
separately by dividing the Ψis with the highest numerical value of a Ψi within the group of
satellites one wants to compare. This leads to 0 < Ψi ≤ 1 and thus, because of the exponents
α, β, γ, · · · ≤ 1 to a MPI with a value range between zero and one. The quality of the results
is in any case the same and a useful consequence of the normalization is also the regain of a
non-dimensional MPI or Ψi. Important for the normalization as well as general remark is the
Page 88
Dimensional Analysis of a Single Satellite
fact that the numerical value of the MPI when using the Top-Down related Ψs is different
from the MPI when using the Bottom-Up related Ψs. Thus the same mission with exactly the
same parameters leads to two different numerical values for the MPI. This difference is based
on the nature of the Ψs as they are differently composed for the two approaches. However, to
avoid as much misunderstandings as possible, it is recommended to always indicate if a MPI
is calculated with the Top-Down or the Bottom-Up approach. Furthermore, normalizations
are recommended to be seperately done for Top-Down and Bottom-Up MPIs. Other than
this, the normalization can be done within a group of satellites one wants to compare, or in a
wider database, taking ideally all ever flown satellites into account. A normalization within a
mission type, however, is recommended, taking only the satellites with same mission type into
account to compare the numerical results more easily. The only disadvantage of that kind of
normalization is the fact that all the normalized MPIs have to be recalculated when a new
mission happens within a group whose MPI or Ψi is bigger than the recent ones. In order
to avoid the recalculation as often as possible, it is recommended to normalize the MPIs not
with the highest MPI in the group but with a figure which is bigger than even the highest MPI
as already suggested above. A wide application of the MPI will help to crystallize a suited
normalization figure.
In the end, when enough data on flown satellites is available so that a whole database of MPIs
can be created, it is assumed that it will be possible to assign MPI ranges to satellite classes.
Two further applications of the MPI than simply qualitative comparisons of the missions will
be possible. The assignment of MPI ranges to satellite classes will firstly enable the user
of the database to see if its finished satellite design with the corresponding calculated MPI
fits the ranges of the original satellite class or if its design is actually in another satellite
class. Secondly, being redirected to another satellite class with the intended MPI, means to
be redirected within the non-dimensional parameters which are also different from satellite
class to satellite class. With a definition of ranges for the input parameters of the Top-Down
and Bottom-Up approach for the different satellite classes, the user can than recalculate its
satellite design by means of the new non-dimensional parameters.
3.6. Application of the Results
It is now important to see how the ratios, non-dimensional parameters and the Mission Per-
formance Index can be practically used. First of all, it is important to determine numerical
values for the ratios and non-dimensional parameters of a satellite class. This procedure neces-
sitates the validation of the parameters with data from real flown satellites as already shown
for the ratios and the top-level approach for the NGSO-satellites. Once the non-dimensional
parameters are determined for a satellite class, they can be used for designing a spacecraft
with as little as four main input parameters and some further secondary design inputs such as
the type of the solar cells. Figures 3.69 and 3.70 show the interdependencies of the quantities
in the system which are expressed in the results of sections 3.3 and 3.4. The figures shall
help calculating the unknown quantities of a given system by indicating the equation number
Page 89
Application of the Results
and the input and output quantities of the equations. The Mission Performance Index of the
satellite can finally be calculated, based on the results of the previous steps and indicating the
mission accomplishment which is achievable with the satellite. In this section, it will be shown
how the calculations are to be done by means of two different satellite types, the CubeSat
and the NGSO satellite. Hereby, hypothetical input values will be used, modeling the ”most
likely” CubeSat and NGSO-satellite. Most of the assumptions for the CubeSat architectures
will be based on the few data which was gathered together in an Excel-Sheet during this
work. The information originates from personal communication with the different student
CubeSat team around the world, the component database described in Appendix B as well as
two main websites on CubeSats [Thomsen, 2010] and [Amsat, 2010]. The calculation of the
non-dimensional parameters as well as the computation of the unknown quantities by means
of the non-dimensional parameters will be done together as the determination of most of the
non-dimensional parameters requires a number of assumptions. A third example, namely the
application of the equations for a larger 2U CubeSat, will be done to conclude this section in
order to show the sole calculation of the unknown quantities and the MPI for a satellite with
a few inputs.
3.6.1. The CubeSat - a Bottom-Up Approach
In order to make all the exemplary calculations for a 1U CubeSat, it is important to determine
the most important input quantities of the system in the beginning. According to figures 3.2
and 3.70 the main input quantities are mS/C , xeq or VS/C and PS/C . Based on the 1U CubeSat
standard and the available information on the flown CubeSats, we assume for the calculations
• mS/C = 1.33kg
• xeq = 104.31mm
• PS/C = 1W .
Furthermore, we will already assume an Earth Observation mission for the satellite, an esti-
mated satellite lifetime tSatellite of 1 year and an orbital altitude hOrbit of 700 km although
this is strictly not a pure Bottom-Up approach as presented in figures 3.2 and 3.70. However,
for a first order approximation as this is here the case, this estimations shall be allowed.
The determination of the Πs and the calculation of the unknown quantities
1. The mass ratios can be determined by the information about the COTS components
(see Appendix B) and the flown CubeSats in table 3.3 and 3.4. We chose the numer-
ical values of the mass ratios as presented in table 3.21 and 3.22 and calculate the
corresponding subsystem masses, also presented in the tables.
Page 90
Dimensional Analysis of a Single Satellite
Table 3.21.: Mass ratios of the payload and the subsystems AOCS, Power and Structure build
with mS/C CubeSat and the corresponding subsystem masses for the given mass
ratios for the hypothetical 1U CubeSat
P/L AOCS Power Structure
Mass ratio [%] 25 15 15 25
Subsystem mass [g] 332.5 199.5 199.5 332.5
Table 3.22.: Mass ratios of the subsystems Thermal, C&DH and Communication build with
mS/C CubeSat and the corresponding subsystem masses for the given mass ratios
for the hypothetical 1U CubeSat
Thermal C&DH Com Prop
Mass ratio [%] 0 3 17 0
Subsystem mass [g] 0 39.9 226.1 0
2. The volume ratios or packing factor are defined according to (3.5) and (3.7). The
collected data of the already flown CubeSats and by own estimations, a packing factor
p = VusedVS/C
of 80% can be expected for the 1U CubeSats. The ratios VBus
VS/Cand
VP/L
VS/C
cannot fall back on given data, so estimations are necessary. We assume that(VBusVS/C
)CubeSat
= 0.6− 0.7(VP/LVS/C
)CubeSat
= 0.1− 0.2
(3.107)
With VS/C = 1.135 · 10−3 m3 this leads us to
VBus CubeSat = 6.81 · 10−4 − 7.945 · 10−4 m3
VP/L CubeSat = 1.135 · 10−4 − 2.27 · 10−4 m3(3.108)
which equals equivalent edge lengths of
xeq Bus CubeSat = 8.80 · 10−2 − 9.26 · 10−2 m
xeq P/L CubeSat = 4.84 · 10−2 − 6.1 · 10−2 m(3.109)
Page 91
Application of the Results
3. The power ratios can be determined with (3.13). After the consultation of the Exo-
planet student team from MIT and the MOVE team from TUM and by own estimations,
it can be expected that (PP/LPS/C
)CubeSat
= 0.3(PBusPS/C
)CubeSat
= 0.7
(3.110)
thus with PS/C = 1W leading to PP/L = 0.3W and PBus = 0.7W .
4. The power and mass ratios (3.14) to (3.17) can be thus calculated to
rMass Power 1 CubeSat =mS/C
mP/L
·PS/CPP/L
= 4 · 1
0.3= 13.333
rMass Power 2 CubeSat =mP/L
mS/C
·PP/LPS/C
= 0.25 · 0.3 = 0.075
rMass Power 3 CubeSat =mS/C
mP/L
·PP/LPS/C
= 4 · 0.3 = 1.2
rMass Power 4 CubeSat =mP/L
mS/C
·PS/CPP/L
= 0.25 · 1
0.3= 0.833
(3.111)
5. The top-level approach with the non-dimensional parameters (3.19) and (3.20) can
be calculated for the hypothetical 1U CubeSat to be
Π1 CubeSat = Π2 CubeSat =PS/C · t3Orbitx2eq ·mS/C
= 308.260 (3.112)
with tOrbit calculated to be 1.64hours. The orbit time is chosen to be inserted in hours
in order to keep the numerical value of Π1 CubeSat as small as possible for manipulation
reasons.
6. The Power subsystem is expressed with the non-dimensional parameters (3.34) for
the battery and (3.46) for the solar array. Both equations as well as the upcoming
parameters of the remaining subsystems possess still unknown exponents expressing the
importance of the corresponding ratios for the non-dimensional parameter. However, as
this is only a first order approximation, we assume all unknown exponents to be equal
to one. An example of this simplification shall be given in terms of ΠBat3 . It is generally
determined by
ΠBat3 =CBattery
PEclipse · tEclipse·(
PS/CPEclipse
)a·(mS/C
mPower
)b(3.113)
Page 92
Dimensional Analysis of a Single Satellite
The exponents a and b being simplified to one, we get
ΠBat3 CubeSat =CBattery
PEclipse · tEclipse·PS/CPEclipse
·mS/C
mPower
=36000Ws
1W · 600s· 1W
1W· 1
0.15
= 400
(3.114)
assumingPEclipse
PS/C= 1 and CBattery = 36000s based on technical data on the corre-
sponding COTS products.
For the solar array (body-mounted) equation (3.46) leads us to
ΠSA5 CubeSat =PSAbol
x2eq · SSolar · d · tSatellite
·
· PSAbol
PSAeol
· mPower
mS/C
· PSAbol
PS/C· tOrbittEclipse
=1.2W
(0.10431m)2 · 1367 Wm2 · 0.0275 · 1year
·
· 1.2W
1.1W· 0.15 · 1.2W
1W· 98.77min
10min
= 5.690
(3.115)
Please note that the maximum eclipse time in an orbital altitude of 700km would be
35 min. We assume 10 min eclipse time for unpredicted conditions as CubeSats are
generally launched in sun-synchronous orbits where no eclipse is expected. We also as-
sume a maximum incident sun angle of 90◦ and a solar constant of 1367 Wm2 . Another
assumption is that Gallium Arsenide cells are used with a yearly degradation d of 2.75%.
Furthermore, with an assumed eclipse time of 10 min and equation (3.45), we can cal-
culate PSAeol= 1.1W . With d = 0.0275 and and estimated satellite lifetime tSatellite
of one year, PSAbolcan be computed to 1.13W with [Larson and Wertz, 1999, p.417],
rounded up to 1.2W.
7. The AOCS subsystem is represented by the calculation of a reaction wheel by means
of equation (3.60). Assuming the reaction wheel RW1 Type B of the company Astro-
und Feinwerktechnik GmbH to be used on the hypothetical CubeSat as tested on BeeSat
we get
ΠRW3 CubeSat =HRW · tOrbitmS/C · x2
eq
· mRW
mS/C
· rRWxeq· mAOCS
mS/C
=0.0001Nms · 5926.2s
1.33kg · (0.10431m)2· 0.012kg
1.33kg· 21mm
104.31mm· 0.15
= 0.011
(3.116)
Page 93
Application of the Results
8. The Communication subsystem is represented by equation (3.71). With the collected
data on the flown CubeSats, is it assumed that mAntennaT = 0.01kg, R = 9.6kbps,
PTransmitter = 0.5W , the frequency of the signal f = 437Mhz which leads to λCom =
0.6858m and BER = 10−4. This leads to
ΠCom2 CubeSat =mAntennaT · h2
Orbit ·R3
PTransmitter· λComDTransmitter
·
· λComDReceiver
·PS/C
PTransmitter· mCom
mS/C
·BER
=0.01kg · (700km)2 · (9.6kbps)3
0.5W· 0.6858m
0.3m·
· 0.6858m
1m· 1W
0.5W· 0.17 · 10−4
= 462.160
(3.117)
9. The Payload subsystem of an Earth Observation mission can be represented by equa-
tion (3.79). Based on explanations in [Larson and Wertz, 1999, Chapter 9] and data of
real flown CubeSats, we assume b = 8 bitpixel
, λP/L = 7 · 10−7m and D = 0.02m and a
ground-pixel resolution X of the instrument of 60 m. This leads to
ΠP/LEarthObservation CubeSat =R · tOrbit
b·λP/LD·mS/C
mP/L
·PS/CPP/L
=9.6kbps · 5926.2s
8 bitpixel
· 7 · 10−7m
0.02m· 4 · 1W
0.3W
= 3.319
(3.118)
The Mission Performance Index
With equations (3.86) and (3.88) to (3.96) and table 3.20 we can now calculate the Mission
Performance Index of the 1U CubeSat.
It is for the payload
(ΨP/L CubeSat)α =
(R · kP/L
)α= (9.6 · 2.893)1
= 27.773
(3.119)
The AOCS system is represented by
(ΨAOCS CubeSat)β = (HRW · kRW )β
=(0.0001 · 8.96 · 10−3
)0.5
= 9.47 · 10−4
(3.120)
Page 94
Dimensional Analysis of a Single Satellite
As already mentioned in section 3.5, the modeled Power subsystem consists of batteries and
solar arrays, thus
(ΨPower CubeSat)χ =
(1
ΨBattery
·ΨSolar Array
)χ=
(1
CBattery · kBattery· P 3
SAbol· kSolarArray
)χ=
(1
36000 · 4000· 1.23 · 0.304
)0.8
= 1.777 · 10−7
(3.121)
The Structure subsystem as well as the C&DH subsystem are presented simply by their mass
ratios.
(ΨStructure CubeSat)δ =
(mS/C
mStructure
)δ=
(1
0.25
)0.3
= 1.516 (3.122)
(ΨC&DH CubeSat)φ =
(mS/C
mC&DH
)φ=
(1
0.03
)0.6
= 8.198 (3.123)
We assume that the hypothetical CubeSat does not have a Thermal subsystem. Thus the rule
applies, that its Ψ is equal to 10−20.
(ΨThermal CubeSat)ε =
(mS/C
mThermal
)ε= (10−20)0.2 = 10−4 (3.124)
The Communication subsystem is calculated with
(ΨCom CubeSat)ϕ =
(R3 · kCom
)ϕ=(9.63 · 1.914
)0.8
= 382.885
(3.125)
And finally, as in case of the Thermal subsystem, the Propulsion subsystem is represented by
a Ψ of 10−20.
(ΨPropulsion CubeSat)γ =
(mS/C
mPropulsion
)γ= (10−20)0.5 = 10−10 (3.126)
Thus, with a choice of the exponent α, β, χ, . . . from table 3.20, we can write
MPICubeSat = (ΨP/L CubeSat)α · (ΨAOCS CubeSat)
β · (ΨPower CubeSat)χ·
· (ΨStructure CubeSat)δ · (ΨC&DH CubeSat)
φ · (ΨThermal CubeSat)ε·
· (ΨCom CubeSat)ϕ · (ΨPropulsion CubeSat)
γ
= 27.773 · (9.47 · 10−4) · (1.777 · 10−7) · 1.516 · 8.198 · 10−4 · 382.885 · 10−10
= 2.224 · 10−5 · 10−4 · 10−10
(3.127)
Page 95
Application of the Results
and separately showing the factors 10−4 and 10−10 due to the lack of the Propulsion and
Thermal subsystem.
Please note that a Mission Performance Index only makes sense in comparison to the MPI
of other satellites. This is the reason why a normalization is recommended for the MPI. The
normalization can be only done when another satellite can be used for comparison.
3.6.2. The NGSO-Satellite - a Top-Down Approach
In a Top-Down-approach for a Technology Demonstration mission the payload parameters
mP/L, xeqP/L or VP/L and PP/L and tSatellite are the input variables of the design approach.
As figure 3.69 deals with an Earth Observation mission, the parameters R, X and Y are
shown as further input variables. However, the payload of an hypothetical NGSO-satellite is
in general its Communication subsystem, especially the antenna. Thus the Communication
subsystem is used in two ways, as provider of the communication link to the ground station
and other satellites of the constellation but also as payload, providing different services for the
user. This double function will be reflected in the exponent γ when calculating the MPI. In
terms of the input variables this means that X and Y are unnecessary parameters, only the
datarate R counts. However, R and hOrbit are connected by the relatively complex equations
of the link budget [Larson and Wertz, 1999, p.550 ff.]. In order to avoid those equations as
this is only a first order approximation, we assume that hOrbit and R are given simultaneously.
Based on the available information about NGSO-satellites, we assume for the calculations
• mP/L = 100kg
• xeq P/L = 0.5m
• PP/L = 500W
• R = 0.65Mbps
• hOrbit = 1000km
• tSatellite = 5years
The determination of the Πs and the calculation of the unknown quantities
1. The mass ratios can be determined by means of the results of the empirical validation
of the ratios in section 3.3.1. The ratios are set equal to the arithmetical averages of
the mass ratios of the database satellites presented in tables 3.7 and 3.8.
Page 96
Dimensional Analysis of a Single Satellite
Table 3.23.: Mass ratios of the payload and the subsystems AOCS, Power and Structure build
with mP/L and the corresponding subsystem masses for the given mass ratios for
the hypothetical NGSO-satellite
S/Cwet S/Cdry AOCS Power Structure
Mass ratio [%] 371.9 304.9 26.2 90.6 61.9
Subsystem mass [kg] 371.9 304.9 26.2 90.6 61.9
Table 3.24.: Mass ratios of the subsystems Thermal, C&DH, Communication and Propulsion
build with mP/L and the corresponding subsystem masses for the given mass
ratios for the hypothetical NGSO-satellite
Thermal C&DH+TT&C Com Propulsion
Mass ratio [%] 22.5 14.7 114.7 28.4
Subsystem mass [kg] 22.5 14.7 114.7 28.4
Please note that not the whole Communication subsystem is considered as payload.
That is the reason why the C&DH+TT&C-ratio is not zero. However, the communica-
tion subsystem is hereby defined to consist of the payload and the C&DH and TT&C
subsystem. That is the reason for the high mass ratio of 114.7 % for Communication,
emerged simply because an addition of the payload and the C&DH+TT&C percentage.
This figure will especially be necessary when designing the Communication subsystem.
2. The packing factor or volume ratios are defined according to (3.7) and (3.5). As
no data is available on the packing factor for the NGSO-satellites, we assume that
pNGSO is about 60% to 70% as the satellites are bigger than CubeSats so that the size
restrictions are less critical. The ratios VBus
VS/Cand
VP/L
VS/Ccannot fall back on given data,
either, so estimations are necessary. We assume that(VBusVS/C
)NGSO
= 0.3− 0.5(VP/LVS/C
)NGSO
= 0.2− 0.3
(3.128)
With VP/L = 0.125 m3 this leads us to
VS/C NGSO = 0.417− 0.625 m3
VBus NGSO = 0.156− 0.260 m3(3.129)
Page 97
Application of the Results
which equals equivalent edge lengths of
xeq Bus NGSO = 0.539− 0.639 m
xeq S/C NGSO = 0.747− 0.855 m(3.130)
Please note that the Bus volume VBus NGSO was build with VS/C NGSO = 0.521m3
which is the arithmetical average of the two results in equation (3.129).
3. The power ratios can be determined with (3.13) and the results of the empirical
validation of the power ratios by means of the NGSO-satellite database in subsection
3.3.3. The ratios are set equal to the arithmetical averages of the power ratios presented
in table 3.11. Please remember that the power ratio results in section 3.3.3 were based
on data for PSAeoland not as intended in equations (3.13) for PS/C .
(PSAeol
PP/L
)NGSO
= 1.451(PBusPP/L
)NGSO
= 0.335
(3.131)
thus with PP/L = 500W leading to PSAeol= 725.5W and PBus = 167.5W and thus
PS/C = PP/L + PBus = 667.5W . This numerical result can be confirmed with equation
(3.45) since
PS/C = PSAeol· 1
1 +tEclipse
tDaylight
PS/C = 725.5W · 1
1 + 8.73min105.12−8.73min
= 665.25W
(3.132)
which is remarkable as it proves the simplicity and the validity of working with non-
dimensional relationships.
4. The power and mass ratios (3.14) to (3.17) can be thus either read in table 3.15
what leads to
rMass Power 1 NGSO =mS/C
mP/L
· PSAeol
PP/L= 5.624
rMass Power 2 NGSO =mP/L
mS/C
·PP/LPSAeol
= 0.212
rMass Power 3 NGSO =mS/C
mP/L
·PP/LPSAeol
= 2.609
rMass Power 4 NGSO =mP/L
mS/C
· PSAeol
PP/L= 0.451
(3.133)
Page 98
Dimensional Analysis of a Single Satellite
or be calculated with the results of tables 3.23 and 3.24 and equation (3.131) to
rMass Power 1 NGSO =mS/C
mP/L
· PSAeol
PP/L= 3.719 · 1.451 = 5.396
rMass Power 2 NGSO =mP/L
mS/C
·PP/LPSAeol
= 0.269 · 0.689 = 0.185
rMass Power 3 NGSO =mS/C
mP/L
·PP/LPSAeol
= 3.719 · 0.689 = 2.562
rMass Power 4 NGSO =mP/L
mS/C
· PSAeol
PP/L= 0.269 · 1.451 = 0.390
(3.134)
5. The top-level approach with the non-dimensional parameters (3.26) can be calculated
for the hypothetical NGSO-satellite with tOrbit instead of tdutycycle to be
ΠP/L 1 NGSO =PP/L · t3Orbitx2eq P/L ·mP/L
=500W · (1.752hr)3
(0.5m)2 · 100kg
= 61.39
(3.135)
with tOrbit calculated to be 105.12min = 1.752hours. The orbit time is chosen to
be inserted in hours in order to keep the numerical value of ΠP/L 1 NGSO as small as
possible for manipulation reasons.
6. The Power subsystem is expressed with the non-dimensional parameters (3.35) for
the battery and (3.47) for the solar array. Both equations as well as the upcoming
parameters of the remaining subsystems possess as in the case of the CubeSat still
unknown exponents expressing the importance of the corresponding ratios for the non-
dimensional parameter. However, as this is only a first order approximation, we assume
all unknown exponents again to be equal to one.
The battery subsystem can be calculated with equation (3.35) to
ΠBat4 NGSO =CBattery
PEclipse · tEclipse·PP/LPEclipse
·mP/L
mPower
=5.4 · 106Ws
667.5W · 523.8s· 500W
667.5W· 1
0.906
= 12.769
(3.136)
with PEclipse = PS/C = 667.5W and tEclipse for hOrbit = 1000km and an inclination of
i = 50◦ is calculated with Appendix A to 8.73min. CBattery was calculated based on
the assumption that 15kg of Lithium-ion batteries with a specific energy of 100Whkg
are
Page 99
Application of the Results
used on-board. Thus the accumulated capacity of the batteries CBattery is 1500Wh,
thus 5.4 · 106Ws.
The solar array is described by equation (3.47) and (3.49). As PSAeolis given by equation
(3.131), we use (3.49):
ΠSA8 NGSO =PSAeol
· d · tSatellitex2eq · S
·
· PSAeol
PSAbol
· mPower
mP/L
· PSAeol
PP/L· tOrbittEclipse
=667.5W · 0.0275 · 5years
(0.8m)2 · 1367 Wm2
· 667.5W
767.4W· 0.906 · 667.5W
500W· 105.12min
8.73min
= 1.330(3.137)
We chose an average equivalent spacecraft edge length of xeq S/C NGSO = 0.8m based
on the results (3.130). We assume a maximum incident sun angle of 90◦ and a solar
constant of 1367 Wm2 . Another assumption is that Gallium Arsenide cells are used with
a yearly degradation d of 2.75%. With d = 0.0275 and and estimated satellite lifetime
tSatellite of 5 years, PSAbolcan be computed to 767.4W with [Larson and Wertz, 1999,
p.417].
7. The AOCS subsystem of an NGSO-satellite in LEO does imply reaction wheels very
rarely. However, the calculations shall be done here for demonstration purposes. The
reaction wheels of the satellite can be designed by means of equation (3.65) in the
Top-Down approach. This leads to
ΠRW4 NGSO =HRW · tOrbitmS/C · x2
eq
· mRW
mP/L
· rRWxeq P/L
· mAOCS
mS/C
=0.7Nms · 6307.2s
371.9kg · (0.8m)2· 2kg
100kg· 0.1m
0.5m· 0.262
= 0.0194
(3.138)
HRW , mRW and rRW are all calculated, based on the assumptions that the overall
disturbance torque is in the order of 10−5. Even the small picosatellite reaction wheel
tested on Beesat can compensate this torque. What is important, however, is the mo-
mentum storage of the reaction wheel. Taking all the disturbances into consideration
for the given hypothetical NGSO-satellite, a reaction wheel with HRW = 0.7Nms is
best suited for our needs. The mass mRW for such a reaction wheel is estimated to be
about 2kg. The radius of the reaction wheel rRW will be about 100mm.
Page 100
Dimensional Analysis of a Single Satellite
8. The Communication subsystem can be best described with equation (3.72), leading
to
ΠCom3 NGSO =mAntennaT · h2
Orbit ·R3
PTransmitter·
· λComDTransmitter
· λComDReceiver
·PP/L
PTransmitter· mCom
mP/L
·BER
=40kg · (1000km)2 · (0.65Mbps)3
50W· 0.1875m
1.5m·
· 0.1875m
5m· 500W
50W· 1.147 · 10−5
= 0.116
(3.139)
As already mentioned in the mass ratio section above, mCom consists of the payload
and the C&DH and TT&C subsystem, that is the reason why mCom
mP/L> 1. The unknown
quantities are estimated based on the NGSO-database and for the antenna mass also
with [Richharia, 1999] and an estimation for DTransmitter.
9. The Payload is the Communication subsystem. Its influence as payload will become
clearer when determining the MPI and especially the exponent ϕ in the next section.
The Mission Performance Index
With equation (3.86), (3.97) to (3.105) and table 3.20 we can now calculate the Mission
Performance Index of the NGSO-satellite. It is
(ΨAOCS NGSO)β = (HRW · kRW )β
= (0.7 · 36.01)0.7
= 9.573
(3.140)
Please note that β has risen from 0.5 to 0.7 in comparison to the CubeSat. This is due to the
higher importance of pointing accuracy for communication satellites.
The Power subsystem is again represented by batteries and solar arrays, leading to
(ΨPower NGSO)χ =
(1
ΨBattery NGSO
·ΨSolar Array NGSO
)χ=
(1
CBattery · kBattery· P 3
SAeol· kSolarArray
)χ=
(1
5.4 · 106 · 4.229 · 105· 667.53 · 2.238 · 108
)0.8
= 3729.48
(3.141)
Page 101
Application of the Results
The Structure subsystem is represented simply by its mass ratio
(ΨStructure NGSO)δ =
(mP/L
mStructure
)δ=
(1
0.619
)0.3
= 1.155 (3.142)
Communication satellites are provided with Thermal Control Systems. This is the reason why
ΨThermal is not 10−20 in comparison to the 1U CubeSat investigation.
(ΨThermal NGSO)ε =
(mP/L
mThermal
)ε=
(1
0.225
)0.2
= 1.348
(3.143)
The Communication subsystem can be calculated to
(ΨCom NGSO)ϕ =(R3 · kCom
)ϕ=(0.653 · 2.37
)1+0.8
= 0.461
(3.144)
As NGSO-satellites are usually equipped with a Propulsion subsystem, ΨPropulsion has to be
considered. Its exponent γ is chosen to be equal to β since AOCS and Propulsion subsystem
fulfill similar functions once the satellite is in orbit.
(ΨPropulsion NGSO)γ =
(mP/L
mPropulsion
)γ=
(1
0.284
)0.7
= 2.414
(3.145)
Please note that the C&DH subsystem is already considered in the Communication subsystem
by addition of its mass ratio to the payload ratio which leads to the communication ratio.
Therefore the C&DH subsystem is not considered separately in the Ψs.
Finally we can write with (3.86) and with the choice of the exponents α, β, χ, . . . from table
3.20
MPINGSO = (ΨAOCS NGSO)β · (ΨPower NGSO)χ · (ΨStructure NGSO)δ · (ΨThermal NGSO)ε·· (ΨCom NGSO)ϕ · (ΨProp NGSO)γ
= 27.773 · (9.47 · 10−4) · 1.155 · 1.348 · 0.461 · 2.414
= 4.559 · 10−2
(3.146)
As in the case of the CubeSats the MPI only makes sense in comparison to further satellites
with MPIs as it is a quantity expressing the value of a mission in relation to other missions.
Page 102
Dimensional Analysis of a Single Satellite
3.6.3. A New CubeSat Standard? - a Bottom-Up Approach
As it was partly seen in section 3.3.1 with the mass ratios of the structure and Communication
subsystem, the 1U CubeSat is assumed to be too small to be advantageous. A closer look shall
therefore be given here in this section to a 2U CubeSat in order to quantify its performance
by means of the Mission Performance Index. Again it is important to determine the most
important input quantities of the system in the beginning. According to figures 3.2 and 3.70
the main input quantities are mS/C , xeq or VS/C and PS/C . Based on the 2U CubeSat standard
and the available information on the flown CubeSats, we assume for the calculations
• mS/C = 2.66kg
• xeq = 131.42mm
• PS/C = 2W
Furthermore, as in the case of the 1U CubeSat, we will already assume an Earth Observation
mission for the satellite, an estimated satellite lifetime tSatellite of 1 year and an orbital altitude
hOrbit of 700 km. This is the same orbital altitude as it was supposed for the 1U CubeSat in
order to change as less quantities as possible in comparison to the 1U CubeSat. By this, the
comparison between the two different sized satellites is simplified. With the same argumenta-
tion, we have also chosen tSatellite.
The calculation of the unknown quantities
With the knowledge of the numerical values of the non-dimensional parameters for the CubeSat
satellite class, the computation will focus on the determination of the unknown quantities of
the design.
1. The mass ratios can be determined by the results presented in tables 3.21 and 3.22.
Although those values are supposed to be valid for a whole satellite class, it is assumed
that the mass ratios will change for bigger CubeSats as it was already seen for the mass
ratios in section 3.3.1. However, we will continue with those values and show nonetheless
that the 2U CubeSat is more performant than the 1U CubeSat.
Thus with mS/C = 2.66kg we find(mP/L
)CubeSat
= 0.25 · 2.66kg = 0.665kg
(mAOCS)CubeSat = 0.15 · 2.66kg = 0.399kg
(mPower)CubeSat = 0.15 · 2.66kg = 0.399kg
(mStructure)CubeSat = 0.25 · 2.66kg = 0.665kg
(mThermal)CubeSat = 0 · 2.66kg = 0kg
(mC&DH)CubeSat = 0.03 · 2.66kg = 0.0798kg
(mCom)CubeSat = 0.17 · 2.66kg = 0.4522kg
(mPropulsion)CubeSat = 0 · 2.66kg = 0kg
(3.147)
Page 103
Application of the Results
The tables 3.25 and 3.26 summarize the results.
Table 3.25.: Mass ratios of the payload and the subsystems AOCS, Power and Structure build
with mS/C CubeSat and the corresponding subsystem masses for the given mass
ratios for the hypothetical 2U CubeSat
P/L AOCS Power Structure
Mass ratio [%] 25 15 15 25
Subsystem mass [g] 665 399 399 665
Table 3.26.: Mass ratios of the subsystems Thermal, C&DH and Communication build with
mS/C CubeSat and the corresponding subsystem masses for the given mass ratios
for the hypothetical 2U CubeSat
Thermal C&DH Com Prop
Mass ratio [%] 0 3 17 0
Subsystem mass [g] 0 79.8 452.2 0
Please remember that we expect lower mass ratios for the subsystems in real - except
for the payload, where we expect a higher ratio. However, we will continue with the cal-
culation with the above values in order to fulfill the condition that the non-dimensional
parameter is constant for a satellite class.
2. The volume ratios are defined according to (3.107). With VS/C = 2.27 · 10−3 m3 this
leads us to
VBus CubeSat = 0.6− 0.7 · VS/C = 1.362 · 10−3 − 1.589 · 10−3 m3
VP/L CubeSat = 0.1− 0.2 · VS/C = 2.27 · 10−4 − 4.54 · 10−4 m3(3.148)
which equals equivalent edge lengths of
xeq Bus CubeSat = 1.108 · 10−1 − 1.167 · 10−1 m
xeq P/L CubeSat = 6.10 · 10−2 − 7.69 · 10−2 m(3.149)
3. The power ratios can be determined with (3.110).
PP/L CubeSat = 0.3 · PS/C = 0.3 · 2W = 0.6W
PBus CubeSat = 0.7 · PS/C = 0.7 · 2W = 1.4W(3.150)
Page 104
Dimensional Analysis of a Single Satellite
4. The power and mass ratios (3.111) can be used as an alternative for the calculation
of the power quantities. For example, taking rMass Power 3 as this ratio showed the least
deviation for the NGSO-satellites, and assuming that we are looking for the numerical
value of PP/L we can compute
PP/L = rMass Power 3 CubeSat ·mP/L
mS/C
· PS/C
= 1.2 · 0.25 · 2W = 0.6W
(3.151)
thus the same result as in (3.150).
5. The Power subsystem is defined by (3.114) for the battery and (3.115) for the solar
array. Thus with similar assumptions as for the 1U CubeSat, namely tEclipse = 10min
andPEclipse
PS/C= 1, we can determine CBattery by
CBattery = PEclipse · tEclipse · ΠBat3 CubeSat ·PEclipsePS/C
· mPower
mS/C
= 2W · 600s · 400 · 2W
2W· 0.15
= 72000Ws
(3.152)
Concludingly, the battery system of the 2U CubeSat will lead to a lower MPI in com-
parison to the 1U CubeSat.
For the solar array (body-mounted) equation (3.115) with PSAeol= 2.2W calculated by
(3.45) with GalliumArsenide cells and a assumed satellite lifetime of 1 year, leads us to
P 3SolarArraybol
= ΠSA5 CubeSat · x2eq · S · d · tSatellite·
· PSAeol·mS/C
mPower
· PS/C ·tEclipsetOrbit
= 5.690 · (0.13142m)2 · 1367W
m2· 0.0275 · 1year·
· 2.2W · 1
0.15· 2W · 10min
98.77min
= (2.25W )3
(3.153)
So, in terms of the solar array, however, the 2U CubeSat seems to be more performant
than in case of the 1U CubeSat.
Page 105
Application of the Results
6. The AOCS subsystem can be calculated by means of (3.116). As there are three
unknowns which cannot be as easily calculated as PSAeol, we firstly determine the product
of unknowns to
HRW ·mRW · rRW =ΠRW3 CubeSat
tOrbit· (mS/C)2 · x3
eq ·mS/C
mAOCS
=0.011
5926.2s· (2.66kg)2 · (0.13142m)3 · 1
0.15
= 1.987 · 10−7Nms kg m
(3.154)
giving HRW in Nms, mRW in kg and rRW in m. Assuming the mass and the radius of
the reaction wheel to be 0.03kg and 0.004m based on the data for the Beesat reaction
wheel, we can detemine HRW to be 1.656 · 10−3. Concludingly, the AOCS system of
the 2U CubeSat will lead to a higher MPI in comparison to the 1U CubeSat.
7. The Communication subsystem can be designed by means of equation (3.117).
Assuming mAntennaT = 0.02kg, PTransmitter = 1W , the same frequency of the sig-
nal f = 437Mhz as for the 1U CubeSat, which leads to λCom = 0.6858m, and
BER = 10−4, we can determine the data rate R with
R3 =ΠCom2 CubeSat
mAntennaT · h2Orbit
· PTransmitter ·DTransmitter
λCom·
· DReceiver
λCom· PTransmitter
PS/C·mS/C
mCom
· 1
BER
=462.160
0.02kg · (700km)2· 1W · 0.5m
0.6858m· 1m
0.6858m· 1W
2W· 1
0.17· 1
10−4
= (11.38kbps)3
(3.155)
with DTransmitter = 0.5m and DReceiver = 1m. This is an increase in comparison to the
1U CubeSat of 1.78kbps or 18.5%, thus leading to a higher MPI.
8. The Payload subsystem of an Earth Observation mission can be represented by equa-
tion (3.118). With the assumptions as for the 1U CubeSat b = 8 bitpixel
and λP/L =
7 · 10−7m, we can determine firstly D and then a ground-pixel resolution X. First of
all, D can be determined with the data rate calculated in (3.155)
D =R tOrbit
ΠP/LEarthObservation CubeSat · b· λP/L ·
mS/C
mP/L
·PS/CPP/L
=11.38kbps · 5926.2s
3.319 · 8 bitpixel
· 7 · 10−7m · 4 · 2W
0.6W
= 0.023m
(3.156)
Page 106
Dimensional Analysis of a Single Satellite
With equation (3.73) we can compute the ground-pixel resolution X to
X = 2.44 ·hOrbit · λP/L
D
= 2.44 · 700 · 1000m · 7 · 10−7m
0.023m
= 51.98m
(3.157)
thus leading to a better resolution than for the 1U CubeSat. The result is probably
too overoptimistic but this can be due to previous assumptions which were too high.
However, the tendency to better scientific results with larger spacecrafts shall be in the
focus here and not the numerical results which should only show how the equations can
be applied.
The Mission Performance Index
The results for the unknown quantities of the design let suppose that the MPI for the 2U
CubeSat will be higher than for the 1U CubeSat. With equation (3.86) we can finally calculate
the Mission Performance Index of the 2U CubeSat. It is for the Earth Observation payload
(ΨP/L CubeSat2)α =(R · kP/L
)α= (11.38 · 3.43)1
= 39.02
(3.158)
The AOCS subsystem is represented by a reaction wheel. Its Ψ can be calculated by
(ΨAOCS CubeSat2)β = (HRW · kRW )β
=(1.656 · 10−3 · 0.151
)0.5
= 0.016
(3.159)
The power subsystem consists again of batteries and solar arrays:
(ΨPower CubeSat2)χ =
(1
ΨBattery
·ΨSolar Array
)χ=
(1
CBattery · kBattery· P 3
SAbol· kSolarArray
)χ=
(1
72000 · 90· 2.253 · 2.00
)0.8
= 4.337 · 10−5
(3.160)
As the same mass ratios are used for the 2U CubeSat as for the 1U CubeSat since they are in
one satellite class, (ΨStructure)δ and (ΨC&DH)ε have the same values as for the 1U CubeSat.
(ΨStructure CubeSat2)δ =
(mS/C
mStructure
)δ=
(1
0.25
)0.3
= 1.516 (3.161)
Page 107
Application of the Results
(ΨC&DH CubeSat2)φ =
(mS/C
mC&DH
)φ=
(1
0.03
)0.6
= 8.198 (3.162)
The 2U CubeSat is designed without Thermal subsystem as in the case of the hypothetical
1U CubeSat.
(ΨThermal CubeSat2)ε =
(mS/C
mThermal
)ε= (10−20)0.2 = 10−4 (3.163)
The Communication subsystem is described by
(ΨCom CubeSat2)ϕ =(R3 · kCom
)ϕ=(11.383 · 3.189
)0.8
= 866.26
(3.164)
The 2U CubeSat is designed without Propulsion subsystem as in the case of the hypothetical
1U CubeSat.
(ΨPropulsion CubeSat2)γ =
(mS/C
mPropulsion
)γ= (10−20)0.5 = 10−10 (3.165)
Thus we can write
MPICubeSat2 = (ΨP/L CubeSat2)α · (ΨAOCS CubeSat2)β · (ΨPower CubeSat2)χ·· (ΨStructure CubeSat2)δ · (ΨC&DH CubeSat2)φ · (ΨThermal CubeSat2)ε·· (ΨCom CubeSat2)ϕ · (ΨPropulsion CubeSat2)γ
= 39.02 · (0.016) · (4.337 · 10−5) · 1.516 · 8.198 · 10−4 · 866.26 · 10−10
= 0.292 · 10−4 · 10−10
(3.166)
This is a performance gain in comparison to the 1U CubeSat which has a MPI of 4.627 ·10−4 ·10−4 · 10−10. A comparison in percentages is intentionally not given as both calculations, for
the 1U and 2U CubeSat, were based on assumptions and not on a wider database. Thus a
certain inaccuracy is the consequence due the modelling nature of the assumptions. However,
a clear tendency to a higher performance for larger satellites can be noticed. This is a very
important result as it can provide guidelines to a small satellite which is still in the range of
CubeSats but with higher performance, thus combining the advantages of the CubeSat with
a higher mission performance.
As we have now two different MPIs for different satellites in one satellite class, we can exemplary
normalize the results. Our group of missions we want to compare consists of two Earth
Page 108
Dimensional Analysis of a Single Satellite
Observation missions. The MPI of the 2U CubeSat being the larger numerical value, we
define the 2U CubeSat mission to be the mission with a MPI of 100. This means that the 1U
CubeSat mission has a normalized MPI of
MPICubeSat normalized =MPICubeSatMPICubeSat2
· 100
=4.627 · 10−4 · 10−4 · 10−10
0.292 · 10−4 · 10−10· 100
= 0.16
(3.167)
thus showing clearly a significant difference between the 1U and the 2U CubeSat mission.
Page 109
Application of the Results
Figure 3.69.: Top-Down network flow diagram for an Earth Observation mission
Page 110
Dimensional Analysis of a Single Satellite
Figure 3.70.: Bottom-Up network flow diagram for an Earth Observation mission
Page 111
4. Dimensional Analysis of Clusters
As dimensional analysis can be in general applied to systems of any possible kind, also satellite
clusters can be investigated with it. The intention of this chapter is therefore an introduction
to the analysis of satellite clusters with the Buckingham-Π-Theorem similar to the investigation
of the single satellites in Chapter 3. The ratios, non-dimensional parameters and especially the
Mission Performance Indices will make it possible to easily compare the performance of the
cluster with the performance of an equivalent monolithic satellite and facilitate the decision
whether a cluster or a single satellite will best suit the mission requirements.
4.1. The F6-Project and the Concept of Fractionated
Spacecrafts
As this work was intended to be embedded in the current F6-program of the Defense Advanced
Research Projects Agency (DARPA) in collaboration with Orbital Sciences Corporation, the
program shall be described here in some sentences.
F6 is the abbreviation for ”Future, Fast, Flexible, Fractionated, Free-Flying Spacecraft United
Information Exchange”, a program that intends to demonstrate a new approach in space
architecture design: instead of using one sizable monolithic satellite for a mission, several
smaller satellites are set in orbit, each fulfilling one functionality of the subsystems of the
monolith satellite, organized in a cluster, sharing resources and communicating via a wireless-
network. Advantages of this new approach are among other benefits essentially a gain in
flexibility and robustness: the new architecture is more flexible than the traditional one as
elements of the cluster can be easily replaced when they become obsolete. Furthermore new
elements can be integrated into the cluster when necessary. The cluster’s robustness is based
on the fact that a mission is, contrary to an architecture with one monolith satellite, not lost
when the launch fails or a subsystem of the satellite does not work properly since the elements
of the system are replaceable. However there are also potential downsides to a fractionated
cluster of small satellites such as the duplication of subsystems across satellites and an increased
complexity of the inter-spacecraft communication system. Dimensional analysis can help to
facilitate the decision between a cluster architecture and a monolithic satellite for a mission.
The project was in Phase 2 before the collaboration between Orbital Sciences Corporation and
MIT was paused in August 2010. Phase 2 means that DARPA’s industrial partner for the
project, Orbital Sciences Corporation, was in the phase of detailed design. MIT supported
their activity by further investigations. Therefore the MIT Space Systems Laboratory planned
Page 112
Dimensional Analysis of Clusters
to launch a prototype of an F6-CubeSat cluster in 2012/13, which would been designed and
built by MIT students in the coming years.
Although the collaboration is set on pause, cluster architectures have risen the attention of
scientists and industries worldwide and will definitely be a challenge in the coming years.
Therefore first investigations to find scaling laws and Mission Performance Indices for whole
clusters are believed to be foresighted.
4.2. Ratios and Non-Dimensional Parameters
Similar to a single spacecraft which can be investigated on a top-level approach, considering
only the spacecraft as a whole, or on a more detailed approach on a subsystem level, a cluster
can be investigated on two different levels, too. The top level is the cluster itself whereas the
cluster’s ”subsystems” are the different single satellites which collectively build the cluster.
For the single satellites of the cluster, no further investigation in terms of dimensional analysis
is necessary as the same non-dimensional parameters and ratios that are presented in chapter
3 can be applied to them. A further step, however, is required to characterize the cluster as
a whole. One approach to do so is to express the non-dimensional parameters and ratios of
chapter 3 with quantities describing the whole cluster. Those quantities can simply be the
sums of the corresponding quantities of the single satellites. Thus, for example, the total mass
of the cluster mCluster can be used which is defined as the sum of the masses of all the single
satellites of the cluster together. Similarly, further possible quantities are the total mass of
the power subsystem of the cluster mClusterPower, the total volume of the cluster VCluster, the
total power need of the cluster PCluster or the total data rate of the cluster transmitted to the
Earth RCluster.
The two different design approaches, the payload-centric Top-Down and the spacecraft-centric
Bottom-Up approach as presented in section 3.1, are both also valid for the design of the
clusters.
Hence, similar to the ratios (3.4) for a single satellite, we can rewrite the mass ratios in a
Bottom-Up approach for a cluster
rCluster Mass BU1 =mClusterP/L
mCluster
rCluster Mass BU2 =mClusterAOCS
mCluster
rCluster Mass BU3 =mClusterPower
mCluster
rCluster Mass BU4 =mClusterStructure
mCluster
rCluster Mass BU5 =mClusterThermal
mCluster
rCluster Mass BU6 =mClusterC&DH+TT&C
mCluster
rCluster Mass BU7 =mClusterPropulsion
mCluster
(4.1)
Page 113
Ratios and Non-Dimensional Parameters
with mClusteri expressing the sum of the i subsystem masses of all the satellites generating
the cluster.
The same proceeding as for the above mass ratios can be done for the Top-Down mass ratios
(3.3), the power ratios (3.12) and (3.13), the mass and power combining ratios (3.14) to
(3.17) and the top-level non-dimensional parameters (3.19) to (3.24) and (3.28) and (3.29).
In terms of the non-dimensional parameters for the subsystems, the results of section 3.4 can
be changed to be applicable to clusters, however, the physical information of the results is
reduced as the subsystems are especially designed to fulfill the requirements of a single satellite.
This means, for example, that we can build a non-dimensional parameter for the design of a
reaction wheel for the whole cluster based on the results in section 3.4.3 but this information
makes no physical sense as every single satellite requires a reaction wheel which is especially
customized for it. Nevertheless, when dealing with the cluster as if it is a monolithic satellite,
also on subsystem level, comparisons to real monolithic satellites are possible and this can
help to decide whether a cluster architecture or a single satellite is best suited for the mission
requirements. Subsystems which can be treated as one representative subsystem for the whole
cluster and still make physical sense, are the subsystems Power, Structure, Communication and
C&DH. For the power subsystem the results of section 3.4 - (3.33) to (3.35) for the battery
and (3.38), (3.40), (3.41), (3.42), (3.46) and (3.47) for the solar array - can be conveyed for
the cluster by using the corresponding quantities describing the cluster as a whole, such as
the total power need of the cluster during eclipse PClusterEclipseor the capacity of the batteries
on-board CClusterBattery. For the Structure subsystem no further analyses are necessary as
this subsystem is already covered by the corresponding mass and volume ratios. In terms of
the Communication subsystem, additional intersatellite links has to be taken into account as
they play a significant role in the functionality of a cluster. The Communication subsystem
does not only enable the communication with the ground stations but also the data transfer
between the satellites of the cluster so that all resources of the cluster can be used for data
processing. Thus the Communication subsystem becomes more important and complex but
can be still expressed by the non-dimensional parameters (3.70), (3.71) and (3.72). However,
the numerical values of the non-dimensional parameter ΠCom 1, ΠCom 2 and ΠCom 3 will be
different according to the kind of link they will represent: an intersatellite link will show
different characteristics than a link to the ground. Furthermore, only an overall representation
of the satellites’ Communication subsystems for a ground link, and not for a intersatellite link,
will make physical sense. This will be done with equivalent antenna diameters which are not
equal to the real antenna diameters. However, this investigation is especially interesting to
get information about the cluster’s transmitter output power PClusterTransmitterand the data
rate to the ground RCluster. In terms of the C&DH subsystem, one can surely state that this
subsystem will increase in complexity in comparison to a single satellite. Also more power and
more storage will be necessary to process the data. However, its non-dimensional parameter
is supposed to combine the same quantities as for a single satellite.
The subsystems AOCS and Thermal as well as the payload are believed to be too spacecraft
specific to be investigated for the cluster as a whole. Therefore no cluster-specific non-
dimensional parameters except for the corresponding mass and power ratios are supposed for
Page 114
Dimensional Analysis of Clusters
this three subsystems. However, their weighting in the Mission Performance Index will change
as, for example, the AOCS system in a cluster is more critical for the mission success as for a
single satellite since the intersatellite links require a certain station keeping and accuracy. The
different payloads in a cluster can also be interrelated to each other so that their performance is
also more critical in a cluster than when separately used on single satellites. All these changes
and more differences between clusters and single satellites will be issue of the next section
when dealing with a cluster’s Mission Performance Index.
4.3. Mission Performance Index
When investigating the advantages and disadvantages of a cluster in comparison to a single
satellite, the Mission Performance Index is one method which facilitates the decision between
these two mission architectures. The Mission Performance Index of a cluster architecture can
be calculated in two steps. In a first order approach only the characteristics of the cluster
are considered, thus MPICluster only combines ΠP/L Cluster, ΠAOCS Cluster, ΠPower Cluster,
ΠStructure Cluster, . . . . It is calculated as shown in section 3.5 and also normalized to become
a figure between 0 and 100. In a second step also the performance of the single satellites
generating the cluster is taken into account. For that, the MPISingle Sat for every single
satellite is calculated as shown in section 3.5. Afterwards, an arithmetical average is build
over all the satellites which gives us the quantity MPIall/Sats. A weighing can come in here,
too, in case some satellites are considered to be more critical for the mission accomplishment
than others. MPIall Sat is then normalized to a number between 0 and 1 and then multiplied
with MPICluster to create MPISat+Cluster, again a number between 0 and 100. As the
importance of the different subsystems for the mission success changes now in comparison
to the ones for a monolithic satellite, a closer look at the exponents for MPICluster follows
in form of the table 4.1. An increase in the numerical values of the ranges can be seen for
the payload and the subsystems Communication, C&DH, AOCS and Propulsion in comparison
the the exponents for the monolithic satellites. The increase for the latter two subsystems
is due to the more important requirement in satellite station keeping within the cluster to
enable the flawless communication and data transmission between the satellites. An increase
for the Communication and C&DH exponents is caused by the higher complexity of the two
systems. Intersatellite links has to be made possible as well as data processing on-board and
resource sharing. Without all that, the advantages of the cluster architecture are massively
diminished. Finally, the payload increases in importance as the interdependency between the
different payloads generally increases in cluster architectures. The same exponent ranges are
allocated to the subsystems Power, Structure and Thermal here for the cluster architectures
as for the monolithic satellites presented in 3.20 since their importance is believed to stay in
a similar range.
Page 115
Mission Performance Index
Table 4.1.: Ranges for the subsystem exponents α, β, χ, . . . for cluster architectures, including
the tendency of the change in the ranges compared to the single satellite exponents
Subsystems EOM SSM TDM Tendancy
Payload 0.8-1.0 0.8-1.0 0.8-1.0 ↑
Power 0.7-0.9 0.7-0.9 0.6-0.9 const
Communication 0.8-0.9 0.8-0.9 0.7-0.9 ↑
Structure 0.2-0.5 0.2-0.5 0.2-0.5 const
AOCS 0.7-0.9 0.7-0.9 0.6-0.9 ↑
Thermal 0.1-0.3 0.1-0.6 0.1-0.6 const
C&DH 0.7-0.9 0.7-0.9 0.6-0.9 ↑
Propulsion 0.7-0.9 0.7-0.9 0.6-0.9 ↑
Please note that the current form of the Mission Performance Index only implies physical
quantities which directly influence the design of a spacecraft such as the mass and volume of the
spacecraft or the orbit altitude. But a mission consists of more influencing parameters than only
those describing the physics of a spacecraft. Further parameters such as launcher availability,
costs, development time of the design or robustness and flexibility of the architecture are not yet
implied in the Mission Performance Index. Especially the latter two quantities, however, express
the main advantage of cluster architectures in comparison to monolithic spacecraft mission
architectures. Thus, a further development of the MPI shall include also those quantities
expressing the feasibility, usability and reliability of the mission design by taking all stakeholders
of a mission into account. That means a turn from a systems engineering to a engineering
systems approach. This development is necessary in order to do justice to the mission as a
whole. With that alteration, a change in the numerical values of the MPIs is expected with a
tendency to higher MPIs for the clusters.
Page 116
5. Conclusion and Future Prospects
The application of dimensional analysis by means of the Buckingham-Π-Theorem on satellite
design was shown in this work. Ratios and non-dimensional parameters were developed as
well as a possibility for the quantification of mission accomplishment by means of the Mission
Performance Index. The validation of several ratios has been done with data from non-
geosynchronous communication satellites, showing promising results, limited by the nature of
the given data. Numerical applications of the theoretical results for three different hypothetical
satellites have proven that the ratios, non-dimensional parameters and the Mission Performance
Index can be used in practice. Taking especially the application on a hypothetical 1U and
2U CubeSat into consideration, dimensional analysis is a possibility to quickly design a new
spacecraft based on former designs and to examine the mission performance of a satellite in
comparison to other designs. Therefore, the results of this work can be used to optimize given
satellite standards in terms of their mission performance. In this context, it is believed that
the 1U CubeSat standard is too small to be efficient in terms of mission performance. A
further development of the MPI in terms of parameters i.e. costs and launcher availability is
supposed to prove this assumption and to provide the foundation for the definition of a new,
high-performant CubeSat standard.
Future work based on this thesis might include the further improvement and development
of the non-dimensional parameters for every subsystem as well as the validation of them by
means of extensive databases for several satellite classes. Beside the implementation of further
stakeholders into the MPI, the development of an ”universal MPI” which combines all mission
types as well as the two design approaches Top-Down and Bottom-Up, is also desirable.
Furthermore, as clusters gain increasing importance in the space sector, the detailed application
of the theoretical results of the dimensional analysis on cluster architectures is also supposed
to be worth the investigation.
Finally, the combination of the component database with the theoretical results into a design
tool for CubeSats is also desirable. The design tool will ideally enable the quick design of a
spacecraft based on the theoretical results of this work and indicate the COTS-components
which are best suited for the design. A continuously updated database is therefore necessary.
Page 117
A. Calculation of the Eclipse Time
tEclipse
For circular orbits the eclipse time tEclipse for a specific position of the Earth in the ecliptic
can be calculated with the following equation [Larson and Wertz, 1999, p. 107 f]
tEclipse = tOrbit ·φ
360◦(A.1)
thus with a direct proportionality between tEclipse and tOrbit. However, a change in tOrbit means
that the orbital altitude of the spacecraft hOrbit has changed because of equation (3.52). A
change in orbital altitude hOrbit, in turn, also influences the quantity φ in equation (A.1) which
can be calculated by
cos
(φ
2
)=
cos(ρ)
cos(βs)(A.2)
with ρ as the angular radius of the Earth, calculated by
sin(ρ) =REarth
REarth + hOrbit(A.3)
Thus an increase in the orbital altitude hOrbit leads to an decrease in φ which reacts in
direct opposition to the development of tOrbit with increasing orbital altitude. The maximum
eclipse time for different orbital altitude can be found listed in common literature (see Errata
of [Larson and Wertz, 1999]): the maximum eclipse time decreases with increasing orbital
altitude only until about 1500 km. Increasing the altitude further causes the eclipse time to
rise again. That is why it was decided to leave the eclipse time tEclipse and not the orbital
period tOrbit as influencing parameter of the battery system in section 3.4.2.
Please notice that CubeSats often play a special role in terms of these calculations. They
are often launched in sun-synchronous orbits which means their orientation to the sun is the
same over the whole revolution of the Earth around the Sun. This in turn leads to constant
eclipse times during the whole year. Furthermore, as sun-synchronous orbits in LEO have an
inclination of 96.0 to 98.0◦, orbits can be achieved which do not have any eclipses at all. This,
however, also depends on the choice of the Right Ascension of the Ascending Node (RAAN).
Thus for the sake of completeness of this topic, the following further explanations are given.
The only quantity in equation (A.2) not yet explained is βs as it adds a further complexity to
the calculations. βs, also called the β-angle, represents the position of the orbital plane to
the Sun: it is the angle of the Sun above the orbital plane, thus the angle between the orbital
plane and the incident sun. Its calculation is more complex than in case of the angular radius
Page 118
Calculation of the Eclipse Time tEclipse
of the Earth ρ as it generally changes during the revolution of the Earth around the Sun.
During a year, βs ranges from ±(23.5◦ + i). From the design point of view, important values
are the maximum and minimum eclipse time: tEclipse reaches its maximum for βs = 0◦,
i.e. when the sun vector is in the orbital plane, and its minimum when βs takes its extreme
values of ±(23.5◦ + i). Hence, the calculation for the extremes of the eclipse time tEclipsedoes not seem complicated. However, as already stated, CubeSats are often launched in sun-
synchronous orbits where the β-angle is fixed during the year but dependent on the orientation
of the orbit to the sun which is not necessarily βs = 0◦ or βs = i+ 23.5◦. Thus a closer look
at the calculation of βs is recommended.
βs can be calculated in general as∣∣∣90◦ − βs
′∣∣∣ with βs
′being the angle between the normal of
the orbital plane ~nOrbit and the sun vector ~s.
In a fixed heliocentric coordinate system with the vernal equinox as the x-axis, ~nOrbit can be
calculated by
~nOrbit =
sin(RAAN) · sin(α)
−cos(RAAN) · sin(α)
cos(α)
(A.4)
with RAAN being the Right Ascension of the Ascending Node and α being i + 23.5◦. The
sun vector ~s can be expressed by
~s =
cos(δ)
sin(δ)
0
(A.5)
assuming a perfectly circular Earth orbit around the Sun, with δ being 180◦ + δ′. The 180◦
enables the sun vector to ”start” at spring and is necessary because of the definition of the
x-axis (vernal equinox). δ′
represents the actual position of the Earth relatively to the vernal
equinox, thus δ′
= 360◦
365,25TPosition with TPosition being the day for which the eclipse time shall
be calculated, counting from the vernal equinox on.
Thus with the scalar product for ~nOrbit and ~s, βs′
can be calculated as
cos(βs′) = sin(RAAN) · sin(α) · cos(δ)− cos(RAAN) · sin(α) · sin(δ) (A.6)
and with βs =∣∣∣90◦ − βs
′∣∣∣, one can calculate the beta angle βs which in turn determines the
eclipse time with equations (A.2) and (A.1).
For the CubeSats only one calculation is actually necessary - for example in autumn because of
the simplicity of the sun vector ~s - and βs is constant over the whole mission lifetime assuming
a constant orbital altitude over the mission duration.
Page 119
To sum up, the complexity of the calculation of the eclipse time tEclipse has been highlighted.
It is dependent on a number of factors and especially for the CubeSats the following function
applies
tEclipse = f(hOrbit, i, RAAN) (A.7)
Thus for a more detailed investigation, i and RAAN should be taken into consideration for
the calculation of the eclipse time.
Page 120
B. CubeSat-Specific Commercial
off-the-Shelf Component Database
Commercial off-the-shelf (COTS) components are already very often used in satellite engineer-
ing as COTS components are normally less expensive than in-house solutions and usually faster
available. So, the development time of a satellite can be effectively shortened and thus the all
in all costs decreased. However, as CubeSats have strong weight and size limitations, many
”normal” sized components are not suitable for the use on picosatellites. Thus, with the rising
of the importance of CubeSats and new minituarizing technologies, a market has established
itself which is specialized on CubeSats specific off-the-shelf components. Most companies like
Pumpkin Inc., Clyde Space Ltd. and ISIS - Innovative Solutions in Space are only specialized
in components for pico- and nanosatellites, few others like Astro- und Feinwerktechnik Adler-
shof GmbH offer also products for other satellite classes. In order to facilitate and speed up
the choice for a COTS component, a database was created during this work which lists many
of the currently available solutions with their costs, availabilities, technical characteristics and
also the internet links where the product can be find online. The list is created under Microsoft
Office Excel so that it is guaranteed that its use is familiar to an extensive number of satellite
developers. It is organized in several sheets, including CubeSat kits, Energy Power System
solutions, batteries, solar panels, CubeSat platforms, components for the AOCS, structures,
transceivers, receivers, transmitters, antennas, on-board computers, software, harnesses, elec-
trical and mechanical ground support equipment, ground station related products, possible
payloads, the latest research results in propulsion, GPS receivers, various miscellaneous prod-
ucts and also a short list of forthcoming products. Available data sheets of the listed products
are provided on the CD attached to this work and can be consulted for further information. It
is also intended to use the list in a future MATLAB program which shall, based on inputs from
the engineer like the mission performance of the satellite, recommend COTS components for
the mission.
Some universities already showed their interest in such a database and mentioned existing
plans to create one by themselves. The realization though has never been accomplished so
far, in most cases due to a lack of time.
Currently many subsystems can fall back on a wide range of specialized CubeSat COTS compo-
nents which are highly matured. Especially the Attitude and Orbit Control subsystem seems to
have been in the focus of many companies so that the customer can choose between an exten-
sive number of AOCS components. Other subsystems like power, communication and structure
Page 121
are in the product range of a smaller number of companies but interested customers can still
choose between specific flight-proven components. Beside the usual satellite components such
as magnetorquers, batteries and antennas, it is worth mentioning that whole CubeSat kits can
be purchased which promise shorter development times and affordable projects, thus are espe-
cially suited for student missions. The kits of the company Pumpkin Inc., for example, consist
of a complete lightweight structure for the CubeSat as well as a flight module, power supplies
for the use on Earth, programming adapters, cables and tools and a software. In the easiest
case, the customer has only to add a power supply (i.e. batteries, solar panel) and a payload
and the satellite is ready to be launched. Depending on the requirements of the mission, a
communication system consisting of a transceiver and an antenna as well as an attitude and
orbital determination system can also be added. However, what the users of the kits complain
about is the loss of the flexibility in design. More sophisticated missions for the CubeSats re-
quire higher accuracies and longer satellite lifetimes. Thus, picosatellite propulsion is a current
research area in many institutions and companies worldwide. However, many approaches still
need time to mature and existing technologies and components require a decrease in costs to
be used by universities with a restricted mission budget. It also became obvious during the
research, that many universities passed on COTS components in their design as they are still
regarded to be too expensive and especially not suitable for the intended purposes. Higher
development times are therefore required but still preferred to the COTS components. On
the other hand, also not flight-approven components are often used in the satellite designs.
Hence, extensive testings are necessary in this cases but still preferred to much more expensive
COTS components. In the end, it will be one of the future main tasks of a successful CubeSat
component industry to develop components which are much less expensive and meet the needs
of the customer more precisely.
Page 122
CubeSat-Specific Commercial off-the-Shelf Component Database
Figure B.1.: Screenshot of a part of the AOCS-sheet of the CubeSat COTS component
database
Page 123
Bibliography
[Aas et al., 2009a] Christina Aas, Barry T. C. Zandbergen, Rob J. Hamann, and Eberhard
K. A. Gill. Development of a System Level Tool for Conceptual Design of Small Satellites.
In 7th Annual Conference on Systems Engineering Research, Loughborough, UK, April 2009.
[Aas et al., 2009b] Christina Aas, Barry T. C. Zandbergen, Rob J. Hamann, and Eberhard
K. A. Gill. SCALES – A System Level Tool for Conceptual Design of Nano- and Microsatel-
lites. In 7th IAA Symposium on Small Satellites for Earth Oberservation, Berlin, Germany,
May 2009.
[Amsat, 2010] Amsat. , 2010. last access 12/14/2010.
[Anderson, 2007] John D. Anderson. Fundamentals of Aerodynamics, Chapter 1.7, pages
34–40. McGraw-Hill, 4th edition, 2007.
[Bhaskar and Nigam, 1990] R. Bhaskar and Anil Nigam. Qualitative physics using dimensional
analysis. Artificial Intelligence, 45:73–111, September 1990.
[BIPM, 2006] BIPM. Le Systeme International d’Unites / The International System of Units.
Bureau International de poids et mesures, Organisation intergouvermentale de la Convention
du Metre, 8e edition edition, 2006.
[Bridgman, 1948] Percy Williams Bridgman. Dimensional Analysis. Yale University Press, 5th
edition, 1948.
[Buckingham, 1914] Edgar Buckingham. On Physically Similar Systems; Illustrations of the
Use of Dimensional Equations. Physical Review, 4:345–376, 1914.
[Buckingham, 1915a] Edgar Buckingham. Model Experiments of the Forms of Empirical Equa-
tions. Transactions of the A.S.M.E, 37:263–296, 1915.
[Buckingham, 1915b] Edgar Buckingham. The Principle of Similitude. Nature, 96:396–397,
1915.
[CalPoly, 2007] CalPoly. Poly Picosatellite Orbital Deployer Mk III ICD (P-POD Mk III ICD).
California Polytechnic University, San Luis Obispo, 2007.
[CalPoly, 2009] CalPoly. CubeSat Design Specification (CDS). California Polytechnic Univer-
sity, San Luis Obispo, Rev. 12 edition, 2009.
[Courtney and Lozano, 2010] Daniel G. Courtney and Paolo Lozano. Porous Ionic Liquid
Source Thrusters for Small Satellite Propulsion. unpublished, 2010.
Page 124
Bibliography
[de Jong, 1967] Frits J. de Jong. Dimensional Analysis for economists. North Holland Pub.Co.,
Amsterdam, 1967.
[de St. Q. Isaacson and de St. Q. Isaacson, 1975] E. de St. Q. Isaacson and M. de St.
Q. Isaacson. Dimensional Methods in Engineering and Physics. John Wiley and Sons,
1975.
[Gross et al., 2010] Dietmar Gross, Werner Hauger, Jorg Schroder, and Wolfgang A. Wall.
Technische Mechanik 3: Kinetik, Chapter 3.2.2, pages 140–144. Springer, 11th edition,
2010.
[Hedin, 1991] Alan E. Hedin. Extension of the MSIS Thermospheric Model into the Middle
and Lower Atmosphere. Journal of Geophysical Reasearch, 96:1159–1172, 1991.
[Heidt et al., 2001] Hank Heidt, Jordi Puig-Suari, Augustus S. Moore, Shinichi Nakasuka, and
Robert J. Twiggs. CubeSat: A new Generation of Picosatellite for Education and Industry
Low-Cost Space Experimentation, August 2001. AIAA Paper SSC00-V-5.
[Institute, 2010] Van Karman Institute. http://www.vki.ac.be/QB50/project.php, 2010.
last access 08/25/2010.
[ISO-TC12 and IEC-TC25, 2009] ISO-TC12 and IEC-TC25. ISO/IEC 80000. International
Organization for Standardization and International Electrotechnical Commission, 2009.
[JCGM/WG2, 2008] JCGM/WG2. International vocabulary of metrology – Basic and general
concepts and associated terms (VIM) / Vocabulaire international de metrologie – Concepts
fondamentaux et generaux et termes associes (VIM). Working Group 2 of the Joint Com-
mittee for Guides in Metrology, Bureau International de poids et mesures, Organisation
intergouvermentale de la Convention du Metre, 2008.
[Kiesling, 197172] John D. Kiesling. A technique for modelling communication satellites.
Comsat technical Review, 2:73–103, 1971/72.
[Langhaar, 1951] Henry L. Langhaar. Dimensional Analysis and Theory of Models. John Wiley
and Sons, New York, 1951.
[Larson and Wertz, 1999] Wiley J. Larson and James R. Wertz, editors. Space Mission Anal-
ysis and Design. Space Technology Library, 3rd edition, 1999.
[Lehman and Craig, 1963] P. Joseph Lehman and Eugene A. Craig. Dimensional Analysis in
Applied Psychological Research. The Journal of Psychology, 51:223–226, 1963.
[Macagno, 1971] Enzo O. Macagno. Historico-critical Review of Dimensional Analysis. Journal
of the Franklin Institute, 292(6):391–402, 1971.
[McMahon and Bonner, 1983] Thomas A. McMahon and John Tylor Bonner. On Size and
Life. Scientific American Library New York, 1983.
[NASA, 2010] NASA. Announcement of CubeSat Launch Initiative. NASA Goddard Space
Flight Center Headquarters Acquisition Division, July 2010.
Page 125
Bibliography
[Pritchard, 1984] Wilbur L. Pritchard. Estimating the Mass and Power of Communications
Satellites. International Journal of Satellite Communications, 2:107–112, 1984.
[Rayleigh, 1915] Lord Rayleigh. The Principle of Similitude. Nature, 95:66–68, 1915.
[Riabouchinsky, 1911] Dimitri Riabouchinsky. Methode des variables de dimensions zero et
son application en aerodynamique. L’Aerophile, 19:407–408, 1911.
[Richardson, 1919] Lewis F. Richardson. Convective Cooling and the Theory of Dimensions.
Proceedings of the Physical Society of London, 32:405–409, 1919.
[Richharia, 1999] Madhavendra Richharia. Satellite Communications Systems: Design Princi-
ples, Chapter 9.5, pages 313–319. McGraw-Hill, 1999.
[Saleh et al., 2002] Joseph H. Saleh, Daniel E. Hastings, and Dava J. Newman. Spacecraft
Design Lifetime. Journal of Spacecraft and Rockets, 39(2):244–257, 2002.
[Sedov, 1993] Leonid Ivanovich Sedov. Similarity and Dimensional Methods in Mechanics.
CRC Press, 10 edition, 1993.
[Springmann and de Weck, 2004] Philip N. Springmann and Olivier de Weck. Parametric
Scaling Model for Nongeosynchronous Communications Satellites. Journal of Spacecraft
and Rockets, 41(3):472–477, 2004.
[Stahl, 1961a] Walter R. Stahl. Dimensional Analysis in Mathematical Biology I. Bulletin of
Mathematical Biology, 23:355–376, 1961.
[Stahl, 1961b] Walter R. Stahl. Dimensional Analysis in Mathematical Biology II. Bulletin of
Mathematical Biology, 24:81–108, 1961.
[Strutt and Rayleigh, 1894] John Willian Strutt and Baron Rayleigh. The Theory of Sound.
New York Dover Publications, revised and enlarged second edition, 1894.
[Taylor, 1950a] Geoffrey Taylor. The Formation of a Blast Wave by a Very Intense Explosion. I.
Theoretical Discussion. Proceedings of the Royal Society of London. Series A, Mathematical
and Physical Sciences, 201(1065):pp. 159–174, 1950.
[Taylor, 1950b] Geoffrey Taylor. The Formation of a Blast Wave by a Very Intense Explosion.
II. The Atomic Explosion of 1945. Proceedings of the Royal Society of London. Series A,
Mathematical and Physical Sciences, 201(1065):pp. 175–186, 1950.
[Taylor, 1974] Edward S. Taylor. Dimensional Analysis for Engineers. Oxford University Press,
1974.
[Tennekes, 1996] Hendrik Tennekes. The Simple Science of Flight. MIT Press, 1996.
[Thomsen, 2010] Michael Thomsen. http://mtech.dk/thomsen/space/cubesat.php,
2010. last access 11/28/2010.
[Toorian et al., 2005] Armen Toorian, Emily Blundell, Jordi Puig-Suari, and Robert J. Twiggs.
CubeSats as Responsive Satellites, August 2005. AIAA Paper 2005-6828.
Page 126
Bibliography
[TUBerlin, 2010] TUBerlin. BeeSat. http://www.raumfahrttechnik.tu-berlin.de/
beesat/v-menue2/beesat-1/news/, 2010. Accessed 11/15/2010.
[Vaschy, 1892] Aime Vaschy. Sur les lois de similitude en physique. Annales telegraphiques,
19:25–28, 1892.
[Zlokarnik, 2006] Marko Zlokarnik. Scale-up in Chemical Engineering. VILEY-VHC, 2nd edi-
tion, 2006.
Page 127
Image References
[NASA, 2010] NASA, http://www.nasa.gov/directorates/somd/home/
CubeSats_initiative.html, last access 12/13/2010
[Cal Poly, 2010] Cal Poly, http://www.cubesat.org/index.php/media/pictures/
54-p-pod-mk-i, last access 12/13/2010
[MOVE, 2010a] MOVE, Manuel Czech, lrt, TUM, written communication, July 2010
[MOVE, 2010b] MOVE, Claas Olthoff, lrt, TUM, written communication, December 2010
Page 128