Design Manufacturing and Characterisation of A Water Fed CubeSat
Design Manufacturing and Characterisation of A Water Fed CubeSat
Design Manufacturing and Characterisation of A Water Fed CubeSat
By
Master of Science
in Aerospace Engineering
Chair of Space System Engineering
The gas generator based cold gas system have however been shown to be unable to produce the
necessary total impulse within the stringent mass and volume constraints. The formation flying
objective requires a new high performance propulsion system, which is designed during this master
thesis. In order to design this propulsion system a number of tests were conducted on the candidate
components and assemblies. After a survey of all commonly mentioned reaction engines, a resistojet
was found to best meet the needs of this mission. Mainly because resistojets are relatively
uncomplicated, do not require combustion and have a relatively low power consumption per amount
of impulse generated. The best propellant for resistojets, selected from a large number of
possibilities, was found to be water. Water was chosen because it is liquid around room
temperature, It has a high potential impulse bit per unit of mass, it is benign, cheap and readily
available.
Based on the requirements on the propulsion system the following necessary design parameters
were found: a water propellant mass of 50 [g], a nitrogen pressurant mass of 0.2 [g], a storage
pressure of 4.5 [bar] to 2.5 [bar], a nozzle throat diameter of 25 [μm], a nozzle area ratio of 20 [-] and
a final propellant heating temperature of 500 [°C]. The resulting propulsion system performance is: a
total velocity increase ΔV of 21.01 [m/s] (with a satellite mass of 3 [kg]), a total thrusting time of
17 [h] and 56 [min], a power consumption of 6.8 [W] to 3.7 [W] and a thrust force of 1.4 [mN] to
0.8 [mN].
A resistojet propulsion system requires pressure to transport the propellant from the storage tank to
and through the thruster. Because pre-pressurisation is generally not allowed in CubeSats, gas
generators were found to be the best option. These gas generators can be electrically initialised in
orbit, where they produce gas from a solid. Gas generators used for airbag initiation in the
automotive industry were found to be unsuitable for this mission because of the high shock pressure
of possibly up to 95 [bar]. Additionally, three out of four gas generators that were tested
malfunctioned. The reason for this malfunctioning is still unknown. For the DelFFi mission a waiver
was granted, allowing pressurisation before launch. Therefore pre-pressurisation is considered to be
the best option for propellant pressurisation.
Both the manufacturing process and the resulting heating chambers show great potential. The
heating elements have been shown to be robust during the rough testing process. The propellant
channels have very favourable rounded edges which reduce pressure and heat losses, while
maximising the heating contact and propellant mixing. Due to the experimental nature of the
manufacturing process the fluidic channel dept was found to be too large and the nozzle throat
diameters were too large. The performed analysis of these discrepancies will be used to improve the
manufacturing process. The tested resistance values of the resistive heater modules were found to
be 200 to 600 times larger than designed. This increases the required input voltage from 5 [V] to the
range from 70 [V] to 120 [V]. For the current testing phase this is not a problem since these supplies
are available. The flight models will however have to be redesigned in order to comply with the 5 [V]
requirement of the satellite power supply. The main reason of this discrepancy was found to be a
calculation error. Some adjustments have already been listed with which this resistance can be
decreased by a factor of 144. Further design efforts will have to increase this factor to obtain the real
design value of the resistance.
There are some tasks that still have to be performed by a succeeding master student to consolidate
the work done in this master thesis to a propulsion system flight model. The developed thrusters
have to be performance tested with propellant in multiple operating conditions. The results of these
tests lead the final redesign of the developed thrusters. The propellant storage system needs to be
built and tested. Finally the complete in flight propulsion systems has to be defined, built and
integrated into the satellites.
- Author unknown -
“Wat men moet leren doen, leert men door het te doen”
(Anything one needs to learn to do, is learned by doing it)
- Aristotle -
Furthermore I owe my gratitude to a number of people who have aided me towards the completion
of this thesis. Most notably my supervisor Dr. Angelo Cervone, giving valuable advice when
necessary, while granting me the freedom to act on my own accord. My second supervisor ir. Barry
Zandbergen, with his relentless interest and enthusiasm, giving the right tips to help me form my
own judgement. DelFFi propulsion system colleagues Quirino Bellini and Ivan Krusharev for their
humour and their views. Dr. Henk van Zeijl for his seemingly endless stream of innovative MEMS
manufacturing ideas and his help in generating design possibilities. Admirable was the last day of
production, where he worked from sunrise to sunset on the next day in an effort to finish the
thruster manufacturing. The men at the workshop of DEMO, always ready for discussions on
production possibilities, and a help in the manufacturing of the test equipment. Jos van Driel with his
swift help with sensor equipment and necessary software. Nuno, who we all know as the ruler of the
cleanroom and test equipment, always ready to help out students with their novice electrical
questions. Erwin van Dooren from airbagbank, who was kind enough to discuss and demonstrate the
possibilities of using airbag initiators in satellites.
And a final special thanks goes to all my colleagues in the master graduation rooms. Amidst
occasional chaos and disappointment they have always been able to lighten the spirit and put things
into perspective.
Latin
Symbol Unit Description
2 Acceleration
a [m/s ]
A [m2] Cross-sectional area
c [m] Chord length of an aerofoil shape
c [m/s] Characteristic velocity
Cp [J/kg/K] Heat capacity of the fluid at constant pressure
Cv [J/kg/K] Heat capacity of the fluid at constant volume
Cv [-] Flow coefficient
D [m] Diameter
f [-] Fanning friction factor
F [N] (Thrust) force
g0 [m/s2] Gravitational acceleration at standard sea level, 9.81 [m/s2]
G [-] Specific gravity
h [m] Height
H° [J/mol] Enthalpy
I [A] Electrical current
Isp [s] Specific impulse
Kn [-] Knudsen number
L [m] Length
m [kg] Mass
m [kg/s] Mass flow
M [-] Mach number of the flow
M̂ [kg/mol] Molar mass
n [mol] Amount of pressurant
[Pa] ([bar] where Pressure
p
indicated)
p [m] Periphery or inner wall length of a cross-section
P [W] Power
r [m] Radius
R [J/kg/K] Specific gas constant
R [Ω] Electrical resistance
RA [J/K/mol] Universal gas constant
Re [-] Reynolds number
t [K]/1000 Temperature
t [m] Thickness
T [K] Temperature
U [m/s] Velocity
V [m3] Volume
Greek
Symbol Unit Description
[/K] The Temperature Coefficient of Resistance (TCR)
[-] Aspect ratio
[-] Increment or change
[-] Ratio of specific heats
[-] Vandenkerckhove constant
[-] Pressure loss coefficient
[-] Efficiency
[m] Molecular mean free path between collisions
[kg/s/m] Dynamic (absolute) viscosity
[m2/s] Kinematic viscosity
[kg/m3] Density
[Ω·m] Resistivity
[Pa] Stress
[-] Volume fraction
Glossary
Aspect ratio Ratio of the length of a suspended heating element over the width of this element.
The higher the aspect ratio, the thinner the suspended heater is.
Bond wires Ultra fine, highly conductive and nonreactive electrical wires. Made to make
electrical connections to very small electrical components.
Choked flow A compressible flow effect, in a narrower section of a flow channel, whereby the
mass-flow is at a maximum for a given downstream pressure.
CubeSat Nano-satellite built up out of 10 [cm] cubes, called 1U, each having a maximum
mass 1.33 [kg].
Delta-V or ΔV The total velocity increase that a propulsion system can deliver to the spacecraft in
which it is integrated during its entire lifetime. Could also be considered to be the
total impulse the propulsion system can deliver, while taking the satellite mass into
consideration.
Dondersteen The name of the MEMS thruster developed for this master thesis and described in
this document. This Dutch word means “Thunderstone” in English. Chosen because
the thruster is made mainly out of silicon, the main component in stones. And
because it is powered by electricity.
Enthalpy The thermodynamic state or potential of a system, which depends on its internal
energy, pressure and volume.
Flight model The complete hardware package of a satellite subsystem that will be used in the
actual satellite.
Knudsen number The ratio between molecular mean free path length and the length scale of the
problem. In this report used to assess if gas continuum hypotheses are valid.
Micro-satellite Satellite with a mass in the range of 10 [kg] to 100 [kg].
Acronyms
DAQ Data AcQuisiton device (by National Instruments).
DASML Delft Aerospace Structures and Materials Laboratory.
DEMO Dienst Elektronische en Mechanische Ontwikkeling. Which is Dutch for the
electronic and mechanical support division at the TU Delft.
Dimes Delft Institute for Microsystems and Nanoelectronics, the manufacturer of the
Dondersteen MEMS thruster described in this thesis.
DUR-1 Delft University Resistojet - 1.
MEMS Micro Electro-Mechanical System.
NI National Instruments.
NovAM Novel Aerospace Materials, a chair at the faculty of aerospace engineering at the
TU Delft.
PCB Printed Circuit Board.
SEM Scanning Electron Microscope.
SiC Silicon Carbide.
SSE Space Systems Engineering, the chair under which this master thesis was executed
at the TU Delft.
TU Delft Delft University of Technology (Dutch: Technische Universiteit Delft).
List of Figures
FIGURE 2.1 - SCHEMATIC REPRESENTATION OF A TYPICAL MONOPROPELLANT ROCKET ENGINE [12] ........................................... 27
FIGURE 2.2 – A TYPICAL COLD GAS THRUSTER [13] .......................................................................................................... 28
List of tables
TABLE 2.1 - RELEVANT PROPERTIES OF COMMON SUBSTANCES WHICH ARE LIQUID AROUND ROOM TEMPERATURE, ORDERED BY MOLAR
MASS. GATHERED IN ORDER TO ASSESS POTENTIAL PROPELLANTS. MOST PROPERTIES COLLECTED ON THE NIST DATABASE [23]
AND USING [24]. ............................................................................................................................................... 34
TABLE 3.1 - CRITICAL REYNOLDS NUMBERS IN FLOW CHANNELS WITH DIFFERENT GEOMETRIES. THE ASPECT RATIO Α IS THE RATIO
BETWEEN CHANNEL HEIGHT AND CHANNEL WIDTH OF THE FLOW CHANNEL Α = HCH/WCH. .................................................. 40
TABLE 3.2 - THE ASSUMPTIONS MADE IN THE PROPULSION SYSTEM PERFORMANCE CALCULATIONS ............................................ 51
TABLE 3.3 - FINAL SELECTION OF THE DESIGN PARAMETERS FOR THE PROPULSION SYSTEM........................................................ 57
TABLE 5.1 - INTENDED REQUIREMENT COMPLIANCE INFORMATION GATHERING...................................................................... 75
TABLE 5.2 - ACTUAL REQUIREMENT COMPLIANCE INFORMATION GATHERING......................................................................... 97
TABLE 7.1 - CHARACTERISTIC DIMENSIONS AND PERFORMANCE OF A WATER FED MICRO-EVAPORATOR [63], THE MEANINGS OF THE
ABBREVIATIONS ARE GIVEN IN FIGURE 7.34. ......................................................................................................... 123
TABLE 7.2 - MOST IMPORTANT DESIGN AND PERFORMANCE PARAMETERS OF THE HEATER MODULES........................................ 127
TABLE 8.1 - COMPARISON BETWEEN THE DESIGN VALUES OF THE THRUSTER GEOMETRY AND THE ACTUALLY PRODUCED VALUES AS
CHECKED WITH THE OPTICAL MICROSCOPES. .......................................................................................................... 144
TABLE 8.2 - MEASUREMENT RESULTS OF THE CHANNEL DEPTH IN SPECIFIC AREAS, MULTIPLE VERSIONS WERE MEASURED WHERE
MULTIPLE DIMENSIONS ARE LISTED. CHANNEL DEPTH AND RIDGE HEIGHT FROM THE SAME CHANNEL ARE LISTED IN THE SAME
COLUMN. ....................................................................................................................................................... 148
TABLE 8.3 - THE SUMMARY OF THE COLD ELECTRICAL RESISTANCE TESTS ON THE DIFFERENT SIC HEATING ELEMENTS OF THE
DONDERSTEEN THRUSTER. ALL RESISTANCES HAVE BEEN MEASURED ON AT LEAST 4 DIFFERENT MODULES, AND MOST OF THE
VALUES ARE AN AVERAGE OF MULTIPLE MEASUREMENTS ON THE SAME MODULE .......................................................... 148
TABLE 12.1 - CRITICAL REYNOLDS NUMBER FOR RECTANGULAR CHANNELS, DEPENDING ON ASPECT RATIO [31] [32] .................. 183
1.2 Objectives
This operational demonstration will be undertaken by the DelFFi mission, by having two nano-
satellites demonstrating formation flying capabilities, such as inter-satellite distance maintenance
[5]. The formation flying objective requires a new high performance propulsion system. This system
will be partially based on experiences with the T3μPS flown on the Delffi-N3Xt [6]. The DelFFi
satellites will operate in the QB50 framework. This is a network of 50 nano-satellites, with the goal to
take multi-point in-situ measurements of the lower thermosphere and research re-entry [7].
Design a propulsion system to meet the propulsive requirements of the two DelFFi mission
formation flying satellites.
A rivalling thruster is being developed by Mr. I. Krusharev, using different manufacturing techniques
and design philosophies. The thruster designed by the author is produced using Micro Electro-
Mechanical System (MEMS) manufacturing, in cooperation with research institute Dimes. Mr.
Krusharev focuses on commercially available products and traditional mechanical manufacturing
technique. Mr. Krusharev therefore collaborates mainly with the mechanical production workshops
of the university. Both of these thrusters are required to be able to operate on the same propellant
storage and feed system. This enabled the author and Mr. Krusharev to cooperate in the
development of this propellant system. And it makes it possible to eventually chose the best
performing thruster design, to be integrated in the final flight model. It was decided that the
responsibility of the propellant storage and pressurisation lies with the author. Mr. Krusharev is on
the other hand responsible for the propellant feed valve and the necessary drive electronics.
The subsequent chapters will describe the individual sections of the propulsion system. The
necessary calculations are documented, design options are outlined and the final selection is
described and explained. This will start with the description of the propellant storage system in
chapter 4. Which is followed by a slight excursion in chapter 5. Here the extensive tests are
documented, which were performed in order to test one of the pressurisation system options,
described in chapter 4. Chapter 6 will briefly describe the necessary propellant feed system. The
most important and intricate part of the propulsion system is the resistojet thruster itself. All the
integrated sections in this thruster will be elaborated upon in chapter 7. In chapter 8 the
documentation on the necessary test campaign of the newly developed thruster follows. This
chapter will also describes the analysis of the results of the tests conducted thus far. Finally chapter
9 will fulfil the vital task of describing the future tasks that still need to be completed in order to go
from the current prototype to the final flight model of the propulsion system.
PROP-PERF-100: The total Delta-V provided by the propulsion system shall be at least
15 [m/s].
PROP-PERF-200: The thrust provided by the propulsion system shall be above 0.5 [mN].
PROP-PERF-205: The thrust provided by the propulsion system shall be below 9.5 [mN].
PROP-PERF-400: The propulsion system shall have a lifetime of at least 1 year under
operational conditions in space.
PROP-PERF-410: The propulsion system shall be able to withstand at least 5000 on-off cycles
without losing its capability to meet any other performance or system requirement.
PROP-SYST-100: The total wet mass of the propulsion system at launch shall be not higher
than 459 [g].
PROP-SYST-200: The total size of the propulsion system shall be within
90 [mm] x 90 [mm] x 80 [mm]
PROP-SYST-310: The peak power consumption of the propulsion system during ignition or
heating shall not be higher than 10 [W].
PROP-SYST-320: The total energy consumption of the propulsion system shall be no more
than 100 [kJ] per day (i.e. 1.1574 [W] daily average).
PROP-SYST-410: The nominal geometrical axis of the nozzle shall be perpendicular to the
plane of one of the smaller faces (100 [mm] x 100 [mm]) of the satellite.
PROP-SYST-500 to PROP-SYST-560: The propulsion system shall be able to withstand the
launch loads.
PROP-SYST-600: The internal pressure of all propulsion system components shall not be
higher than 10 [bar].
PROP-SYST-610: The propulsion system shall not include any pyrotechnic devices.
PROP-SYST-620: The propellant(s) used by the propulsion system shall not be hazardous for
the operators or the other satellite sub-systems.
In addition to this, there are also two design guidelines which are not in the mentioned requirements
document.
The resistojet housing shall fit within the centre of the antenna board, which is in between
the propulsion system and the exterior bottom panel of the satellite. The cut-out in this
antenna board, and subsequently the bottom panel, has maximum dimensions of 2 [cm] by
2 [cm].
The propulsion system shall be able to operate with a power supply voltage of 5 [V].
Next to these propulsion system specific requirements, the system shall also abide by the standard
CubeSat requirements. These requirements are stated in the CubeSat design specification
document [9]. Any one of these specifications is only to be disregarded when an official waiver has
been granted. The procedure of requesting such a waiver is described in the same document.
Propellant storage
Propellant pressurisation system
Propellant feed system
Reaction or heating chamber
A rocket nozzle
From Newton’s second law of motion F m a , we know that for the same amount of expellant
mass, the reaction force is higher, when the acceleration is higher. Thus, the higher the velocity of
the expellant is, the higher the thrust force. There are many different propulsion system concepts. A
literature study was conducted as a prelude to this master thesis [10]. The purpose was to identify
the best miniaturisation opportunities for reaction engines in the nearby future. The requirements of
the DelFFi mission where used as a reference for nano satellite propulsion system capabilities. In the
literature study, five main groups of reaction engines were identified, according to their method of
thrust production. Within these groups a total of 29 individual thruster designs were evaluated. Of
these 29 concepts, eight were found to be applicable for CubeSat propulsion. A more in-depth trade-
off was therefore conducted on these remaining concepts. The final trade-off grading for these
These three lie too close together to say which concept is best for this mission. This will have to be
evaluated in more detail. The fourth highest scoring concept has a score of 381, which is clearly
distinguishable from the top three scoring concepts.
In the following subsections, the different propulsion system options will be discussed. After an
elimination of all non-optional systems, only the final propulsion system remains.
The main problem for monopropellant systems is the propellant handling and storage [13]. The most
common monopropellant is hydrazine (N2H4). This is highly toxic, corrosive and the freezing point is
relatively high at around 2 [°C]. Another monopropellant is hydrogen peroxide (H2O2). This is not
toxic, it is however corrosive, harmful and a strong oxidant. Hydrogen peroxide also auto-
It can therefore be concluded that the simplicity, low cost and relatively high performance of the
monopropellant thrusters are favourable for university projects. This development however
necessitates more safety facilities, procedures, expertise and experience than which is available at
the chair at this moment. Therefore it cannot be selected as the propulsion system for the DelFFi
mission satellites.
As stated before, cold gas propulsion systems are mainly used in applications, where simplicity and
reliability is more important than performance. In the DelFFi mission however, a considerable
amount of total impulse is required from the propulsion system. In a cold gas thruster there is no
temperature rise due to combustion or heating of some kind. This means that the vessel normally
has to constantly be at the same pressure as the required chamber pressure. Traditionally this
pressurised gas was carried in high pressure storage tanks, which were pre-pressurised before
launch. In CubeSats pre-pressurisation above 1.2 [atm] is however not allowed [16]. Therefore
pressurisation has to be done dynamically in orbit. For example by using cold gas generators [17]. As
suggested by Dr. Cervone, a cold gas thruster is a feasible option for the DelFFi satellites, when
improvements such as a resistojet are in place [18].
The initial design of the DelFFi mission propulsion systems was a scaled up version of the one used in
the Delfi-n3Xt. This involved increasing the size of the gas generators, shown in Figure 2.4, from
18.7 [mm] x 6.7 [mm] [19], to about 80 [mm] x 20 [mm]. The propellant load of each gas generator
will then grow from 0.3 [g], to 11.4 [g]. This will in its turn result in an increase of nitrogen gas
production from 0.125 [g] to 4.77 [g]. When increasing the amount of gas produced is increased by,
the plenum internal volume has to grow with the same factor, when the same settling pressure is
desired. This will then have to grow from 28 [cm3] to 1067 [cm3]. This volume is equivalent to a cube
with inner sides of 10.2 [cm]. This plenum volume alone would not even fit the allocated DelFFi
propulsion system. The internal plenum volume could be as little as half the size, which would double
the settling pressure and increase the mass of the plenum. Even still it would not fit the propulsion
system volume budget, together with the gas generators, valve, nozzle and drive electronics.
3
Figure 2.4 - T μPS Cool gas generator used on board the Delfi-N3xt
The key advantage of this propulsion system is its flexibility. This is due to the fact that the energy
consumption and thrust manoeuvres are partially decoupled. Therefore it can produce the
pressurised gas by electrolysis, whenever there is available excess electrical power. And it can release
this pressurised gas, in order to produce thrust, at any required moment. Since it operates on gas
alone, there is no risk of incomplete combustion or vaporisation, which may lead to performance
reduction.
The propulsion system has been calculated, to require 6.5 times the electrical energy input of an
equivalent resistojet, in order to produce the same amount of thrust. These calculations however
only consider the theoretical optimal behaviour. The efficiency of the resistojet is assumed to be
lower than that of an electrolysis propulsion system. Therefore it is assumed that this difference will,
in reality, be less pronounced. The biggest disadvantage of the proposed technology is on the other
hand the technology readiness level. This propulsion system would still require a great deal of
research, which is not feasible within the timeframe of the DelFFi mission. Therefore this design
option has to be discarded as well.
At the SSE chair there is a large amount of experience in resistojet development, as will be shown in
section 7.1. Preliminary calculations of a resistojet propulsion system using water as a propellant
were performed. These showed that such a system can indeed fulfil the DelFFi propulsion system
requirements, with respectable margins for error. It was therefore decided to conduct this master
thesis, to develop a propulsion system for the DelFFi mission satellites using a resistojet. Section 2.4
will describe the possible layout of such a propulsion system. After which section 2.5 is devoted to
showing which propellant is most suitable for such a propulsion system.
There are also a couple of support components necessary for the operation and monitoring of the
propulsion system. There has to be a fill valve to insert the propellant and possibly the pressurant
into the storage tank. A temperature sensor is required, to monitor the temperature of the resistojet
heating chamber. It is also desirable to have a pressure sensor in the propellant storage. The
monitoring of this pressure will aid in the assessment of the operation and effectiveness of the
propulsion system. The propulsion system has to be operated and monitored by electronics on a
Printed Circuit Board (PCB). All the mentioned propulsion system sections will be discussed
individually in the coming chapters of this report. In Figure 2.5 a schematic overview of a proposed
propulsion system for the DelFFi mission satellites is shown. This is only meant as an overview of the
relevant components and their relation to each other. It is not meant to be a correct structural
drawing of the final propulsion system.
The numbers added to this image indicate the following parts of the propulsion system:
Not included in Figure 2.5 are the necessary sensors for monitoring the system, the propellant fill
valve and a possible pressurant fill valve. Contrary to the design in the figure, the most optimal
design would have the resistojet in the centre of the propellant storage system. The propellant could
then be in a toroidal tank around the thruster itself. In this configuration the principal heat loss of the
thruster can be absorbed by the stored propellant. This will help in the effort of keeping the water
propellant from freezing. The resistojet itself could then even be used as a heater to melt the water
propellant, when it freezes. The elevated temperature of the stored propellant will also result in a
lower necessary heat input for vaporisation in the thruster. It will therefore decrease the power
consumption of the thruster, by using the waste heat of the thruster. A toroidal propellant tank
shape is relatively complex. This shape will also possibly increase the complexity of propellant
pressurisation and collection, which will be discussed later in this document.
As will be shown in equation 3-24, the thrust produced by a resistojet depends on a number of
different factors. In order to compare the effect of using different propellants, the behaviour is
idealised. When it is assumed that the nozzle is able to ideally expand the flow to a vacuum, the
thrust in equation 3-24, only depends on the propellant mass flow and exhaust velocity. In the ideal
case of adiabatic expansion of the accelerated flow to vacuum conditions, the exhaust velocity equals
the limiting velocity UL[22]:
RA
UL 2 T 2-1
1 Mˆ c
Where:
UL Limiting velocity [m/s]
Ratio of specific heats [-]
RA Universal gas constant 8.314510 [J/mol/K] [4]
M̂ Molar mass [g/mol]
Tc Heating/combustion chamber temperature [K]
The universal gas constant is the same for every substance. The height of the limiting exhaust
velocity, for a certain combustion chamber temperature, therefore depends on the following ratio:
ˆ 1
M
The ratio of specific heats is difficult to define, as it depends on the temperature of the substance.
The difference in ratio of specific heats between different substances is very small compared to the
difference in molar mass. Therefore the maximum exhaust velocity of the substances depends mainly
on their molar mass. The lower the molar mass, the higher the exhaust velocity. The higher the
exhaust velocity, the higher the thrust for a certain mass flow, as will be shown in chapter 3. In this
case idealisation of the behaviour is done to simplify the equations. Even when the behaviour is not
ideal or when it is operating in non-vacuum conditions, this relation between molar mass and
exhaust velocity still holds.
Since the DelFFi mission is mainly conducted by students, the safety of the potential propellants also
plays a big role. As students do not always have the time for lengthy safety courses, and generally
lack experience with dangerous substances. Also because the mission is of a highly experimental
Table 2.1 - Relevant properties of common substances which are liquid around room temperature, ordered by molar
mass. Gathered in order to assess potential propellants. Most properties collected on the NIST database [23] and on [24].
As can be seen from Table 2.1, water is the substance with the lowest molar mass, which is still liquid
at room temperature. It is therefore capable of producing the highest total impulse for a given
amount of fuel. One disadvantage of using water as resistojet propellant, is that the melting
temperature is relatively high. Therefore there is the risk of propellant freezing. As said before, this
might be circumvented by adding ammonia to the water. A second disadvantage is that it has a
rather high enthalpy of vaporisation. It will therefore require relatively large amounts of energy to
vaporise the liquid.
Hydrogen cyanide is the substance with the second lowest molar mass in the list. It will therefore not
be able to attain the exhaust velocities and thus total impulse of water. The advantage is that the
enthalpy of vaporisation is significantly lower. It will therefore require less energy to vaporise. The
liquid temperature range is also more favourable, because it does not pose the risk of propellant
freezing. Hydrogen cyanide is however extremely toxic and is therefore not suitable for student
handling and highly experimental projects.
Alcohols have a lower heat requirement and a more favourable temperature range than water. They
are also readily available and cheap. The lightest one, methanol, is very toxic, even though the molar
mass is already almost double that of water. Therefore water is still favoured above alcohols as
resistojet propellant.
Hydrogen peroxide still has a relatively low molar mass, and a good liquid temperature range. It is
however very difficult to handle and to store. This is due to its instability, corrosiveness and
harmfulness. In the presence of a proper catalyst and/or temperature, it will decompose. This
produces its own heat, and exhaust products with a lower molar mass. It would however then form a
chemical monopropellant thruster, which was discarded in subsection 2.3.2.
Hydrochloric acid deserves a special mention because of its very low enthalpy of vaporisation. It
therefore requires the least amount of energy input to vaporise it, out of all here listed liquids. It
does however have a significantly high molar mass compared to water. Out of all substances listed
here hydrochloric acid is, in addition to that, very hazardous to work with. Low concentrations of
<10% of the hydrochloric acid, can be introduced in the water propellant to lower the melting point
down to -20 [°C]. Again reducing the risk of propellant freezing. But this will increase the average
molar mass.
Hydrocarbons are readily obtainable, very cheap and well documented. However the simple
hydrocarbons with the lowest molar mass are all gaseous at room temperature. The lightest
hydrocarbon which is liquid at room temperature is Pentane (C5H12). This still has a relatively low
Essential oils are usually terpenes, and all have a high molar mass, due to the large molecular
structure with 10-40 carbon atoms [25]. Their liquid phase temperature range does not pose a risk of
either boiling or solidifying during storage. They however have a large enthalpy of vaporisation, and
the molar masses are too large to be effective as resistojet propellant
Bromine and Mercury are the only elements that are liquid at standard room temperature and
pressure. They are however the liquids with the highest molar mass in this list. Therefore they are
not desirable as propellant in resistojets.
To conclude this propellant selection, water is found to be the best resistojet propellant, due to its
low molar mass alone. In addition to this, it is also the most benign and readily available, cheap
substance in this investigation. Other substances can be added in order to reduce the melting
temperature. This reduces the chance on propellant freezing, during cold operational periods. The
specific heat was not taken into consideration in this investigation. In future efforts this might be
considered together with the enthalpy of vaporisation, in order to form a view on which fuel would
be most energy efficient to use.
The performance of the system is assessed by iterating over different design parameters, until the
requirements are met. The behaviour of the propulsion system is simulated using equations for
partially idealised situations. For these calculations, the system starts at the initial propellant and
pressurant mass. At which point all the flow parameters from propellant storage to expulsion are
calculated. Time steps are taken from a full tank to an empty tank. At each time step, all flow
parameters are calculated throughout the system. This data is then finally integrated over time to
obtain the total performance. In the following sections of this chapter, the calculation of the flow
properties of each section of the propulsion system will be described.
mpropi mpropi1 mi 1 t
Where:
mprop The total stored propellant mass [kg]
i The current time step [-]
m Mass flow [kg/s]
t The duration of the last step [s]
If this calculation shows that the propellant supply has been exhausted, i.e. mpropi 0 , the
calculations stop and display the total performance values. The total stored propellant volume is the
propellant mass divided by its density:
mpropi
Vpropi
prop
The new pressurant volume, i.e. the volume that contains the gas used for pressurisation, is the total
storage volume minus the total stored propellant volume.
The pressurant for the current time step can be calculated using the ideal gas law, shown in equation
3-1 below.
n RA Tpres
ppresi 3-1
Vpresi
Where:
n Amount of pressurant [mol]
ppresi Pressurant pressure [Pa]
RA Universal gas constant [J/K/mol]
Tpres Pressurant temperature [K]
Equation 3-1 only holds for hypothetical gasses where molecules are negligible in size and have no
intermolecular forces [22]. Due to the relatively low temperature, pressure and flow rate of the
pressurant gas, this assumption is valid.
As will be explained in chapter 6, there is some difficulty in extracting the propellant from the tank, in
weightless conditions. Therefore it is highly likely that not all propellant can be extracted from the
tank during the lifetime of the propulsion system. The percentage of the total initial amount of
propellant, that can actually be expelled from the propellant storage is called the “expulsion
efficiency”. This expulsion efficiency is in the order of 95% to 99%, for normal liquid propulsion
systems [22]. The propulsion system described in this master thesis is however of a more
experimental nature. It is also mostly designed by relatively inexperienced engineers. Therefore an
expulsion efficiency of only 90% was assumed for the calculations in this thesis.
Propellant tank exit in the centre of the face which faces the nozzle location
Straight connection piping
Propellant control valve
Fluidic connection piece to the thruster
These sections would ideally all be straight, smooth and have the exact same diameter. This would
lead to a system with only the losses due to a relatively short section of straight, smooth piping. Such
a loss will later be shown to be negligible. Only the flow constriction loss of the flow from the
propellant tank to the feed system will induce some extra losses.
For the propulsion system engineering model tests, the Lee company INKX0511400A A solenoid valve
is used for propellant supply control. This valve is selected for its similarity to the valve which will be
used on the actual satellites. Therefore the inside diameter of 0.89 [mm] [27] [28], of this
engineering model valve is used as a reference for flow calculations.
m
U 3-2
A
In the given channel this corresponds to a propellant flow velocity of U = 3.21 · 10-3 [m/s]. The
kinematic viscosity is a measure of a fluids resistance to deformation, which is calculated using
equation 3-3.
3-3
Where:
Kinematic viscosity [m2/s]
Dynamic (absolute) viscosity [kg/s/m]
The dynamic viscosity for water at room temperature (25 [°C]) is μ = 891·10-6 [kg/s/m] [29]. One of
the most indicative flow parameters used throughout the aerospace research spectrum, is the
Reynolds number. This dimensionless quantity aids in the prediction of flow behaviour. The Reynolds
number based on tube diameter is usually defined as shown in equation 3-4 below.
U Dh
Re 3-4
Where:
Re Reynolds number [-]
U Flow velocity [m/s]
Dh Hydraulic diameter of the tube [m]
Turbulence in the flow will induce height convectional losses, due to the chaotic behaviour. The
transition between regimes occurs at critical Reynolds numbers. Typical critical Reynolds numbers.
depending on the flow channel geometry can be found in Table 3.1.
Table 3.1 - Critical Reynolds numbers in flow channels with different geometries. The aspect ratio α is the ratio between
channel height and channel width of the flow channel α = h ch/wch.
16
f 3-5
Re
Where:
f Fanning friction factor [-]
Which can be calculated to be f = 4.98 [-]. Using all previously calculated flow properties, the
pressure drop across a straight tube is calculated using the following formula [31]:
L
p 2 U2 f 3-6
Dh
Where:
p Pressure drop [Pa]
L Tube length [m]
The pressure drop of 90° bends can be calculated using equation 3-6, with an equivalent length of 57
times the inner diameter [31]. This equivalent length adds to the estimated 17 [cm] of piping.
Therefore the total equivalent length of the feed system becomes 0.424 [m]. Using equation 3-6 with
all the calculated flow parameters, results in the following pressure drop:
0.89 103
This can clearly be seen to be negligible. Even if the propellant is pressurised only to atmospheric
pressure of 1 [bar], this is still only 0.049% pressure loss.
2
q
p G f 3-7
N1 Cv
Where:
Gf Liquid specific gravity, water = 1.0 [-]
q Flow rate [L/min]
Later in this chapter, a mass flow rate of 2 [mg/s] is shown to be above the highest value which will
be encountered in operation. This mass flow rate results in a volume flow rate of 120·10-6 [L/min].
This yields a pressure drop of 4.33·10-6 [bar]. This pressure drop across the valve can thus be seen to
be negligible, compared to the feed system pressure in the order of 5 [bar] to 7 [bar].
1
p U2 3-8
2
Where:
p Pressure drop [Pa]
Pressure loss coefficient [-]
3
A2 4
contr 0.5 1
3-9
A1
Where:
A2 Secondary, smaller piping cross-sectional area [m2]
A1 Primary, larger piping cross-sectional area [m2]
2
A
exp 1 1 3-10
A2
Where:
A1 Primary, smaller piping, cross-sectional area [m2]
A2 Secondary, larger piping, cross-sectional area [m2]
To simulate the transitions between the different feed system sections, they are represented by
sudden diameter changes. To have an order of magnitude pressure loss estimation, an increase in
diameter of a factor two is assumed. After which it immediately returns back to the general piping
diameter. This agrees with the transition for the valve fittings [28]. The pressure drop of one such a
transition will then be:
2 4 4 2
p 4.99 10 Pa
3
Due to the very low flow velocities and modest diameter change, this pressure drop is clearly
negligible. Even for the estimated four section transitions, this amounts to a mere 0.020 [Pa]
pressure drop. This is many orders of magnitude lower than the operating pressure of the propulsion
system, which is in the order of 5 [bar] as will be shown later in this chapter.
t2 t3 t4 E
H A t B C D F 3-11
2 3 4 t
Where:
H The enthalpy of the fluid at the given temperature [J/mol]
A to H Empirical constants from the NIST database, shown in Appendix B
t Temperature [K/1000]
The temperature of the flow at the end of the vaporisation section, is equal to the temperature at
which the pressure in the flow is equal to the vapour pressure. Which means that the temperature is
high enough to allow the propellant to be vaporised under the given pressure. This boiling point
temperature is found using the Antoine equation from the NIST database [23]:
When this vaporisation temperature or boiling point at the given pressure is known, the enthalpy at
the end of the vaporisation section can be determined. This enthalpy is calculated with equation 3-11
using the coefficients for gaseous water, which are also documented in Appendix B. The power
required for vaporisation is then calculated using equation 3-13 below.
mprop
Pvap Hvapexit Hvapin 3-13
ˆ prop
M
Where:
Pvap Power required for vaporisation [W]
mprop Propellant mass flow [kg/s]
ˆ prop
M Propellant molar mass [kg/mol]
For the heater efficiency calculations, it is assumed that the liquid water density change because of
the heating up until the vaporisation is negligible. This means that the volume flow at the beginning
of the vaporisation section is equal to the volume flow just before the actual vaporisation. The
electrical power required for the propellant vaporisation section, depends on the efficiency of the
heater to convert the electrical energy to thermal energy in the propellant:
This efficiency depends on the heat transfer or loss to the surroundings of the thruster. In these
initial system level calculations, a heating efficiency of 50% was assumed to do the full calculations. It
was later attempted to estimate the true heater efficiency, using heat transfer calculations, as will be
shown in chapter 7. A detailed calculation proved to be too intensive for the timeframe of this
research. The necessary assumptions to make a simplified approximation of the heat loss posed too
much uncertainty to be valid. Eventually an approximate simplified solution has been calculate to
verify the assumption.
mprop
Pheat Hheatexit HEvapexit 3-14
ˆ prop
M
Where:
Hheatexit Enthalpy of the gaseous propellant, with the desired final temperature, at
gaseous propellant heating section exit [J/mol]
The electrical power required for the gaseous propellant heating section, again depends on the
efficiency of the heater:
PElecheat heat Pheat
Where:
PElecevap Electrical power required for the gaseous propellant heating section [W]
evap Heater efficiency [-]
Again a heater efficiency of 50% was assumed. Which means that 50% of the inserted electrical
energy is lost to the surroundings.
The nozzle chosen for the Dondersteen micro-resistojet is a converging-diverging nozzle, as shown in
Figure 3.1. This nozzle type is the most commonly used type for rocket engines that are based on
heated gas expulsion. It functions on the principle of the different behaviour of gas between subsonic
and supersonic flow. A property of subsonic flow is that the flow speed increases when the flow
channel cross-sectional area decreases. The subsonic flow is therefore first accelerated due to the
contraction of the converging section. If the pressure difference between the heating chamber and
the nozzle exit are high enough, the flow will be choked in the narrowest section, called the throat.
This means that the flow attains sonic velocity. A property of supersonic flow is that the velocity
Figure 3.1 - Schematic cross-section of a converging-diverging nozzle - Flow direction from left to right
The flow can theoretically be expanded until the ambient pressure is reached. It can also be
expanded to a pressure even lower than the ambient pressure. This is excessive expansion is called
overexpansion. Since the Dondersteen thruster will operate in vacuum conditions, this is however
not possible or relevant. To expand the exhaust flow to vacuum ambient conditions, the exit has to
be infinitely large, which is physically impossible. Therefore there is a point where the increase in
exhaust velocity, does no longer weigh up to the increase in nozzle size. The exhaust flow will
therefore always be under expanded. This under expansion reduces the efficiency of the nozzle to
convert the potential and heat energy of the flow to kinetic energy.
In subsection 3.4.1 the theoretical nozzle performance will be discussed. This is followed by a
prediction for the translation of this theoretical performance to the true performance, in subsection
3.4.2.
prop pheatexit At
mcrit 3-15
R Theatexit
Where:
mcrit Critical mass flow [kg/s]
prop Vandenkerckhove parameter [-]
RA
R 3-16
Mˆ
The above mentioned Vandenkerckhove constant is calculated using equation 3-17 [22].
p 1
2 2 p 1
3-17
prop p
1
p
Where:
p Is the ratio of specific heats Cp/Cv [-]
Cp Heat capacity of the fluid at constant pressure [J/kg/K]
Cv Heat capacity of the fluid at constant volume [J/kg/K]
If the mass flow corresponds to the critical mass flow, the flow is choked and the velocity in the
throat is sonic. In that case the Mach number of the flow at the exit of the nozzle can be calculated,
using equation 3-18 [22].
p 1
Ae 1 2 p 1 2 p 2
3-18
1 Me
At Me p 1 2
Where:
Ae Cross-sectional area of the exit of the nozzle [m2]
Me Mach number of the flow at the exit of the nozzle [-]
This equation cannot be solved analytically. Therefore the MATLAB’s “fsolve” function is used. This
function needs a first estimate of the parameter to be solved, to be able to operate. If this estimation
is too far from the value that needs to be found, there is the risk that the function finds other
solutions. Therefore the exit Mach number was estimated to lie around Me=4 [-].This was later
verified with the results. Assuming isentropic flow, the pressure ratio in the propellant flow, between
the nozzle exit and nozzle throat is [35]:
p
pe p 1 2 1 p 3-19
1 Me
pt 2
Where:
pe Pressure of the flow at the exit of the nozzle [Pa]
pt Pressure in the flow at the nozzle throat [Pa]
Using previously defined parameters, the characteristic velocity is calculated using equation 3-20
[22]. This property reflects the energy level of the propellant, similar to the specific impulse. The
1
c R Theatexit 3-20
p
Where
c Characteristic velocity [m/s]
Theatexit The (desired) temperature at the exit of the heating chamber [K]
The characteristic velocity is subsequently used in equation 3-21 [35], to calculate the specific
impulse. This is done by taking the pressure ratio into account.
1
p 1
p 1 2
c 2 2 p 1
p p
Isp p 1 e
3-21
g0 p 1 p 1 pt
Where:
Isp Specific impulse [s]
g0 Gravitational acceleration at standard sea level 9.81 [m/s2]
The definition of the specific impulse, see 3-22, is then used to calculate the exhaust velocity.
Ue Isp g0 3-22
Where:
Ue Exhaust velocity [m/s]
Since the flow in the throat is sonic, the pressure is equal to the critical pressure shown in equation
3-23 [22].
p
2 p 1 3-23
pt pheatexit
1
p
Where:
pheatexit The pressure at the end of the heating chamber [Pa]
Using this throat pressure and the pressure ratio calculated with equation 3-19, the pressure in the
exhaust flow at the nozzle exit can be determined. Using the now calculated parameters, the axial
thrust force can be calculated using equation 3-24 [36]. In this equation, the ambient pressure is
assumed to be vacuum i.e. 0 [Pa].
The Reynolds number based on tube diameter, is usually defined as shown in the previously stated
equation 3-4. The kinematic viscosity is calculated with equation 3-3. The flow velocity is calculated
with equation 3-2. The area of a circle is A 1 D . Using these equations, the formula to
2
4
calculate the Reynolds number, shown in equation 3-4, can be rewritten as following.
m Dh 4m
Re
Dh
Dh2
1 3-25
4
In the case of non-circular tubing, the equivalent hydraulic diameter Dh has to be calculated. The
hydraulic diameter Dh is a measure to relate the area of any cross-section to the diameter of a
circular cross-section with the same area. This parameter is defined as follows:
4A
Dh 3-26
p
Where:
A Cross-sectional area of the tube [m2]
p Perimeter or inner wall length of the cross-section of the tube [m]
Federico La Torre has thoroughly investigated the flow behaviour through micro-nozzles [37]. For the
calculations in this paper a flow of gaseous nitrogen (N2), at 300 [K] and 10 [bar], in the inlet is
assumed. It was found that nozzles with a thrust of larger than ∼1 [mN], rarefaction and wall slip
effects can be neglected. This assumption can be made due to the very low value of de Knudsen
number. The Knudsen number is the ratio between molecular mean free path length and the
characteristic length scale, of a problem in physics. In the case of gas flowing through a channel, this
is used to assess if gas continuum hypotheses are valid. The characteristic length in this case is
therefore the hydraulic diameter Dh [31], as shown in equation 3-27. When this Knudsen number
rises above 0.01 [-], the molecular density is too low for traditional ideal continuum flow equations.
The flow is then called rarefied, and special flow property calculation methodologies apply.
Kn 3-27
Dh
The mean free path of the gas molecules in the flow can be calculated using equation 3-28 [31].
3-28
2 p
In the research by la Torre the highest encountered Reynolds number was 105 [-], at a thrust of
1 [N] [37]. At this Reynolds number the difference between laminar flow property approximations,
and models using turbulent effects, was less than 1.5%. Therefore it was concluded that the flow in
all investigated nozzles was laminar. Probably because, due to the short nozzle length, the
turbulence had neither time nor space to develop. In this study by la Torre throat diameters between
0.94 [μm] and 860 [mm] were modelled, with a thrust of 1 [μN] to 1 [MN] respectively. As will be
shown later, the nozzles developed for the Dondersteen thruster fall well within this range.
In the throat of the nozzle, there can be a sharp angle between the converging and diverging
sections, as seen in Figure 3.1. This throat angle can also be rounded off with a radius of Rc, see
Figure 3.2 below. This rounding is applied while retaining the same throat diameter Rt. The radius of
this throat edge chamfer will have an effect on the thrust efficiency of the nozzle, as can be seen in
Table C.1 in Appendix C.
Figure 3.2 - Longitudinal cross-section of a conical nozzle, including its geometrical parameters [37]
As will be shown in section Fout! Verwijzingsbron niet gevonden., the thrust of the nozzles
described in Table C.1 is very close to the thrust that the Dondersteen thruster is designed for.
Therefore the thrust efficiency behaviour is assumed to be indicative of the behaviour expected from
the Dondersteen thruster. It can therefore be concluded that it is most optimal to have a sharp edge,
between the converging and diverging section of the nozzle. This means that the curvature radius
Rc = 0 [m].
In a study to optimise the combination of diverging and converging half angles of micro-nozzles,
many different combinations were attempted [31]. The most relevant resulting data is summarised in
Table C.2. In this study the combination of θ1 = 15 [°] and θ2 = 20 [°] yields the highest efficiency.
With this ratio, the converging angle is relatively low, which generally has a value in the region of
30 [°] [22]. The diverging half angle is relatively high, compared to range of 12 [°] to 18 [°], normally
considered optimal. When deviating from the conical shape, the long nozzle length due to these low
angles can be decreased. The length of the nozzle can be reduced to about 75% of the conical length,
while having the same area ratio and thrust coefficient CF, using a bell shaped nozzle [22].
The etching process used for the production of the Dondersteen micro-thruster, generally has a sub-
micrometer dimensional accuracy. The throat diameter itself is in the order of 25 [μm], as will be
shown in section 7.5 of this report. This could therefore cause a maximum surface roughness with an
amplitude of 4% of the throat diameter. It is therefore possible, that some weak shocks waves are
formed in the nozzle area. However the roughness is relatively uniform, in contrary to the very
pronounced wavy surface roughness in the simulations by la Torre. And since the roughness is also
generally lower, than the maximum of 1 [μm], the losses due to this roughness are expected to be
limited.
The general efficiency that MEMS nozzles can achieve can be seen in the results of the research by la
Torre [37]. As seen in the data in Appendix C, the nozzle efficiency can easily lie in the region of 90%.
This efficiency indicates the ratio between ideal theoretical thrust of the nozzle, and the thrust which
is actually delivered. This is also called the nozzle quality. The nozzle efficiency is therefore assumed
to be 90% for these calculations.
The conclusions derived from the above mentioned geometrically dependent nozzle efficiency, will
be described in section 7.5.
Table 3.2 - The assumptions made in the propulsion system performance calculations
Parameter Assumption
System temperature 293.15 [K]
MEMS thruster flow channel height 100 [μm]
Total satellite mass 3.0 [kg]
Pressurant gas Nitrogen
Percentage of electrical input power lost to the 50%
surroundings while heating
Pressure loss in the heating chamber 20%
Expulsion efficiency 90%
Nozzle efficiency/quality 90%
Total ΔV
Thrust
Pressurant pressure
Electrical power consumption
Knudsen number in the nozzle throat
Reynolds number in the nozzle throat
21.8
21.6
21.4
V
21.2
21
20.8
20.6
0 100 200 300 400 500
Number of steps [-]
Figure 3.3 - Effect of the number of integration steps, on the resulting total ΔV after the calculations.
The deviation between calculations at a low number of steps is relatively large. The behaviour
stabilises relatively quickly around 100 steps. When changing the design variables, this stabilisation
duration will be Different. Therefore a minimal of 300 iteration steps are chosen for a reliable output
of the calculations. In this case the calculated total ΔV only fluctuates with 0.24 % at 300 steps. This
will therefore not further decrease the inaccuracies of the calculations.
10
Total V [m/s]
20.9
20.8
5
20.7
20.6 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Initial pressurant mass [kg] -4
x 10 Initial pressurant mass [kg] -4
x 10
-3 5
x 10 x 10
3 10
2
6
1.5
4
1
2
0.5
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Initial pressurant mass [kg] -4
x 10 Initial pressurant mass [kg] -4
x 10
Figure 3.4 - Effect of initial pressurant mass on propulsion system performance
These graphs show the effect that the total ΔV of the system is only limited above 0.1 [g] of initial
pressurant mass. This curve seems to oscillate, which is an artefact of the accuracy of the
measurements. The oscillation has an amplitude of less than 0.5%, which is acceptable. Due to the
higher mass flow, the thrust and power consumption do increase with increasing pressurant mass. It
can therefore be concluded that the pressurant mass has to be chosen, such that the force and ΔV
requirements are met, for the entire operational range. Further increasing the pressurant mass will
lead to a larger power consumption and larger pressures.
3.5
Maximum power consumption [W]
2.5
10 2
1.5
5 1
0.5
0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Nozzle throat diameter [m] -5
x 10 Nozzle throat diameter [m] -5
x 10
Figure 3.5 - Effect of nozzle throat diameter on propulsion system performance
As is to be expected, the total ΔV remains constant with increasing throat diameter and is therefore
not shown. Due to the increased critical mass flow rate, the power and thrust increase linearly. The
nozzle throat will therefore be sized to provide the necessary thrust level throughout the service
range. The diameter will not be increased much further, in order to maintain an acceptable power
consumption.
1.8
10.2
1.7
10
1.6
1.5
9.8
1.4
9.6
1.3
1.2 9.4
400 500 600 700 800 900 1000 1100 400 500 600 700 800 900 1000 1100
Final propellant temperature [K] Final propellant temperature [K]
25 12
P
vap
P
20 6
15 0
400 500 600 700 800 900 1000 1100 400 500 600 700 800 900 1000 1100
Final propellant temperature [K] Final propellant temperature [K]
Figure 3.6 - Effect of final vaporised propellant temperature on propulsion system performance. Pvap is the electrical
power needed to vaporize the propellant, Pheat is the electrical power required to heat the vaporised propellant to the
desired final temperature’, Ptot is the total required electrical power.
As the temperature increases, the critical mass flow rate decreases. This decreased mass flow rate
results in a slightly lower power consumption. The power consumption increases again at higher
temperatures. This is due to the increased prominence of the propellant heat capacity to heat it to
higher temperatures, compared to the enthalpy of vaporisation, which dominates the power
consumption at lower temperatures. This change in prominence is shown in the bottom right hand
corner of Figure 3.6. This graph also shows that the absolute electrical power consumption variation
is nearly negligible. The maximum thrust force with increasing propellant temperature is constant
and therefore not shown. Due to the increased exhaust velocity, and thus specific impulse of the
thruster, the total ΔV increases with increasing temperature. It is therefore desirable to have the
highest final temperature possible, in order to maximise the ΔV. Due to the restrictions of the heater
material and the surroundings the maximum propellant temperature has been set to 500 [°C], as will
be discussed in chapter 7.
20
Maximum thrust force [N]
2
18
Total V [m/s]
16
1.5
14
12
1
10
0.5 8
0 10 20 30 40 50 0 10 20 30 40 50
Area ratio [-] Area ratio [-]
Figure 3.7 - Effect of nozzle area ratio on propulsion system performance.
The mass flow rate and power consumptions remain constant with increasing nozzle area ratios and
are therefore not shown. The thrust however increases due to the increased exhaust velocity, and
therefore so does the total ΔV. This increase in thrust force with increasing area converges quite
rapidly to the theoretical maximum. The theoretical maximum thrust is the condition that the nozzle
exit pressure equals the ambient pressure, which] in this case is vacuum i.e. 0 [Pa]. At higher area
ratios, there is only a slight thrust force increase with an increase in the area ratio. This increase of
the area ratio will however lead to an increase in the nozzle length. It will also lead to an even
greater increase in the nozzle wall area, which increases the heat loss, and decreases efficiency. An
area ratio of 10 can be seen to more than double the total performance of the system and is without
a doubt beneficial. Above an area ratio of 20, the increased theoretical performance no longer
weighs up to the mentioned disadvantages. Therefore an area ratio of 20 is selected for the thrusters
developed during this thesis.
Table 3.3 - Final selection of the design parameters for the propulsion system
The calculations mentioned earlier in this chapter have been performed with these design values.
The resulting magnitude of the relevant performance parameters, for the time it takes to empty the
propellant tank, is shown in Figure 3.8.
4.5 10
Propellant storage pressure [Pa]
3.5 8
3 7
2.5 6
2 5
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
time [s] 4
x 10 time [s] 4
x 10
7 13000
6.5 12000
Electrical power consumption [W]
5.5 10000
5 9000
4.5 8000
4 7000
3.5 6000
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
time [s] 4
x 10 time [s] 4
x 10
-4 -3
x 10 x 10
5.5 1.6
5
Knudsen number in the throat [-]
1.4
4.5
Thrust force [N]
1.2
4
3.5
1
3
0.8
2.5
2 0.6
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
time [s] 4
x 10 time [s] 4
x 10
Figure 3.8 - The performance of the chosen propulsion system over time
The final total velocity increase or ΔV supplied by the propulsion system for these design choices is
21.01 [m/s]. The total operational time of the system is 64,573 [s], i.e. 17 [h] and 56 [min]. For the
requirement of 5000 on-off cycles, this leads to a thrust time of 13 [s] per cycle. In terms of energy
consumption and orbital availability that is a very acceptable duration. The pressure, electrical power
consumption, thrust force and total ΔV abides to the requirements mentioned in chapter 2. These
calculations already include relatively conservative efficiency for all components. The actual
performance of the thruster is therefore not expected to be much lower than what is seen here. The
Reynolds number indicates that, according to the work by la Torre [37], the flow through the nozzle
will be laminar. The Knudsen number is far below the value for rarefied flows. Therefore the
simplifications and assumptions, made to calculate the flow behaviour through the nozzle, hold for
These options will be discussed in the subsequent subsections. After which a preliminary selection is
conducted in section 4.2.
There are some difficulties with a piston fed system. Firstly there is a risk of leakage at the moving
seal on top of the plunger. The system complexity and risk will be increased due to a number of
moving parts and possible electronic drive circuits. An optional spring loaded system would impose
pre-pressurisation problems on the system. This storage system would therefore necessitate
elaborate design and testing activities. For this reason, this is not a preferred propellant storage
system option.
The problem with such a system is, that it costs a large amount of energy to heat the entire
propellant mass. Especially since it has to be heated to a temperature where the vapour pressure is
sufficient, for the required propulsion system pressure. Due to the relatively large size of the
propellant tank, there will be large heat losses to the surroundings. An additional problem is how to
separate the liquid and the gaseous propellant. If a liquid/gaseous mix is supplied to the thruster, the
Figure 4.2- A cross-sectional view of the SKF single point automatic lubricator, electrolysis powered lubricant feed system
[39]
Assuming a grease density of 1 [kg/l], the dry mass is around 70 [g] or 75 [g] depending on the
capacity. That fits within the DelFFi propulsion system mass budget. Due to the relatively low price
compared to normal spacecraft hardware, it is easy to acquire one lubricator to perform tests with.
When taking it into consideration for the real flight hardware, SKF has to be contacted if they are
willing to cooperate. They can then be asked if it is possible to supply the part without the grease
inside. The following important preliminary tests have to be performed:
Leak tightness and operational characteristics when used in combination with water
The possibility of interfacing with a feed system
The possibility of actuation in space (the lubricator is meant to be manually actuated and
controlled)
The behaviour under discontinuous outflow. Because the lubricator is meant to produce a
constant outflow of grease, and the thruster operates in pulses.
It did not fit within the timeframe of this thesis to contact SKF for possibilities of cooperation, or to
test the part for applicability as a propellant feed system.
Due its simplicity and reliability, the propellant tank is chosen to be a bladder containing 50 [mg] of
water. This bladder separates the water inside from the pressurant outside the bladder. For the
bladder mass estimation a rubber bulb, usually used in medical applications, is used as an example
[41]. This has a diameter of 54.5 [mm], which gives a slightly larger volume than necessary. The mass
of this bladder is 25 [g] - 35 [g] and the price is €0.74 in large volumes. A more reasonable price for
smaller numbers is probably €6.- [42]. The bladder is surrounded by a spherical stainless steel tank,
with an internal volume of 100 [ml], which contains the pressurant and the propellant filled bladder.
Figure 4.3 - A rubber bulb, normally used for medical applications, envisioned to be a possible propellant storage
bladder [42].
The optimal shape for an internally or externally pressurised vessel is a sphere. This shape will result
in the smallest wall thickness for a vessel that can sustain a certain pressure, and therefore the
lowest mass. The thickness that a spherical vessel shall have in order to contain the design pressure,
will therefore serve as a benchmark for the necessary storage system wall thickness. The stress in the
wall of a thin-walled spherical pressure vessel is [43]:
pr
4-1
2t
Where:
Stress in the wall [Pa]
p Internal pressure [Pa]
m A t 4 r 2 t 4-2
Where:
m Mass of the pressure vessel [kg]
A Surface area of the vessel [m]
Density of the vessel material [kg/m3]
Combining equation 4-1 and 4-2, the minimum mass required for a spherical pressure vessel, in order
for it not to yield under the necessary overpressure is:
2 r 3 p
m 4-3
y
Where:
y Yield stress of the vessel material [Pa]
Since the radius r is dependent on the necessary internal capacity, this value is fixed. The internal
pressure p is also already determined, in order to have the right performance of the propulsion
system. According to the theory of material selection by Ashby [43], the material index that needs to
be optimised for this problem will therefore be y . The higher the value of this material index is,
the better the given material suits the design of the propellant tank. The graphical method for finding
the best material for the propellant storage tank is shown in Figure 4.4.
Figure 4.4 - Material selection chart for the optimisation of the ratio between density and strength. The farther the
materials in the search region are from the dotted line, the better they would serve as propellant tank materials [43].
Other options are the standard high performance aerospace metals, magnesium and titanium alloys.
These are however difficult and expensive to produce. The price of production of these metals has to
be weighed up to the reduced cost, and only limited reduced performance, when using high strength
aluminium alloys. Therefore if there is a high strength aluminium, which can be manufactured to the
necessary design, that would be the preferred material.
A very important consideration in this propellant tank design, is that H2O propellant freezing must be
taken into account. Especially during storage, launch and initial operational mode of the satellite,
when the thermal control subsystem is not working completely yet. Water has a relatively high
melting point of 273.15 [K] and the ambient temperature during these phases easily drops below this
temperature. When containing the liquid propellant in the flexible bladder, this problem is mitigated.
This is because the bladder has the capacity to expand and contract when necessary.
Up to this point, it was assumed that the propellant storage vessel will be spherical. From a
manufacturing and integration point of view, this is however not the most optimal shape. Due to the
stacked squares build-up of a CubeSat, the most logical shape would be a square based prism. The
preliminary design, worked out during this thesis, features a five sided body of the tank. This body is
closed off with a flat lid, which is fastened by a number of screws. The envisioned body and lid can be
seen in Figure 4.5 and Figure 4.6 respectively. The leak tightness of this design is guaranteed by
equipping the body of the tank with a groove, along the whole interface plane with the lid. This
groove is not shown in Figure 4.5. This groove can then accommodate an o-ring. This o-ring seals the
gap between the body and lid, when both are pressed against each other by the fastening screws.
In the tank lid in Figure 4.6, an inner row of holes is visible for attachment to the tank body. Four
other holes in the corners of the lid are present to fasten the propellant tank to the internal
mounting rods of the satellite. Smaller holes are present to accommodate the electrical feed through
for the pressure sensor. There are also two threaded cylinders positioned on top of the lid. The
central one is the interface for connecting the propellant storage bladder on the one side, and the
feed system on the other. The off centre cylinder is the interface for the pressurant fill valve.
If calculations prove that this structure is not capable of containing the required pressure, it can be
adapted to feature support posts. An example of such a support post in a simple propellant tank
design is shown in Figure 4.7. This has to be compatible with the parts which separate the propellant
and pressurant. In case of a propellant storage bladder, this is more difficult. Alternatively, the
propellant tank can also be made cylindrical. A cylindrical pressure vessel is a good compromise
between rigidity under internal pressure and easy CubeSat integration.
Figure 4.7 - JPL Micro-Inspector Multifunctional Tank Design featuring Liquid and Plenum Tank [44]
No pressure supply
A mechanical pump
A pressurant gas
A spring
To be applicable to the DelFFi propulsion system, the pressurisation system must abide by the
following requirements:
1. Any chemical reaction and resulting heat shall not damage the rest of the propulsion system.
2. The pressurant shall not corrode the storage system.
3. The pressure variation due to volume change must be well characterised. This is needed in order
to know the pressure variation during the emptying of the propulsion system.
4. Possible residue of the pressurant system shall not hinder the operation of the propellant feed
system.
5. Possible electrical power consumption for initiation shall not be higher than 9 [W] , for no more
than 10 [min] per orbit. (This takes the propulsion system maximum of 10 [W] into account and
leaves some room for the most vital electronics on the PCB)
6. The mass shall not exceed 75 [g].
7. It shall be able to withstand the maximum acceptable launch loads.
8. It shall be able to operate throughout the operational temperature range of -20 [°C] to 80 [°C].
9. A possible initial shock pressure shall not damage the surrounding structure, to a point where it
will hinder its operation.
10. The pressure shall be kept between the minimum and maximum design pressures throughout
the operational temperature range.
11. The volume of the pressurant and its support structure shall be smaller than 100 [ml].
12. The pressurisation system shall work on 5 [V] or it shall include a voltage converter.
In section 4.4, these requirements will be applied to the design options, in order to yield a final
design choice.
Airbag inflators are generally quite large compared to the dimensions of a CubeSat. Cold gas curtain
inflators from Key Safety Systems for example, are 30.0 [mm] to 33.3 [mm] in diameter and
100 [mm] to 328 [mm] in length [47]. They have a mass of 500 [g] to 700 [g] and produce 1.2 [mol] to
3.6 [mol] of helium. One of the smallest airbag inflators is used for side- and knee airbag applications.
An example of such a gas generator from Key Safety Systems, has a diameter of 30 [mm] and a length
of 160 [mm] [48]. It releases 0.92 [mol] of a gas mixture of 50% argon and 50% helium and has a
mass of 260 [g]. These show that the gas generators generally produce too much gas, are too heavy
and relatively large. There are however very small generators, which are generally used only to
initiate the main stage or to actuate seatbelt pretensioners.
In order to estimate the mass generated by the gas generator, the “SPECIAL DEVICES INC” FGI (OCH)
initiator, with a total output of 260 [mg] nitrogen [49], is used. This gas generator is depicted in
Figure 4.8. The maximum diameter is 11.1 [mm] and the total length including the connection pins is
19.3 [mm]. The price of this gas generator is indicated to be $ 1.48 per generator for large quantities.
A more reasonable price of € 25.- from Dutch suppliers can be assumed for smaller quantities.
When the generator is initiated, the gas is released in the pressurant chamber with a volume of
50 [ml], which will settle to room temperature (20 [°C]). Using the ideal gas law (equation 3-1) the
system is calculated to have a final stabilised pressure of 4.5 [bar]. The energy consumption of this
exact generator is not stated. Many times larger generators have however been seen to require
around 6 [W] for up to 2 [s]. This smaller generator will therefore certainly not consume more power.
This one-time initiation power consumption easily fits within the 10 [W] budget for the propulsion
system.
Therefore it can be concluded that the automotive gas generators meet the pressurisation, size, mass
and power consumption requirements for the DelFFi propulsion system. Therefore the research can
proceed into a physical testing stage, to assess if the gas generators meet the other propulsion
system requirements as well.
The available models are mainly steering wheel airbags, which are too big and do not the
right shape for CubeSat application.
All modules are mixed up and badly searchable.
All modules are different, so there is no reliability in reproducibility.
Only entire modules are present, which have to first be taken apart. This is not desirable
without professional supervision, since they contain explosives.
Therefore it was decided not to test any of the airbag modules that are available at the scrapyard.
There are a number of companies which manufacture airbag gas generators. One of them is Autoliv,
which advertises with “micro gas generators”. These are designed to be used in seat belt
pretentioners. E-mails were sent to both the French and Dutch representatives of Autoliv. The French
representatives never responded to requests. The Dutch representative called “Van Oerle Alberton
BV” provided contact details of R,D&E Director Global Webbing, Mr. S. Valkenburg. He warned about
the many regulations the inflators impose on storage, handling and volume. Mr. Valkenburg then
referred to Senior Director - Inflator Global Development Mr. T. Bertacini. Mr. Bertacini however
A local car dealer explained that only specialised companies are allowed to handle and sell airbag
parts. This is partially due to the fact that they contain explosives, and partially because airbags are
human safety systems. One such a company is Airbagbank.eu, which specialises in airbag
manufacturing. This company was able to extensively discuss the different automotive gas generator
possibilities. The most applicable gas generator type was then provided and tested. This whole
procedure and its results are described in chapter 5. The conclusion is that the tested gas generators
cannot be used on the DelFFi satellites propulsion system. The main reason for this is the extremely
high shock pressure and the questionable reliability. It can therefore be concluded, that either pre-
pressurisation or cooperation with “CGG technologies”, is needed to pressurise the propulsion
system.
The obtained hot gas generators that have been characterised are shown in Figure 5.1. These gas
generators are simple metal casings containing a solid chemical explosive. This solid explosive is
ignited using an imbedded resistive heater element, which is connected to a DC power supply with
two electrical pins. Due to the simple nature of the heating element both pins can be connected to
either the positive or the negative supply terminal. After ignition the intentionally weakened top of
the gas generator will rupture, releasing the hot gas.
A technical drawing of the gas generator, including approximate relevant dimensions in [mm] can be
seen in Figure D.1 in Appendix D. The cross-section of a gas generator after the explosion is shown in
Figure 5.2 below.
In Figure 5.2 the following internal sections within the gas generator can clearly be distinguished:
In the automotive industry these gas generators are used for either seatbelt pre-tensioners or for
airbag initiators. For a seatbelt pre-tensioner, the hot expanding gas is used to drive a piston inside a
cylinder. This piston will drive a steel cable fastened to the seat belt; in effect tightening it around the
user. When used as an airbag initiator, the hot gas is used to activate a second stage. Either by
rupturing a tank filled with pressurised gas, or to ignite a larger solid explosive charge. Mainly the gas
produced by these second stages will fill the airbags, whereas the smaller gas generator is only used
for initiation.
Requirement Preparation GG 1 GG 2 GG 3 GG 4
1 - Combustion temperature Yes Yes Yes Yes
2 - Corrosion Yes Yes Yes Yes
3 - Volume change Yes Yes Optionally Optionally
4 - residue Yes Yes Yes Yes
5 - Power Yes Yes Yes Yes
6 - Mass Partially
7 - Launch loads
8 - Operational temperature Partially Partially partially partially
9 - Shock pressure Yes Yes Yes Yes
10 - Temperature change Partially Partially Optionally Optionally
partially partially
11 - Volume Partially
12 - Voltage Optionally Optionally
Since three out of four gas generators did not properly ignite, the information gathering in Table 5.1
could unfortunately not be realised. The details of this will be elaborated upon in the following
sections.
The final assembled configuration of the test chamber with the pressure test valve is shown in Figure
5.3. The mentioned sensors and valves are vissible in this figure as well.
Figure 5.3 - The test chamber including sensors and gas generator inlet on the left, and test valve on the right
The test chamber including the water release valve can be seen in Figure 5.4. This valve assembly
adds an additional 12 [ml] of volume to the total internal test chamber volume.
There was also an external thermometer to measure the ambient temperature of the room. This
aided calibration of the thermocouple and determination of when the system has stabilised close to
the room temperature. This stabilisation is important in order to be sure that effects which are not
important to the measurement are stabilised. Variables such as thermal expansion of the test
chamber.
To characterise the behaviour of the gas across the temperature range, a temperature controlled hot
air blower was used to elevate the temperature. The distributed hot air flow can heat the chamber as
evenly as possible. This air blower can heat the air to up to 600 [°C], so the necessary 80 [°C] internal
temperature was easily attainable.
The DAQ that will be used is the NI USB-6008, which has a sampling frequency of 10 [kHz], a
maximum voltage (drop) of 10 [V], the absolute accuracy at full scale is 14.7 [mV] at 25 [°C] and
maximally 138 [mV] across the temperature range. Using Ohm’s law, the current across the shunt
resistor, and thus the gas generator, can be calculated when the voltage difference is known.
V
I 5-1
R
Where:
I Electrical current [A]
V Electrical potential difference [V]
R Electrical resistance [Ω]
The voltage across the gas generator is the total potential difference, from the power source, minus
the voltage measured across the shunt resistor. The power consumption by the gas generator can
then be calculated using Joule’s first law. However the setup was unbalanced and the measuring
circuit took too much voltage away from the gas generator for it to be initiated. Additionally, the
other difficulties encountered in gas generator testing were more severe and required all attention.
The capturing of the voltage drop using LabVIEW also did not work properly. Therefore this
measurement was not conducted during this test campaign. If full characterisation is needed, this
test has to be conducted in the future.
The final positioning of all hardware as it was tested is schematically represented in Figure 5.6 below.
Each of these stages is described in the consecutive subsection below. As mentioned, four gas
generators have been acquired for testing. All stages should have been conducted for at least two
gas generators. If these two tests would have been successful, the gas characterisation would have
been complete enough for a first order investigation. The remaining two gas generators could have
been used to supplement the tests where something went wrong, to do additional tests, or to assess
the initial pressure after initialisation. However, as stated before, three gas generators
malfunctioned.
After the leakage tests, the test chamber was fitted with the water release valve, and filled with
500 [ml] of water. This finished setup is shown in Figure 5.7. After the whole setup was been built up
it was left untouched for a minimum of 15 [min] in order for it to attain environmental temperature
and pressure. After which all the measurement equipment were switched on and checked in a dry
run, without the gas release. This procedure determines if they are rightly calibrated to
environmental conditions and if the test volume attained room temperature again.
Figure 5.7 - The test setup just before gas generator initiation. Pressure sensor and gas generator interface on top, water
release valve on the bottom.
The gas generators are designed to be initiated by using 12 [V] of a car battery. In the DelFFi satellites
however, a 5 [V] power supply is used for the propulsion system. This supply can be converted on the
Printed Circuit Board (PCB) to deliver 12 [V]. However this conversion would cost power, volume and
After capturing the initiation phase, the system was left to rest until the internal temperature settles
close to room temperature. Only then the actual pressurant pressure can be determined. When this
settling pressure was measured, the following test stages commenced.
All sensors are situated on top of the pressure vessel, above the water. The water release valve is
positioned in the bottom of the test chamber. Gravity therefore assures that only water comes out
when opening the valve, and all the pressurant remains inside. Small quantities of water can be
released using a manual release valve. The water was collected in a measuring cup, and the release
valve was closed after the release of 50 [ml] of water every step. These water release steps are
repeated until 50 [ml] water remains in the pressure vessel. This remainder is left in, in order not to
run the risk of releasing the pressurant gas together with the water.
The cooling of the system can be achieved by using the liquid nitrogen which is available in the
manufacturing hall of the faculty. If the gas generators are selected as an option for the DelFFi
propulsion system, the operation in temperatures down to -20 [°C] will also have to be physically
tested.
The highest possible pressure is achieved when 100% of the chemical mass is transformed into
nitrogen gas. The maximum temperature of this gas is expected to be in the order of the flame
temperature of fireworks, which is 2000 [K] [51]. Since all gas is assumed to be nitrogen, the average
temperature is simply determined as shown in equation 5-2.
Where:
Tavg Average internal temperature [K]
Tgen Temperature of the gas from the gas generator [K]
mgen Mass of the gas from the gas generator [kg]
Tair Temperature of air in the test chamber [K]
mair Mass of air in the test chamber [kg]
Using the ideal gas law, shown in equation 3-1, at this average temperature of 1176 [K]the pressure
inside the chamber will reach 8.6 [bar]. However in the more likely low pressure case, an estimated
10% of the chemical mass is transformed into nitrogen gas. The maximum flame temperature of the
reaction is assumed to be a low 373.15 [K]. Using the ideal gas law, the pressure in the chamber in
this case is calculated to be 1.2 [bar]. In reality the value will be somewhere in between, but probably
closer to the lower bound. If the pressure is too low, the sensitivity of the pressure sensor will not be
high enough to get accurate data. Therefore the first test will be performed without water, in order
to make sure it will be able to withstand the maximum pressure. If the pressure during this test is
found to be very low, the amount of water in subsequent tests will be increased, in order to increase
the accuracy of the measurements.
The pressure is expected to settle at a room temperature of 298.15 [K]. The settling pressures can
again be calculated using the ideal gas law. With the estimated gas mass of the highest pressure
estimations, the settling pressure will be 2.19 [bar]. With the estimated gas mass of the lowest
pressure estimations, the settling pressure will be 1.17 [bar]. Again the real value of the settling
pressure will lie in between these two values. This is because the low pressure case is too
conservative and it is physically impossible for 100% of the explosive to be converted to nitrogen gas,
as assumed in the high pressure case.
Due to the fact that explosives were tested, safety had to be considered thoroughly. With the high
pressure and high temperature which were encountered, all components had to be able to withstand
the loads at all times. The rods have been calculated to be able to withstand the tensile force exerted
by the end plates. The walls were calculated not to yield under the internal pressure. These
necessary safety calculations are shown in the subsections below.
p b2
max
b
6
5-3
2t 2 0.623 1
a
Where:
max Maximum stress [N/m2]
p Uniform compressive surface pressure on plate [Pa]
b Minor length of rectangular plate [m]
t Plate thickness [m]
a Major length of rectangular plate [m]
Figure 5.8 - Rectangular Flat Plate , uniform load ,edge clamped [52]
When the maximum stress is taken to be the yield stress of the material, the pressure will
correspond to the maximum pressure that the vessel can hold before yielding. The yield strength for
aluminium 2024 is 324 [MPa] [53], and for aluminium 7075 is 503 [MPa] [53]. However, since the
type of aluminium is not known, a lower value of 200 [MPa] is used, to be on the safe side. Using
equation 5-3, the maximum pressure before yielding in that case is 202 [bar] of overpressure.
2 Ar
pin 5-4
Aend
Where:
pin Internal test chamber overpressure [Pa]
r Yield strength of the rod material [Pa]
Ar The cross-sectional area of the rod [m2]
Aend The internal area of the end plate [m2]
This will result in a maximum overpressure of 21.6 [bar]. Which means that there is a safety factor of
2.8, with respect to the maximum theoretical pressure that could possibly be encountered in the
tests. This calculation does not take the decrease in rod diameter due to the screw thread into
account. However, since the worst case is assumed which in reality will never occur, and there still is
a high safety factor, the rods will certainly be able to handle all realistic test conditions.
The hot-melt glue was then removed completely. As a second attempt, the thermocouple was
secured and sealed using cyanoacrylate adhesive. Due to the long drying time of larger amounts of
this adhesive, it was left for 3 days to dry. After this drying period the leakage of the test chamber
was characterised using the pressure and temperature sensors. The pressure variation observed
during the test period is shown in Figure 5.9.
3.5
Absolute internal pressure [bar]
2.5
1.5
0.5
0
0
8
15
23
30
38
45
53
60
68
75
83
90
98
143
105
113
120
128
135
150
158
165
173
180
188
195
Time [s]
It is visible that the leakage is unacceptably high for the longer duration tests. With a pressure drop
of 1 [bar] in only 2 minutes, the settling behaviour cannot be properly characterised. The volume and
temperature alteration tests take a significant amount of time. With this amount of leakage, the gas
would escape before proper characterisation could be achieved. The pressure drop due to leakage
could also interfere with the pressure drop and rise due to characterisation procedures. The leakage
itself is probably caused by air bubbles in the adhesive, due to the excessive shrinkage which was
observed while solidifying.
Following the second-glue test “Siliconen kit - bouw”, a silicon sealant, was used. To be sure of
proper solidification and optimal sealing the sealant was left to set for 2 days. The chamber was
again pressurised and left to settle to determine the leakage rate. The result of one of these tests is
shown in Figure 5.10 below.
2.5
1.5
0.5
0
0
60
20
40
80
100
120
140
160
180
200
220
240
260
280
300
321
341
361
381
401
421
441
Time [s]
The leakage rate can be seen to be slightly lower, with 1 [bar] of pressure loss in 3 minutes. However
as stated before, this is still not acceptable for the pressurisation durations encountered in this test
campaign.
Next, an interlocking Swagelok threaded fitting was used together with a byrolock. Rubber was
placed between the two parts, after which the thermocouple was lead through. When tightening the
Swagelok components the rubber effectively clamped the thermocouple. This chamber setup was
tested under pressure, the result of which are shown in Figure 5.11.
3.5
Absolute internal pressure [bar]
2.5
1.5
0.5
0
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300
Time [s]
A pressure drop of 1 [bar] in 2 minutes was observed, which is higher than the previous method with
the silicone sealant. For the before mentioned reasons, this leakage rate would again be too large for
the necessary gas generator testing. Test quickly concluded that the leakage yet again originated
from the thermocouple feed-through. After careful deliberation it has been decided to move the
In this last pressurisation test shown in Figure 5.11, the pressurisation valve was also opened in the
end, to determine if there was any drift of the sensor. As can be seen in the data, during this test,
there has been 0.00 [bar] drift, which is to say none. This was later found to be the case during all
tests.
The thermocouple was thus bonded to the outside wall of the test chamber. The spot on the wall
was chosen to be where the hot gas from the gas generator would impact the wall on the inside. The
hole previously made for the thermocouple was closed with a nut and a bolt with washers, and
rubber sealing on both sides. The test volume was then again pressurised with the sensors initiated.
The pressure and temperature in time can be seen in Figure 5.12 and Figure 5.13 respectively.
4
Absolute internal pressure [bar]
3.5
3
2.5
2
1.5
1
0.5
0
0 30 60 90 120 150 180 210 240 270 300 331 361 391
Time [s]
22
21.8
21.6
21.4
21.2
21
0 30 60 90 120 150 180 210 240 270 300 330 360 390
Time [s]
As can be seen in the temperature variation in Figure 5.13, the test chamber was actually still settling
to environmental temperature. However this 0.6 [°C] difference does not affect the measurement
significantly. At 120 [s] the pressure sufficiently stabilised at 3.676 [bar]. At 400 [s] the pressure
dropped down to 3.576 [bar]. This means a drop of 0.1 [bar] in 280 [s]. As can be seen in the
tendency of the graph, this leak rate also decreases with decreasing overpressure. For the application
of testing the gas generators, this leakage rate is acceptable, even though not desirable.
The leakage test included a submersion in water, to determine where the leakage is situated. The
leakage was thereby seen to be concentrated at the gas generator adapter. However since this is a
specialised component, which has already been isolated as much as possible with rubber sealing,
there is not much else to be done. In subsequent tests the seal was closed with even more force, in
order to minimise the gas loss.
In the future, the pressure dependent leakage rate can be determined, for example in [mg/s/Pa]. This
will help in the calculation of the influence of leakage on the readings of the tests.
0
0
30
60
90
240
570
120
150
180
210
270
300
330
360
390
420
450
480
510
540
600
630
660
690
Time [s]
45
Test chamber wall temperature [°C]
40
35
30
25
20
15
10
5
0
0
30
60
90
120
150
180
210
240
270
300
330
360
390
420
450
480
510
540
569
600
630
660
690
Time [s]
Both the shock pressure stage and the stabilisation pressure stage cannot be clearly distinguished
from these images. Therefore these stages are separated and analysed individually. Firstly the shock
pressure stage is separated from the other data in Figure 5.16.
0
53.5 53.7 53.9 54.1 54.3 54.5
Time [s]
Figure 5.16 - First gas generator test - Pressure variation - Shock phase
As can be seen from the indicated data points, the sampling frequency of the pressure sensor was
not high enough to unambiguously capture the shock pressure. A curve has been fitted to match the
data points. However in reality some of the points could be part of an initial oscillation and probably
none of the data points represents the absolute maximum shock pressure. The experiment should
therefore be executed multiple times to ensure that the behaviour is repeatable. In this first test the
maximum shock pressure was measured to be 10.413 [bar]
The sensor itself could have influenced the readout. Possibly some oscillatory behaviour occurred
within the mechanics of the sensor, due to the sudden significant increase in pressure. This could
have lead to an unrealistically high reading if the sensor mechanism overshoots, or an unrealistically
low reading if there was some internal friction. However, the reading seems to look relatively clean,
and oscillation free. Additionally, the maximum measured value is only at 26% of the sensor range.
Therefore the reading is assumed to be sufficiently reliable.
The shock temperature stage is separated from the whole temperature data range in Figure 5.17
below. Due to the placement of the temperature sensor on the outside of the test chamber, the
response time depends on the mass of the entire system. The duration of this phase is therefore
significantly larger than that of the shock pressure.
40
35
30
25
20
53 55 57 59 62 64 66 68 70 73 75 77 79
Time [s]
Figure 5.17 - First gas generator test - Temperature variation - Shock phase
Before gas generator initiation the temperature had stabilised at 23.7 [°C]. The temperature rise of
the chamber can be seen to have a maximum at 39.05 [°C]. His indicates a temperature rise of only
15.4 [°C]. This maximum temperature is very low compared to the maximum operating temperature
of e.g. aluminium at 150 [°C] to 200 [°C] or of steel at 200 [°C] to 700 [°C] [43].
1.4
1.39
Absolute internal pressure [bar]
1.38
1.37
1.36
1.35
1.34
1.33
1.32
1.31
1.3
600 620 640 660 680
Time [s]
Figure 5.18 - First gas generator test - Pressure variation - Stabilisation phase
The only jumps in the data points which can be observed is due to the accuracy of the pressure
sensor, with a resolution of 0.02 [bar]. The average pressure measured during this time is
1.344 [bar].
The temperature alteration is however partially assessed with the first gas generator test. If the
initial pressure estimation and the final settling pressure estimations are both accurate with the
findings, the estimations on gas behaviour are verified. This will be shown in subsection 5.7.1.
The absolute amount of residue was rather small, as can be seen in Figure 5.20. The residue that was
captured has a mass of only 71.3 [mg]. It seemed to mainly consist of the primary charge coating
shown in Figure 5.2.
The residue of this generator will therefore not pose a large threat to a possible propulsion system. It
will only have to be kept from entering the small fuel lines and valves. This can be achieved by
separating the pressurant from the propellant by means of a piston, membrane or bladder. Or by
filtering the propellant before exiting the tank.
After the first successful test, all three subsequent gas generator did not fully ignite. This resulted in
the expulsion of the largest part of the actual propellant grain, which is ignited to generate the gas.
The malfunctioning was possibly due to the large clamping force on the gas generators. This had to
be done slightly different to the automotive integration standard, to integrate it with the test
chamber, and to improve long duration leak tightness. An alternative explanation is that gas
generators might have come into contact with the water, which was used to dynamically alter the
test chamber volume. From a reliability point of view, this is disastrous, but it grants some insight
into the explosive that is used.
The residue collected from the three failed tests can be seen in Figure 5.21. This consists of green
multi-tubular explosive grains, similar to those seen in gun propellants. The diameter is 2 [mm] and
the lengths are 1 [mm] to 3 [mm] and they have seven circular perforations. These perforations are
common in solid propellant grains, and ensure a progressive burn rate. Which means that the area,
where the chemical reaction can take place, increases during the burn time of the grain. A schematic
representation of the propellant grain can be seen in Figure 5.22. Next to those grains, some residue
from the initial stage coating, shown in Figure 5.2, is also present.
The total mass of the residue that was captured of all three failed tests is 2,590 [mg], i.e. an average
of 863 [mg] each. However, some of the propellant already ignited to cause the bursting of the gas
generator, not all residue is actually propellant and that not all residue could be captured. Therefore
the actual explosive charge per gas generator is estimated to be at least 900 [mg].
When introduced to heat, the decomposition reaction shown in Figure 5.23 takes place.
The substances will therefore react in the molar ratios shown in equation 5-5.
Which means that the average molar mass of the gasses after the explosion, when all products are
gaseous, is:
As stated in subsection 5.6.5, there is approximately 900 [mg] of nitrocellulose in each gas generator.
Assuming 100% of the explosive charge combusts, This leads to the production of 37.04 [mmol] of
hot gas. The internal test chamber volume is 1002 [ml]. Assuming an air density of 1.225 [kg/m3], and
the molar mass of air to be 0.0287 [kg/mol], the amount of air present in this volume is
42.77 [mmol]. The molar heat capacity of all species in the gas from the generator, and in the air, is
similar [23]. Therefore the average temperature of the gas in the test chamber can be calculated
using equation 5-2. Since it is known that the shock pressure must be at least 10.413 [bar], the flame
temperature of the reaction can be determined. Using the ideal gas law shown in equation 3-1, the
gas must have a temperature of 3517 [K]. This is relatively high, since the flame temperature of
nitrocellulose is generally in the range of 2400 [K] and 3100 [K].
However, in reality, the hot gas and the present air will not mix instantaneously. The gas will actually
be a lot hotter near the gas generator, than in the rest of the testing chamber. This will locally
increase the pressure, potentially also causing a pressure wave. When the gasses mix, the
temperature is averaged, and the pressure drops, as seen in Figure 5.16. It is therefore probable that
the actual flame temperature was indeed in between 2400 [K] and 3100 [K].
After cooling down, the water will condensate, and thus the remaining molar mass of the gasses will
be:
Due to the condensation of the water, the amount of gas from the gas generator is decreased by:
As mentioned in subsection 5.6.1 the leakage rate was found to be 0.1 [bar] in 280 [s]. This leakage
rate will go down with lower pressure, and the test chamber was actually more leak tight during the
gas generator tests. But due to the lack of an overpressure dependent leakage rate the worst case
leakage during the time of the test is assumed to be:
Taking all of this into account in the ideal gas law shown in equation 3-1, the settling pressure would
be 1.394 [bar]. This is close to the actually measured 1.344 [bar]. The reason why in reality it is a little
bit lower could be that not all the explosive charge fully reacted, or that it contained other
constituents next to nitrocellulose. The leakage rate can also have been higher than previously
measured due to the high shock pressure.
Using the properties of the released gas deduced in subsection 5.7.1, the pressures can be
determined using the pressurant volume of 100 [ml]. Using the ideal gas law, the shock pressure
would then be 95 [bar]. However, the ideal gas law doesn’t hold anymore at these high pressures
and temperatures. Therefore the actual pressure peak will be lower. The shock load would again be
of only very small duration, in the order of 0.1 [s] as seen from Figure 5.16. For the settling pressure
calculation, it is assumed that the volume is built such that there will be no initial leakage. The
settling pressure at room temperature will then be 6.49 [bar].
The temperature rise of the propellant storage system due to the combustion process is thus
acceptable. However the very high estimated shock pressure would require a relatively heavy tank.
Despite of this high shock pressure the final settling pressure will still not be the required 10 [bar] for
the DelFFi propulsion system.
For a pressurant volume of 100 [ml], the shock pressure is estimated to rise to a possible
maximum of 95 [bar]. The settling pressure will then be 6.49 [bar], which is in range for the
DelFFi mission propulsion system.
The reliability was found to be low, with 3 out of 4 gas generators malfunctioning.
The flame temperature is estimated to be on the higher end of the 2400 [K] to 3100 [K]
range, which is hard to contain in the small volume.
A voltage converter might be required convert the 5 [V] satellite power supply to the 12 [V]
for which the gas generator was built.
When the shock pressure, flame temperature and reliability issues can be solved, the automotive gas
generators could prove to be effective gas generators for pressurisation purposes. For now however,
these gas generators have proven to be incompatible with the DelFFi propulsion system
requirements.
As shown in section 5.2, a number of requirements were meant to have been checked for
compliance to the DelFFi propulsion system. Some of these envisioned requirement checks were
actually not carried out due to the malfunctioning of the gas generators. In
Table 5.2 below, the requirements compliance table, shown in Table 5.1, is depicted once more. This
time it indicates which requirements were actually checked and which were not.
Requirement Preparation GG 1 GG 2 GG 3 GG 4
1 - Combustion temperature yes failed failed failed
2 - Corrosion yes partially partially partially
3 - Volume change no failed failed no
4 - residue yes yes yes yes
5 - Power partially partially partially partially
6 - Mass Partially
7 - Launch loads N.A. N.A. N.A. N.A. N.A.
8 - Operational temperature partially partially partially partially
9 - Shock pressure yes failed failed failed
10 - Temperature change Partially failed failed failed
11 - Volume Partially
12 - Voltage partially no no no
It needs to be found out why three out of four gas generators failed, as they should be very
reliable.
o Possible causes are the clamping forces, causing warping or premature damaging of
the burst section of the generators.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 97
o Another cause could be the possible contact with the water, which was used for
decreasing the test chamber internal volume.
More gas generators shall be tested using the method described above, to fully characterise
the expelled gas.
The mass of the gas generators needs to be measured accurately, before and after
detonation. This will help in estimating the amount of propellant in each generator.
Different types of gas generators could be tested.
In parallel to this, there is also a connection to connect the pressurant fill valve on top of the tank.
This filling has to be done, as close to the launch of the satellite as possible, in order to minimise the
possible leakage. Once the pressurant is filled into the storage tank, the connection can be welded
shut, after which the fill valve can be removed. The welding will minimise the risk of pressurant
leakage. And the removal of the pressurant fill valve will decrease the mass of the propulsion system
during launch. In this case, there also have to be no compromises on the fill valve selection. Any best
applicable valve can be chosen, since it will be removed afterwards.
6.2 Piping
As said before, the valve will be connected to the propellant tank interface horizontally. Ideally, the
valve inlet will be directly welded to this interface. This would then also minimise the risk of leakage.
If direct welding is not possible, a section of rigid piping will be welded between the two points. The
thruster section of the propulsion system will actually be mounted to the bottom panel of the
satellite, as shown in Figure 2.5. Therefore the valve and the thruster will be connected using a
flexible hose, which is screwed in place. This flexible hose will ease the integration, and allow for
misalignment of the propulsion system components.
For the rigid piping the industrially manufactured capillary tubes by “Salomon’s metalen” can be used
[26]. Here a whole range of orifice size and wall thicknesses can be found for RVS capillary tubing. For
the propulsion system, the size which best fits the valve is chosen. This is tubing with an outside
diameter of 2.4 [mm] and an inside diameter of 1.6 [mm]. The inside diameter therefore connects
Swagelok flexible tubing will be used to connect the valve and the thruster propellant flows.
Swagelok produces a great number of flexible hose types. The tubing must be chosen such that it will
withstand the pressure and temperature requirements. The connection has to be chosen such that it
best fits the valve which is used. Therefore the same Swagelok flexible piping which is used in the
engineering model feed system is envisioned for use in the flight model. This piping will be
elaborated upon in the master thesis documentation by Mr. Krusharev.
In Figure 6.1 below, a cross-section of a solenoid valve is shown. This example is in fact the Lee
company IEP series valve, which will be used in the DelFFi mission flight model of the propulsion
system. As can be seen, this valve closes the flow using a plunger which presses a seal to the outlet
port. This plunger is moved, to open the flow path, by running electricity though the coil.
Contamination in the propellant flow can cause improper sealing and therefore leakages. In order to
prevent contamination of the seal, a filter will be placed in front of the valve. All valves of the Lee
Company have the recommended filtration levels stated in the product charts. For the IEP series
solenoid valve filtration down to 10 [μm] is recommended [56].
Figure 6.1 - Longitudinal cross-section of the Lee company IEP series solenoid valve [56], which will be used in the DelFFi
propulsion system flight model.
The feed valve is chosen for its similarity to the valve that was used in T3μPS on the Delfi-n3Xt
satellite. This was the Lee company Extended Performance Solenoid Valve (EPSV) depicted in Figure
6.2 below. These have a piping diameter of 1.6 [mm], a power consumption of 0.5 [W], a total length,
including piping of 33 [mm] and a maximum diameter of 6.4 [mm]. The mass of the valve is under 6
[g]. Due to the fact that The Lee Company EPSV valve has already proven itself in spacecraft
propulsion in the T3μPS, from a reliability point of view it is favourable to use the same type, or one
that closely resembles this flight proven one.
100 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
Figure 6.2 - The Lee Company EPSV fluid flow control valve [57]
Figure 6.3 - Nitrogen fed MEMS resistojet by Tittu Varghese Mathew, showing the needle bonded to the thruster [31]
For the production process of the MEMS resistojet it is important that the inlet channel width is as
small as possible. This thruster and its manufacturing process will be described in more detail in
chapter 7. If the width of this channel is not equal to the depth of the heating section flow channel,
an extra etching step has to be executed. The possible depth of the inlet channel is also limited to the
maximum etching that can be done, while maintaining structural integrity of the silicon piece.
Therefore a survey of generally available needles was conducted. The results of this survey are
shown in Table F.1 in Appendix F. The smallest needles available have an outside diameter of
0.3 [mm]. Since multiple suppliers of these needles are available, the chair preferred supplier called
Farnell was chosen. The selected needle is called “WELLER - KDS3012P - DISPENSING NEEDLE, GA30,
ID 0.15MM”. It has an inner flow channel diameter of 0.15 [mm].
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 101
The selected needle has a Luer-Lock connection interface. This is an internationally used
standardised leak tight connection for pressurised needles. The connection has a male cone on one
side, and a female cone on the other side. The male cone is slid into the female cone, and fastened
into place by a screw thread on the outside of the needle. This connection is best visualised in Figure
6.4.
To connect it with the flexible hose, two connection pieces have been acquired. One connector is the
Luer-Lock connector shown in Figure 6.5. The second connector connects this Luer-Lock connection
to the standard Swagelok flexible hose connection.
102 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
7 Thruster
This chapter will deal with the most extensive and novel part of the propulsion system, the resistojet
thruster itself. During the research, a number of thruster geometries were assessed. The ones which
are perceived to be most optimal for the DelFFi propulsion system resistojet have been produced.
This production method consists of etching the necessary channels in the centre of a silicon wafer.
Part of this channel is designated as the heating chamber. Suspended in the middle of the heating
chamber is a series of silicon carbide resistive heating elements. This heating element vaporises and
heats the propellant flow, after which it is accelerated and expelled through a nozzle. This MEMS
resistojet has been given the name “Dondersteen”, which is Dutch for “thunderstone”. This name
was chosen because thunderstones were once thought to be produced by lightning, i.e. electricity,
which is what powers a resistojet. Secondly the thruster is mainly made out of silicon, which is the
main compound in most stone.
The manufacturing of the thruster was sponsored by the Delft Institute for Microsystems and
Nanoelectronics (Dimes). The design process was therefore conducted in cooperation between the
author and the Dimes representatives Dr Henk van Zeijl and Bruno Morana. The design methodology,
materials and accuracies were all chosen to fit the capacities of Dimes, while meeting the
requirements of the DelFFi propulsion system.
In section 7.1 some of the previous work done on resistojets at the TU Delft is summarised. This is
followed by the method by which the design and subsequent manufacturing were undertaken in
section 7.2. The individual sections of the thrusters are dealt with next. The inlet in section 7.3, the
heating part in section 7.4 and the nozzle in section 7.5. The extensive heat transfer calculations
were conducted, in an effort to verify the heat loss assumptions made in chapter 3. These heat
transfer calculations are shown in section 7.6. The thruster manufacturing process is then explained
in section 7.7. Finally the integration of the thruster on the electronic circuit is described in section
7.8.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 103
power consumption of 300 [W] during the tests. An additional problem was the low efficiency of the
heating process of only 36%.
Figure 7.2 - heater module vacuum tube and radiation shielding of the DUR-1.1 [60]
104 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
Figure 7.3 - Heating section of the DUR-1.2 [60]
After the application of insulation as shown in Figure 7.5, a thermal efficiency of 86% was obtained.
During this research, the theoretical approximation of the heat loss at stable conditions was proven
to be accurate. The pressure drop was proven to be substantially less significant than theoretical
predictions indicated. The major point of caution found during this research was the very low nozzle
efficiency. This was observed to be due to condensation in the nozzle. The lesson for future liquid fed
resistojets is, to make sure that the conditions in the nozzle are designed such that the expellant is
gaseous throughout the length of the nozzle.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 105
needle forms the fluidic interconnect with the propellant feed system. The propellant is inserted into
the thruster via this needle, which is bonded inside the thruster inlet using epoxy adhesive. The
propellant is then guided though the straight rectangular heating channel, where it is heated to an
elevated temperature. This heat comes from an aluminium electrical resistive heating layer, which is
deposited on top of the channel. This heat is then conducted to the flow via the silicon body and
channel walls of the thruster. Finally the propellant is accelerated and expelled via a converging
diverging nozzle. This internal channel structure is clearly depicted in Figure 7.7. Several chamber
geometries were tested, with different heating chamber geometries and different nozzle sizes.
Heating chambers with 150 [μm] and 50 [μm] channel width were tested, in both single and multiple
parallel channels concepts. All channels had a length of 2 [cm], with a uniform etching depth of
50 [μm]. Nozzle throat diameters of 5 [μm] and 10 [μm] were tested. Valuable experience was also
gained in the necessary test setup for MEMS thrusters. Calculations on the result of this research
showed a lower nozzle efficiency for lower flow Reynolds numbers in the nozzle throat. This indicates
the prominence of viscous losses in micro fluidics. The three mechanisms of heat loss were studied
for this thruster: free convection, radiation and conduction. The conduction through the structure
proved to be the most prominent of the three.
The power consumption of this device is less than 3 [W], in order to heat the propellant with a mass
flow rate of 1 [mg/s], to a temperature of 350 [°C]. The discharge factor, i.e. the ratio between
experimental and theoretical mass flow rate, of the nozzle was found to be 50% to 80%, across the
range of mass flow rates. The heating temperature in this design was limited by the low maximum
operating temperature of the aluminium heater and epoxy glue. When these materials are replaced
by materials with a larger temperature range, the maximum temperature of the flow can be
substantially increased. The propellant heating efficiency was found to be around 13%, which is very
low compared to reference resistojets. This loss was mainly due to the high conductive losses within
the silicon structure and the metallic fluidic interconnect to the feed system. There was also no
radiative heat shield in place, even though the heating element was positioned on top of the
channel. This meant that the heating element lost a lot of its heat, even before it could reach the
propellant flow. During the testing the input current was manually controlled to achieve a certain
heater temperature. From this experience it was advised to develop a PID controller to control this
current in future projects.
106 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
Figure 7.6 - Outside view of the nitrogen fed MEMS micro-resistojet [31]
Figure 7.7 - Cross-section of the flow channel of one of the nitrogen fed micro-resistojet geometries [31]
Figure 7.8 - Schematic depictions of the top view of all sections of the dondersteen thruster sections, the intended
propellant flow direction is from left to right.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 107
The numbers depicted in Figure 7.8 indicate the following sections of the thruster.
1. Propellant inlet
2. Divergent inlet section
3. Heating section
4. Nozzle convergent area
5. Nozzle throat
6. Nozzle diverging area
For testing purposes, it is desirable to vary the dimensions and geometries of some of these sections,
in order to optimise the performance. The manufacturing process only allows for a square window of
22 [mm] by 22 [mm] to fit all designs on. This is the size of the mask with which the etching process is
executed. Different sections on this masking window can be used sequentially, to build up thrusters
with the desired composition of sections. Therefore the overall design of the thruster has been
divided into three different sections. The inlet, the heating area and the nozzle. Each of these
sections, from now on called “modules” can have multiple designs. Any number of modules can then
be placed behind one another to form the necessary thruster cross-section. This is best explained
with images. Figure 7.9, Figure 7.10 and Figure 7.11 show simplified examples of what a thruster
module can look like. As shown in Figure 7.12 multiple of these modules can then be etched one
behind the other, to form the desired thruster design. The actual masking window including the
thruster modules is shown in Figure E.1 in Appendix E.
Figure 7.9 - Example thruster inlet Figure 7.10 - Example heating Figure 7.11 - Example nozzle module
module chamber module
Figure 7.12 - Example of multiple thruster modules forming a complete thruster, A combination of one inlet module, six
heating chamber modules and one nozzle module. The white boundaries in this image are only to indicate the different
modules which were used. In reality the modules are etched close enough together, that no boundaries are formed.
The different thruster modules will be discussed in the three subsequent sections of this chapter.
The previous TU Delft MEMS resistojet had an aluminium heating element. It became immediately
clear that for the new design an aluminium heater is not preferred, due to its low melting point. The
concept of a new type of heating chamber evolved from one of the heating elements that Dimes has
produced before, shown in Figure 7.13. It consists of a titanium nitride (TiN) heating element
108 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
suspended in a flow channel on top of a silicon nitride membrane. Such an element is suspended in
the centre of the flow channel. That positioning maximises the contact with the fluid flow and
minimises the heat loss from the heating element to the surroundings. The resistive element can also
be made of platinum, which a maximum operating temperature of 800 [°C]. An advantage of
platinum is that it is the least reactive metal, ensuring that oxidation will not take place due to the
water.
Figure 7.13 - A titanium nitride (TiN) heating element suspended in a flow channel on top of a low stress, thin silicon
nitride membrane
Mr. Morana recently developed silicon carbide (SiC) resistive heaters at the Dimes facilities. SiC is a
very promising material for resistive heating elements due to its inertness, high strength, relatively
low density and high maximum operating temperature. As can be seen in Figure 4.4 it is one of the
materials with the highest strength to density ratio. This means that a SiC heater would not need to
be placed on top of a membrane to be placed in the flow. The heater element itself can be the
membrane.
The necessary propellant flow channels are etched out in silicon, because this is the easiest and
cheapest production method for these kinds of structures. Due to the highly developed and
standardised method of processing silicon by etching, the accuracy is also high. The inaccuracy of this
process at Dimes is below 10% of the channel depth. Another option for the channel manufacturing
is sand blasting the shape into glass. This would be desirable because glass has a very low heat
conductivity and would minimise heat loss through the structure. This method was however
discarded by Dimes due to the high labour intensiveness and cost.
A new development, called greyscale lithography, could have been used for approximating 3-D
shapes. This would enable the production of truly conically shaped nozzles and rounded flow
channels. This method deposits the photoresist layer with different thicknesses. In the plasma
etching phase, the photoresist layers will then subside after different exposure times, thereby
attaining a differential etching depth. Due to the still very experimental phase of this technology this
was not chosen for this time-critical design and production.
The thrusters were designed to be enclosed on both sides with an anodically bonded layer of glass
for conductive insulation. A reflecting metal layer would be applied to the outside of this glass layer
for it to work as a radiation shield as well. Finally the thruster would be encased in an approximately
1 [cm] layer of epoxy bonding compound, to give some extra rigidity, insulation and protection. All
the intended layers of the thruster are indicated in Figure 7.14, from the heater in the centre to the
outside resin encasing. The industrial silicon insulator compound is readily available at Dimes and
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 109
inexpensive. This final layer of epoxy bonding compound was later reconsidered. This compound is
actually designed to conduct heat very well, and would therefore be very unsuitable to insulate the
thruster. A proper material for this encasing still needs to be found. Since the production process was
rushed the external layers are still unsure.
Figure 7.14 - A schematic representation of the intended layers for the dondersteen resistojet. From conductive heating
element at the centre of the thruster (shown on the bottom), to the resin encasing on the outside (shown on top). This
stratification is thus mirrored in the layer of the heating element.
No obstructions
Fins
A wall with small orifices
Flow channels
The more constricting the concepts are, the higher the pressure drop is. The injector will have to be
designed such that the pressure drop is minimal, while having a homogenous propellant distribution
over the heating chamber entrance.
The flow pattern in a MEMS thruster was qualitatively evaluated using the COMSOL software. The
mass flow was chosen to be 1 [mg/s], as is necessary for the thruster. In the inlet of the thruster the
water propellant is still liquid and therefore the flow velocity is 0.021 [m/s], which is relatively low.
Therefore laminar flow can be assumed. The flow behaviour was analysed for the thruster geometry
with both circular and diamond shaped fins. Since the flow velocity in the inlet and heating chamber
width are the same for all varieties, this behaviour is representative for all designs. In the Figure 7.15
and Figure 7.16 the resulting flow velocity patterns are shown for two heating chamber geometries.
110 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
Figure 7.15 - Top view of the flow velocity magnitude Figure 7.16 - Top view of the flow velocity streamlines
through the empty inlet of a diamond pillar thruster. Flow through the empty inlet area of a thruster with circular
direction from left to right. Red indicates high velocity, shaped fins. Flow direction from left to right. Each line
blue indicates a low velocity. represents the flow path of a particle starting at the initial
position on the left.
As can be seen in Figure 7.15 the colour of the flow at the entrance of the heating section, where the
pillars are, is the same from side to side. This indicates that the magnitude of the flow velocity across
the inlet is the same. In this case the flow velocity in the centre of the heater channel is no more than
9% larger than that near the sides of the heating chamber. Therefore the empty inlet section is well
capable of dividing the flow without additional structures or channels. The same conclusion can be
drawn from Figure 7.16. Here it can be seen that all the streamlines, and thus the water propellant,
are homogeneously distributed across the width of the thruster heating section entrance. Therefore
there is no need for an injector structure to distribute the water propellant flow across the width of
the heating section.
The inlet section starts with a 1 [cm] long straight section, with a square cross-section with 400 [μm]
sides, to insert the fluidic interconnect. It will then feature a divergent section with a divergent half
angle of 45°, to convert the inlet width to the 3000 [μm] of the heating section. This geometry is
shown in Figure 7.17. Since the transition from the small inlet diameter, to the larger vaporisation
section is gradual, there is no significant pressure loss due to the expansion.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 111
propellant, while minimising the pressure drop and heat loss to the surroundings. As calculated in
chapter 3, the propellant mass flow needs a certain power input in order to be vaporised and heated
to the desired temperature. When the heat transfer from the resistive heating elements to the
propellant flow is maximised, the length is minimised, aiding in a low heat loss to the surroundings.
In all design efforts of the heating section in this report a SiC resistive heating element is suspended
in the propellant flow channel. The geometry of both the flow channel and the heating elements are
the design parameters. In the following subsections the different design options for this thruster
section are first described. After which a selection is made from the most promising geometries,
based on performance estimations. These geometries are the ones which are taken in production.
Therefore the reasoning for the dimensions within the heating chamber designs will be discussed. At
the end of this section the estimated performance of the heater sections is calculated.
In these heating sections a balance must be found between the optimal heating element designs,
and the flow channel geometry. The flow channels must be designed such that the flow vaporisation
and heating is as homogeneous as possible. The heating elements need to have enough contact area
with the flow to conduct the necessary amount of heat, while maintaining structural integrity. The
heaters need to be sized such that they produce the right amount of resistance at the operating
temperature, in order to dissipate the set amount of power at the fixed potential difference of 5 [V].
When using conventional machining techniques, there are only a limited number of options for the
propellant flow channel geometry, depending on the product availability. However, since the heating
chamber will be a MEMS structure, the geometry and internal features are totally up to the designer
and only bounded by the precision constraints of the MEMS manufacturing. Infinite possibilities exist
to design the structures inside the heating chamber section. But the most logical choices can be
separated in the following main categories:
The open rectangular cross-section in Figure 7.8 can also be seen as a straight flow channel design, as
seen in Figure 7.18, but with only one channel. Therefore two main heating chamber geometry
groups can be identified. One group has flow channels, which are shown in the images below. Of
course, all kinds of differently shaped channel geometries can be thought of. However in the interest
of time and common sense, only the most basic line types where chosen for this comparison.
112 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
Figure 7.18 - Top view of a propellant heating chamber Figure 7.19 - Top view of a propellant heating chamber
design with straight flow channels. Flow direction from design with Semi-circular serpentine flow channels. Flow
bottom to top. direction from bottom to top.
Figure 7.20 - Top view of a propellant heating chamber Figure 7.21 - Top view of a propellant heating chamber
design with triangular serpentine flow channels. Flow design with sinusoidal serpentine flow channels. Flow
direction from bottom to top. direction from bottom to top.
The second group of heating chamber geometries has differently shaped, free standing pillar or fin
structures in the heating chamber. These are shown in the images below.
Figure 7.22 - Top view of a propellant heating chamber Figure 7.23 - Top view of a propellant heating chamber
design with alternating rectangular pillars in a single flow design with alternating circular pillars in a single flow
channel. Flow direction from bottom to top. channel. Flow direction from bottom to top.
Figure 7.24 - Top view of a propellant heating chamber Figure 7.25 - Top view of a propellant heating chamber
design with alternating diamond or rhombus pillars in a design with alternating aerofoil pillars in a single flow
single flow channel. Flow direction from bottom to top. channel. Flow direction from bottom to top.
The shape of the aerofoil pillars are defined using the equations for the symmetrical 4-digit NACA
aerofoil:
t x
2 3 4
x x x x
yt c 0.2969 0.1260 0.3516 0.2843 0.1015 7-1
0.2 c c c c c
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 113
Where:
yt Half thickness of the chord to surface [m]
t Maximum thickness as a fraction of the chord length [-]
c Chord length [m]
x Position along the chord from 0 to c [m]
Using this formula, the thickness at the trailing edge does not totally converge to zero. This is
problematic when using this equation in the computer modelling. Therefore the last coefficient is
changed from −0.1015 to −0.1036. This has the smallest effect on the overall geometry, while making
the thickness converge to zero at the trailing edge.
The partially liquid flow shown in Figure 7.26 would potentially severely reduce the performance of
the thruster. If the internal heat transfer within the flow is therefore not fast enough, the geometry
of the heater chamber has to be optimised to increase the contact area between the hot chamber
walls and the flow. This can be done by having winding serpentine channels, roughing up the channel
walls or placing obstructions (fins) in the flow.
114 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
In the section where the propellant is gaseous, it needs to be ensured that there is no part of the
geometry which could act as a throat nozzle. This would make the flow sonic in this section and lead
to large pressure losses. Therefore the narrowest section in the heater must be at least 3 times larger
than nozzle throat [37]. This will ensure that the flow speed will not become sonic before the throat.
For the sinusoidal serpentine heating channels shown in Figure 7.21, this might become an issue, due
to the large decrease in width in the narrow sections of this flow. The greatly fluctuating channel
width will also result in cycles of large flow acceleration and deceleration. This will always be
accompanied by losses in the flow. Due to the complexity of this channel, it was unfortunately not
possible to model the flow properties through it, during the scope of this thesis. Therefore the
sinusoidal serpentine channel design is also not further evaluated.
In order to evaluate the differences in flow properties in the triangular and the semi-circle serpentine
design, the flow through these sections was modelled using the COMSOL software. The result of
which is shown in the two figures below. The mass flow rates through these sections were chosen to
be in the range of 1 [mg/s] across all heating channels combined. As was shown in chapter 3, that is
the maximum expected mass flow rate for the Dondersteen thruster. The working fluid was chosen
to be liquid water at room temperature.
Figure 7.27 - Flow velocity magnitude through a semi- Figure 7.28 - Flow velocity magnitude through a triangular
circular serpentine channel. Isometric view. Red indicates serpentine channel. Top view. Red indicates high velocity,
high velocity, blue indicates low velocity. Flow direction blue indicates low velocity. Flow direction from bottom to
from bottom to top. top.
As can be seen from the images above, the flow pattern through both channels is quite different. The
Semi-circular serpentine channel in Figure 7.27 shows a very smooth velocity profile, with relatively
small boundary layers on the sides. The triangular serpentine channel in Figure 7.28 shows a far more
irregular flow pattern. The flow can be seen to stagnate in the corners of the channels, increasing
the velocity in the centre of that section. It can also be seen that there is a far larger boundary layer
of stagnated flow on the inside triangle of the flow. This is seen by the fact that the blue layer, of low
velocity flow, is thicker on one side of the channel than on the other. Therefore the losses in the flow
of the semi-circular design are assumed to be lower, and the heat transfer more optimal. The semi-
circular channel design is therefore the flow channel design which will be produced for the thruster
tests. To further verify this assumption, this can also be compared to the flow around aerodynamic
bodies. Where the semi-circular serpentines resemble the flow around a cylinder, and the triangle
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 115
shape resembles the flow around an angled cube or rhombus. For the same Reynolds number of the
flow, the drag coefficient of the flow around a rhombus is about 1.5 or 2 times as high as that around
a cylinder. Therefore the angular channel design is expected to induce more drag.
As a first analysis of the heating module designs with the pillar elements in the flow channel, the
necessarily for alternating pillars was assessed. In previous research, designs have also included non-
alternating pillar designs. The flow velocity patterns through both alternating and non-alternating
designs are assessed using the COMSOL software. The result of which is shown in the images below.
In Figure 7.29 the flow can clearly be seen to stagnate behind the pillars. The flow velocity magnitude
also shows that the flow velocity is very irregular. The flow constantly decelerates where there are
no pillars, and accelerates in between pillars. As can be seen from the streamlines, virtually no
particles flow in the space behind the pillars, resulting in a lot of useless space. Some of the
streamlines are also virtually straight from beginning to end. This means that flow stratification and
incomplete vaporisation can occur. In Figure 7.30 some flow stagnation can be seen in front of and
behind the pillars. However the velocity magnitude profile between the pillars is very uniform. This
means that there are no phased flow accelerations and decelerations. In the streamline figure it is
also visible that the channel is a lot more optimally used. Additionally, none of the streamlines passes
through in a straight line, which avoids stratification.
The conclusion from this is that alternating pillars are required for a more stable flow pattern, a more
efficient use of the flow channel and a more secure vaporisation. In the following figures the flow
pattern investigation of circular and diamond shaped pillars is depicted.
116 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
Figure 7.31 - Propellant flow patterns, though a heating Figure 7.32 - Propellant flow patterns, though a heating
section with alternating circular fins. On top, the velocity section with alternating diamond (rhombus) shaped fins.
magnitude where blue is low velocity flow and white is On top, the velocity magnitude where black is low velocity
high velocity flow. On bottom velocity streamlines along flow and white is high velocity flow. On bottom velocity
the flow. Flow direction from left to right. Top side view. streamlines along the flow. Flow direction from left to
right. Top side view.
As can be seen from Figure 7.31, the flow pattern around the circular pillars is quite irregular. There
is a lot of flow stagnation before and behind the pillars. And in between the pillars there is flow
acceleration. During the simulations a rhombus shaped flow stagnation region was identified around
the circular pillars. In Figure 7.32 a very constant velocity flow pattern can be seen around the
diamond shaped pillars. There is a slight flow deceleration in between pillar rows. As can be seen
from the streamline figure, the particle flow is almost uniformly divided across the entire channel
space. Comparatively very little flow stagnation is encountered in front of or behind the pillars and
the observed boundary layers are a lot smaller than for the other geometries. This also gives
reasoning behind why the diamond shaped pillars were found to perform best in micro evaporator
structures previously investigated at Dimes [63].
Therefore it can be concluded that the diamond shape pillars create the best flow pattern. This
results in small boundary layers, minimal flow stagnation and homogeneous flow velocities.
Therefore these pillar designs were chosen for the production of the physical test models of the
thruster.
Pheat I 2 R 7-2
Where:
Pheat
Dissipated heat [W]
I Electrical current through the resistive element [A]
R Electrical resistance of the material [Ω]
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 117
The electrical resistance in its turn is determined by resistor geometry and resistivity [63]:
L
R0 7-3
w h
Where:
R0 The electrical resistance at the reference temperature [Ω]
Resistivity of the resistive material [Ω·m]
L Length of the resistor [m]
w Width of the resistor [m]
h Height of the resistor [m]
Since the resistivity of a material depends heavily on temperature, the temperature dependent
resistance is calculated with the following formula [63]:
R T R0 1 T 7-4
Where:
The Temperature Coefficient of Resistance (TCR) [/K]
T The temperature rise compared to the reference temperature [K]
Unfortunately a simple mistake has been made in this part of the calculations, which has profound
consequences for the resulting thrusters. The value of the resistivity of the SiC was wrongly
converted from the paper by Mr. Morana. This led to a wrong design of the resistive elements, which
were now designed to yield 100 times the required electrical resistance. SiC, has a resistivity of
ρ = 1.27·10-4 [Ω·m] (the calculations were made with a value of ρ = 1.27·10-6 [Ω·m]) when doped
with 15 [sccm] NH3 [64]. The resistivity decreases to 0.38·10-4 [Ω·m] (the calculations were made
with a value of ρ = 0.38·10-6 Ω·m]) when doped with 55 [sccm] NH3. The calculations which will
follow that depend on the resistivity will therefore not have the correct values. The TCR of SiC is α = -
0.83·10-3 [/K] when doped with 15 [sccm] NH3 and α=-0.5·10-3 [/K] when doped with 55 [sccm] NH3.
When connected in series, the total resistance is simply the sum of the resistances of the individual
elements, as shown in equation 7-5.
Rtot R1 R2 Rn 7-5
When the resistive elements are placed in parallel, the reciprocal of the total resistance will be equal
to the sum of the reciprocals of the individual elements, as shown in equation 7-6.
1 1 1 1
7-6
Rtot R1 R2 Rn
Therefore this simplifies to equation 7-7, when all n individual elements have identical resistance.
R1
Rtot 7-7
n
118 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
7.4.4 Design and lay-out of the resistive heating element
For the prototype model all heating elements will span the entire with of the heating chamber. The
suspended heating elements are laid on top of the pillar, in order to structurally support them across
the heating chamber. Therefore the pillars or flow channel walls have a dual purpose. On the one
hand they divide and mix the flow, and on the other, they support the heating elements. A
longitudinal cross-section of such a heating chamber module is shown in Figure 7.33. All the
important elements can be identified. The flow channel which is filled with the diamond shaped
pillars. The heating layer on top has three different sections:
1. On the sides there are the bond surfaces, where the electrical connections are bonded to the
power supply.
2. In between the pillars are the suspended heating elements, which dissipate most of the heat,
and conduct it to the passing flow.
3. On top of the pillars is a wider continuation of the heating layer. This will dissipate some heat
to the pillars, which in turn will heat the passing flow.
Figure 7.33 - Longitudinal cross-section of one heating module with diamond shaped pillars. Including the suspended
heating elements and the bond surfaces on the side. The indicated gaps show the physical separation between different
heating elements.
In discussions with the manufacturers at DIMES a heater membrane thickness of 100 [nm] to
300 [nm] was said to be achievable. This was confirmed when making the heater geometries, which
was designed for a SiC layer with 200 [nm] thickness. During the week of production this thickness
range indication was suddenly changed to 300 [nm] to 500 [nm]. Calculations were quickly conducted
to investigate how much the resistivity should be changed to accommodate for this new thickness,
while maintaining the same geometry.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 119
Due to the modular design it is easy to put multiple heating modules behind one another. Each
heating module is therefore designed to produce the practical amount of 1 [W] of heating power.
The heating elements will be connected to an external power source using bond pads. Based on their
actual performance, multiple heating elements will be placed in parallel or in series in a multitude of
different configurations. Since a new set of thrusters will be manufactured for the flight model this
can incorporate these inter-element connections on the silicon wafer itself. Instead of having seven
separate connections going to each heater module, this will only require one electrical connection to
the control system. This will greatly reduce the time required for manufacturing. Due to the fact that
fewer wires need to be bonded, and the configuration is already implemented in one production
step. To increase the reliability the single electrical bond pad can still be using several bond wires. If
one bond wire breaks, the others will still be able to supply the necessary electricity to the element.
The advantage of these heating elements is that they can not only be used for heating, but also as a
thermistor. As told before, the resistivity of materials depends on the temperature in the relationship
shown in equation 7-4. When the TCR of the material is experimentally determined, the temperature
of a material can be derived from the electrical resistance. This characteristic is exactly what will be
used in the thrusters to assess the temperature of the heaters. During the testing campaign this will
enable the use of the unpowered heating elements as temperature sensors, in order to measure the
flow temperature. For the final flight model of the propulsion system, these measurements will be
used to keep the heating elements at the desired 500 [°C].
Figure 7.34 - MEMS heater with diamond shaped fins [63]. k is the pillar length, m is the minimum distance between
pillars, h is the maximum distance between pillars, l is the width of the pillars.
This heating chamber configuration consists of a rectangular chamber, with diamond (rhombus)
shaped fins, of which the largest diameter is placed in parallel with the flow direction. One could
already look at Figure 7.39 and Figure 7.40 for a reference. The heater layer spans between these
pillars in direction perpendicular to the flow direction. The layer occupies the centre of the pillars and
120 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
is suspended between these pillars in a rectangular flat shape. Since there are multiple rows of fins
placed behind one-another, the flow then also passes through the rows of these heating filaments.
To calculate the resistance across one of these resistive heating element rows, the geometry is
simplified. The most basic repetitive shape is a combination of a rectangular suspended section and a
trapezoidal section on the pillars, as shown in Figure 7.35.
The rectangular section is comprised of half of the suspended heating element, and the corner of the
diamond shaped pillar. The trapezoidal section is the remaining part of half of the diamond element.
Therefore:
1 w w
Lr Lsus s fin 7-8
2 2 Lfin
Where:
Lr
Length of the rectangular section [m]
Lsus Length of the suspended element (distance between two pillars) [m]
ws
Width of the suspended element [m]
w fin
Width of the fin with respect to the flow [m]
Lfin
Length of the fin with respect to the flow [m]
1 w w
Lt w fin s fin 7-9
2 2 Lfin
Where:
Lt
Length of the trapezoidal section [m]
For the rectangular section, equation 7-3 is used to calculate the resistance of the element at the
reference temperature. For the trapezoidal section, the change in width has to be taken into
account. The width of the trapezoidal section from left to right, as seen in Figure 7.35, can be
calculated using simple trigonometric rules.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 121
L ws
wt x fin x ws 7-10
Lt
Where:
wt x Width of the trapezoidal section as a function of the distance of the narrow section
of the trapezoid [m]
x Distance from the narrowest section of the trapezoid [m]
Taking into account the variable width of the trapezoidal section equation 7-3 can be rewritten as
follows:
x
Lt Lt Lt
x x
R0 dx dx dx
wt x h h 0 wt x h 0 L ws 7-11
0
ws fin x
Lt
The integral can be solved using the following standard integral solving method [65].
u du
a b u du b a b u a ln a b u C
1
2
7-12
Lfin w s L ws
w s w s ln w s
1
R0trap w s Lt w s ln w s fin Lt
h Lfin w s 2
Lt Lt
Lt 7-13
R0trap
1
h Lfin w s 2
Lfin w s ln Lfin w s w s ln w s
Lt
The total resistance of electrical resistors in series is the sum of the resistances of the individual
resistors. Therefore the resistance of the section shown in Figure 7.35 is the sum of the resistances of
the straight part and the trapezoidal part. The resistance of the heater per fin is then two times this
amount. And the resistances per row of fins, is the amount of fins, times the resistance of the heater
of each fin. i.e.:
R0row 2 R0straight R0trap norows 7-14
Where:
R0row Resistance of the resistive heater for each row of fins [Ω]
R0straight Resistance of the straight part in Figure 7.35 [Ω]
R0trap Resistance of the trapezoidal part in Figure 7.35 [Ω]
122 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
The resistance at operational temperature is calculated using equation 7-4. This actual resistance,
together with the known system voltage of 5 [V] is used to calculate the power consumption of the
resistor using Joule’s law in equation 7-2 and Ohms law.
The results of the calculations were checked by running simulations on the same geometries with the
COMSOL program. For these diamond shaped resistors, the resistance was found to be 1.65 times
higher in the simulations ran on COMSOL than what the simplified calculations showed. When
looking at the current density through the diamond elements it was immediately clear why this is the
case. In the simplified calculations, it was assumed that the current density would be homogeneously
distributed across the width of the diamond heaters. In the simulations however it showed that there
was virtually no flow in the tip of the elements, and most of the current went through the centre of
the diamonds following the line of the suspended heating elements. The conversion coefficient of
1.65 was therefore used to convert the simplified calculations into the, more accurate, simulated
results. This conversion coefficient is only valid when the ratio between the pillar length, pillar width,
heating element width and heating element length is in the region as which is used in the design
shown here. To make the calculations more accurate, the final flight model design will have to be
simulated with COMSOL in order to get the most exact results possible.
Heating and subsequent heat losses of the resistive heater layer electrical connection bonding area
are neglected. It is assumed that since the width of this section is significantly larger than the heating
elements in the flow, the losses in the sides are negligible. This section will also be kept as short as
possible, in order to further minimise losses.
The pillar dimensions could not be theoretically optimised within the time schedule of this thesis.
Therefore the dimensions are taken from the previously tested micro-evaporators, by Marko
Mihailović [63]. These dimensions were derived within the much larger timeframe of a PhD thesis,
and were shown to be effective. The dimensions used for this design are shown in Table 7.1, and are
clarified in Figure 7.34. The documented performance is also shown in this table, some of which is
clarified in Figure 7.36.
Table 7.1 - Characteristic dimensions and performance of a water fed micro-evaporator [63], the meanings of the
abbreviations are given in Figure 7.34.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 123
Figure 7.36 - performance of a diamond fin micro-evaporator [63]
The maximum vaporised mass flow rate through this reference evaporator is in the same range as
the required mass flow for the MEMS resistojet. The exit temperature is a lot lower than required for
the Dondersteen thruster. In the MEMS resistojet, the flow will therefore be heated in gas phase for
a longer duration of time. Gaseous propellant with the same mass flow rate as the liquid propellant
will need more volume. The maximum input power will also be around 1.8 to 2.3 times higher.
For these reasons it is likely to assume that the heating volume and heater contact area also have to
be at least twice that of the micro-evaporator. In order to keep the design as comparable as possible,
and for reasons of nozzle performance, the channel height is kept at the same 100 [μm] as for the
micro-evaporator. Propellant “spitting” can occur when liquid water is propelled forward by boiling
flow behind it. If the liquid is propelled with such a velocity that it cannot be vaporised anymore
within the subsequent flow path, this could flow through the nozzle as a liquid. This drastically
reduces the efficiency of a resistojet. In order to prevent this, the propelled liquid, needs to be forced
to make contact with the heater area. Therefore the width of the heating section is kept the same as
for the micro-evaporator. Only the heating chamber length, of the nominal design, is increased by at
least a factor two. The heating section will therefore have a length of at least 8 [mm]. This ensures
that the propellant flow path is long enough to vaporise and heat the propellant completely and
relatively uniformly.
Since the length of each fin is 160 [μm], and the total length has to be 1.00 [cm], there are 57 Rows
of fins in the nominal design. The heating chamber designs are built up out of sections consuming
1 [W] of electrical power. Since in the nominal design, the power consumption is 7 [W], each section
will have 8 rows. This also complies with the fact that each section should have an even number of
rows, since the pillars have to be alternating. This will make the total length of the heater section
8,960 [μm] long. The width of the heating section is chosen to be 3,000 [μm], for convenience, since
that means that there are 50 fins per row.
Using the above mentioned equations, the power consumption in this section can be calculated as a
function of the suspended heating element width. This relation is shown for the rows connected in
parallel in Figure 7.37 and for the rows connected in series in Figure 7.38.
124 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
1.4 1.4
1.2 1.2
Total power consumption [W]
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 0.2 0.4 0.6 0.8 1 0 0.5 1 1.5
Suspended heating element width [m] -5
x 10 Suspended heating element width [m] -7
x 10
Figure 7.37 - Modular section power consumption - Rows Figure 7.38 - Modular section power consumption - Rows
connected in parallel connected in series
As can be seen, the width of the suspended element in parallel connection has to be around 8 [μm],
to achieve a total power consumption of 1 [W]. Since the distance between the fin tips is 20 [μm] the
aspect ratio of the main span is 2.5. This aspect ratio is assumed to be adequate for a structurally
sound design. This has also been discussed with Mr. Morana, who is the most experienced expert on
SiC heating elements available at Dimes. The width of the suspended element in series connection
however has to be around 0.11 [μm], to have a total power consumption of 1 [W]. This means an
aspect ratio of the main span of 182. This aspect is too high for a structurally sound design, which can
withstand the necessary loads. Additionally it will most probably pose a problem for manufacturing
as well. Therefore all rows in one heater section will be connected in parallel, which is also easier for
connecting and testing purposes. In addition to that, from a redundancy point of view parallel
connections are a lot better. Since if one fails, the others can still produce power. When the rows of
heating elements would be connected in series, a failure at any point in the chain of heating
elements will render the whole heating module useless.
In order to have enough space to bound the power supply cables to on the bound pads, the wall
thickness is increased to 2000 [μm]. This also makes sure that the side walls will be able to cope with
the required chamber pressures. The final nominal design of one heating section with diamond
shaped fins is shown in Figure 7.39 and Figure 7.40.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 125
Figure 7.39 - Heating channel etching indication with diamond shaped fins - Nominal design - 1 [W] power consumption
Figure 7.40 - Heating channel heating layer with diamond shaped fins - Nominal design - 1 [W] power consumption
In order to compare the effects of pillar size on the performance of the heater module, a second
diamond pillars heating chamber module was designed. To give a proper base of comparison the
diamond pillars were increased in size by a factor four. This should be significant enough to show
some consistent difference with the nominal design. The important dimensions and performance of
the two diamond pillar designs are shown in Table 7.2.
126 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
Table 7.2 - Most important design and performance parameters of the heater modules
The most important difference between this heating chamber design and the diamond pillar one is
the flow patterns it will induce. These flow patterns will determine the pressure drop, heat transfer
rate within the flow and the heat losses to the surroundings. The pressure drop and heat losses need
to be as small as possible, for an as large as possible heat transfer rate. To be able to objectively
assess the performance differences between both designs, the total module size and the cross-
sectional channel area are chosen to be as similar as possible.
The volume fraction of the channel in the heating section of the diamond shape is can be calculated
easily. The area of the diamond shaped fins is:
1
Afin w fin Lfin 7-15
2
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 127
The heating chamber area per fin is:
The volume fraction of the channel in the semi-circular design is relatively easily defined for a section
with the length of one wave. This is shown in Figure 7.42 below.
Figure 7.42 - A section with 1 wave across the entire width of the heating chamber
From the geometrical properties, it can be seen that the volume fraction of the channel in such a
section is:
rout
2
rin2
channel 7-17
4 rout rout rin
Where:
channel Volume fraction of the channels [-]
rout Outer radius of the semi-circle [m]
rin Inner radius of the semi-circle [m]
The volume fraction of the channel in the diamond shaped fin heating chamber design can be
calculated to be 67%. For simplicity, the element length for the semi-circular sections is taken to be
equal to the diamond fin length of 160 [μm]. If the outer radius is then chosen, the length of the
inner radius follows from equation 7-17. In Figure 7.43 the resulting volume fraction of the channel is
shown for a number of outer radii.
128 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
90
80
Channel volume fraction ϕ [%]
70
60
50
40
30
20
10
0
4.00E-05 5.00E-05 6.00E-05 7.00E-05 8.00E-05
Outer radius rout [m]
Figure 7.43 - Channel volume fractions for a range of semi-circle outer radii, assuming 160 [μm] total wave length
A channel volume fraction of 67% can be seen to be achieved at an outer radius of 70 [μm] and a
resulting inner radius of 10 [μm]. This gives a total width per wave of 140 [μm]. For a total heater
width of 3000 [μm], there will be 21 channels next to one-another. And to make the semi-circle
design as comparable with the diamond fin design as possible, the length of the heating section
module will also be chosen to be 1280 [μm]. This means each channel will make a total of 8 waves,
per heater module.
For simplicity, the heating elements in this design were chosen to be straight rectangular beams from
one side of the heating chamber to the other. They are supported throughout the heating chamber
by the walls of the serpentine channels. As said before, due to the fact that the electrical potential
difference is set to be 5 [V], the resistance of each modules must be 25 [Ω] in order to produce 1 [W]
of electricity. This information is used in equations 7-3 and 7-4 in order to calculate the necessary
heating element width. This heating element design is shown in Figure 7.44. The finally
manufactured geometry is shown in Figure 7.45. The calculations of the resistance of these heating
elements were also checked by simulating the current in COMSOL. In this case the simulations and
the simplified equations matched 100%.
Figure 7.44 - Top side view of the designed heating elements for the semi-circular serpentine heating chamber geometry.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 129
Figure 7.45 - Optical microscope image of the top view of one module of the nominal semi-circular serpentine channel
design. Including the electrical connection pads to the sides and the suspended heating elements. Flow direction is from
bottom to top.
Also for this geometry a roughly four times large design was made. This was done in order to test the
influence of the size of the serpentines on the thruster performance. The smaller serpentines are
expected to have a larger heat transfer to the propellant flow, due to their larger contact surface.
However this increased contact area is also expected to induce a larger pressure loss. The geometry
for this larger serpentine design is shown in Figure 7.46.
Figure 7.46 - Optical microscope image of the top view of one module of the large semi-circular serpentine channel
design. Including the electrical connection pads to the sides and the suspended heating elements. Flow direction is from
bottom to top.
The relevant design parameters for both semi-circle serpentine heating modules are shown in Table
8.1. In this table a comparison is made between the designed dimensions of the heating modules and
the actual dimensions after production.
7.4.7 Electromigration
Small scale electronics in MEMS such as in this heater run the risk of failure due to electromigration,
when operating at high current density [66]. Electromigration is the physical movement of
conductive material, on an atomic-scale, due to the electrical current. The collisions between the
electrons and atoms in the Ohmic conduction regime, results in a momentum exchange between the
electrons and atoms. For accurate calculations on the impact of this phenomenon on the lifetime of
electronic parts, both extensive analytical and numerical calculations are used. However, they quickly
go a lot more into depth than is required for this design. The goal is to keep far away from the
current densities where electromigration becomes an issue. Therefore the traditional black’s
equation is usually used to get an order of magnitude estimate on the Mean Time To Failure (MTTF):
130 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
w t kaT
E
MTTF A e 7-18
j2
Where:
A Constant depending on technology features
w Width of the conductor [m]
t Thickness of the conductor [m]
j Current density [A/m2]
Ea Activation energy [eV]
k Boltzmann’s constant, 1.3806488·10-23 [J/K]
T Absolute temperature [K]
The necessary property is however not very well defined or known for the doped SiC which is used in
the heating layers of the Dondersteen thruster. Thus to get an indication, some reference values of
maximum allowable current densities need to be used. The maximum current density calculated for
the designs that were discussed in the previous subsections is 2·106 [A/cm2]. As a reference,
aluminium can withstand a current density of up to 105 [A/cm2] and copper can withstand
106 [A/cm2]. This is over a period in the order of 10 years. Therefore the much harder SiC should not
have problems for a relatively short duration of time at 2·106 [A/cm2], during less than one month of
operation.
The pressure drop across the section will be calculated using the Ergun pressure drop correlation
[62].
1 1 U2 L
2
Where:
P Pressure drop [Pa]
Ratio between the gas volume over the total volume (packed bed porosity) [-]
Dp Particle diameter [m]
Fluid viscosity [N·s/m2]
Us Superficial or free stream fluid velocity [m/s]
L Packed bed length [m]
Fluid density [kg/m3]
All these parameters speak for themselves; except for the superficial fluid velocity vs. This is the
velocity which the stream would have if the packed bed volume would not contain the pebbles. For
the liquid phase this free stream velocity is easily calculated.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 131
m
Us 7-20
A
Where:
m Propellant mass flow rate [kg/s]
A Empty channel cross-sectional area [m2]
The mass flow rate is assumed to be 1 [mg/s], as calculated in chapter 3. The porosity of the diamond
shaped heating geometry is 0.67 [-] as shown in subsection 7.4.6. The particle diameter is assumed to
be the width of the pillar, which is 40 [μm]. The viscosity of liquid water at 100 [°C] is 0.282·10-3
[N·s/m2]. The length of the chamber is 8.96 [mm]. The density of liquid water at 100 [°C] is
958.4 [kg/m3]. This all results in a pressure drop of for a flow through of liquid water only. The
pressure drop is then calculated to be 127,569 [Pa] = 1.28 [bar]. This connects relatively well with the
assumption made of 20% of the pressure, which is 0.9 [bar]. Due to the much smoother and
regulated flow patterns, the proposed resistojet is however expected to witness less of a pressure
drop. The same can be done for a full flow of gaseous water. If the results are then averaged, it will
come even closer to the true value. Due to time constraints no further elaboration was attempted.
As stated before, all the flow channels besides the nozzle throat should have a diameter of at least
three times the throat diameter. With a diameter larger than this value, no premature shockwaves or
nozzle behaviour will be formed [37]. Therefore it was decided that the divergent angle would be
chosen to be a steep 45°, in order to minimise the length. At a diameter of three times the throat,
the converging angle will then change to 15°, in an effort to have the maximum nozzle efficiency. As
mentioned in chapter 3 the length of the nozzle can be reduced to about 75% of the conical length,
while having the same area ratio and thrust coefficient CF, using a bell shaped nozzle [22].
The bell shape is a simple type of parabola of which the function is derived by the theory explained
by ir. Zandbergen [22]. The diverging angle behind the throat is a standard 60° [22] and the exit angle
is chosen to be the mentioned optimal for MEMS nozzles of 15° [37]. This reduced length will further
reduce the possible heat loss through the nozzle walls. This nozzle is therefore designed for optimum
performance in a slightly experimental shape. A comparison between the diverging section of a
conical nozzle and a bell nozzle, with the same exit angle and throat diameter, was made. This
comparison is shown in Figure 7.47, clearly showing the smaller length of the bell shaped nozzle. The
final shape of this optimised nozzle is shown in Figure 7.48 and the design parameters are summated
in Table 8.1.
132 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
Figure 7.47 - Divergent nozzle sections showing a Figure 7.48 - Shape of the flow channel of the final design
comparison in length for conical and bell shaped nozzles, of the optimised bell nozzle design
with the same throat diameter and exit angle.
The primary design will however be a more conventional conical shape. By choosing a converging
half-angle of 30° and a diverging half-angle of 20°, this nozzle should still be able to achieve a nozzle
efficiency of up to 91.1% [37]. The disadvantage however is the longer length, with might lead to
more losses in the flow. The final design of this nozzle is shown in Figure 7.49 and the design
parameters are summated in Table 8.1.
Figure 7.49 - Shape of the flow channel of the final design of the sharp edge conical nozzle design
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 133
All these attempts could not be conducted successfully within the timeframe of this thesis. The
numerical calculations attempted with the aid of literature required too many assumptions. These
assumptions could not be made sensibly at the time the calculations were conducted. The
calculations as far as they were conducted are documented in Appendix H. Another attempt was
made by using the COMSOL software. This software was actually successfully used to make some
qualitative liquid flow trade-offs and to verify the resistive heating element calculations, both in this
chapter. The attempts to simulate the heating of the flow were halted as they were too time-
consuming. In order to produce the thrusters in time a more qualitative approach was taken and
results of earlier research served as a reference.
7.7 Manufacturing
In order to assess the influence that the different design parameters have on the performance a
number of different module configurations were used. How the desired combinations of different
thruster modules form the test thrusters, is best shown in Table E.1 in Appendix E. The reasoning
behind the selection is also shown in this table. The manufacturing had to be rushed, due to the
imminent maintenance closure of the production facilities. This led to the fact that the desired
combinations of thrusters were not all produced. Some of the combinations were devised by the
manufacturers on the spot. The main distinction between the designs is the length. One selection of
thrusters is designed to produce 7 [W] of heating power. These are the ones that are designed to the
specifications of the DelFFi mission propulsion system. The other thrusters are designed to study the
influence of length on the performance of the thrusters. These have a series of 14 heater modules as
a heating chamber, capable of producing 14 [W] of heat. These will however not all be used to heat
the propellant flow. This length and number of heating elements does facilitate the possibility of
conducting elaborate tests. Hereby some of the elements can be used as temperature sensors and
others as heaters. In this way the flow behaviour and heat loss can be well characterised.
The first step of the production process is to deposit a layer of SiC on a silicon wafer. Due to the fact
that the silicon wafer is highly conductive an oxide layer is placed in between the silicon substrate
and the SiC heater layer for electrical isolation. This layer will later be manipulated such that it
becomes the suspended heating element in the flow channel. This layer was deposited by the SiC
expert at Dimes, Mr. Morana. For the production of the flow channels, the masking process
described in section 7.2 is used. Each thruster module has a specific place in the masking window.
Each desired thruster module is then etched sequentially to build up the desired thruster design. The
unique method in this fabrication is that the heating membrane is not produced separately. The SiC is
already deposited on the silicon wafer, and the heating elements are etched free in the same etching
steps in which the flow channels are produced.
The etching process itself is done in three stages. The first stage of the etching process is to etch
away the part of the SiC layer which is not needed for the heating. This includes the separation slits
between the electrical contact pads of the heating modules, shown in Figure 7.33. Also the straight
heating elements which are on top of the semi-circular flow channel walls, shown in Figure 7.44,
need to be etched in this step.
The second stage of etching is called Deep reactive-ion etching (DRIE). This is a highly anisotropic
etching process, which means it only etches straight down from the surface. An example pattern that
134 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
is used is shown in Figure 7.50 below. The sections that need not be etched are covered with photo
resist material. The wafer is then bombarded with ions. Under this bombardment the unprotected
areas are excavated, and the areas protected by the photoresist are unharmed. After this etching
step the resistive heating elements are still on top of a silicon ridge, so they are not suspended in the
flow yet.
Figure 7.50 - Pattern which was used for the nominal diamond shape design. In black is the section that is etched, in
white is the section that is covered by photoresist.
The third step of the etching process is to use isotropic etching to achieve under-etching below the
heating elements, in order to make them suspended in the flow. This process excavates the silicon
walls which were still supporting the heating elements. The under etching below a heating element
can be seen in Figure 7.51. In this image, the inherent problem of this method can clearly be seen.
The fact that a ridge of silicon remains where the under-etching fronts from both sides of the heating
element meet.
Figure 7.51 - SEM image to clearly illustrate the under-etching process below the heating elements, in order to suspend
them in the flow channel. This is an image of the nominal semi-circular flow channel design.
This is where the manufacturing process ended during the timeline of this master thesis. The
manufacturers at Dimes were unfortunately not able to finish the thrusters before their facilities
closed for major overhaul. The following steps indicate what will be done in the future, in order to
finalise the Dondersteen thrusters.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 135
After the channels have been etched, and the heating elements are suspended in above the flow
channel, the thrusters can be closed. Parallel to the above mentioned process, one wafer is produced
by etching with a mirror image of only the flow channels. This wafer is then placed on top of the
other wafer, with the flow channels facing each other. These wafers are then anodically bonded to fix
them together and seal the gap between them. The thruster is then a closed design with the heating
elements suspended in the middle of the flow.
Since the thrusters are then closed, the SiC heating layer is in the centre of the silicon wafer
sandwich. The electrical connections have to be made on the bond areas shown in Figure 7.33. In
order to connect this heating layer to the power supply, a hole is etched on top of the silicon wafer.
This etching reaches from the top of the silicon wafer down to the SiC heating layer in the centre. The
holes have a square area with sides of 1 [mm]. Since the electrical bond wires cannot be bonded on
SiC directly, a thin aluminium layer is first deposited on top of the exposed SiC. The bond wires can
later be bonded to this aluminium layer by a special process, conducted at Dimes, which fuses the
aluminium to the bond wire.
The final step in the manufacturing process there is 100 [μm] loss by cutting the thrusters apart,
when diamond saws are used. The outset is however cut all thrusters out of the wafers by etching on
the back of the wafer. This could potentially cut a narrower line. This etching is done in the same
etching step as the opening of the electrical connection areas.
The manufacturing could unfortunately not be fully finished by Mr. van Zeijl before the closure of the
Dimes production facilities. Therefore it was agreed to finalise the production of one of the wafers,
to such a stage that the geometrical and electrical characterisation could already be completed. This
enables vital preliminary investigations on the deviation between the design and the practice of this
novel manufacturing method. This test wafer is shown in Figure 7.52.
Figure 7.52 - Test wafer filled with the Dondersteen micro-thruster designs, used for geometrical and electrical
characterisation.
136 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
Due to the necessity of the under-etching, the heater elements cannot be wider than 9 [μm].
Otherwise the etching process will not be able to remove enough material from underneath the
heating elements. Heating elements that were calculated to be wider than 9 [μm] have been divided
into smaller elements of 9 [μm] wide.
In conversations with Dimes, the idea came up to do the fluidic thruster tests on the halves of the
thrusters that were now actually available. To this end a glass layer would be bonded to the open
side of the wafer to close the thruster structures. A number of steps would have had to be taken to
make this possible. First of all, the protective oxide layer would have to be removed from the SiC, at a
different etching facility on the campus. After which a specialised company would have to be ordered
to produce a glass covering, with a recess for the heating elements. This is because the heating
elements stick out from the silicon wafer slightly. So if a glass layer would be placed on top it would
only touch the heating elements. This glass plate with recesses would then have to be accurately
bonded to the silicon wafer, without the adhesive clogging any part of the flow channel. The
thrusters would have had to be separated by using a diamond saw. This gives the risk of material
chippings clogging the channels. A special inlet cap would have to be produced to serve as a fluidic
interconnect, in order to feed the propellant flow to the thruster. This is because there is no needle
that would fit the too small inlet opening. Then the SiC electrical connection pads, which would then
be sandwiched between the glass layer and the silicon wafer, need to be made accessible. This could
possibly be by adding holes through the glass plate which is bonded on top.
This idea was however discarded; because it would take a few months to get this all commissioned
and build at external companies. Next to that there are all kinds of uncertainties if it would have
worked. After which the actual tests would still have had to commence. By the time all these
arrangements could have been made, the Dimes manufacturing facilities would be up and running
again. At Dimes the rest of the production would only take another day. Therefore this temporary
solution was seen as not worth the trouble, cost and manpower.
It was first understood that he PCB would be designed, built and integrated at Dimes. However at a
later stage it was said that this was not the case. The Dimes representatives said that the university’s
electronic and mechanical support division, Dienst Elektronische en Mechanische Ontwikkeling
(DEMO), will be able to do it. To this end, the representative Mr. B. Schelen was contacted. He
explained that DEMO does not produce PCBs anymore. They are however available and willing to
cooperate in the outsourcing of the production to the company Eurocircuits.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 137
Mr. Schelen also recommended that if very high PCB temperature resistance is required, micro
Electronics bv. is able to produce ceramic PCBs. The material which is normally used for PCBs will
start delaminating and disintegrating at temperatures in excess of 180 [°C]. He did mention that this
only happens after hours of continuous high temperature operation. Experiences with the previous
MEMS resistojet has shown that temperatures up to 400 [°C] did not significantly harm the PCB
material [31].
The PCB was designed to be as small and functional as possible. Mr. N. Dos Santos has been a great
help in designing the eventual PCB layout shown in Figure 7.53. A three dimensional rendering of the
thruster mounted on the PCB is shown in Figure 7.54.
Figure 7.54 - Three dimensional representation of the thruster mounted on the PCB
The PCB is designed to have copper lined, circular holes in the area on which the PCB is mounted.
There are in place to make sure that the PCB will not be able to absorb all of the heat coming from
the thruster. The holes will reduce the heat loss by conductivity through the PCB material. The
copper lining will make sure that the heat which is absorbed by the PCB is quickly transported away.
This will reduce the risk of heat damage to the PCB material. Careful though has gone in to the
placement of the golden bond pads, in order to make sure they can be used by both the 7 and the 14
module long thrusters. The PCB has four screw holes in the corners in order to fasten it to the test
rig.
138 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
Because many different thrusters have to be tested, the connection between the PCB and the test
setup has been made easily interchangeable. The easiest and cheapest way to connect it is using
headers. Because some boards only require 7 connections on each side and others require 14, it
would be easiest to use the 7 pin single row headers. And put on one on each side of the board for
most setups and 2 headers on each side for the 14 connection thrusters. A small trade-off between
optional connector pins was conducted to find the most appropriate one. The best candidates in this
trade-off are shown in Table F.2 in Appendix F. The pins that were eventually selected are also shown
in Figure 7.54. The “TE CONNECTIVITY / AMP - 215307-7 - SOCKET, VERTICAL, 2ROW”, is the female
connector which is linked to the electrical system. This simplifies the testing of multiple thrusters
sequentially. The female connector is fixed to the electrical test system and each PCB has its own
male headers to connect to. When swapping different thrusters they can be disconnected and
reconnected simply by clicking the connection in and out.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 139
8 Thruster testing
The physical testing of the MEMS micro-thruster Dondersteen is conducted to validate the
theoretically predicted performance. This verification consists of both static and dynamic testing. The
static testing is conducted to assess how much the final product differs from the theoretical design.
In the dynamic tests, the thrusters are operated in several different methods, to determine the
performance of the mission relevant parameters. This chapter is only meant to give a rough overview
of these tests. The actual test plan was written by the designated test engineering Quirino Bellini in
cooperation with the author and the other propulsion system engineer Mr. Krusharev [67]. The
reader is referred to this document for further information.
The characteristics of the SiC heating elements need to be known. This in order to better understand
the effect of geometry on the electrical conduction properties, and to assess the specific properties
of the SiC material used. To this end, the temperature dependent electrical resistance needs to be
determined.
The performance parameters, for the thruster in operation, are required to assess the differences
between the theoretical and the true performance. The performance parameters which need to be
known are:
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 141
Properties of the separate heating elements
o Electrical input power
o Temperature
All these parameters need to be known for different propellant pressure, and electrical input powers.
After the propulsion tests have been executed, the thrusters have to be inspected for
malfunctioning. The PCB, thruster exterior and the heating membrane have to be inspected for
defects or alterations.
142 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
the DASML and can magnify up to 100 times, which is sufficient to inspect for cracks, molten or burnt
sections and other malfunctions. The digital cameras attached to these microscopes will be used to
produce images depicting the thruster state after testing.
They shall have a maximum operating pressure of more than 7 [bar], preferably even 10 [bar]
( The valve for the engineering model has a maximum operating pressure of 7 [bar],
therefore the engineering model tests will not go higher than 7 [bar]. However, if the sensors
used for the engineering model have a maximum operating pressure of 10 [bar] they can be
reused in the flight model tests)
They shall be vacuum compatible.
They shall be compact (maximally about 10 [cm] by 10 [cm] in order for it to fit inside the
vacuum chamber)
They shall have an operational temperature range of 20 [°C] to 80 [°C] (The actual
temperature of the propulsion system lies between -20 [°C] to 80 [°C], but the vacuum
chamber can only heat, and not cool, so temperatures below room temperature are
physically impossible to test in the present facilities)
The Omega engineering inc. PX181-200G5B has thereafter been integrated into the feed system of
the test setup of the thruster engineering models. If these at any point appear to be damaged or
otherwise inadequate the following sensor can be acquired from Farnell:
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 143
Especially due to its low mass and correct supply voltage, this is also a pressure sensor which can be
included in the flight model of the propulsion system. The necessary thermocouples for temperature
measurements are already available in the vacuum oven which will be used for the tests. If these
appear to be inadequate, one of the following sensors can be acquired from Farnell:
Table 8.1 - Comparison between the design values of the thruster geometry and the actually produced values as checked
with the optical microscopes.
144 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
Module Parameter Design Actual value Unit Deviation [%]
value
module Pillar diamond width 160 143 [μm] -10.63%
diamond Heater diamond length 640 578 [μm] -9.69%
shapes large Pillar diamond length 640 508 [μm] -20.63%
Wall roughness 0 0.60 [μm] n/a
Suspended heating element 90 99.7 [μm] 10.78%
average length
Suspended heating element 38 32.6 [μm] -14.21%
width (total)
Total chamber width 3000 3021 [μm] 0.70%
Heating layer separation 10 11.4 [μm] 14.00%
Heating Wall roughness 0 0.56 [μm] n/a
module Channel width 60 57.5 [μm] -4.17%
Serpentines Inner channel curvature 20 23.4 [μm] 17.00%
nominal diameter
Outer channel curvature 140 139 [μm] -0.71%
diameter
Heating element total width 29 22.0 [μm] -24.14%
Number of channels 21 21 [-] 0.00%
Heating layer separation 10 11.0 [μm] 10.00%
Heating Wall roughness 0 0.33 [μm] n/a
module Channel width 232 226 [μm] -2.59%
Serpentines Inner channel curvature 88 77.14 [μm] -12.34%
Large diameter
Outer channel curvature 552 557 [μm] 0.91%
diameter
Heating element total width 77 80.8 [μm] 4.94%
Number of channels 5 5 [-] 0.00%
Heating layer separation 10 10.2 [μm] 2.00%
Nozzle sharp Convergent angle 30 30.2 [°] 0.67%
edge Throat diameter 25 34.4 [μm] 37.60%
Divergent angle 20 20.5 [°] 2.50%
Exit diameter 500 499 [μm] -0.20%
Nozzle bell Convergent angle 1 45 45.3 [°] 0.67%
shape Angle change diameter 75 90.7 [μm] 20.93%
Convergent angle 2 15 18.3 [°] 22.00%
Throat diameter 25 30.2 [μm] 20.80%
Divergent angle 60 59.8 [°] -0.33%
Exit angle 15 13 [°] -13.33%
Exit diameter 500 498 [μm] -0.40%
Nozzle Ivan Convergent angle 45 45.1 [°] 0.22%
Krusharev Throat diameter 30 43.7 [μm] 45.67%
Divergent angle 30 29.9 [°] -0.33%
Exit diameter 800 795 [μm] -0.63%
In Table 8.1 all the design parameters have been mentioned and checked for completeness. Some of
them, such as the geometrical angles, are exactly correct. Many of them are however far removed
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 145
from their design value. The most important and most deviating values will be discussed. As
mentioned in section 7-3, due to the rushing of the manufacturing process, the inlet could not be
designed to specifications. The width and length of the small diamond shaped pillars deviate quite
significantly from the design value. This is due to the under-etching of the isotropic etching process.
This effect is actually very beneficial as it still supports the heating elements, but there is less
material, so less conductivity from the pillars to the silicon structure of the thruster. The wall
roughness can be seen to be well below 1[μm] for all different designs. This is exactly as specified by
Dimes before production and is acceptable since it remains below 4% deviation of the most critical
dimension, which is the nozzle throat. The suspended heating element length can be seen to be
larger due to the under-etching of the pillars. The width of the suspended elements for most of the
designs is off by a relevant amount. This is also one of the causes for the resistance of the heating
elements, which is higher than it was designed to be.
Finally the most important deviation, the nozzle throats. Depending on the magnitude of the
converging and diverging angle ratios, the nozzle throat diameter significantly deviates from the
design values. This has a major impact on the performance of the thruster. A larger nozzle throat
area leads to a larger mass flow, which in turn requires a higher electrical power input in order to
vaporise it. This widening of the nozzle throat is an artefact of the experimental production process.
The results from this design will help in fine-tuning the production in such a way that the nozzle
throats can be made more accurate for the final flight models of the thruster.
Subsequent to these tests, the Scanning Electron Microscope (SEM) was used, in order to assess the
profile of the channel walls and floor at an angle. As can be seen in the following images, side walls of
the flow channels are still relatively straight. The corners of the flow channels are rounded off. And at
the points where there are heating elements or small flow passages, substantial ridge formation is
observed.
Figure 8.1 - SEM image of the nominal thruster Figure 8.2 - SEM image of the nominal diamond shaped
design throat area - Straight walls, chamfered heater element design - Under-etching of the diamond
elements, chamfered corners and ridge forming can be
corners and ridge forming can be observed in the observed in the flow channel.
flow channel.
146 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
Figure 8.3 - SEM image of the inlet - Straight walls and Figure 8.4 - SEM image of the nominal serpentine shaped
chamfered corners can be observed in the flow channel. heater chamber design - Under-etching of the heating
elements, chamfered corners and ridge forming can be
observed in the flow channel.
The measurements on the channel depth were assessed using an optical microscope. An optical
microscope can only focus on one plane perpendicular to the viewing direction. Anything that is
closer to or further away from the lens is blurred. The method of channel depth measurement is
therefore very simple. The microscope is focussed on layer A of material. The platform on which the
sample rests is then moved upwards or downwards until layer B comes in to focus. The height with
which the platform had to be moved to shift the focus from layer A to layer B is then equal to the
vertical distance between the two layers.
This is illustrated in the following figures. In Figure 8.5 the microscope is focussed on the top surface
of the thruster wafer. The platform is then moved up by 70 [μm] to put the ridge in the throat into
focus, as shown in Figure 8.6. Finally the platform is moved up by another 20.5 [μm] to focus on the
flow channel bottom, shown in Figure 8.7. This means that this flow channel has a total depth of
90.5 [μm].
Figure 8.5 - Bell nozzle throat depth Figure 8.6 - Bell nozzle throat depth Figure 8.7 - Bell nozzle throat depth
measurement - Top surface measurement - Flow channel ridge measurement - Flow channel bottom
During these tests, the platform movement could be read out with an accuracy of 0.5 [μm] and the
focussing was seen to be accurate to about 2 [μm]. The measurement results on the channel depth
and ridge height are displayed in Table 8.2. All the channel depths were designed to be 50 [μm] deep
on one side. It can clearly be seen that this is not the case
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 147
Table 8.2 - Measurement results of the channel depth in specific areas, multiple versions were measured where multiple
dimensions are listed. Channel depth and ridge height from the same channel are listed in the same column.
The characterisation tests have so far only included the electrical conductivity of the heating
elements. This has been proven to be far removed from the design values. In order to test the
resistance, miniature electrical probes were placed on the sides of the SiC heating elements.
Different electrical potential differences where then applied between the two heater terminals. The
result of these tests then forms a current-voltage characteristic, or I-V curve for the tested heating
element. From this characteristic, the resistance is found using Ohm’s law. The results of these
characterisation tests are shown in Table 8.3. As shown in this table, for each heating element type,
at least four different tests were conducted, in order to show the divergence in the measurements.
Table 8.3 - The summary of the cold electrical resistance tests on the different SiC heating elements of the Dondersteen
thruster. All resistances have been measured on at least 4 different modules, and most of the values are an average of
multiple measurements on the same module
148 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
Section Parameter Design value Actual value Unit
Average 25 12,200 [Ohm]
Diamond shaped heaters Electrical resistance 1 25 12,500 [Ohm]
large design Electrical resistance 2 25 17,000 [Ohm]
Electrical resistance 3 25 7,500 [Ohm]
Electrical resistance 4 25 14,500 [Ohm]
Average 25 12,900 [Ohm]
Serpentine heating Electrical resistance 1 25 4,038 [Ohm]
channels nominal Electrical resistance 2 25 3,500 [Ohm]
Electrical resistance 3 25 4,500 [Ohm]
Electrical resistance 4 25 5,500 [Ohm]
Average 25 4,400 [Ohm]
Serpentine heating Electrical resistance 1 25 7,000 [Ohm]
channels large Electrical resistance 2 25 12,000 [Ohm]
Electrical resistance 3 25 13,500 [Ohm]
Electrical resistance 4 25 15,000 [Ohm]
Average 25 11,900 [Ohm]
As can be seen from the measured values in Table 8.3, the measured values of the electrical
resistance differ from the design values with 2 to 3 orders of magnitude. This is such a large
deviation, that it completely alters the behaviour. As can be seen from the combination of Joule’s
and Ohm’s law in equation 5-1, when the resistance increases, the voltage also has to increase in
order to yield the same power dissipation. As can be seen from this formula, an increase in
resistance, will lead to a necessary increase in voltage of the square root of the magnitude increase
of the resistance, in order to get the same output power. Therefore the 100 to 1000 times increase in
the resistance, necessitates a 10 to 32 times increase in voltage, in order to achieve the same power
dissipation. For the case of these heating elements, this means that the input voltage will have to be
increase to 50 [V] to 160 [V] depending on the type of heating element. For the thruster tests, this
might be possible, but for the flight model this is highly undesirable. As to increase the voltage to
such a high level will induce a lot of electrical losses in the necessary boost- or step-up converter.
The hot characterisation procedure in order to find the TCR of the SiC heaters can also be performed
on the test structures. However since there are no electrical connections bonded to the heating
elements yet, this will also have to be done using probes. To this end the thermal characterisation
chamber, i.e. oven, has to be equipped with miniature electrical probes on the inside. This is not
present yet, and it is too time-consuming to still implement it during the course of this thesis.
However due to the very high divergence of the resistance that was measured, it is very desirable to
also measure this TCR, before proceeding with the production process. If this value also differs as
much from the design value, this will have to be taken into consideration in future design iterations.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 149
One significant problem is that the size of the throat differs substantially from the design value. This
would lead to a larger critical mass flow, which in turn leads to a larger power consumption. Because
the power consumption is already designed to be near the maximum set by the requirements it is
not acceptable for it to become even higher. The manufacturing crew has been notified of this and
will fine-tune the production process to yield more accurate results in the future. Due to the process
of first anisotropic etching and subsequent isotropic etching, it was difficult to find the balance. The
masks had to be intentionally made smaller, in order to cover for the sideways etching of the
isentropic etching. Apparently however the mask was either not reduced in size enough, or the wafer
was exposed to the isotropic etching process for too long. It could also be a combination of both.
During the production process, the masks will therefore be further reduced in size, or the isotropic
etching process is reduced in duration.
This increase in nozzle throat area could have been slightly diminished by the formation of a ridge in
the nozzle throat, which reduces the area. This ridge relative to the bottom surface of the flow
channel is due to isotropic etching process, which has less effect on narrower sections. Unfortunately
the channel has also been etched too deep, which immediately nullifies the area shrinking effect of
the ridge. In the end the nozzle throat is both larger in height and larger in width/diameter for all
nozzle designs. This significantly increases the critical mass flow rate, and therefore the necessary
electrical power input. For the nominal nozzle design, the nozzle throat channel depth is increased by
a factor 1.43 and the diameter is increased by a factor 1.37. This increases the nozzle throat area by a
factor 1.96. As shown in Figure 3.5, both the thrust force and the power consumption scale linearly
with nozzle throat area. Both values will therefore be almost doubled. For the thrust, that is not a
problem, as it still falls well within the range specified in the requirements. The power consumption
was already designed to be close to the maximum allowed value. Therefore the resulting power
consumption will be almost twice the maximum allowed value.
The largest problem that was immediately detected when doing the electrical resistance
measurements is that the resistance is a factor 200 to 600 off of the design value. Most of the
discrepancy was quickly traced back to a design error. The resistivity that was used for the design is a
factor 100 lower than that of SiC in reality. This value was used due to a mistake in copying the
results from earlier research, whereby the decimal point shifted two places. For the engineering
model this is not such a large problem. This will require 63 [V] to 113 [V], depending on the type of
heating module, to produce the same amount of heating power. Power supplies with these voltages
are readily available at the university.
The flight models however do have to operate using a voltage of 5 [V]. The heating elements
therefore need to be redesigned. From the decision making process it can be concluded that the
following parameters can be changed:
The resistivity of the SiC can be reduced by a factor 4 by doping it with more NH3
The thickness of the silicon carbide layer can be increased by a factor 2
The length of the heating elements can be decreased by a factor 2
The width of the heating elements can be increased by a factor 3
The length of the heating module can be increased by a factor 3
These changes alone would already decrease the resistance of the elements by a factor 144. In
addition to this the throat area can still be reduced slightly, which will reduce the critical mass flow,
150 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
and thus the power consumption. This is however difficult because it has already proven to be
difficult to manufacture the small size of the nozzle as it was. These suggested changes will result in a
total length of the thruster of 34.6 [mm] and it will reduce the total width of the thruster to
5.5 [mm]. This still fits within the geometrical constraints of the DelFFi propulsion system.
A more drastic change would be to change or enhance the heating elements using platinum. This is
also highly temperature resistant and inert, but has a far lower resistivity.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 151
9 Future work to finalise the DelFFi propulsion system
This chapter will describe the steps that have to be taken, from the current development stage of the
propulsion system development, to the final flight model. The steps that still need to be taken can be
subdivided into different categories, which are manifested as sections of this chapter. The main
function of this chapter is to serve as an initial, straightforward roadmap for the succeeding student
on this research. Therefore it will state where the necessary material can be found and who needs to
be asked.
The thrusters will then have to be integrated with the already manufactured PCBs. These PCBs are
stored by Mr. Cervone and await integration. The electrical connections called headers will first have
to be soldered to the PCBs. These headers have been ordered and received; they are stored by Mr.
Dos Santos. After the headers are soldered the PCBs can be given to Dimes. They will bond the
thrusters to the specified area on the PCB. Dimes has specialised staff which will connect the
electrical connection pads on the thruster to the golden bond pads on the PCBs using bond wires.
Finally the needles will have to be bonded to the thrusters, in order to provide the necessary fluidic
connection. These needles have been ordered and received from Farnell and they are stored by Mr.
Cervone. The needles will be bonded using the same method and epoxy adhesive used by Mr.
Mathew [31]. This is best done using the microscope available in the cleanroom. The needle is then
inserted slightly into the inlet cavity. The adhesive is applied to outside of the needle and around the
inlet cavity after which the needle is fully inserted into the inlet cavity. After the hardening of the
adhesive, the thruster assembly is finished and ready for testing.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 153
The feed system for both Mr. Krusharev’s thruster and the Dondersteen thrusters is the same. It ends
in a flexible Swagelok tube with a 062 MINSTAC interface, which needs to be connected to thruster.
To that end a connection piece was acquired that connects the Luer Lock interface of the thruster
inlet needle, to the Swagelok 062 MINSTAC tube interface. This connection piece is stored by Mr.
Cervone. The Dondersteen engineering model thruster assembly can be connected to the thrust
stand using the specially made connection plate, which is stored by Mr. Cervone. Four screws in the
attachment holes of the PCB are used to connect the assembly to this plate.
The main work that still needs to be done before the testing can commence is the connection of the
electrical system. The connections have to be fed through the vacuum chamber walls with the
already present com port connections. This feed-through is firstly used to supply the electrical power
to the heating elements. Secondly they are used to measure resistance of the heating elements in
order to determine their temperature. Preferably a heating chamber power supply regulator will be
developed. The power supply can then be cut off or current regulated when the temperature of the
heater reaches a certain percentage above the target temperature. The power can then be switched
on again when the heater reaches a temperature, a certain percentage below the target
temperature.
It preferably needs to be evaluated what the minimum impulse bit of the thruster is. When the
propellant valve is opened, the propellant will fill the entire space in between the valve and the
exhaust nozzle. After the valve is closed, this section is still an open system filled with propellant. It
will however not have the backpressure of the propellant pressurisation system anymore. The
behaviour of this phase has to be properly characterised. It could results in the expulsion of relatively
large amounts of propellant, without producing the necessary amount of thrust. A propulsion system
redesign might have to follow. This could be as simple as taking a certain percentage of extra fuel
into account.
154 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
geometrical design restrictions of the DelFFi propulsion system. If both heater sections have a
different optimal heating chamber module, the heating chamber will be manufactured with both
types of heater modules.
Due to the novelty of the production method, the resulting geometries were found to deviate from
the design values. The acquired data on the magnitude of deviations is being used at Dimes to fine-
tune the manufacturing process. Especially the nozzle throat section will be investigated, so the
diameter can better resemble the design value. It will also be attempted to further minimise the
ridges in some of the smaller flow channels. Finally the channel depth will be reduced, in order to
more closely resemble the intended 50 [μm].
The most important redesign that has to be done is to rectify the mistakes made on using the wrong
resistivity. In addition to this it was found that the resistance values were even more off than this
calculation error alone would have caused. This means that the resistance of the heating elements
has to be reduced by a factor 200 to 600, depending on the module type. The suggestions on how
this can be done are given on subsection 8.4.3. It is strongly advised to use software do a more
accurate analysis of the resistance behaviour of the resistive elements. COMSOL can readily be used
for this purpose. Easy tutorials are available on electricity flow simulations, it is relatively easy to do
and it is not extremely computationally intensive.
Possibly some other discrepancies will be found during the performance tests, which might also
necessitate a certain redesign of one or more of the thruster modules.
It needs to be investigated which pressurant gas can best be used. At this moment, nitrogen is
assumed to be used, which has the advantages that it is cheap inert. There might however be other
gasses which are also suitable, but have larger molecule. The larger the molecules are, the less
problem there is with leakage. The largest anticipated problem for the complete propulsion system is
leakage. If the leakage of a pre-pressurised system is found to be too high, collaboration options with
CGG technologies need to be investigated. Chief Technical Officer of the company, ir. Berry Sanders
has been found to be interested in such collaboration.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 155
During and after the operation of the propulsion system in orbit, the data has to be analysed to see if
the propulsion system worked as expected. This will lead to the final conclusions on the functionality
of the system.
156 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
10 Conclusions
The aim of this master thesis research was: “Design a propulsion system to meet the propulsive
requirements of the two DelFFi mission formation flying satellites”. In order to design this propulsion
system a number of tests were conducted on the candidate components and assemblies. After a
survey of all commonly mentioned reaction engines, a resistojet was found to best meet the needs of
this mission. This was mainly selected because resistojets are relatively uncomplicated, do not
require combustion and have relatively low power consumption per amount of impulse generated.
These advantages were even more applicable to a cold gas thruster system. A cold gas thruster based
system was however shown to be incapable to meet all requirements and restrictions of the DelFFi
mission propulsion system. The best propellant for resistojets, selected from a large number of
possibilities, was found to be water. Water was chosen because it is liquid around room
temperature, it has a high potential impulse bit per unit of mass, and it is benign, cheap and readily
available.
Based on the requirements on the propulsion system the following necessary design parameters
were found: a water propellant mass of 50 [g], a nitrogen pressurant mass of 0.2 [g], a storage
pressure of 4.5 [bar] to 2.5 [bar], a nozzle throat diameter of 25 [μm], a nozzle area ratio of 20 [-] and
a final propellant heating temperature of 500 [°C]. The resulting propulsion system performance is: a
total ΔV of 21.01 [m/s], a total thrusting time of 17 [h] and 56 [min], a power consumption of 6.8 [W]
to 3.7 [W] and a thrust force of 1.4 [mN] to 0.8 [mN].
A resistojet propulsion system requires pressure to transport the propellant from the storage tank to
and through the thruster. Because pre-pressurisation is generally not allowed in CubeSats, gas
generators were found to be the best option. These gas generators can be electrically initialised in
orbit, where they produce gas from a solid. Gas generators used for airbag initiation in the
automotive industry were found to be unsuitable for this mission because of the high shock pressure
of possibly up to 95 [bar]. Additionally, three out of four gas generators that were tested
malfunctioned. The reason for this malfunctioning is still unknown. For the DelFFi mission a waiver
was granted, now allowing pressurisation before launch. Therefore pre-pressurisation is now
considered to be the best option for propellant pressurisation.
The most innovative part of this thesis is the development of the MEMS micro-thrusters called
Dondersteen. These consist of an integrated fluidic inlet channel, heating chamber and rocket nozzle,
which are all etched in silicon. The propellant is heated using heating elements made out of the very
strong material silicon carbide. These elements are suspended in the middle of the fluid flow in order
to maximise the heat transfer to the propellant, and minimise the heat loss to the surroundings. The
developed thruster consists of the following three distinct sections: inlet, heating chamber and
nozzle. Multiple designs were made for each of these sections. The production process was chosen
such that any combination of these sections can be etched behind one another. This creates the
flexibility of being able to produce and test multiple thruster layouts, in order to find the one which
best fits the requirements. Different heating chamber geometries were designed and tested. These
will enable the assessment of the geometrical influence on heat loss, pressure drop and propellant
heating capabilities. Due to the experimental nature of the manufacturing process, some of the
geometrical parameters were not exactly made to specifications. Namely the channel dept was
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 157
found to be too large and the nozzle throat diameters were too large. The performed analysis of
these discrepancies will be used to improve the manufacturing process.
The tested resistance values of the resistive heater modules were found to be 200 to 600 times
larger than designed. This increases the required input voltage from 5 [V] to the range from 70 [V] to
120 [V]. For the current testing phase this is not a problem since these supplies are available. The
flight models will however have to be redesigned in order to comply with the 5 [V] requirement of
the satellite power supply. The main reason of this discrepancy was found to be a calculation error.
Some adjustments have already been listed with which this resistance can be decreased by a factor
of 144. Further design efforts will have to increase this factor slightly to obtain the real design value
of the resistance.
There are some tasks that still have to be performed by a succeeding master student to consolidate
the work done in this master thesis to a propulsion system flight model. The developed thrusters
have to be performance tested in multiple operating conditions. The results of these tests lead the
final redesign of the developed thrusters. The propellant storage system needs to be built and
tested. Finally the complete in flight propulsion systems can be defined, built and integrated into the
satellites.
Therefore it can be said that the aim of the master thesis has been fulfilled to a large extent. A design
has been made of the propulsion system by weighing the different component candidates against
each other based on test results. Some components however still require testing and redesign, after
which the final flight version design can be produced.
158 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
11 Recommendations
This chapter will include some recommendations on different aspects of a master thesis. Section 11.1
lists a few research topics that could advance the performance of the Dondersteen thrusters. In
section 11.2 some of mistakes I have made are formulated as requirements for future students, so
they are warned for making the same mistakes. Finally in section 11.3 some recommendations are
made to the university, to better facilitate the conduction of master thesis research.
For interfaculty cooperation, members of both sides need to co-operate on the design. In this
research it was a too one-sided approach. Of the author working mainly on the design and Dimes
staff mainly on production. A design produced more in a direct cooperation between both fields
would make for a better balance between optimal manufacturing, and necessary design lay-out.
The MEMS thruster could be combined with other MEMS developments. The most mentionable is to
integrate a MEMS valve into the same component. This would reduce integration problems and
further minimise the size of the propulsion system. The reason that this was not developed for the
Dondersteen thruster is that the technology is too immature to handle the necessary pressures for
long durations of time without leakage.
First completely finish documenting the previous phase, before going to the next. Otherwise there
will never be concrete numbers and figures of what went wrong/right. And it is difficult to show what
you’ve actually spent your time on. Do not only focus on project deadlines. An extensively
documented master thesis research document is required. If you only follow project deadlines which
often have to do with manufacturing and testing, it is very easy to do more work than befits a master
thesis. And you will then also face the problem of documenting all of these steps, and cramming it
into the page limit.
Do not be tempted to calculate all the necessary behaviour yourself. Some of the research takes
years to perform correctly. It is sometimes better to get a right source for than to try and approach
everything with crude and idealised methods. In this case I spent too much time on heat conduction
calculations and trying to model the behaviour in the heating chamber. Such CFD analysis would be a
separate master thesis already. The values from experiences at Dimes should have been used earlier
instead.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 159
11.3 Recommendations for thesis supervision
Databases of available materials and test equipments within the whole university should be built and
maintained. It is the author’s belief that all necessary testing equipment is generally available at the
university, but there is no knowledge of what is where. A database with specialities of people within
the entire university would help in the same way. That way if there are any questions on a specific
topic, the students will know who to turn to.
It would be beneficial to have some more general technical staff for the entire university. The
meetshop is a very good example and should to my opinion surely be extended.
160 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
12 Bibliography
[1] J. Mueller, R. Hofer and J. Ziemer, “Survey of propulsion technologies applicable to cubesats,”
Jet Propulsion Laboratory California Institute of Technology, Pasadena, CA, USA, 2010.
[3] J. Bouwmeester, G. Brouwer, E. Gill, G. Monna and J. Rotteveel, “Design status of the Delfi-n3Xt
nanosatellite project,” in 61st International Astronautical Congress, Prague, Czech Republic,
2010.
[4] W. Larson and J. Wertz, Space Mission Analysis and Design, E1 Segundo, California, USA: Kluwer
Academic Publishers, Microcosm Press, 1999.
[5] J. Guo and E. Gill, “DelFFi: formation flying of two CubeSats for technology, education and
science,” Int. J. Space Science and Engineering Vol. 1, no. 2, pp. 113-127, 2013.
[8] A. Cervone, “Propulsion system requirements for the DelFFi satellites,” TU Delft, The
Netherlands, 2013.
[9] “CubeSat Design Specifications (CDS) Rev. 12,” California Polytechnic State University, USA,
2009.
[10] R. M. A. Poyck, “Literature study - Miniaturisation of satellite propulsion systems,” Delft, The
Netherlands, 2013.
[11] J. R. Wertz, Space Mission Engineering: The New SMAD, Microcosm Press, 2011.
[12] “Arthur Valentine OBE: Val Cleaver the pioneer of UK space program,” 06 11 2011. [Online].
Available: http://bowshooter.blogspot.nl/2011/11/arthur-valentine-obe-val-cleaver.html.
[Accessed 20 11 2013].
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 161
[14] B. Reed, “HAN-Based Monopropellant Technology Development,” Nasa_Ntrs, USA, 10-12-2002.
[16] “CubeSat Design Specifications (CDS) Rev. 12,” California Polytechnic State University, USA,
2009.
[19] A. Migliaccio, “Test campaign of T3uPS: a microthruster based on solid propellant cool gas
generators - Master thesis,” TU Delft, Delft, 1-12-2010.
[20] R. Poyck, “AE4S06 Mechatronics - Essay on: A self-pressurising integrated MEMS electrolysis
propulsion system,” TU Delft, the Netherlands, 11-06-2014.
[21] J. Garai, “Physical model for vaporization,” Fluid Phase Equilibria, vol. 283, no. 1-2, pp. 89-92, 15-
09-2009.
[22] B. Zandbergen, Thermal Rocket Propulsion - AE4-S01 - Reader (version 2.04), Delft, the
Netherlands: TU Delft, 29-01-2010.
[25] P. J. Marriotta, R. Shelliea and C. Cornwell, “Gas chromatographic technologies for the analysis
of essential oils,” Journal of Chromatography A, p. 1–22, 2011.
[26] Salomon's metalen, “CAPILLAIRE BUIS, hardgetrokken AISI 304,” [Online]. Available:
http://www.salomons-metalen.nl/nl/producten/rvs/buis/capillair. [Accessed 19 06 2014].
[27] The Lee Company, “INKX0511400A A valve technical drawing,” Essex, USA, 16-09-2013.
[28] The Lee Company, “062 MINSTAC - 062 Boss Configuration,” [Online]. Available:
http://www.theleeco.com/electro-fluidic-systems/manifolds-tubing-
162 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
accessories/MINSTAC/MINSTAC-062.cfm. [Accessed 19 06 2014].
[30] R. Ferreira, “Heat Transfer & Pressure Drop in Single-phase and Boiling Flow,” TU Delft, the
Netherlands, 20-05-2009.
[33] “062 MINSTAC Inlet / 062 MINSTAC Outlet (M/M),” The Lee Company, [Online]. Available:
http://www.theleeco.com/electro-fluidic-systems/solenoid-valves/vhs/062-minstac-inlet_062-
minstac-outlet-mm.cfm. [Accessed 18 07 2014].
[35] R. Humble, G. Henry and W. Larson, Space propulsion analysis and design, McGraw-Hill, 1995.
[36] M. Tajmar, Advanced space propulsion systems, Austria: Springer-Verlag Wien New York, 2003.
[37] F. L. Torre, “Gas Flow in Miniaturized Nozzles for Micro-Thrusters,” TU Delft, the Netherlands,
22-06-2011.
[39] Reliability direct, “System 24 LAGD 10 Pack Single Point Automatic Lubricators,” [Online].
Available: http://www.reliabilitydirectstore.com/SKF-System-24-LAGD-p/skf-lagd.htm.
[Accessed 07 07 2014].
[40] SKF, “Gas driven single point automatic lubricators - SKF SYSTEM 24,” [Online]. Available:
http://www.skf.com/group/products/lubrication-solutions/lubrication-systems /single-point-
automatic-lubricators/gas-driven-single-point-automatic-lubricators/index.html. [Accessed 07
07 2014].
[42] “Equipment for physician and hospital,” Hecht Assistent, [Online]. Available: http://www.hecht-
assistent.de/en/product-field/hospital-equipment/equipment-for-physician-and-hospital.html.
[Accessed 20 07 2014].
[44] J. Mueller, R. Hofer and J. Ziemer, “Survey of propulsion technologies applicable to cubesats,”
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 163
Jet Propulsion Laboratory California Institute of Technology, Pasadena, CA, USA, 2010.
[45] M. Tajmar, Advanced space propulsion systems, Austria: Springer-Verlag Wien New York, 2003.
[47] Key Safety Systems (KSS), “Cold Gas Curtain Inflator - HPH S,” 2008. [Online]. Available:
http://keysafetyinc.com/pdfs/Cold_Gas_Curtain_Inflator_HPH_S.pdf. [Accessed 06 07 2014].
[48] Key Safety Systems (KSS), “Cold Gas Inflator - CGI,” 2008. [Online]. Available:
http://keysafetyinc.com/pdfs/Cold_Gas_Inflator_CGI.pdf. [Accessed 06 07 2014].
[49] Special devices inc (SDI), AUTOMOTIVE PRODUCTS 2014 LIST PRICING, USA, 2014.
[50] Daicel group, “Gas generator/initiator price and performance list,” USA, 12-12-2013.
[51] M. S. Russell, The chemistry of fireworks - 2nd edition, Cambridge, UK: The royal society of
chemistry, 2009.
[53] Aerospace Specification Metals (ASM) inc., “Aerospace materials properties,” [Online]. Available:
http://www.aerospacemetals.com/. [Accessed 2014 04 08].
[55] “Gun Cotton,” The Florida state university, 28 04 2010. [Online]. Available:
http://www.chem.fsu.edu/outreach/gun-cotton.php. [Accessed 28 04 2014].
[56] “IEP SERIES SOLENOID VALVES,” The Lee Company, [Online]. Available:
http://www.theleeco.com/electro-fluidic-systems/special-products/iep/iep-series.cfm.
[Accessed 18 07 2014].
[57] DirectIndustry, “Mini solenoid dispensing valve,” The lee company, [Online]. Available:
http://www.directindustry.com/prod/the-lee-company/mini-solenoid-dispensing-valves-5322-
620684.html. [Accessed 22 07 2014].
164 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
[61] R. Ferreira, “Development and testing of a water microresistojet,” TU Delft, the Netherlands,
2008.
[62] T. J. Lawrence, “Research into resistojet rockets for small satellite applications,” University of
Surrey, Guildford, UK, 1998.
[63] M. Mihailović, MEMS monocrystalline-silicon based thermal devices for chemical and
microfluidic applications, Delft, the Netherlands: TU Delft, 2011.
[64] B. Morana, F. Santagata, L. Mele, M. Mihailović, G. Pandraud, J. Creemer and P. Sarro, “A silicon
carbide MEMS microhotplate for nanomaterial characterization in TEM,” in MEMS 2011,
Cancun, Mexico, 2011.
[66] C.-U. Kim, Electromigration in thin films and electronic devices, Woodhead publishing, 2011.
[67] Q. Bellini, “Test Plan - DelFFi Propulsion System Test Plan,” TU Delft, Delft, 2014.
[68] H. Petersen, P. Zevenbergen, B. Benthem and K. Sudmeijer, Theory and analysis - v.4.1 - course
AE4S20 - Satellite thermal control, Delft, the Netherlands: TU Delft, 2012.
[69] R. W. Knight, D. J. Hall, J. S. Goodling and R. C. Jaeger, “Heat sink optimization with application to
micro-channels,” IEEE transactions of components, hybrids, and manufacturing technology, vol.
15, no. 5, October 1992.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 165
Appendix A - Propulsion system concept trade-off
Table A.1 - Grading of different concepts for the trade-off of the DelFFi mission satellites propulsion system [10]
manufacturability
propellant safety
Electrical power
scientific value
consumption
availability
Property
storability
reliability
In orbit
Mass
Cost
ranking
TRL
total
Weight 7 4 8 4 7 7 8 8 9 7
Cold gas 0 8 8 1 7 7 7 6 7 7 414 1
solid 1 8 7 4 7 8 4 5 4 7 373 5
chemical
Mono- 3 9 6 4 8 4 5 6 7 8 412 3
propellant
chemical
Concept
bipropellan 1 9 6 4 8 5 4 5 4 4 334 7
t chemical
Resistojet 2 7 7 3 4 7 7 7 8 6 413 2
Arcjet 4 7 7 3 2 7 7 7 6 4 381 4
Colloid 0 7 7 2 1 8 8 4 4 3 308 8
Solar 2 6 6 5 7 7 7 3 3 4 339 6
thermal
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 167
Figure A.1 - Schematic overview of components for a proposed electrolysis propulsion system, those indicated with * are
optional [20]
Table B.1 - The parameters for the Shomate equation to calculate the enthalpy change in fluids. Including the parameter
validity range between the minimum temperature (Tmin) and the maximum temperature (Tmax) [23].
Fluid
Water (liquid) Water (gaseous) Water (gaseous)
Parameter Tmin [K] 274 374 1700
validity range Tmax [K] 500 1700 6000
Shomate A -203.6060 30.09200 41.96426
equation B 1523.290 6.832514 8.622053
properties C -3196.413 6.793435 -1.499780
D 2474.455 -2.534480 0.098119
E 3.855326 0.082139 -11.15764
F -256.5478 -250.8810 -272.1797
G -488.7163 223.3967 219.7809
168 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
Table B.2 - The parameters for the Antoine equation, to calculate the vapour pressure or the boiling temperature at a
certain pressure. Different values belong to different temperature ranges between the minimum temperature Tmin and
maximum temperature Tmax [23].
Table C.1 - Divergent section length Ldiv and thrust efficiency η, for micro-nozzles with different throat curvatures. The
nominal (non-viscous) thrust is Tideal = 1.14 [mN], and the diverging and converging angles are θ1 = θ2 = 15°, for all
designs [37].
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 169
Table C.2 - The thruster efficiency η for micro-nozzles with varying combinations of convergent half-angle, θ1, and
divergent half-angle, θ2. All combinations were calculated for a nominal (non-viscous) thrust of ∼1 [mN] and a sharp
(Rc=0) nozzle throat curvature [37].
Figure D.1 - Technical drawing of the gas generator exterior, with approximate dimensions in [mm]
170 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
1.05
Absolute internal pressure [bar]
1.04
1.03
1.02
1.01
1
0.99
0.98
0.97
35 36 37 38 39 40
Time [s]
23
Test chamber wall temperature [°C]
22.5
22
21.5
21
20.5
35 48 61 75 88 101 114 127 141
Time [s]
There was a failure in capturing the pressure during the third gas generator test. Therefore there is
no graph of the test chamber pressure behaviour during this test. The temperature was captured
successfully during this test, and is displayed in Figure D.4 below.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 171
23
Test chamber wall temperature [°C]
22.5
22
21.5
21
20.5
24 33 42 51 59 68 77
Time [s]
1.1
Absolute internal pressure [bar]
1.05
0.95
0.9
0.85
5 7 9 11 13 15 17 19
Time [s]
172 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
24.5
Test chamber wall temperature [°C]
24
23.5
23
22.5
8 17 26 35 43 52
Time [s]
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 173
Module
174
Total length Number of
I-1 D-1 D-2 S-1 S-2 N-1 N-2 N-K Reason
[μm] thrusters
T2 1 7 1 16639 3 Test the effect of enlarging the diamond shapes (decreasing the contact area)
T4 1 7 1 16639 3 Test the effect of enlarging the serpentine channel width (decreasing the contact area)
T5 1 7 1 16639 3 Testing the nominal design with an different shaped nozzle design
T6 1 14 1 25599 3 Testing the effect of heating chamber length on diamond shaped pillars
Appendix E - Manufacturing information
Thruster
T7 1 14 1 25599 3 Testing the effect of heating chamber length on serpentine heating channels
Production of a nozzle for Ivan's thruster, to test the difference between a MEMS slit
T10 1 1 7679 3
nozzle and a nozzle produced using traditional machining techniques
Table E.1 - Different desirable thruster module configurations which would be crucial or interesting to test. Explanation
Table E.2 - Explanation of the abbreviations used in Table E.1
I Inlet
N-1 Nominal nozzle design (45 [deg] to 20 [deg], sharp edge, 25 [μm] throat)
T Thruster
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 175
Figure E.1 - The photoresist mask that was used in the etching process of the fluidic channels and heating elements of the
Dondersteen prototype thrusters.
176 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
Name Type Diameter Length Website
[mm] [mm]
Ultradunne canules Luer 0.40 18 http://www.praxisdienst.nl/nl/Injectie
recht +infusie/Naalden/Wegwerpnaalden+s
tandaard+maten/Ultradunne+canules
+recht+0+40+x+18.html
Sterican canules voor Luerlock 0.5 25 http://www.praxisdienst.nl/nl/Injectie
de dentale anesthesie +infusie/Naalden/Wegwerpnaalden+s
tandaard+maten/Sterican+canules+vo
or+de+dentale+anesthesie+0+5+x+25
mm+oranje.html
Sterican®- Luerlock 0.45 12 http://www.praxisdienst.nl/nl/Injectie
insulinenaalden +infusie/Naalden/Wegwerpnaalden+s
tandaard+maten/Sterican+insulinena
alden+0+45+x+12+bruin.html
BD 30Gx 1/2" - ? 0,3 13 http://www.medicosupply.nl/bd-
0,3x13mm Injectienaald 30gx-1-2-03x13mm-injectienaald-
geel geel.html
Terumo Botox Met spuit 0.3 12 http://www.medicosupply.nl/terumo-
spuit/naald 1ml U-100 botox-spuit-naald-1ml-u-100-29g-
03x12mm.html
LLA4656300 Luer lock 0.3 12 http://www.mls.be/html/catalogus.as
p?lang=nl&group=4.2.2&detail=1
Luer lock Multiple Multiple http://www.schippers.nl/varkens/luer
MS injectienaald luer -lock/ms-injectienaald-luer-
lock lock;pgid=8dBgSxs3eN1SR0zOXCSTBu
Vz0000Vl699KsH;sid=qCpv8QRRUImp
8lfIF-bqhgRbuMQAaRS_x7I=
Luer lock Multiple Multiple http://www.schippers.nl/varkens/luer
Delvo injectienaald luer -lock/delvo-injectienaald-luer-
lock lock;pgid=8dBgSxs3eN1SR0zOXCSTBu
Vz0000Vl699KsH;sid=qCpv8QRRUImp
8lfIF-bqhgRbuMQAaRS_x7I=
MS Alu-hub Luer lock Multiple Multiple http://www.schippers.nl/varkens/luer
injectienaald luer lock -lock/ms-alu-hub-injectienaald-luer-
per 100 stuks lock-per-100-
stuks;pgid=8dBgSxs3eN1SR0zOXCSTB
uVz0000Vl699KsH;sid=qCpv8QRRUIm
p8lfIF-bqhgRbuMQAaRS_x7I=
WELLER - KDS3012P - Luer lock 0.30 12.7 http://nl.farnell.com/weller/kds3012p
DISPENSING NEEDLE, /dispensing-needle-ga30-id-0-
GA30, ID 0.15MM 15mm/dp/3975678
Table F.2 - Different electrical connector options between PCB and electrical power supply
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 177
Image Name Price [euro] Availability Link
TE 0.975 9.450 (UK) http://nl.farnell.com/te-
CONNECTIVITY / connectivity-amp/1241050-
AMP - 1241050-7 7/connector-header-14pos-
- CONNECTOR, 2row-2-54mm/dp/2399735
HEADER, 14POS,
2ROW
TE 0.93 744 (UK) 111 http://nl.farnell.com/te-
CONNECTIVITY / (EU) connectivity-amp/215307-
AMP - 215307-7 - 7/socket-vertical-2row-
SOCKET, 14way/dp/3419230?MER=en-
VERTICAL, me-pd-r2-acce-con
2ROW, 14WAY
MOLEX - 10-89- 2.52 149 (US) http://nl.farnell.com/molex/10-
7140 - 89-7140/connector-header-
CONNECTOR, 14pos-2row-2-
HEADER, 14POS, 54mm/dp/2112307
2ROW, 2.54MM
FCI - 67997- 0.77 1745 (US) http://nl.farnell.com/fci/67997-
216HLF - BOARD 216hlf/board-to-board-header-
TO BOARD, 16pos-2row/dp/1923919
HEADER, 16POS,
2ROW
178 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
Appendix G - Microscope images for geometry validation
Figure G.1 - Inlet - width measurement Figure G.2 - Heating module - Large diamonds - heater and
pillar width, length measurements
Since the heating element is suspended in the flow, nearly all the heat from the element will have to
pass through the propellant. There will be conduction to the exterior via the heating membrane
itself. Due to the membrane thickness of only 300 [nm] this is only minimal. The largest part of the
heat loss will occur through the contact between the water and the silicon wall of the heating
chamber.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 179
The heat flow from the heater to the water.
The heat flow of the water to the surrounding silicon.
The heat flow through the silicon.
The radiation of the silicon to its surrounding.
The heat transfer from the heater to the propellant will determine the effectiveness of the heater.
The heat transfer characteristics will determine the length and geometry of the heating channel, for
it to heat the propellant to the desired temperature.
For the conduction through a part the conductive coupling can be determined by:
kA
C ij 12-2
L
Where:
k Conductivity of the medium [W/m/K] (temperature and material dependent)
A Cross-sectional area [m2]
L Length of the conductive path [m]
Cij hc A 12-3
Where:
hc contact conductance [W/m2/K]
A Contact area [m2]
180 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
When looking at the shape dependence of Cij the following standard shapes are worked out as
following:
The heater section will be approximated by this last shape of the disc in radial direction. α will then
be 2π, d will be the length of the heating section and R1-R0 is the thickness of the layer.
Cross-section of the heating section, from the central heater to the outside casing
Each of the heat couplings now has to be determined. First of all there is a conductive coupling from
the heater to the water.
The optional glass layer has a thickness of 500 [μm]. And the metal reflector which is chosen is
aluminium.
The convective heat transfer from a surface to a fluid is shown in equation 12-4 below.
Q h A Tw Tf 12-4
Where:
h Convection coefficient [W/m2/K]
A Contact area [m2]
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 181
Tw Channel wall temperature [K]
Tf Average flow temperature
For the calculation of this heat transfer it is crucial to know the flow regime. Whether the flow is
laminar, in transition or turbulent has a big impact on the convection coefficient h. The Reynolds
number gives an indication in which flow regime a flow is. The Reynolds number based on tube
diameter is usually defined as determined equation 12-5 below.
U Dh
Re 12-5
Where:
Re Reynolds number [-]
U Flow velocity [m/s]
Dh Hydraulic diameter of the tube [m]
Kinematic viscosity [m2/s]
The equivalent hydraulic diameter Dh is a measure to relate the area of any cross-section to the
diameter of a circular cross-section with the same area. This parameter is defined as follows:
4A
Dh 12-6
p
Where:
A Cross-sectional area of the tube [m2]
p Periphery or inner wall length of the cross-section of the tube [m]
12-7
Where:
Dynamic (absolute) viscosity [kg/s/m]
Density of the flow [kg/m3]
m
U 12-8
A
Where:
m Mass flow [kg/s]
182 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
1
Since the area of a circle is A D2 , the Reynolds number in equation 12-5 can be rewritten
4
using equations 3-3 and 3-2.
m Dh 4m
Re
Dh
Dh2
1 12-9
4
Assuming that the rectangular flow channel has smooth walls, the critical Reynolds numbers can be
h
approximated using the aspect ratio of the flow channel ch . The relation between the aspect
wch
ratio and critical Reynolds number can be seen in Table 12.1. For values of the aspect ratio which are
not mentioned, the critical Reynolds number will be linearly interpolated from the known points.
Table 12.1 - Critical Reynolds number for rectangular channels, depending on aspect ratio [31][32]
For the heat transfer calculations, the following additional assumptions are made:
The duct is long enough for the flow to be fully developed over most of its length L.
The entering flow has a uniform temperature T0 at x=0
The heat will flow from the wall to the propellant with a heat flux qw , due to the wall temperature Tw.
Both the heat flux and wall temperature can be functions of longitudinal direction of the duct. In the
heating region, a thermal boundary layer will start close to the wall, as can be seen in Figure 12.1.
This boundary layer will grow in thickness in the downstream longitudinal direction, until the flow is
fully thermally developed at x = XT.
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 183
Figure 12.1 - Development of the thermal boundary layer to thermally fully developed flow through a tube [29]
The thermal entrance region has different thermal behaviour, and therefore calculation methods
than the thermally fully developed region. It is therefore necessary to see how significant this region
is, in order to assess whether it needs to be considered or can be neglected. The flow entrance
length, for low and high Prandtl numbers, can be estimated using equation 12-10 [29].
XT
0.05 ReDh Pr 12-10
Dh
Where:
XT Thermal entrance length [m]
ReDh Reynolds number based on tube diameter [-]
Pr Prandtl number [-]
The Prandtl number can be calculated with the following formula [62]:
Cp
Pr 12-11
kb
Where:
Cp Heat capacity of the fluid [J/kg/K]
kb Conductivity of the fluid [W/m/K]
For the initial calculations, the water shown in Table B.3 in Appendix B is used. The assumption is
made that the local pressure changes are negligible.
Conductive coupling from heater to propellant assuming uniform wall heat flux
184 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
When the local wall heat flux qw is assumed to be constant, the Nusselt number can be used to
determine the heat transfer coefficient. The definition of the Nusselt number based on hydraulic
diameter is shown in equation 12-12.
h Dh
NuDh 12-12
k
Where:
NuDh Nusselt number based on hydraulic diameter [-]
For laminar flow and slug flow, the Nusselt number is in the order of 1. For turbulent flow, the
Nusselt number is generally in the order of 100-1000. For the case of uniform wall heat flux, with
fully developed laminar flow, an empirical relation for the Nusselt number is shown in equation
12-13 below [69].
2
1
1
G 12-14
2
1
1
Where:
hch
Aspect ratio of the channel
wch
The Reynolds number will thus first have to be calculated to show that the flow regime. After that
the relevant Nusselt number can be found in the literature. Using the Nusselt number, equations
12-12, 3-26 and the known values of the conductivity k from, the convection coefficient h can be
determined. The flow temperature calculations can be conducted for small segments in the
longitudinal direction of the stream channel, with a length of dx. Under the given assumption of
constant wall heat flux, with a known total heat input of the wall section Q can be used to calculate
the temperature difference between the flow temperature and wall temperature can be calculated
using equation 12-4. For this small flow section it is clear that the temperature increase due to the
heat flux can be calculated using equation 12-15 below.
dT C p mf Q 12-15
Where:
mf Mass flow of the flow [kg/s]
This only holds when the length of the channel section is smaller than the flow velocity. Therefore
the section length dx has to be chosen to be adequately small to ensure that dx 1 s V .
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 185
In these calculations, instead of a constant heat flux, a constant wall temperature is assumed. This
results in a slightly different Nusselt number. The channel is again subdivided into small section with
a length of dx. The flow temperature in the section is assumed to be equal to the flow temperature
at the exit of the previous flow section. Using equation 12-12 and 12-4 the heat transfer to the
channel section can be determined. Using this heat transfer, equation 12-15 can be used to calculate
the flow temperature at the end of the channel section. Calculating these properties for every
channel section is calculated. The sum of all these section will yield the total heat transferred and the
flow temperature at the end of the channel.
For the calculations, the differentiation has to be made between “fully developed flow” and “slug
flow”. The latter is a fluid flow with a Prandtl number approaching zero. For that kind of flow the
viscosity is so much smaller than the thermal diffusivity, that the longitudinal velocity profile remains
uniform.
In the actual satellite, the thruster will be suspended in a cylinder which guides it through the
structure, to end up in
Alternative method for calculating the heat transfer from heater to the flow
For high power density and small particle size, the heat transfer in packed beds is usually calculated
using the Achenbach correlation (from source: B-09 page (85)). The heat transfer coefficient can be
expressed by:
kb
ht Nu 12-16
Dp
Where:
ht Heat transfer coefficient [W/m2/K]
Nu Nusselt number [-]
kb Bulk gas conductivity [W/m/K]
Dp Packed bed particle diameter [m]
1 Re 0.33
2.32 3
Re 4
Nu 0.622926 6.44603 10 Pr 12-17
1 1
Where:
Bed porosity [-]
Re Reynolds number [-]
Pr Prandtl number [-]
186 Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014
Dp v s
Re 12-18
Where:
Fluid density [kg/m3]
Dp Particle diameter [m]
vs Superficial or free stream fluid velocity [m/s]
Fluid viscosity [N·s/m2]
Cp
Pr 12-19
kb
Where:
Cp Heat capacity of the fluid [J/kg/K]
Knowing the heat transfer coefficient, the heat transfer to the surrounding can be determined. Since
the heating element is totally surrounded by the propellant, it can reasonably be assumed that all
conductive heat is transferred between the heater and the silicon via the water. The radiation
between the heater and the silicon though the water needs to be determined later.
Knowing the transfer coefficient, the minimal length of the heating element can be determined
Design, manufacturing and characterisation of a water fed CubeSat micro-resistojet - 28-07-2014 187