Reliability analysis and optimization of the LISA Phasemeter Frequency Distribution System

Proyecto Fin de Carrera Ingeniería de Telecomunicación Formato de Publicación de la Escuela Técnica Superior de Ingeniería Autor: F. Javier Payán Somet Tutor: Juan José Murillo Fuentes Dep. Teoría de la Señal y Comunicaciones Escuela Técnica Superior de Ingeniería Universidad de Sevilla Sevilla, 2013 Trabajo Fin de Grado Grado en Ingeniería Aeroespacial Reliability analysis and optimization of the LISA Phasemeter Frequency Distribution System Autor: Alicia López Jaén Tutor: Rafael Vázquez Valenzuela Dpto. Ingeniería Aeroespacial y Mecánica de Fluidos Escuela Técnica Superior de Ingeniería Universidad de Sevilla Sevilla, 2022 Trabajo Fin de Grado Grado en Ingeniería Aeroespacial Reliability analysis and optimization of the LISA Phasemeter Frequency Distribution System Autor: Alicia López Jaén Tutor: Rafael Vázquez Valenzuela Profesor Titular Dpto. Ingeniería Aeroespacial y Mecánica de Fluidos Escuela Técnica Superior de Ingeniería Universidad de Sevilla Sevilla, 2022 Trabajo Fin de Grado: Reliability analysis and optimization of the LISA Phasemeter Frequency Distribution System Autor: Alicia López Jaén Tutor: Rafael Vázquez Valenzuela El tribunal nombrado para juzgar el trabajo arriba indicado, compuesto por los siguientes profesores: Presidente: Vocal/es: Secretario: acuerdan otorgarle la calificación de: El Secretario del Tribunal Fecha: Agradecimientos Hay oportunidades que llegan para cambiarte la vida, y esta se la tengo que agradecer enormemente a Pablo. Gracias por confiar en mí, valorar mi trabajo y enseñarme tanto. Gracias a Esteban, por el cariño, la paciencia y la confianza desde el primer día. Ha sido un placer aprender de personas como vosotros. Gracias a todas las personas que han hecho de esta aventura una experiencia inolvidable, especialmente a Reshma, Johny y Sergio, por ser hogar en un sitio tan lejos del mío. A mis amigas, por apoyarme y escucharme desde el primer día hasta el último, por animarme siempre y confiar en mí. Especialmente gracias a Elena, Laura y Ana, gran parte de lo que soy ahora os lo debo a vosotras. Gracias a ARUS y a la Escuela, por toda la gente maravillosa que me ha dado y que sigue a mi lado, y por enseñarme cosas mucho más allá de lo técnico. Y como no, a mi familia, por apoyarme incondicionalmente y creer en mí cuando ni yo misma era capaz de hacerlo. Gracias a mi hermana Cristina, por ser un apoyo imprescindible para mí y tener siempre las palabras exactas para animarme a seguir. Gracias a mi padre, por inspirarme y motivarme siempre a retarme y mejorar. Y gracias en especial a ti, mamá, por ser un ejemplo a seguir, por inculcarme tanta valentía y perseverancia a luchar por lo que quiero. Estoy segura de que sin vosotros no habría llegado hasta aquí. I Resumen Global Título: Análisis de fiabilidad y optimización del Sistema de Distribución de Frecuencia del Fasómetro de LISA. Resumen del trabajo: Tras introducir la misión espacial LISA y sus objetivos, se expone la metodología empleada para el análisis de fiabilidad del módulo electrónico objeto de este trabajo, así como el modelo térmico asumido para el mismo. Como punto de partida para su optimización, se obtiene un resultado de fiabilidad preliminar del módulo, cuyos resultados se buscan mejorar respaldados por los análisis térmicos realizados con distintos parámetros. Además, se propone una modificación estructural del módulo. Finalmente se extraen conclusiones de los resultados obtenidos y se propone una línea de trabajo futuro. Palabras clave: Fiabilidad, redundancia, diagrama de bloques de fiabilidad, tasa de fallo, temperatura de operación, modelo térmico. Conclusiones: El resultado final de fiabilidad arrojado por los cálculos es muy favorable, lo que sugiere no implementar el módulo redundante propuesto, ahorrando así masa y coste. Por otro lado, la mejora de fiabilidad tras la optimización ha sido muy reducida dada la baja tasa de fallo de los componentes de partida. A pesar de ello se ha conseguido una reducción de peso del módulo de un 17.84%. III XList of Figures 3.17 Crystals transient response 43 3.18 PCBs transient response 44 3.19 Optimised enclosure design 44 3.20 PMS assembly structure 45 3.21 Boundary conditions on displacements 46 3.22 PMS assembly mesh 46 3.23 Final thermal distribution 47 List of Tables 2.1 Activation energy for failure mechanisms 14 3.1 VFDS Reliability calculations 28 3.2 CFDS Reliability calculations 33 3.3 FDS Devices Temperature 34 3.4 Devices Temperature 39 3.5 Devices Temperature. Different PCBs configuration 41 3.6 Structural results 46 3.7 Final working temperatures 47 3.8 Final reliability results 48 3.9 VFDS Final reliability calculations 48 3.10 CFDS Final reliability calculations 49 A.1 Devices reliability data 53 A.2 Devices thermal data 54 A.3 VFDS Power Budget 55 A.4 CFDS Power Budget 55 XI Contents Resumen Global III Resumen V Abstract VII List of Figures VIII List of Tables XI Notation XV Constants XV Acronyms XVI 1 Introduction 1 1.1 Nature and detection of Gravitational waves 2 1.2 LISA 3 1.3 Phase Measurement System 5 1.3.1 Frequency Distribution System 6 1.4 Objectives of this work 8 1.5 Structure 8 2 Theoretical background 11 2.1 Reliability methodology 11 2.1.1 Requirements 12 2.1.2 Reliability analysis 12 2.1.3 Reliability equations 13 2.2 Thermal model 14 2.2.1 Implementation of the thermal model in COMSOL Multiphysics 15 2.2.2 Geometry and materials 15 Boundary conditions 18 Stationary analysis 20 Time dependent analysis 21 2.3 Structural model 22 3 Results 25 3.1 Preliminary reliability analysis 25 3.1.1 VCXO 25 Functional diagram 25 Reliability diagram 27 XIII XIV Contents Reliability data 27 3.1.2 Clock Sidebands 29 Functional diagram 29 Reliability diagram 31 Reliability data 33 3.2 Thermal results 33 3.2.1 Stationary results 33 Layers and thickness optimisation 40 3.2.2 Transient results 41 3.2.3 Structural optimisation 43 Structural analysis 44 3.3 Final results 47 4 Conclusions and future work 51 4.1 Conclusions 51 4.2 Future work 51 Appendix A Data and additional calculations 53 A.1 Reliability data 53 A.2 Thermal data 54 A.3 Power Budget 54 Bibliography 57 Notation ∆Increment Ea Activation energy θJC Junction to case thermal resistance ∆TTemperature increment hLayer conductance qHeat flux by conduction uVelocity vector of translation motion Constants Notation Description Value SI-Unit MMass of the Sun 1.988 ·1030 kg cSpeed of light 299792458 m·s−1 kBBoltzmann’s constant 0.862 ·10−4eV ·K−1 ρAl Density of Aluminium 2700 kg ·m−3 ρCHO Density of CHOTHERM 1550 kg ·m−3 ρCu Density of Copper 8930 kg ·m−3 ρFR4Density of FR4 1850 kg ·m−3 kAl Thermal conductivity of Aluminium 1.67 ·102W·m−1K−1 kCHO Thermal conductivity of CHOTHERM 2.6W·m−1K−1 kCu Thermal conductivity of Copper 3.98 ·102W·m−1K−1 kAl Thermal conductivity of Aluminium 1.67 ·102W·m−1K−1 kFR4xThermal conductivity of FR4 in ˆx9.35 ·10−1W·m−1K−1 kFR4yThermal conductivity of FR4 in ˆy9.35 ·10−1W·m−1K−1 kFR4zThermal conductivity of FR4 in ˆz3.17 ·10−1W·m−1K−1 CpAl Specific heat of Aluminium 8.96 ·102J·kg−1K−1 CpCHO Specific heat of CHOTHERM 1000 J·kg−1K−1 CpCu Specific heat of Copper 3.85 ·102J·kg−1K−1 CpFR4Specific heat of FR4 7.95 ·102J·kg−1K−1 σyield Al Yield strength of Aluminium 276 ·106Pa σultimate Al Ultimate tensile strength of Aluminium 310 ·106Pa XV XVI Chapter 0. Notation Acronyms AEI Albert Einstein Institute ADC Analog to Digital Converter BEE Back-End Electronics CAD Computer Aided Design CFDS Clock side bands Frequency Distribution System CTM Compact Thermal Model DHLC Data Handling and Laser Control ESCC European Space Components Coordination EOM Electro-Optic Modulator ESA European Space Agency FBD Functional Block Diagram FDS Frequency Distribution System FIT Failure In Time GW Gravitational Wave IFO Interferometer LIGO Laser Interferemeter Gravitational Wave Observatory LISA Laser Interferometer Space Antenna MTTF Mean Time To Failure NASA National Aeronautic and Space Administration PCB Printed Circuit Board PMS Phase Measurement Subsystem PSM Power Supply Module PT Pilot Tone REF Reference RF Radio frequency SPF Single Point Failure TM Test Mass TFP Temperature Reference Points USO Ultra-Stable oscillator VFDS VCXO Frequency Distribution System 1 Introduction The detection of the first gravitational wave in 2015 was a revolution for the scientific world, opening a whole new window to explore the Universe. This occurrence, GW150914, was observed by the onground gravitational waves detectors LIGO (Laser Interferometer Gravitational-Wave Observatory). These waves were caused over a billion light years away by the merger of a pair of black holes around 36 M and 29 M . The resulting black hole had a mass of 62 M , leaving 3.0 M of energy radiated away as gravitational waves [ 1 ]. The obtained frequency of the signal rises, ranging from 40 to 250 Hz, as shown in Figure 1.1. Nevertheless, the frequency of gravitational waves that can be sensed from Earth is limited to a specific range of 40-7000 Hz [ 2 ], owing to the effects of seismic noise, such as the atmosphere or the Earth’s own gravity. In addition, the size of the detectors is limited due to the planet’s curvature, which also limits the bandwidth. Figure 1.1 Frequency data from the event GW150914 at the Washington and Louisiana observatories on 14 September 2015 [1]. This limitation in the frequency range was the motivation for the mission LISA (Laser Interferometry Space Antenna), on which this work is focused, becoming the first space-based observatory for gravitational waves. LISA will allow access to a band of gravitational waves from about 0.1 mHz to 100 mHz. This low-frequency range will provide information about how binary stars formed in our Milky Way, the formation of the first seed black hole or even the cosmic background gravitational radiation [ 3 ]. LISA will provide information from a part of the Universe that will forever remain undetectable with light. By ‘listening’ to gravity, it will be possible to see further than ever before. 1 2Chapter 1. Introduction 1.1 Nature and detection of Gravitational waves Figure 1.2 GWs generated by binary neutron stars [4]. General Relativity replaces the Newtonian conception of gravitation as instantaneous action at a distance with a geometric one, relating that space and time do not have an independent existence but are in intense interaction with the physical world. The essence of this theory is that matter and energy create a curve four-dimensional in the fabric of space-time, and this curvature influences the movement of matter. Gravitational waves are a consequence of this basis. If a mass distribution has a changing asymmetric movement (like acceleration spinning), then gravitational waves appear as space-time indentations travelling outwards as ripples in space-time [5]. This concept is illustrated in Figure 1.2 For a better understanding of this idea, they are compared to the well-known electromagnetic waves, from which almost all information about the Universe has been obtained so far. Gravitational waves are the solution for Einstein’s field equations, that determine the interaction between mass and curvature of space-time, in the same way as electromagnetic waves are the Maxwell’s equations solution that relates the relationship between electric charges and electromagnetic fields. Nevertheless , gravitational waves are fundamentally different. Electromagnetic waves propagate in space and time, but gravitational waves are waves of the space-time fabric itself. They stretch and squeeze any distribution of matter or energy by which they go through, acting transverse to the direction of wave propagation. The propagation of GWs in the two polarizations is presented in Figures 1.3 and 1.4. In particular, the graph in Figure 1.4 presents the travelled distance by the gravitational wave, described by a strain amplitude h, while the rest shows the four phases of one particle deformation cycle: the strain design contracts space in the transverse dimension, whereas it expands along the orthogonal direction within the transverse plane. Furthermore, they have a different nature. On one side, electromagnetic waves are created by the acceleration of electric charges. As charge only exists in two polarities, electromagnetic radiation is caused by a dipole moment of the charge distribution. On the other side, gravitational waves are generated by accelerated masses. Due to mass and momentum conservation, gravitational waves can only be created by a variable mass quadrupole . Figure 1.3 GWs propagating along the z axis in the two different polarisations (“+”and “x”) [3]. 1.2 LISA 3 Figure 1.4 Passage of a GW through a ring of masses [3]. Therefore, it can be said that the nature of gravitational waves is a change in the quadrupole moment of the mass distribution. This is why spherically symmetrical gravitational effects do not generate gravitational radiation. An example of this is a perfectly symmetric collapse of a supernova, while a binary system will always emit GWs [5]. Gravitational waves’ effects are apparent. They bend space-time, changing the distance between macroscopic free objects. When a gravitational wave passes through the Solar System it induces a time-varying strain in space that periodically changes the distances between all bodies in the Solar System in a direction perpendicular to the direction of wave propagation. The main problem is that the relative change in length due to the passage of a gravitational wave is exceedingly tiny. In order to detect gravitational waves, it is necessary to measure these variations in the distance between objects with extremely high precision (pm to nm). The measurement process used is laser interferometry. For that, a laser beam pulse is established between a free-floating source and a free-floating reflecting mirror. The basis of interferometry is to constantly measure the time it takes the laser beam to return after being reflected in the mirror. When a gravitational wave crosses the laser path, it will stretch space and alter the distance between the source and the mirror. As a result, the laser light will return at a different time. LISA’s measurement system is explained in section 1.2. 1.2 LISA The principal aim of the Laser Interferometry Space Antenna (LISA) is the detection of gravitational waves in a frequency range from 0.1 mHz to 100 mHz, which is impossible to measure in groundbased experiments. LISA project is based on the success of LISA Pathfinder, whose goal was to pave the way for LISA mission. More precisely, its main objective was to test the core assumption of gravitational wave detection and general relativity: that free particles follow the waves of space-time curvature, as well as demonstrate the new technologies in a space environment that will be used in LISA, such as inertial sensors or high precision laser interferometry to free-floating mirrors [6]. The LISA mission will be made up of three spacecraft flying in a triangular formation, (one ‘mother’ at the vertex and two ‘daughters’ at the ends) with a nominal separation distance of 2.5 million km, which follow an Earth-trailing heliocentric orbit between 50 and 65 million km from Earth. The distance between the spacecraft, the interferometer arm length, determines the frequency 2 Theoretical background This section establishes a theoretical basis that will serve as support for the development of this document and the achievement of its objectives. 2.1 Reliability methodology Given the enormous scientific importance and amount of money invested in space missions, reliability prediction significantly impacts the design of any space-applicable electronic architecture. Reliability is understood as the probability that equipment can perform continuously, without failure, for a specified interval of time when operating under stated conditions. For a better understanding of the process to be carried out, some key concepts for conducting reliability analysis are also introduced: • Failure: an event when equipment is not available to produce parts at specified conditions when scheduled or is not capable of producing parts or performing scheduled operations. For every failure, an action is required. • MTTF (Mean Time To failure): measures the mean time a component functions before failing, an essential basic measure of system reliability for non-repairable items [hours/failure]. • Failure rate: number of failures per unit of gross operating period in terms of time, events, cycles, or number of parts. •Single Point Failure (SPF): Failure that can stop the entire system from working. • Redundancy: is a commonly applied reliability improvement technique to enhance the system reliability and availability of safety-critical systems, or operational impact systems in the railroad and mass transit industry. • Hot Redundancy: redundant components are continuously energized and operate simultaneously. • Cold Redundancy: some or all of the redundant items are not operating continuously but are activated only upon failure of the primary item performing the function. • Cross-strapping: two identical units that are interconnected with the remaining system in such a way that either unit can provide the required functionality. 11 12 Chapter 2. Theoretical background 2.1.1 Requirements One of the main aspects sought in this thesis is to achieve the requirements imposed by the mission. In particular, the requirements associated with reliability exposed in [13] are, •The overall reliability of the PMS shall be above 97% over the in-orbit operational lifetime. •An equipment failure shall not propagate to another unit or subsystem. • No single hardware or software failure of the PMS shall lead to irreversible degradation in performance or loss of mission. 2.1.2 Reliability analysis As mentioned in the Space Standards ECSS-Q-ST-30C, this analysis will be used to predict the system’s life service, identify the criticalities of the baseline architecture and optimise the design to meet the reliability requirements while considering limitations such as mass and cost [14]. The variables that have a major influence on the system reliability are given below: • Duration of use of the electronics, which in LISA’s case can be translated in a lifetime of seven years. • Reliability block diagram based on the system’s operation, establishing the redundancy architecture in terms of serial or parallel blocks. •Failure rates of every electronic device and the number of them that make up the system. •Working temperature value. In terms of obtaining the failure rate, there are three types of reliability prediction methods [ 15 ]: • Bottom–up statistical methods. This method is built on the assumption that the failure rate of a device is a consequence of the failure rates of its internal components, which are supposed to be independent of each other. The failure of the component as a whole is not studied but built as a result of applying different mathematical models to the prediction. This type of method uses mainly two models of reliability prediction. These methods vary in the degree of information needed to apply them. – Part count method: requires less information, generally part quantities, quality level and the application environment. The Parts Count Methods usually results in a more conservative estimate of system reliability than the Parts Stress Method. – Part Stress Analysis Methods: needs a greater amount of detailed information, including the schematics. An example of this method is the widely used military standards MIL-HDBK-217F [16]. • Top–down similarity analysis methods. Based on proprietary databases where failures are recorded of the whole systems or subsystems. • Bottom-up physics-of-failure methods. This method utilises the devices’ reliability results from the physical accelerated test such as temperature cycling, highly accelerated stress test or temperature humidity bias. This data builds the failure rate by using statistical laws, commonly Weibull, using a confidence level of 60% or 90%. The final failure rate of the device is formed with the obtained failure rates from each accelerated test. 2.1 Reliability methodology 13 In this process of reliability prediction analysis to be done, the primary reference for the FDS equipment is the manufacturer data, manufacturer calculation tools and the communication with manufacturers. The methods used by manufacturers varies between them, although most provide failure rates based on bottom-up physics-of-failure methods, commonly expressed in the form of FITs for a better understanding of them. Detailed reliability data of each device is provided in the Annex A.1. 2.1.3 Reliability equations Reliability calculations are ruled by the following equations [17]: •Failure rate: λhours =1 MTTFhours (2.1) •Failures in Time: FIT =λhours ·109(2.2) •Reliability of one element: Ri=e−λhourst(2.3) where t is the mission lifetime in hours. •Reliability of n elements in series: Rns= n ∏ i=1 Ri(2.4) •Reliability of n identical elements in hot redundancy: Rnhr =1−(1−Ri)n(2.5) • Reliability of n identical elements in cold redundancy for non-repairable systems as demonstrated in [18]: Rncr =1−(1−e−λhours t 2)n=1−(1−pRi)n(2.6) •Reliability of “m” out of “n” identical elements to survive: Rm/n= n ∑ k=m n! k!(n−k)!Rk i(1−Rn−k i)(2.7) It is important to consider that reliability is highly influenced by other agents apart from time of use and redundancy. Electronic parts are exposed to varied stresses such as electrical, thermal or 14 Chapter 2. Theoretical background ambient throughout the operation. However, temperature is the most influential factor in electronic failures. It has a significant impact on its reliability, as it causes slow progressive deterioration of the materials of which the part is made [ 19 ]. Equipment which works adequately at one temperature level will be less reliable or inoperable as temperatures increase. Hence is crucial to be aware of the effect of the temperature on the reliability of the device. Given the reliability value (or total FIT) of a device at a specific temperature, one can determine what the reliability would be at a different temperature by introducing an acceleration factor [20]: Ri(Tn) = [Ri(Ti)]1/AF (2.8) Where AF is the acceleration factor given by the Arrhenius equation: AF =e −Ea kB(1 Ti−1 Ti)(2.9) Where kB is the Boltzmann’s constant and Ea is the activation energy of the failure mechanism. In Table 2.1 are shown the activation energy values for some typical failure mechanism [20]. Table 2.1 Activation energy for failure mechanisms. Failure Mechanism Ea Corrosion 0.45 eV Electromigration of Al 0.6 eV Contamination 1.0 eV Oxide defects 0.3-0.5 eV Charge inj ection 1.3 eV Mask defects 0.7 eV Assembly defects 0.5 -0.7 eV The activation energy for each component is detailed in Annex A A.1. Consider an assembly that is comprised of the PCB and the electronic devices, both active and passive. High power values consumed by an electronic component entail a source of heat. Heat transfer can imply the heat flux from one hot device to another through different heat transfer mechanisms. If the component receiving the heat is a device sensitive to temperature and the internal heat reaches a high value, the device reliability will decrease consequently. As electrical and thermal stresses are closely interrelated with reliability, it is crucial to study the accurate working temperature of each device in order to have realistic reliability results. This temperature is obtained by simulation in COMSOL Multiphysics, explained in detail in Section 2.2, considering all the effects of heat transmission between the power consumed by each device and how this heat is transmitted to the PCB, the enclosure and the CHOTHERM. 2.2 Thermal model Three different mechanisms can occur in the process of heat transfer between two bodies: • Conduction: this process arises from the molecular vibration of the solids, where the heat is exchanged from the solid with more energy to the solid with less energy. Conduction is 2.2 Thermal model 15 commonly the primary heat transfer mechanism in solids. • Convection: this form of heat dissipation is caused by the movement of a fluid. Due to the vacuum environment operation, convection will not be studied. • Radiation: as radiated heat is transferred by electromagnetic waves, it does not require a physical medium to occur. Consequently, thermal radiation may become considerably relevant in space applications, up to 30% of the heat dissipation of an electronic device to the surrounding environment. As exposed in [ 21 ], thermal conduction problems are solved using an electrical analogy. In these models, known as compact thermal models (CTMs), the thermal contact between the solids of the system is characterized by a thermal conductance h , simulating the electrical equivalent of an electrical resistor, and the heat flow is modelled as the circuit current flow. In order to model the heat transfer by conduction of the whole Frequency Distribution System, the electrical analogy presented in Figure 2.1 has been assumed. This outline shows all the solids involved in dissipating heat from the component to outwards the PMS, whose detailed heat transfer behaviour is explained below in Section 2.2.1. The starting point is the heat ’applied’ to the electronic component. This heat is transferred directly to the PCB, followed by the enclosure. Moreover, a conductive elastomer insulators has been placed at the base of the enclosure, CHOTHERM, whose utility consists in ensuring efficient heat dissipation and preventing local temperature overloads, moreover it must also be considered for the thermal study. PCB ENCLOSURE CHOTTERM T2T3T4T5T6TB T1 DEVICE ADkDxD kD RkPCB R ,,APCB kx ,, PCB PCB AEkxE ,, EACxC ,, C k hPCB RkE RkE RkC R E C Qθ . PD Figure 2.1 Electrical analogy for the FDS. 2.2.1 Implementation of the thermal model in COMSOL Multiphysics This section aims to explain the creation of the electrical analogy shown in Figure 2.1 in the COMSOL Multiphysics simulation program, explaining in detail the assumptions considered, boundary conditions applied, and meshing of the geometry. 2.2.2 Geometry and materials The first step for the implementation of this model is the creation of the geometry to work with. The CAD software used for this purpose is SolidWorks, where an assembly has been created with all the 16 Chapter 2. Theoretical background aforementioned solids: PCBs, enclosure, CHOTHERM and every particular device geometry. For the thermal modelling of each electronic component, several simplifying hypotheses have been considered : • The whole device has been assumed as a unique resistance, simplifying the composition of die, package and finish. • Heat transmission from the device to the PCB will be considered perfect [ 21 ]. Thermal transfer between two solids is dependent on the on the surface’s roughness and the pressure applied between them. In the current design state, it is impossible to know these variables with precision, and experimentation would be needed for each of the component. Given the enormous number of components, and the differences between them, it seems reasonable to model the heat transfer based on the junction-to-case value θJC , provided in the datasheet. Its implementation is detailed explained in this Section. • Radiated heat from every device will not be considered. Despite the fact that radiation in space applications plays a very important role, radiated heat from electrical components depends on many variables that are still unknown at this stage of the design, such as the ambient temperature of the module or the disposition of the different modules along the PMS, as some of them will generate more heat than others and heat that will be conducted within the modules. Therefore, internal heat radiation is not considered for the analysis to be carried out, assuming in that sense a worst-case scenario, where all the heat dissipated by the component will be presupposed to source from conduction. • As the operation of each component generates internal heating which must be dissipated to the PCB, the heat transmission is modelled as a heat flux from the top surface of the device as shown in Figure 2.2. This heat rate is the power consumed by every device, whose value is calculated in the power budget in Annex A.3. RθJC P D Figure 2.2 Electronic device thermal model. These considerations have a great impact the model’s material definition, as each of the solids in the model must have an associated material. When defining materials in COMSOL Multiphysics, three parameters must be designated: density ρ , heat capacity at constant pressure Cp and thermal conductivity k. 2.2 Thermal model 17 In order to simulate the behaviour of each device as closely as possible to reality, the material constituting the component package has been taken as the reference material for the density and heat capacity values, as it constitutes the central part of the entire component. The package generally used for space devices is ceramics, as it performs better against the internal heating of the component. In some cases, the devices’ datasheet are insufficient to know the package materials specifications accurately. Hence, a generic ceramic material for space upgrade has been used for density and heat capacity values, and their properties were also obtained from the material database Matweb [ 22 ]. This simplification is reinforced later with the approach of the heat transfer equations solved in COMSOL in Section 2.2.2. Regarding the thermal conductivity k , it is calculated based on each devices’ characteristics. This property is related to how well the material can conduct heat. Subsequent to consider the perfect heat transmission hypothesis already mentioned, the thermal conductivity of each device has been modelled by its junction-to-case data θJC =K/W , obtained in the datasheet. The process reflected in the Equation 2.10 leads to the derivation of the thermal conductivity: k=W m·K=S θJC ·V(2.10) Where S is the contact surface of the device and V the volume. All the thermal data employed is attached in Annex A.2. Following the diagram in Figure 2.1, the next item is the PCB. To define the whole material that constitutes the PCB, several assumptions have been considered. First, the PCB is formed by several layers of different material superposes (copper and dielectric, generally FR4), as shown in Figure 2.3. Figure 2.3 PCB layers [23]. This implies that the material that makes up the PCB is a composite material. Consequently, their parameters depend on the properties of the constituent materials and the quantity of each one. In a first approximation, a thickness of 1.65 mm and six layers of copper have been assumed. Since the objective of this study is the thermal optimisation of the FDS, it is helpful to introduce in COMSOL these quantities in the form of parameters to facilitate future extraction of results as follows: n=6 ThCu =35 ·10−6mm LPCB =1.65 mm LCu =n·T hCu LFR4=LPCB −LCu Where LFR4 is the remaining thickness of the PCB, n is the number of used copper layers and ThCu is the used copper thickness, whose values will be 6 layers and 35 microns in principle. On 18 Chapter 2. Theoretical background this basis, the density ρ[kg/m3] and the heat capacity at constant pressure of the PCB material Cp [J/(kg ·K)] have been obtained with the following approach: Cppcb =LCu ·CpCu +LFR4·CpFR4 L(2.11) ρpcb =LCu ·ρCu +LFR4·ρFR4 L(2.12) Nevertheless, for the whole thermal conductivity, it must be considered that the FR4 has lower thermal conductivity in the plane perpendicular to the PCB, and consequently, the board cannot be treated as an isotropic material. To implement this in COMSOL, the thermal conductivity of the material defining PCB has been modelled with a diagonal matrix of its values in each direction k=ki,i=  kx0 0 0ky0 0 0 kz  (2.13) Basing this calculation on the same logic as the previous ones: ki=LCu ·kCu +FR4LLFR4 Lf or i =x,y,z.(2.14) This question is simplified in the case of the other solids parameters in the scheme, which are the enclosure and the CHOTHERM. The density, heat capacity and thermal conductivity for these materials had been directly obtained from the database [24] [25]. Boundary conditions One of the main boundary conditions for a thermal study is the definition of the thermal contact areas between the solids. Heat transfer between the enclosure and the PCB is defined by the layer conductance value, though it depends on many unknown variables. Not all the contact between the PCB and the enclosure will be the same throughout the module, as there are areas with higher pressure between the two solids. The example that best illustrates this situation is the screw joint area. In these, the contact area will be larger; therefore, the heat transfer will also be higher than the rest of the interface. In consequence, a good approximation of the thermal contact area is crucial to carry out a realistic thermal study. In the following analysis, the thermal contact is limited to a circular region as presented in Figure 2.4 around the union screws as studied in [23]. Figure 2.4 Thermal contact area. Inner diameter is 4mm and outer 8mm. 2.2 Thermal model 19 The implementation of the thermal boundary conditions in COMSOL looks as displayed in Figure 2.5. The Figure in the left shows the internal surface of the CHOTHERM as a thermal contact boundary condition with the enclosure, where all the surface has been selected as heat transfer interface. Additionally, each screw junction area has been established as a thermal contact between the enclosure and the PCBs, characterised by the contact layer conductance discussed below. Nevertheless, as the thermal contact between these solids is supposed to be limited exclusively to this circular area, the rest of the contact area between the enclosure and the PCB must be set as a thermally isolated area, as displayed in the right-hand Figure. Setting this condition is necessary because COMSOL would default set it as a zone where the thermal contact is perfect between these solids, obtaining very optimistic results far from reality. Figure 2.5 Thermal contact boundaries conditions. In order to get the unknown value of the layer conductance of this circular area, a comparison between experiments and simulations was carried out in the AEI [ 23 ]. By sweeping the values for this layer conductance in the simulations, and then comparing the results with the experimental data, it has been concluded that the appropriate layer conductance value for this area is hPCB→E=1650 [W m2·K] A similar study has been carried out to calculate the layer conductance value between the enclosure and the CHOTHERM. In this case, the contact area has been assumed to be the entire junction area between the two, obtaining the following result hE→C=293 [W m2·K] Additionally, boundary conditions from the outer part of the enclosure had also been defined. On the one hand, the enclosure base is supposed to remain at 20 ° C by design, with a tolerance of ±5oC . With the aim of parameterise this temperature and its possible variations, a variable called Ttable was generated. This temperature condition has been imposed on the outer face of the CHOTHERM, as seen in the left image of Figure 2.6. On the other hand, the right image shows a boundary condition associated with the heat transmitted by radiation. This has been assumed because, although the effects of radiated heat have not been considered at any time given the influence of many unknown internal variables, this surface shown is the only surface not influenced by other factors, and can radiate heat freely inwards the satellite. To implement the radiation condition in COMSOL the study of surface-to-surface radiation has been explicitly used, making it necessary to establish the refractive index of the medium, whose value for vacuum is 1 by definition: n=c v(2.15) As c is the speed of light in vacuum and v is the speed of light in the medium, which coincides in this case. 26 Chapter 3. Results Temp Monitoring) Temp Monitoring) Temp Monitoring) TempTempTemp USO_R USO_N PPS_R PPS_N + - + - Square PI Controller + - + - Square PI Controller USO/PPS X-Strapping USO OUT PPS OUT 80M OUT INPUT SIGNAL VCXO_N VCXO_R Slip-less Clock Switching NG Large FPGA + - + - + - + - ÷ M φ + - + - + - + - + - + - VCXO_N VCXO_R USO PPS 80M ÷ K Clock Distribution & VCXO Locking Electronics Linear Regulator 10 10 USO BUFFER PPS BUFFER 80 MHz BUFFER LNA_N LNA_R LVDS Receiver Quartz_N Quartz_R T° Controller Diode Transistor Figure 3.1 VFDS functional diagram. 3.1 Preliminary reliability analysis 27 the fact that switching from one clock to another requires a transition time in which cycles are lost. Given the criticality of the 80 MHz signal, this slip-less switching point ensures that the 80 MHz signal is constantly being provided to the PMS. This external loop is represented in the right part of Figure 3.1. The green rectangle includes the logical control operator of the FPGA from the DHLC, whose inputs are the USO OUT and the 80 MHz OUT. Regarding the electrical aspects, this board receives 5 volts from the Power Supply Module. However, a linear regulator must convert these five volts to 3.3 volts, as some devices require this specific voltage for their correct operation. Additional functions, such as temperature reading, are also performed in this board. Reliability diagram Based on the functional diagram shown in Figure 3.1 and the reliability methodology explained, the reliability diagram of this PCB is built. In terms of reliability, it is necessary to establish the ’success’ to be achieved. For this PCB, the success will be the correct functioning of the whole PCB, including all the functions it has to perform (PPS, USO and 80 MHz signal distribution, voltage regulation, temperature control). To represent this objective, each component should be considered as a functional unit, as this is where the failure rate is known. Before building this diagram, it is necessary to identify the redundant parts or if, on the contrary, the devices form a serial chain. In the VFDS, it can be found a redundant architecture in the clock chain. This is justified by the fact that the quartz crystal has the highest failure rate on this board. It is also supported due to the criticality of getting the 80 MHz signal. In addition, the LVDS Receiver is a quad-gate device that must perform three functions: convert the sinusoidal signal to square from the nominal string, from the redundant string and establish a junction point where the two strings are combined. However, it can be clearly seen that this junction point is an SPF. If this LVDS receiver fails, the 80 MHz signal is completely lost. For this reason implementing a second cold redundant LVDS receiver was proposed, thus avoiding this problem. Focusing on the reliability of this board, the following Figure 3.2 is presented as a function of the devices that compose the VFDS. The figure shows how the linear regulator and buffers, USO and PPS, are in a series arrangement, followed by the redundant clock chains, whose constituent devices differ. This architecture is justified in an intern design document [ 11 ]. After that, it can be found LVDS driver and receiver again in series. It should also be mentioned that the operation of this board involves the functioning of the modules DHLC (external control loop with the FPGA) and PSM (power supply to the board), marked in red in the diagram. If any of these modules fail, the VFDS will also fail. However, this dependency is not studied in this project since that corresponds to a complete reliability study of all the PMS modules. Reliability data Based on the reliability diagram exposed in Figure 3.2 previously developed and the methodology explained in Section 2.1, the achieved result is: 28 Chapter 3. Results LINEAR REGULATOR PSM DHLC PPS BUFFER USO BUFFER OPAM_N DIODE CRYSTAL_N TRANSISTOR 1 TRANSISTOR 2LVDS RECEIVER OPAM_R CRYSTAL_R LVDS DRIVER 80 MHz BUFFER CLK NOMINAL CLK REDUNDANT LVDS RECEIVER Tº SENSOR Figure 3.2 VFDS Reliability diagram. TOTAL VFDS RELIABILITY 99.85 % Whose calculations are shown in Table 3.1. Detailed reliability data for each component can be found in Appendix A.1. Table 3.1 VFDS Reliability calculations. BLOCK DEVICE FIT RELIABILITY BLOCK RELIABILITY USO BUFFER USO Buffer 2.8 0.999828319 0.999828319 PPS BUFFER PPS Buffer 2.8 0.999828319 0.999828319 VCXO NOMINAL OPAM N 1.17 0.999928258 0.998788515 0.99999695 Diode 0.5938 0.999963592 Crystal N 18 0.998896849 Transistor 1 0.0025 0.999999847 Transistor 2 0.0025 0.999999847 VCXO REDUNDANT OPAM R 1.17 0.999928258 0.99747864 Crystal R 40 0.997550206 RECEIVER LVDS Receiver 5.4 0.999668927 0.999668927 0.99999989 80MHz BUFFER LVDS Driver 2.9 0.999822188 0.999650537 80 MHz Buffer 2.8 0.999828319 REGULATOR Linear regulator 4.29 0.999736972 0.999736972 TºSENSOR T Sensor 2.8 0.999828319 0.999828319 3.1 Preliminary reliability analysis 29 3.1.2 Clock Sidebands Functional diagram In this section, the internal functioning of the Clock Sidebands board is explained based on the diagram shown in Figure 3.3. Similar to the VFDS, this board essentially fulfils two functions. The main one is to distribute the Pilot Tone signal to the rest of the modules (LARM-BEE 1, LARM-BEE 2, TM-BEE and REF-BEE), while the second one is the procurement of a 1 MHz signal, which can be added or not to the output of the REF-BEE, in the convenience of operation. PILOT TONE SIGNAL DISTRIBUTION The main input of this PCB is the 80 MHz signal coming from the VFDS, already locked with the Ultra Stable Oscillator master signal. This input connector is shown on the upper right side in the Figure 3.3. The first device involved after the input is the frequency up-converter, generating the clock sidebands modulation signals. This component can generate two different frequencies using its internal oscillator and different intern dividers or multipliers. These frequencies can be seen on the two different output ports of the controller A and B. Port A is set at a frequency of 2.4 GHz or 2.401 GHz, depending on operating requirements, while port B is set at Pilot Tone. In Figure 3.3, the case where the exact 2.4 GHz signal is generated at PMS 1 has been depicted as an example. The purpose of this frequency difference is to secure two different strings to obtain the Pilot Tone. From port A, 2.4 or 2.401 GHz are output in differential form to suppress the high-frequency harmonics at the output of the frequency up-converter, acting as a shaping filter to convert from a square signal to a sinusoidal output. After this, the signal arrives at power splitter 1, where it is distributed to other parts of the satellite, in addition to the interconnection with the other PMS. This interconnection is shown in Figure 3.3 as the PMS2_Mixer_out and PMS2_PT_out outputs, both marked in blue. It should be noted that these outputs go to a board completely identical to this one, which means that we find these same signals as input also on this PCB, marked in red. The PMS2_PT_in input is found on the next component, switch 1. This device allows the module to choose between 2.4 GHz generated on PMS 1 or PMS 2. After this choice, a chain of dividers converts the 2.4 GHz into Pilot Tone in two steps. From port B, the Pilot Tone signal comes directly out of this port without any transformations up to switch 2, clearly constituting a redundant path for the Pilot Tone signal. However, this chain has an undesired signal degradation, so that the nominal chain will be the one from port A. Both chains converge at switch 2, where the chosen path is programmed from the FPGA. After that, the signal is amplified (LNA 1) due to possible losses that the signal may have experienced, passing then through a filter (LPF 2). Finally, the signal is split in the power splitter 2 to distribute it to all the modules, represented in Figure 3.3 as LARM_BEE 1, LARM_BEE 2, TM_BEE, REF_- BEE outputs in green. For clarity, the redundant modules have not been included in the diagram, but the two redundant modules of the BEE have been taken into account in the design. 1 MHz SIGNAL PROCUREMENT The interaction between the two PMS is also involved in acquiring the 1 MHz signal. The signal is extracted in the mixer, located at the mid-bottom part of the Figure 3.3. As it is represented, one 30 Chapter 3. Results LPF 1 PMS2 _PT_out Splitter 1 PMS 2_Mixer_out Loop Filter (RC) Loop Filter (RC) V tune CPout ÷ N÷ N xRxRxR φφ AA LPF 2 Splitter 2 B 80 MHz Sys ClK LPF 3 ÷L÷L ÷B÷B Temp controller Temp controller Temp controller TempTempTemp Temp controller Temp Temp controller Temp PMS2 _PT_in Linear Regulator LNA 3LNA 3LNA 3 LARM_BEE1 LARM_BEE2 TM_BEE REF_BEE 2.401GHz 2.4GHz 1 MHz PMS2 _Mixer_in 75 MHz Pilot Tone 75 MHz LNA 2 Clock Sidebands Distribution & Electronics AdderAdderAdder LPF 4 LNA 1LNA 1LNA 1 Switch 1 Switch 3 Switch 2 Telemetry controller Mixer Power divider 1 Power divider 22.4GHz INPUT Figure 3.3 CFDS functional diagram. 3.1 Preliminary reliability analysis 31 of the inputs comes from this same board, specifically the 2.4 GHz of splitter 1, while the second input comes from PMS 2, with a frequency of 2.401 GHz. The mixer gets the difference of both signals as the output, obtaining the required 1 MHz signal. This signal is amplified (LNA 2) before entering the mixer due to the possible power losses it may have experienced. Hereafter the mixer, the signal is again amplified (LNA 3) and passes through a low frequency filter (LPF 3) to remove noise. This 1 MHz signal can be added or not to the REE_BEE output from splitter 2, decision implemented by the switch 2. Reliability diagram It is important to note that the reliability diagram represents the probability of success of the entire operation of the board (both Pilot Tone and 1 MHz signal, T control and 3.3V power supply). As explained above, these functions require direct intervention by the similar board on the PMS 2. The PMS interconnection can make the reliability diagram complex to set up. This is why the reliability diagrams of the PCB functionalities explained previously are made independently, and then the diagram of the whole PCB will be constructed. On the one hand, several devices are involved in obtaining the Pilot Tone signal. Depending on the position of Switch 1, the signal will come from either PMS 1 or PMS 2 as shown in Figure 3.4. It is important to stress that one of them will be set up at 2.4 GHz and the other at 2.401 GHz, so this can not be considered a redundant path, as it would not be possible to get the Pilot Tone signal from the port set to 2.401 GHz. However, port B of both phasemeters shall be set to exactly Pilot Tone. Therefore, although this is not a desired path due to the signal degradation, port B is a redundant path on both boards from a reliability point of view. The reliability diagram to get the Pilot Tone signal is represented in Figure 3.4. As in both options the signal is obtained with the same component configuration due to the similarity of the boards, this functionality is studied as if the switch were activated to get the signal from PMS 1. LINEAR REGULATOR TELEMETRY CONTROLLER LPF 1 SPLITTER 1 SWITCH 1 POWER DIVIDER 1 POWER DIVIDER 2 SWITCH 2 LNA 1 LPF 2 SPLITTER 2 PMS 1 LINEAR REGULATOR TELEMETRY CONTROLLER LPF 1 SPLITTER 1 PMS 2 PMS2 _PT_in SWITCH 3 Figure 3.4 CFDS Pilot Tone reliability diagram. Note that the interaction of the PMS 2 is displayed in the diagram just to show the same components that would be involved if the signal comes from PMS 2, but they are not considered in the reliability diagram of this function. On the other hand, for procuring the 1 MHz signal in the mixer, two incoming signals are involved, one from PMS 1 and other from PMS 2. As both signals are necessary to get the desired output 32 Chapter 3. Results in the mixer, all the implicated elements are arranged in series in the reliability diagram. The components that perform after the mixer to get the 1 MHz signal have also been included building the reliability block diagram in Figure 3.5 of this function. LINEAR REGULATOR TELEMETRY CONTROLLER SPLITTER 1 PMS 1 PMS 2 PMS2 _Mixer_in ADDER SWITCH 3 MIXER LNA 3 LPF 3 LNA 2 PMS 1 LPF 4 LPF 1 LINEAR REGULATOR TELEMETRY CONTROLLER SPLITTER 1 LPF 1 Figure 3.5 CFDS 1 MHz reliability diagram. Where the symbol * represents the serial continuation of the reliability diagram. Based on this fragmented study, the reliability diagram of the whole CFDS is presented in Figure 3.6 LINEAR REGULATOR TELEMETRY CONTROLLER LPF 1 SPLITTER 1 SWITCH 1 POWER DIVIDER 1 POWER DIVIDER 2 SWITCH 2LNA 1 LPF 2 SPLITTER 2 PMS 1 LINEAR REGULATOR TELEMETRY CONTROLLER LPF 1 SPLITTER 1 PMS 2 PMS2 _Mixer_in ADDER SWITCH 3 MIXER LNA 3 LPF 3LNA 2 PMS 1 LPF 4 CFDS PSM DHLC Tº SENSOR Figure 3.6 CFDS reliability diagram. 3.2 Thermal results 33 Reliability data The reliability results for the basis diagram presented in Figure 3.6 are detailed in Table 3.2. Table 3.2 CFDS Reliability calculations. BLOCK DEVICE FIT RELIABILITY BLOCK RELIABILITY Linear regulator 4.29 0.999736972 0.99965115 Telemetry controller 1.4 0.999914156 75 MHz OPTION LPF 1 85.216 0.994788184 0.98729689 1 Splitter 1 96.572 0.994095704 Switch 1 26.7 0.998364096 Power divider 1 0 1 Power divider 2 0 1 - 0 1 1 DISTRIBUTION Switch 2 26.7 0.998364096 0.995251041 LPF 2 10.68 0.999345317 LNA 1 0.25 0.99998467 Splitter 2 40 0.997550206 PMS 2 CHAIN Linear regulator 4.29 0.999736972 0.988569677 Telemetry controller 1.4 0.999914156 LPF 1 85.216 0.994788184 Splitter 1 96.572 0.994095704 1 MHz CHAIN LNA 2 1.269 0.999922188 0.995954218 Mixer 3.953 0.999757631 LNA 3 0.25 0.99998467 LPF 3 23.01 0.998590022 Adder 0.25 0.99998467 LPF 4 10.68 0.999345317 Switch 3 26.7 0.998364096 TºSENSOR T Sensor 2.8 0.999828319 0.999828319 Yielding a preliminary reliability result for this board of TOTAL CFDS RELIABILITY 97.93 % 3.2 Thermal results As mentioned in Section 2.2, every electronic device is modelled as Figure 2.2, where the internal heat is characterized by its power consumption. For that reason, the first step to carry out the thermal analysis is to make a power budget for every device’s consumption. The obtained values are listed in Annex A.3. 3.2.1 Stationary results This section focuses on performing stationary analysis to find the temperature of the components and the associated heat maps. Most of the module optimisation process will take place here, modifying parameters such as components layout in the board and PCB features. In addition, the process of constructing the geometry used for the analysis is detailed, as well as the decisions made to place each component in a specific location. Firstly, it is necessary to make an assembly consisting of the PCB and each particular electronic component. For this purpose, a part was created for each device according to the dimensions of its package provided in each of the datasheets, simplifying the physical component details and 34 Chapter 3. Results constructing a prism by taking the outermost dimensions of the devices. Regarding the PCB, its size is imposed by the enclosure dimensions, and a preliminary thickness of 1.65 mm has been assumed as it is a standard value in PCB manufacturing. Note that only the active devices have been considered for this model. Regarding the devices distribution on the PCB, the driving factor of the CFDS layout is the sensitivity of the Pilot Tone signal to temperature gradients. Since there is no information about this at first, a preliminary arrangement very similar to that associated with the functional diagram in Figure 3.3 is adopted, whose result of which can be seen in the Figure 3.7. This will also help in the identification of each component and the Pilot Tone signal path, marked in yellow in the figure. At all times, an approximate spacing has been considered for the remaining electronic components which are not included in this analysis. The starting layout of the devices in the VFDS was done considering the needs of each device. One of the main priorities is the correct functioning of the crystals due to their criticality, and for this reason they must be at the lowest possible temperature. The results from the thermal studies of the enclosure developed at the AEI [ 23 ], yields the conclusion that the base of the enclosure is proved to be the coldest area. That is why the crystals are placed as close as possible to the base of the enclosure. On the other hand, the regulator has a high power consumption, so it was also placed near the base to dissipate more heat and near the power supply input connector. Finally, several aspects have been considered for the arrangement of the buffers. Firstly, the signal path in the PCB and the distribution of the connectors. Secondly, as they are elements that also heat up considerably, they should be located as far apart as possible from each other and close to areas with a greater amount of metal. This distribution is shown in Figure 3.8. Table 3.3 FDS Devices Temperature. VFDS Devices Temperature [°C] CFDS Devices Temperature [°C] USO BUFFER 48.29 Telemetry controller 59.49 PPS BUFFER 51.49 Switch 1 55.09 Driver 42.66 Power divider 1 60.07 80 MHz Buffer 50.65 Power divider 2 60.92 T°controller 42.49 Switch2 55.65 Linear regulator 50.14 Linear regulator 58.71 LVDS receiver N 39.67 LNA 1 57.83 OPAM N 40.05 T°controller 51.80 Crystal 1 39.212 LNA 2 56.52 LVDS receiver R 39.37 Mixer 54.47 Crystal 2 38.61 LNA 3 55.94 OPAM R 41.35 Adder 58.31 - - Switch 3 57.34 After an initial analysis, Table 3.3 lists the temperature results for the Baseline distribution, and Figure 3.9 depicts its thermal map, where the isotherms lines have been included. The temperature of the VFDS components is in a reasonable range, especially that of the crystals, and no additional considerations are necessary. However, CFDS is clearly subject to improvement, primarily seeking that the line that follows the signal crosses the minimum possible temperature gradients, the effect can be minimized simply by changing the distribution of the components. 3.2 Thermal results 35 TELEMETRY CONTROLLER SWITCH 1 SWITCH 2 SWITCH 3 POWER DIVIDER 1 POWER DIVIDER 2 LNA 1 LNA 2 LNA 3 ADDER T° CONTROLLER MIXER LINEAR REGULATOR Figure 3.7 CFDS Device layout. With this aim, different layouts are studied with the purpose of finding one that can minimize this effect as much as possible. It should be noted that this will not be searched for the 1 MHz chain, as this signal is not as sensitive. In the first modification, the aim was to approximate the components of the Pilot Tone chain, seeking to form a more uniform area where the temperature is more regular. The thermal map is depicted in Figure 3.10. To check if there is any improvement due to the influence of the components of the 1 MHz chain on the isotherms of this PCB, these components have been separated in the second optimisation shown in Figure 3.11, maintaining the position of those of the Pilot Tone chain. For the third optimisation, a different arrangement is tested. In this layout, the 80 MHz signal will enter from the bottom right of the CFDS, running across the PCB in the opposite direction to that shown in the baseline 3.7. In addition, the regulator has now been positioned in the top left corner of the PCB. This means that the elements that consume the most power and therefore dissipate the most heat are now as far from each other as possible and close to the areas with the most aluminium. This result can be found in Figure 3.12. Finally, in optimisations four and five, the Pilot Tone signal path is approximated to each other to see if, in this case, a smaller temperature gradient is generated in the area. The thermal distributions of each distribution are presented in 3.13 and 3.14 respectively. It is also possible to detect the hottest spots from every analysis, being them placed in the power dividers. This is mainly caused by the elevated value of power consumed by them as well as a high θJC value. 42 Chapter 3. Results Figure 3.15 Pilot Tone measurement points. Figure 3.16 Pilot Tone measurement points transient response. In the crystals behaviour Figure 3.17, it can be noticed the second-order response similarity to the input function due to the fact that it is very close to the base, where the change of temperature is applied. As it was mentioned previously, the proper functioning of quartz crystals is determined by the operating temperature, which should remain as cold as possible. The highest temperatures for this device could be reached then at the peak of this second-order function. Nevertheless, the graph shows that the response of the crystals is even softer than the input signal, so the peak is not as pronounced. Furthermore, it is proven in this graph that the maximum temperature to which it can be exposed is still low. 3.2 Thermal results 43 Figure 3.17 Crystals transient response. It should be remembered that this input function is the experimental function used in this laboratory, which has been obtained by applying a change of five degrees using a vacuum chamber. However, in reality, the evolution of this five-degree temperature tolerance may be completely different. For additional information, the transient response of the surface of each PCB has been included in Figure 3.18, proving again how the areas furthest from the base have a generally less pronounced result, and then confirming that the layout of both PCBs is appropriate. 3.2.3 Structural optimisation In order to further optimise the module, an adjustment of the enclosure’s shape is carried out. The enclosure can be modified by adding or removing aluminium or even changing its shape to improve the heat dissipation in some areas. However, the thermal improvement is meagre in relation to the weight added. In pursuit of an overall improvement of the PMS, aluminium mass is removed to reduce the weight of the mission. For this purpose, mainly part of the internal ribs of the module have been removed, as the outer part of the module is exposed to more structural stresses and needs to be studied in depth to make significant changes to it. However, it is true that when placing the components, there is some excess space in the direction perpendicular to the plate. For this reason, the height of the module has also been reduced by five millimetres, from 50 mm to 45 mm in the outer area and from 20 mm to 15 mm in the area of the internal ribs. The final shape of the enclosure can be found in Figure 3.19. 44 Chapter 3. Results Figure 3.18 PCBs transient response. Figure 3.19 Optimised enclosure design. The volume decrease from an initial value of 6.5707·10−4m3 to a final volume of 5.3984·10−4m3 . These changes lead to a volume reduction of 17.84%, which can be translated into an aluminium mass saving of 316.521 grams. Structural analysis In the optimisation of the enclosure, it has been proposed to remove internal ribs and base volume to reduce weight. This motivates a structural analysis, as introduced in section 2.3. For the study of the new architecture a new COMSOL physics module was used, Structural Mechanics, in particular 3.2 Thermal results 45 the Solid Mechanics option. The first step for the implementation in COMSOL has been to build the assembly. To get the structural response of this module, the complete assembly of the PMS has been built with the optimised module shown in Figure 3.19, superimposing seven units simulating the rest of the modules. The result of this assembly is illustrated in Figure 3.20. Although the rest of the modules may not have the same internal structure given the needs and geometry of each module, seven identical FDS design units have been arranged. The other modules are more complex and hotter, and therefore generally have a more robust internal structure. The safety factor is being increased by studying the worst case by doing the study only with the module with the lower internal reinforcement. In this assembly, a 2.5mm top and bottom cover and 2.5mm thick triangular stiffeners have also been added to the structure to achieve greater structural integrity. The final appearance of the structure is shown in Figure 3.20. Figure 3.20 PMS assembly structure. Neither the CHOTHERM nor the PCBs had been considered for this analysis since they have no structural function, thus having an assembly made of aluminium solids only. Detailed structural analyses of static, shock, thermal and vibration loads will be done in the future when the development is consolidated. The implementation of the boundary conditions is then considered. Regarding the system’s displacements conditions, the shifting in the perpendicular direction to the TRF surface is impeded, as mentioned in Section 2.3. To impose this restriction the ’Prescribed Displacement’ option has been used in COMSOL, which allows to limit the movement on a single axis. In the applied case, the displacements of the base surfaces in the z-axis are restricted. In addition, as the screws are presupposed to be attached, the movement of the inner area of the screws is restricted in the assembly. This is implemented in COMSOL with the ’FIXED Constraint’ option in the inner cylinder surface of the screw hole. The influenced surfaces by these boundary conditions are illustrated in Figure 3.21. 46 Chapter 3. Results Figure 3.21 Boundary conditions on displacements. To implement the 5G force in COMSOL, the ’Volumetric forces’ option has been used, as it is not a point force or a force that acts on a surface of the solid but on each point of the solid. In particular, the ’Gravity’ option has been used, as it allows the force to be written as a function of gravity. Regarding the meshing, the generic tetrahedral mesh has been used again as shown in Figure 3.22. For the analysis of the assembly, given its more extensive volume, the size of the tetrahedral mesh can be replaced by a normal size by selecting the option ’Normal’. The number of elements generated amounts to 409436 elements, which is a reasonable number for the size of this structure. Figure 3.22 PMS assembly mesh. The analysis applying the 5G load in the different axes yields the results listed in Table 3.6. These values verify that the assembly’s maximum stresses do not exceed the aluminium’s yield strength σyield Al =276 ·106Pa in none of the axes. Therefore, it is demonstrated that the proposed enclosure meets the preliminary structural requirements, allowing the this optimised design to be adopted as the new internal configuration of the module. Table 3.6 Structural results. Load direction Max Stress[N/m2] Max displacement [m] X2.88 ·1061.75 ·10−5 Y1.54 ·1064.12 ·10−6 Z1.79 ·1062.09 ·10−6 3.3 Final results 47 3.3 Final results After replacing the enclosure by the optimised one, a final thermal analysis is carried out. The obtained thermal distribution is presented in Figure 3.23, which clearly confirms that the Pilot Tone signal will not be exposed to a high temperature gradient on its path along the PCB, avoiding the coupling noise effects. Figure 3.23 Final thermal distribution. Table 3.7 Final working temperatures. DEVICES TEMPERATURE [oC] CFDS VFDS Linear regulator 55.59 USO Buffer 46.94 Telemetry controller 56.58 PPS Buffer 48.45 Switch 1 54.63 LVDS Driver 41.91 Power divider 1 58.05 80 MHz Buffer 47.14 Power divider 2 57.23 T°Controller 43.14 Switch 2 53.83 Linear regulator 43.5 LNA 1 55.38 LVDS Receiver N 39.85 T°Controller 52.64 OPAM N 39.85 LNA 2 56.6 Crystal 1 38.63 Mixer 53.56 LVDS Receiver R 40.2 LNA 3 54.72 Crystal 2 39.67 Adder 54.99 OPAM R 40.15 Switch 3 54.13 - - Moreover, the working temperature data from this analysis is reported in Table 3.7. Comparing these results with the ones obtained in the initial distribution, listed in Table 3.3, it can be checked that the operating temperature of all the devices has been reduced. This is of great importance for devices such as the telemetry controller and linear regulator in the CFDS, which are the ones that 48 Chapter 3. Results consume more power; in those there has been a decrease of approximately 3 degrees in the operating temperature. In the case of the VFDS linear regulator, this reduction has been almost five degrees. Lastly, the reliability calculations based on the working temperature of each component are presented for VFDS and CFDS in Tables 3.9 and 3.10 respectively, where the data has been obtained by adjusting the FIT with the acceleration factor caused by the effect of the temperature from Equation 2.9. Table 3.8 summarises the yielded results of the updated model and the ones obtained in the preliminary reliability analysis, showing an improvement in the reliability of 0.03% for the VFDS and 0.05% for the CFDS. Table 3.8 Final reliability results. PCB Final Reliability % Preliminary reliability % VFDS 99.88 99.85 CFDS 97.98 97.93 Table 3.9 VFDS Final reliability calculations. BLOCK DEVICE FIT NEW FIT RELIABILITY RELIABILITY BLOCK USO BUFFER USO Buffer 2.8 1.50175 0.999828319 0.999828319 PPS BUFFER PPS Buffer 2.8 1.69131 0.999828319 0.999828319 VCXO NOMINAL OPAM N 1.17 0.35306 0.999928258 0.99878851 1 Diode 0.59375 0.59375 0.999963592 Crystal N 18 4.9072 0.998896849 Transistor 1 0.0025 0.0025 0.999999847 Transistor 2 0.0025 0.0025 0.999999847 VCXO REDUNDANT OPAM R 1.17 0.3619 0.999928258 0.99747864 Crystal R 40 11.8897 0.997550206 RECEIVER LVDS Receiver 5.4 1.62951 0.999668927 0.99966893 1 80MHz BUFFER LVDS Driver 2.9 1.03679 0.999822188 0.999650537 80 MHz Buffer 2.8 1.52564 0.999828319 REGULATOR Linear regulator 4.29 1.7463 0.999736972 0.999736972 T°SENSOR T°Sensor 2.8 1.10645 0.999828319 0.999828319 3.3 Final results 49 Table 3.10 CFDS Final reliability calculations. BLOCK DEVICE FIT NEW FIT RELIABILITY BLOCK RELIABILITY Linear regulator 4.29 4.48598 0.999724958 0.999628362 Telemetry controller 1.4 1.57578 0.999903378 PT OPTION LPF 1 85.216 85.216 0.994788184 0.98734 1 Splitter 1 96.572 96.572 0.994095704 Switch 1 26.7 25.9581 0.998409518 Power divider 1 0 0.01 0.999999387 Power divider 2 0 0.01 0.999999387 - 0 0 1 1 DISTRIBUTION Switch 2 26.7 24.4365 0.998502676 0.995388752 LPF 2 10.68 10.68 0.999345317 LNA 1 0.25 0.25717 0.999984231 Splitter 2 40 40 0.997550206 PMS 2 CHAIN Linear regulator 4.29 4.48598 0.999724958 0.988547142 Telemetry controller 1.4 1.57578 0.999903378 LPF 1 85.216 85.216 0.994788184 Splitter 1 96.572 96.572 0.994095704 1 MHz CHAIN LNA 2 1.269 1.43115 0.999912246 0.99607329 Mixer 3.953 3.54566 0.999782604 LNA 3 0.25 0.2447 0.999984995 LPF 3 23.01 23.01 0.998590022 Adder 0.25 0.24986 0.999984679 LPF 4 10.68 10.68 0.999345317 Switch 3 26.7 25.001 0.998468111 T°SENSOR T°Sensor 2.8 1.10645 0.999828319 0.999828319 4 Conclusions and future work 4.1 Conclusions The obtained results yield some clear conclusions. First of all, very favorable reliability results have been obtained for both PCBs, this being a value adjusted to reality by considering the effect of heating on each component. This allows to deduce that a second redundant FDS would not be necessary, proposing to use the enclosure space reserved for it for other functions, thus saving mass and cost. Besides, the achieved reliability improvement is very low. This is due to the fact that none of the devices had a high failure rate, whose decrease in the working temperature could lead to a significant improve on the reliability of the whole PCB. However, a decrease up to 5 degrees has been achieve for some of the devices. Although it does not have a relevant impact on reliability, it is a minor source of heating that can influence the rest of the components. Furthermore, an optimal and justified positioning of the quartz crystals, which play a crucial role in the operation of the plate, has been achieved, as well as ensuring that the temperature gradients couples minimally to the Pilot Tone electronic circuitry. This is in addition to a mass reduction of the enclosure of 17.84%, without affecting the performance of both PCBs. 4.2 Future work Once the analysis developed in this thesis has been presented, the main line of work proposed is to carry out the same process for the rest of the modules that form the PMS: DHLC, PSM and BEE. This would lead to realistic reliability results for the whole PMS, allowing to check whether the PMS complies with the requirements imposed by the mission. Further, the physical architecture of the PMS can be defined based on this result, including redundant modules needed to meet the reliability requirements imposed by design or, on the contrary, eliminating them. Therefore, the PMS reliability result becomes an essential design factor for the mission given the close relationship between module redundancy with a more electrical complexity due to a higher number of connections, increased weight, more money investment or difficulties in programming to activate or deactivate modules that are in cold redundancy. Another important aspect be to improve is the uniformity in the method of obtaining the failure rate. As a requirement of the mission, these data shall be provided by the manufacturer. However, 51 58 Bibliography [15] B Foucher, J Boullie, B Meslet, and Digu Das. “A review of reliability prediction methods for electronic devices”. Microelectronics reliability, 42(8), 2002. [16] US Department of Defense. MIL-HDBK-217F-Notice 2. Reliability Prediction of electronic equipment., 1995. [17] Diogo Coutinho. PMS Reliability Assessment. Albert Einstein Institute, 2021. [18] James Li. “Reliability comparative evaluation of active redundancy vs. standby redundancy”. International Journal of Mathematical, Engineering and Management Sciences, 1(3), 2016. [19] Jeff Watson and Gustavo Castro. “High-temperature electronics pose design and reliability challenges”. Analog Dialogue, 46(2), 2012. [20] V Lakshminarayanan and N Sriraam. The effect of temperature on the reliability of electronic components. In 2014 IEEE international conference on electronics, computing and communication technologies (CONECCT). IEEE, 2014. [21] Pablo Martinez Cano. Thermomechanical design and analysis of the LISA Phase Measurement System. Master’s thesis, Luleå University of Technology, 2019. [22] MatWeb. 95% alumina, vitreous body. https://matweb.com/search/ datasheet.aspx? MatGUID=ed1c6a34dfaf4f6cad9f0d32e782bf7d, . [Accessed 2022-04-16]. [23] Richard Johansson. LISA’s Phasemeter Structural-Thermal Model test campaign and computational model calibration. Master’s thesis, Luleå University of Technology, 2022. [24] MatWeb. Aluminium 6061-t6. https:// www.matweb.com/search/ datasheet_print.aspx? matguid=1b8c06d0ca7c456694c7777d9e10be5b, . [Accessed 2022-05-10]. [25] Chomerics. CHO-THERM ® 1671, thermally conductive elastomer insulators. https:// www. spacematdb.com/ spacemat/ manudatasheets/ cho-therm1671.pdf . [Accessed 2022-05.05]. [26] ECSS Secretariat. ECSS-Q-ST-70-12C. Space Product Assurance. Design of printed circuit board, 1999. [27] US Department of Defense. MIL-STD-1835D. Electronic component case outlines., 2004.