The CubeSat is a type of small satellite, initially conceived reduce the cost access to space to universities due to its small and standardised $\SI{10x10x10}{\centi\metre}$ cubic form factor. The total number of CubeSats launched into space is growing exponentially due to their low cost, doubling every $\SI{2.5}{\year}$, however the mission success rate has not increased significantly since 2018, levelling off at 75\%\cite{welle2020overview,bouwmeester2022improving}.
Vibration and shock tests are industry standard procedures which aim to emulate launch conditions, however they cannot perfectly replicate them \cite{gordon2015benefits}. Testing of CubeSats on suborbital high-power rockets (HPR) is a novel qualification method that can potentially replicate launch conditions more accurately than traditional shaker table tests, and therefore better detect issues and improve the likelihood of mission success. While there have been tests of university CubeSats on high-power rockets \cite{slongo2019pre}, there are no direct comparisons to shaker table tests to evaluate their effectiveness as a qualification method.
This paper outlines the construction of a data acquisition system to obtain acceleration data from the HPR launch, the HPR launch and vibration table tests and finally makes a direct comparison of the vibration environment on the HPR launch and vibration table.
I'd like to thank all the people and organisations who have supported me throughout this project. Dilusha Silva for being a wonderful mentor and for coordinating the project. Michal Zawierta for his expertise flying drones for the drone tests of the CubeSat. Jamir Khan for being a wonderful friend and engineer who worked on the mechanical side of this project, including construction of the high-power rocket, and for putting up with all my delays. Timothy Ludovico for designing the camera payload and being all around wonderful to work with. Jeremy Marelich and AVI for providing their shaker table facilities and conducting the tests. UWA Aerospace for being a wonderful institution who has been with me from first year through my growth as an engineer and has supported me through this project. Space Angel for creating this project and providing expertise and connections to the Indian Institute of Space Science and Technology (IIST). Dr. Priyadarshnam Hari and the Indian Institute of Space Science and Technology for providing their launch expertise and opportunity to launch on POEM. International Space Centre for supporting this project with funding.
% The marking is not based on sections right? Some of my design sections has stuff which is related to results like the actual tests used (modified from the original recommendations due to limitations of our machines), but I did not want to split this to prevent confusion.
% It might be better to ask for two weeks in a row, i doubt they would give basically 4 weeks. not sure if i can take the risk this close to the due date.
% What would be the best time or way to get some feedback, is it possible to have two meetings next week since it will probably be the last before submitting?
% Introduction or Background This provides the reader with the context of the project. For example, what is the application area, why is it important, what (in general terms) has been done before?
The University of Western Australia (UWA) Microelectronics Research Group (MRG) is developing a 2U CubeSat to measure the health of vegetation through an infrared camera array \cite{ludovico2024}. The CubeSat is a type of small satellite designed to reduce the cost of access to space for universities and space startups due to its small and standardised $\SI{10x10x10}{\centi\metre}$ cube form factor. This CubeSat will launch on an Indian Polar Satellite Launch Vehicle (PSLV) in the PSLV Orbital Experiment module (POEM), which will host multiple CubeSats in orbit and will provide services including power and communications to the CubeSat.
The total number of CubeSats launched into space is growing exponentially due to their low cost, doubling every $\SI{2.5}{\year}$, however the mission success rate has not increased significantly since 2018, levelling off at 75\%\cite{welle2020overview,bouwmeester2022improving}, which implies a need for novel qualification methods. For most single-launch satellites, increased testing is the optimal strategy to minimise failure \cite{bouwmeester2022improving}. Qualification of the CubeSat is required to maximise mission success and is required by the launch provider to minimise the risk of damage to the launch vehicle or other payloads. The MRG is planning to qualify this CubeSat on a suborbital high-power rocket (HPR) in combination with traditional vibration and shock tests on a single degree of freedom (SDOF) electrodynamic shaker table.
Vibration and shock testing are typical tests for CubeSats which are intended to replicate the conditions of launch \cite{welle2020overview}. Despite their widespread use, SDOF vibration and shock tests do not perfectly replicate the conditions at launch as \cite{gordon2015benefits,nath2022study}:
\item The peak flight responses are not able to be achieved since a vibration table cannot simulate steady-state thrust forces since they only can simulate dynamic forces \cite{gordon2015benefits}.
\item A SDOF test can only excite one axis at a time which is not representative of the launch environment \cite{gordon2015benefits,nath2022study}.
\item A vibration table tests a "fixed-base" case which has different modes compared to the case where the satellite is fixed to the launch vehicle \cite{gordon2015benefits}.
A HPR has a higher total impulse than model rockets but a lower impulse than sounding rockets, with a range of \SI{36}{\newton\second} up to \SI{163840}{\newton\second}, and have a sub-orbital trajectory unlike COTS launch vehicles \cite{pierce2019development}. Suborbital rockets have been used for testing several CubeSats \cite{9316404,minelli2019mobile}, however this qualification method is not in widespread use in the industry.
% Problem Identification What is the problem that you are trying to solve, or the hypothesis that you are intending to test? What is your intended contribution to the state of the art?
For institutions with limited budget, shock and random vibration tests using a SDOF vibration table is the current state of the art (SOTA) method for qualification. HPRs are a potential qualification method which can complement SDOF vibration tests, however there is no prior studies comparing both HPRs and SDOF vibration tests against the qualification level set by the launch provider. If HPRs can produce a vibration environment similar to the qualification level, HPRs may be a useful complement to SDOF vibration tests and may be useful in increasing mission success rates.
Since HPRs have not been frequently used as a test platform, another issue is the lack of tooling for making HPRs an effective test platform. This research will involve design and evaluation of a combined test and data acquisition platform which:
\begin{enumerate}
\item Measures the vibration response of the rocket on the CubeSat required for evaluating the HPR platform and
\item Provides the same power and communications services as the POEM to ensure the payload-under-test has access to the same environment as on launch.
% Literature Review or Previous Work Explain the literature (e.g. refereed research papers) or previous body of work (e.g. previous projects within the research group) on which your investigation is based. This should not simply be a linear account, but rather a synthesis of what is important from what has gone before. It will often be a hierarchical account, moving from a general understanding of the field, to identification and expansion of work that is specifically relevant to your project
This literature review will cover the current testing methods used in CubeSats, the use of suborbital rockets as a qualification method and cover the types of sensors and systems required to record these tests.
Satellites undergo a panel of qualification tests to maximise the chance of mission success, and may be required by the launch provider to demonstrate that there is minimal risk of the satellite to the launch vehicle and other payloads which may be present. There are multiple satellite qualification standards, an example is the NASA General Environmental Verification Specification (GEVS) which is a panel of tests including electromagnetic compatibility (EMC), thermal, acoustic and vibration tests that are required for all NASA Goddard Space Flight Center projects \cite{nasa-gevs}. Other standards include ISO-15864, JERG-2-002, NASA-STD-7002A, ECSS-E-ST-10-03C and SMC-S-01 \cite{cho2012overview}. While these standards have flight heritage, being used on many successful payloads, they were designed for medium or large satellites, and therefore fully complying with these standards are out of the budget of most university CubeSat programs \cite{cho2012overview}. While is no widely used test standard for CubeSats currently in use, since most CubeSat projects perform the minimum panel of tests required by the launch provider to minimise cost, there is a de facto minimum series of tests which are random vibration, shock and thermal vacuum testing \cite{welle2020overview}.
\subsection{Vibration}
Vibrations are experienced by satellites during transportation and loading, and most prominently during launch \cite{brown_elements_2002}. The purpose of vibration testing is to ensure that the satellite will survive transportation and launch conditions, and to find workmanship errors \cite{brown_elements_2002,gordon2015benefits}.
\subsubsection{Random vibration / sine sweep vibration test}
In the random vibration test, a uniform vibration spectrum is applied to the satellite which tests all the resonant frequencies of the satellite \cite{nieto2019cubesat}. This range includes frequencies on the magnitude of $\SI{100}{\hertz}$, since higher frequencies couple to the satellite through acoustic means rather than through the structure \cite{gordon2015benefits}. A sine sweep vibration test is similar, but instead of the frequency being randomly sampled it is swept through sequentially from either low to high frequency or vice versa. An example of a random vibration test is shown in figure \ref{fig:random}, where frequencies up to $\SI{100}{\hertz}$ were evaluated, and higher frequencies above $\SI{100}{\hertz}$ were attenuated proportional to frequency.
\caption{Random vibration test \cite{nieto2019cubesat}}
\label{fig:random}
\end{figure}
The limitations of random vibration tests is that the shaker and table will have different modes than the launch vehicle and payload mount, resulting in the test response not perfectly matching the flight response \cite{gordon2015benefits,aglietti2019spacecraft}. Gordon and Kern argue that this difference is not a factor in practice since shaker tests are "not intended to be a strength test" \cite[p.~7]{gordon2015benefits} and that components "should have been strength qualified prior to integration" \cite[p.~7]{gordon2015benefits}. Component level is argued as a best practice in the CubeSat community \cite{rawsonbest}, however some argue that component level testing is not suited to the short timeline of university CubeSat projects and that more effort should be put into integration testing \cite{decker2016systems}. If a testing program focuses on integration testing, then this mismatch between shaker table and flight response could result in the CubeSat not being properly qualified.
Finally, although 6 degrees of freedom (DOF) vibration tables exist which can replicate the vibrations experienced in all dimensions during launch, most satellites are still tested with single-axis or random input shakers which only provide one dimension \cite{gordon2015benefits,aglietti2019spacecraft,nath2022study}. While Gordon and Kern \cite{gordon2015benefits} state that these limitations are adequately managed by testing in all three orthogonal axes separately, Aglietti and Nath \cite{nath2022study} created a model of three, two and single axis vibration tests and found that to match the 3 DOF response with a single DOF table, the satellite needed to be subjected to 2.5 times the $g_\text{rms}$ forces than in 3 DOF testing, leading to the satellite being over designed \cite{nath2022study}.
\subsubsection{Quasi-static acceleration test (QAT)}
A quasi-static test replicates the liftoff stage of flight, where there is a combination of random vibration from engines and quasi-static axial acceleration from the engine and other external forces on the launch vehicle \cite{nieto2019cubesat,brown_elements_2002}, which are approximated as constant forces at selected frequencies as shown in figure \ref{fig:qatforces}. The QAT is usually compared to results from coupled loads analysis, where all forces are assumed to be applied to the satellite through the launch vehicle as shown in figure \ref{fig:cla}\cite{dickens2001coupled}.
\caption{Quasi-static acceleration test. The input profile high acceleration from $\SI{20}{\hertz}$ to $\SI{21}{\hertz}$, resulting in the response having a force of 10.8 \textit{g} acceleration around this frequency \cite{nieto2019cubesat}}
\caption{Coupled loads model \cite{dickens2001coupled}}
\label{fig:cla}
\end{figure}
The first limitation of a quasi-static acceleration test is that the shaker table cannot apply the peak response evenly on the CubeSat that is predicted by coupled loads analysis (CLA) \cite{gordon2015benefits}. Again, Gordon and Kern state that these limitations are addressed by component-level strength qualification. They also state that applying the peak response evenly is not necessary, since if an item does not fail, the correctly applying the response evenly does not greatly increase its likelihood of failing \cite{gordon2015benefits}.
The second limitation is there is a difference in modes, since a quasi-static acceleration test also contains random vibrations \cite{gordon2015benefits}.
\subsection{Vibroacoustic testing}
As stated, low frequency vibrations from $\SI{0}{\hertz}$ to $\SI{100}{\hertz}$ tend to couple well through the payload mount, however high frequency vibrations above $\SI{100}{\hertz}$ are more efficiently imparted on the satellite acoustically \cite{gordon2015benefits}. These acoustic loads have an effect on payload electronics \cite{casalino2012rocket}, and primarily originate from the highly turbulent engine exhaust \cite{casalino2012rocket}.
Vibroacoustic testing is not necessary for CubeSats due to their small surface area \cite{nasa-gevs}, since the magnitude of the acoustic response is proportional to the satellite's surface area to mass ratio \cite{brown_elements_2002}, therefore the effect of the acoustic loads is negligible. Instead, vibroacoustic testing is more relevant for large and light payloads such as solar panel arrays \cite{brown_elements_2002}, therefore it will not be part of this research.
\subsection{Shock}
Shock is experienced by satellites when pyrotechnics are detonated or deflagrated during events such as staging and ignition, the response appears as a range of decaying sinusoids in the $\SI{100}{\hertz}$ to $\SI{10}{\kilo\hertz}$ frequency range \cite{brown_elements_2002}, which decay in $\SI{5}{\milli\second}$ to $\SI{15}{\milli\second}$\cite{brown_elements_2002}. The spectrum extends up to $\SI{40}{\kilo\hertz}$, however for analysis frequencies above $\SI{10}{\kilo\hertz}$ are assumed to be non-damaging \cite{bement1995manual,nasa-pyroshock}. Pyroshock may cause peak accelerations of up to 10000 \textit{g}\cite{nasa-pyroshock}. High explosives are primarily used for explosive elements on rockets in combination with some low explosives for initiators \cite{bement1995manual}.
Shock is tested using a shock-generating device which is applied to the satellite along all three axes \cite{nasa-gevs,nasa-pyroshock}, the shock generating device for a CubeSat can be an electrodynamic shaker table \cite{nieto2019cubesat} with a half-sine, pulse profile \cite{nieto2019cubesat}. The shock test has similar limitations as the random vibration test, since it also uses a vibration table to affect the satellite.
Shock tests are compared using the shock response spectrum (SRS), which plots the maximum acceleration per frequency bin. The SRS contains an octave slope which rises to the first resonant frequency called the "knee frequency". The octave slope can be approximately 9 dB/octave to 12 dB/octave depending on distance to the source.
Sounding rockets are a class of suborbital rocket used between $\SI{40}{\kilo\metre}$ and $\SI{200}{\kilo\metre}$, above where weather balloons operate \cite{seibert2006history}. While sounding rockets have been used to launch many CubeSats as the primary launch vehicle for suborbital CubeSat missions, such as in the REXUS-25 mission \cite{pont2019rexus}, there has been only one published instance of sounding rockets being used as an additional qualification platform for a CubeSat \cite{slongo2019pre}. The FloripaSat-I CubeSat was tested on a VSB-30 sounding rocket \cite{slongo2019pre} to qualify the CubeSat under launch conditions. This qualification method was intended not to replace, but to complement standard vibration and shock qualification methods \cite{slongo2019pre}. The test measured these launch conditions through the MPU6050 6 DOF inertial measurement unit (IMU) \cite{slongo2019pre}.
\caption{Acceleration in time domain (Left), Angular velocity in time domain (Right) during the launch of FloripaSat-I \cite{9316404}}
\label{fig:accel-rot}
\end{figure}
While this study does show the time-domain accelerometer and gyroscope measurements from the sounding rocket launch in figure \ref{fig:accel-rot}, it does not compare the data to other qualification tests in the FloripaSat-I campaign, such as traditional vibration and shock testing. Additionally, the launch data was not presented in the frequency domain through the boost and coast phases of the flight, meaning they could not be compared to the acceleration spectra which was shown for the shaker table testing in figure \ref{fig:shaker}.
Another shortcoming of the study is that a shock test using a half-sine pulse was not performed. The use of a sounding rocket is a potential method of qualifying the CubeSat's ability to tolerate shocks since there will be shock events when pyrotechnics are lit to stage the rocket, although the forces will have intensity than on a larger launch vehicle.
While sounding rockets have a significantly lower cost compared to an orbital-class launch vehicle, they cost \$1 million USD per launch to launch $\SI{200}{\kilo\gram}$ on average \cite{jurist2009commercial}, resulting in a specific cost of \$5000 USD/kg, which is still a large amount for university CubeSat programs. High-power rockets (HPR) are a lower-performance but cheaper alternative to sounding rockets, which can leverage the design expertise of university rocketry teams while having similar qualification potential as sounding rockets. A single stage level 3 certification rocket can reach altitudes above $\SI{10000}{\feet}$\cite{canepa2005modern} for a cost of only \$1200 USD \cite{canepa2005modern}. Despite the potential cost benefits, there have not been any published instances of a HPR being used to qualify a CubeSat.
\begin{itemize}% TODO: FIND SOURCES FOR THIS SECTION
\item Boost phase: The HPR is being powered by a solid rocket motor. In most HPR launches, this phase only lasts several seconds at maximum.
\item Coast phase: After the rocket motor burns out and produces no thrust the rocket coasts up on a ballistic trajectory to the maximum altitude (the apogee).
\item Apogee: This is the maximum altitude the rocket will reach. At this point the drogue parachute is deployed, which limits the rocket's descent velocity to a reasonable rate %TODO: WHat?
\item Main parachute deployment: At a fixed altitude above ground level the main parachute is deployed. This parachute has a higher surface area than the drogue chute and slows the rocket down to a safe landing velocity. A main parachute should not be deployed at apogee since this would result in the rocket drifting further which complicates recovery efforts.
\item Landing: The rocket lands on the ground and is recovered by the rocketry team for safing (disarming of live energetics) and transportation. While the landing occurs minutes after launch, finding the rocket is a harder task and may occur hours after landing.
One potential issue with HPRs as a qualification platform for shock is that low explosive black powder is used \cite{canepa2005modern} which has different explosive characteristics, such as a subsonic flame front, compared to the high-explosives used in launch vehicles \cite{bement1995manual} and will therefore produce different shock responses. One study \cite{wang2023numerical} performed finite element analysis of igniters filled with low explosives including aluminium potassium perchlorate and boron potassium nitrate and determined the SRS, shown in figure \ref{fig:lowsrs}. Compared to the SRS of high-explosives in figure \ref{fig:pyroshock}, where at a frequency of 1 kHz the acceleration is over $10^2$\textit{g}\cite{nasa-pyroshock}, in these low explosive simulations the acceleration at 1 kHz is only $10^1$\textit{g}\cite{wang2023numerical}. Therefore, it is hypothesised that HPRs will not be useful for shock qualification since the response of low explosives is different from the high explosives used on launch vehicles.
\caption{Shock response spectrum from computer modelling of an igniter based on the low explosive aluminium potassium perchlorate \cite{wang2023numerical}}
This project involved the design of an electronic system to support the testing of a camera payload and to evaluate the usefulness of a HPR as a vibration testing platform. This section outlines in detail the process, shown in figure \ref{fig:design-process-hl}, used to create two designs: an initial design and a final design.
The ultimate goal of the CubeSat project is to launch on the POEM and receive at least one image from orbit. The POEM will remain in low Earth orbit (LEO) for 6 months \cite{jagdale2023sanket}.
Before launch the CubeSat must be tested on the ground to reduce the risk of mission failure. A novel HPR testing method will be performed at two launches (one private and one at the Australian Universities Rocket Competition (AURC) 2024), in addition to traditional shaker table testing methods. To pass testing, the camera payload should capture at least one image on a preliminary drone test flight, the HPR launch and during vibration testing.
\item Since part of the testing campaign uses a novel HPR test flight which has not been compared to standard shaker table testing, the second part of this project is to design an experiment and a data acquisition system (DAQ) to evaluate the effectiveness of this testing method.
\item Design and build an "emulation platform" as part of the DAQ which provides the same services to the camera payload as provided by POEM. The goal of this to reduce the risk of integration failure. This emulation platform will need to emulate most of the specifications of POEM, but some systems will be designed to a minimum standard required for the tests.
The CubeSat design group was made of Peter Tanner, Jamir Khan and Timothy Ludovico. As shown in figure \ref{fig:cubesat-responsibilities}, each person is working on a unique part of the CubeSat and requires specific information to be communicated. The scheduling of the project is agreed upon at this stage, since each part of the CubeSat requires integration before tests.
The dimensions of the DAQ were to be negotiated to fit it inside the CubeSat. While the camera payload should be built as if it were connecting to POEM, some aspects such as the data rate, format and connector pinouts had to be negotiated.
After defining the goals for the project, a list of high-level requirements and constraints were created. A summary is shown in table \ref{tabl:high-level-requirements}.
\hypertarget{req-E1}{\textbf{E1}}& Emulation power and voltage & EPS must emulate a \SI{5}{V} bus with at least \SI{3}{A} current capacity. \\\hline
\hypertarget{req-E2}{\textbf{E2}}& Battery life & DAQ system must sustain camera payload operation for at least \SI{2}{\hour}. \\\hline
\hypertarget{req-P1}{\textbf{P1}}& Shock and vibration & DAQ must pass shock, random vibration, and sine-sweep tests as described in \secref{sec:shaker-table-test}. \\\hline
\hypertarget{req-P2}{\textbf{P2}}& Hot and cold operation & DAQ must pass temperature qualification range of \SIrange{-20}{80}{\degreeCelsius}. \\\hline
\hypertarget{req-P1}{\textbf{P1}}& Physical dimensions & DAQ must fit within 1 CubeSat unit. \\\hline
\hypertarget{req-R1}{\textbf{R1}}& GNSS tracking & Must have GNSS tracking suitable for a \SI{10000}{\feet} HPR launch. \\\hline
\hypertarget{req-R2}{\textbf{R2}}& Radio downlink & Radio link must be stable to receive one image from a \SI{10000}{\feet} HPR launch. \\\hline
\hypertarget{req-V1}{\textbf{V1}}& Testing controlled environment & A direct comparison of the two methods requires the construction of the satellite to be controlled. \\\hline
\hypertarget{req-V2}{\textbf{V2}}& Vibration Sampling Rate & DAQ must sample vibrations at a high frequency ($>\SI{1}{\kilo\hertz}$). \\\hline
\hypertarget{req-V3}{\textbf{V3}}& Maximum Acceleration & DAQ must sample vibrations with a high acceleration range ($>\SI{200}{\gacc}$). \\\hline
\hypertarget{req-A1}{\textbf{A1}}& AURC Regulations & Must comply with AURC 2024 rules to perform second launch on UWA Aerospace competition rocket. \\\hline
\hypertarget{req-A2}{\textbf{A2}}& Budget & DAQ system cost must be below \aud 1500. \\\hline
\hypertarget{req-A3}{\textbf{A3}}& Project Deadline & System and experiments to be finalised by October 18, 2024. \\\hline
A requirement of the EPS is to emulate the voltage and current provided by the POEM to one CubeSat to facilitate testing of the camera payload's power electronics. POEM contains a \SI{28}{\volt} and \SI{42}{\volt} bus, however IIST has informed the design team that a \SI{5}{\volt} connection with a maximum current of \SI{1.5}{\ampere} is provided to CubeSats, with the option to expand to \SI{3}{\ampere}. The EPS will have to emulate at least one of these power busses to validate the camera's power system.
POEM outputs a consistent amount of power to each CubeSat while on orbit due to its solar panels and battery system, however in the testing it will not be possible to deploy a solar panel therefore the DAQ system must have adequate battery life to power the camera payload throughout the length of the HPR test. To account for setup times, this should be a minimum of \SI{2}{\hour}.
If this research is continued, it is possible a future version of this payload will fly with the camera payload on POEM to make a direct comparison between the vibration environment on POEM to the conditions on both a HPR and the shaker table tests. Therefore, the DAQ must go through the same qualification campaign as the camera payload.
The DAQ must remain functional during the vibration environment of the rocket. This means it must pass the IIST recommended qualification procedure, which involves shock, random vibration and sine-sweep tests. These tests are described in more detail in \secref{sec:shaker-table-test}.
The DAQ must be able to survive at temperatures of \SIrange{-20}{80}{\degreeCelsius} as described in \secref{sec:htemp-test-framework} and \secref{sec:ltemp-test-framework}. This will restrict the components able to be used to only those with industrial temperature ranges.
One of the goals of the testing campaign is to receive one image from a drone or HPR flight, to achieve this some systems on POEM need to be implemented in some form. These systems do not need to exactly emulate the systems on POEM since this test is not going to orbit, but they must function to an altitude of at least \SI{10000}{\feet}, which was the intended height of the HPR.
A launch apogee of \SI{10000}{\feet} will result in the rocket drifting away from the launch site, potentially out of sight. Therefore, a GNSS tracking solution is needed to ensure recovery of the CubeSat and rocket.
A radio link needs to be designed to allow reception of at least one image during flight. The link needs to be designed so that data loss still results in a recognisable image.
\subsubsection{Experiment design} To make a direct comparison between the two methods, the physical construction of the CubeSat and accelerometer placement should be controlled.
\subsubsection{Sampling rate} The method used to measure vibration (accelerometer or ADC) used must be able to sample at twice the frequency bandwidth of the tests. This is to avoid aliasing according to the Nyquist criterion. As shown in the literature review, MEMS accelerometers struggle to sample high frequencies however characterising the knee frequency would be sufficient for the experiment, therefore an accelerometer with a range over \SI{1}{\kilo\hertz} would be appropriate.
\subsubsection{Maximum measurable acceleration} Pyroshock events and motor launch are high-\textit{g} events that require accelerometers with measurement scales above these events, otherwise they will saturate at the maximum scale.
As shown in the literature review, MEMS accelerometers struggle to measure high acceleration ranges however characterising the knee frequency would be sufficient for the experiment, therefore an accelerometer with a range over \SI{200}{\gacc} would be appropriate.
\subsubsection{2024 Australian Universities Rocket Competition (AURC) regulations} This payload was intended to fly on the UWA Aerospace rocket \textit{Svengeance} in the AURC 2024 competition, as part of a collaboration with UWA Aerospace. AURC has additional rules for electronics systems, relevant rules include (but not limited to) \cite{ayaa2023specifications}:
\begin{itemize}
\item Lithium-polymer batteries are not allowed (unless using COTS equipment)
\item Connectors must have a positive locking mechanism
\item Electronics must be mounted using rigid fixing methods
The high level requirements and constraints identified in \secref{sec:constraints-and-requirements} determines the high level design and the types of major components to realise the design. Lower level requirements and constraints were created for each class of component or subsystem, since the high-level requirements are quite general.
A high-level design was created, since some subsystems depend on how others are configured. For example, using the batteries in parallel requires additional balancing components to be selected in this stage.
Only the minimum amount of detail was considered at this stage in the design for flexibility. For example, the types of capacitors to be used for power supply decoupling are not part of the high-level design since they can be treated on a subsystem level later in the design process.
Components must be able to be sourced from at least one of the following suppliers, and will not be sourced from other suppliers for reasons including high minimum order quantities or risk of receiving broken or counterfeit products.
JLCPCB is a PCB manufacturing service that will be used to manufacture the PCBs for this project. In addition to manufacturing they provide a service to automatically place and solder components to the PCBs. This component source has advantages over component distributors as they:
JLCPCB cannot be used for expensive components that need fine control over the amount being purchased, since JLCPCB can only provide component assembly for either two full board or five full boards. Additionally, some components cannot be sourced on JLCPCB. This is the case with the chosen GNSS receiver, which has a very high unit price and should not be assembled on all five boards, and is not available in the JLCPCB catalogue.
Mouser and DigiKey are component distributors which allow the purchasing of individual components, with a higher markup compared to JLCPCB. If a part cannot be purchased from JLCPCB for any reason, it will be purchased from Mouser or DigiKey and manually assembled.
Firstly, a power budget was created which accounts for the worst-case current consumption of the payload under test and DAQ components. The power budget is a major requirement for the EPS performance. An example of a power budget, using values for the final design, is shown in table \ref{tabl:eps-power-budget}.
\textbf{Item}&\textbf{Voltage (\si{\volt})}&\textbf{Max unit current (\si{\milli\ampere})}&\textbf{Quantity}&\textbf{Max current (\si{\milli\ampere})}\\
\hline
Payload-under-test & 5 & 3000 & 1 & 3000 \\
Pi Zero & 5 & 1300 \cite{raspberry-pi-hardware-doc}& 1 & 1200 \\
Stable in vacuum & The DAQ system will be vacuum tested, therefore the cells need to be stable under vacuum &\hlreq{P1}\\\hline
Wide temperature range & Batteries must survive temperature testing &\\\hline
Low cost & Due to limited budget, space qualified batteries would be cost prohibitive &\\\hline
High energy density and specific energy & The system needs to support the payload for at least \SI{2}{\hour} at the worst-case power budget without recharging &\\\hline
Reasonable power density and specific power & The overall system needs to be able to provide at least \SI{3}{\ampere} of current in a small space/mass, so the batteries should have similar current capacities. &\\\hline
Must not be a lithium-polymer battery & This is an AURC rule. &\\\hline
Space heritage & Ideally the batteries would be used in space to maximise the chance of passing vacuum and temperature testing. &\\\hline
To create a high-level design, the battery chemistry must be selected to design other parts of the high-level design such as the charging and power conditioning electronics, since these are dependent on the nominal cell voltage. Lithium-ion (\liion) cells were selected since in general they satisfy many of the criteria.
A 2S battery pack would make sense if the electronics operate on a higher voltage (for example, if there are motors or high-power radios). However, in this application the highest voltage used is \SI{5}{\volt} which is not significantly higher than \SI{3.7}{\volt}. Additionally, the $I^2R$ losses between the battery and the power converters is low since they are integrated on the same board. A preliminary analysis shows that a power system using 2S batteries results in a higher cost compared to a 1S system due to the additional battery balancing electronics required. Therefore, a 1S battery pack was selected for the high-level design.
Based on the high level description of the battery pack and the power budget, a boost converter and regulator would be used to condition the battery voltage. These components will have the following requirements:
Input voltage range & It must have an input voltage range of at least $<\SI{2}{\volt}$ and $>\SI{4.4}{\volt}$ to support a \liion battery. & TODO: \\\hline
Output voltage range & It must be able to output \SI{5}{\volt}. & TODO: \\\hline
Output current range & It must be able to output at least \SI{3}{\ampere}. & TODO: \\\hline
High efficiency & A lower efficiency converter would require more batteries and therefore more space and mass to meet the battery life requirement compared to a higher efficiency converter. & TODO: \\\hline
Industrial temperature range & This is a component that must pass temperature testing. & TODO: \\\hline
Dropout voltage & This linear regulator will receive a constant \SI{5}{\volt}, it must have a dropout voltage lower than \SI{1.7}{\volt} so that it can power a \SI{3.3}{\volt} system. & TODO: \\\hline
Output voltage range & It must be able to output \SI{3.3}{\volt}. & TODO: \\\hline
Industrial temperature range & This is a component that must pass temperature testing. & TODO: \\\hline
The POEM contains a radio downlink which allows experiments to transmit data to the ground at a speed of \SI{5}{\kilo\bit\per\second}. The radio cannot be used to control the CubeSats from the ground. Since the HPR test is relatively short, the downlink speed is increased to allow one image to be received over the time span of one flight.
Data rate & Must have adequate data rate to transmit one image over the time of a \SI{10000}{\feet} HPR test flight & TODO: \\\hline
Range & Must have adequate link budget to allow reception of picture data over \SI{10000}{\feet}. & TODO: \\\hline
Common interfaces & Must use a non-proprietary interface, such as UART or SPI to simplify development. & TODO: \\\hline
USB ground station & Must be able to be connected to a laptop on the ground for live visualisation of picture data. & TODO: \\\hline
Class license operation & Radio must be able to be legally operated using a free class license \cite{australia2015radiocommunications}, since obtaining an amateur radio license would require significant time. & TODO: \\\hline
Industrial temperature range & This is a component that must pass temperature testing. & TODO: \\\hline
Note that this radio transceiver will not be used on the space launch, since POEM will provide a radio downlink, but it must pass environmental testing since wireless communication is the only method of monitoring the system during environmental testing.
For the payload communications, a physical layer and data layer need to be specified. The physical layer determines the representation of data as (in this case) electric signals. The data layer determines how data is transmitted and received as frames.
CubeSats use the RS-485 physical layer specification and UART data layer specification to transmit data to POEM at a maximum speed of \SI{5}{\kilo\bit\per\second}. Therefore, this system has no choice but to emulate the physical and data layer specification used on POEM to satisfy design goals.
RS-485 was chosen since it is a differential bus which increases immunity to electromagnetic interference from other elements on the launch vehicle, therefore reducing the bit error rate \cite{cratere2024board}. A combination of RS-485 and UART allows large amounts of data to be transferred since the UART data link specification is simple, especially compared to other protocols such as CAN bus \cite{cratere2024board}.
Since the DAQ needs to emulate POEM services as a constraint, payload communications had to use the combination of RS-485 and UART. The Pi Zero has one hardware UART, this single-ended UART signal needs to be converted to the differential RS-485 signal using an RS-485 transceiver. The SP3485 transceiver was chosen due its low unit cost of \aud 0.32 and high availability on JLCPCB. The transceiver supports data rates up to \SI{10}{\mega\bit\per\second}, which is more than adequate for the \SI{5}{\kilo\bit\per\second} downlink speed. It uses a \SI{3.3}{\volt} logic level on the single-ended side, which is required to interface with the Pi Zero without level shifting \cite{maxlinear2021sp3485}. The EN-L variant was chosen due to its extended temperature range of \SIrange{-40}{85}{\degreeCelsius} required to pass environmental testing \cite{maxlinear2021sp3485}.
The DAQ must be able to track the HPR throughout the full launch to enable recovery as stated in \secref{sec:hpr-test-req}. This will be achieved through a Global Navigation Satellite System (GNSS) receiver, which receives signals from GNSS satellites and determines the position and altitude of the receiver.
Micro-electromechanical systems (MEMS) based accelerometers were chosen for the DAQ due to their low cost and low power consumption compared to traditional piezoelectric accelerometers. %TODO:
High output data rate & Output data rate should be over \SI{1}{\kilo\hertz} to maximise range of the shock response spectrum. & TODO: \\\hline
High acceleration range & Maximum measurable acceleration should be over \SI{200}{\gacc} to maximise range of the shock response spectrum. & TODO: \\\hline
SPI with FIFO and hardware interrupt & To reduce load on the processor SPI should be used over other interfaces (such as \iic), and a FIFO queue with interrupt should be available. & TODO: \\\hline
PCB with adequate mounting points & The accelerometers need to be mounted to a PCB with a strong mounting point to maximise the resonant frequency. & TODO: \\\hline
Industrial temperature range & This is a component that must pass temperature testing. & TODO: \\\hline
The OBDH unit acquires data from sensors and the payload and saves it to a storage device and relays relevant data to the ground station through the radio. The two main requirements of the OBDH is that it has enough resources to be able to process and save the sensor and payload data without losing data, and that it has adequate storage to hold sensor data from one HPR flight.
Data storage & OBDH should be able to store several images for post-flight analysis if the radio downlink underperforms. It should be able to store readings read from the accelerometer at a high rate. & TODO: \\\hline
SPI and UART interfaces & These interfaces are requirements for the GNSS receiver, payload under test and the accelerometers. & TODO: \\\hline
Adequate processor speed & The OBDH must not "drop" image frames or data samples due to inadequate processing speed. & TODO: \\\hline
Industrial temperature range & This is a component that must pass temperature testing. & TODO: \\\hline
Altium Designer 24 (Altium) is an electronics design automation (EDA) tool which is widely used in industry and has been used for design of CubeSat and space hardware \cite{10061409}.
The author chose to use Altium over other EDA tools since they were familiar with this tool having used it in previous projects, which minimises development time.
\subsubsection{Schematic editor}
\label{sec:schematic-editor}
A circuit is first implemented using schematic symbol representations of components in the schematic editor. In the schematic view the connections between the components are abstracted using net labels and wires. The schematic view does not necessarily represent the physical layout of the PCB but is intended to convey the connections between components in a format that can easily be read.
A root schematic contains references to other schematics which are abstracted as sheet symbols with ports. Each sheet symbol represents a particular subsystem of the DAQ. The hierarchical sheet symbol representation has several benefits, including that it facilitates reuse of designs and allows larger systems to be decomposed into multiple schematics which are easier to modify and read. This is shown in figure \ref{fig:altium-schematic-hierarchical}.
\subsubsection{PCB editor}
\label{sec:pcb-editor}
After a design is implemented using schematics, its physical layout defined in the PCB editor.
A full "net list" of connections between each symbol in the schematic representation is transferred to the PCB editor. Each symbol is linked to a footprint, which is the physical representation of the component and contains information such as
\begin{itemize}
\item The land pattern, which is the layout of pads or holes required for mounting the component on the PCB,
\item The component's 3d model
\end{itemize}
The PCB editor contains automated design rule check (DRC) tools which is used in the design process to reduce the likelihood of a faulty PCB. The DRC uses rules set in the project and if a rule is violated, it is reported.
This feature is used to ensure components have adequate clearance, geometry such as vias and tracks have manufacturable dimensions, and to control the impedance of microwave frequency tracks.
From a design process standpoint, this reduces the risk of receiving PCBs with design issues and increases the likelihood of meeting project budget constraints.
Once a PCB is ready to be manufactured, an automated "outjob" ensures that the required design files are automatically generated with the right settings for manufacturing. From a design process standpoint, this reduces the risk of receiving boards with incorrect specifications due to incorrect export settings.
The artifacts generated include:
\begin{itemize}
\item Bill of materials
\item Gerber files
\item Drill location files
\item Pick-and-place component locations
\end{itemize}
\subsubsection{PCB project and version control system (VCS)}
The schematics, PCBs and output jobs are integrated into an Altium PCB project, which is versioned using the Git version control system (VCS) and pushed to a remote Git server on every commit. A VCS is
\subsection{circuit.js}
Circuit.js is a simple browser-based analog circuit simulator \cite{falstad22falstad}. Circuits in this simulator can be edited and interacted with in real-time, whereas in traditional SPICE simulators the circuit cannot be edited once the simulation starts. Circuit.js uses a numerical method which is prone to error however, therefore this simulator was used for rapid, real-time prototyping of designs. After these designs were finalised they were simulated in traditional SPICE-based simulators.
\subsection{LTspice}
The simulation of components is done using LTspice, a freeware circuit simulator which uses the SPICE method
LTspice was used to the DC-DC boost converter for this project, which was required to power the internal DAQ systems and the payload. A simulation was performed to characterise the ripple voltage and to validate its performance over a range of input voltages. LTspice has been used for simulation of boost converters in the past and is free which makes it a suitable circuit simulator for this project \cite{giesselmann2019modeling}.
Ultimately LTspice was chosen over other freeware SPICE simulators such as PSpice since LTspice contains an "alternate" solver which has less error at the trade-off of simulation time \cite{ltspice2022}. The reduced error results in the solver converging on a solution, whereas in PSpice or in LTspice normal solver mode it was not able to converge on a solution and the simulation could not be completed.
% TODO:
% \subsection{SolidWorks 2023}
% SolidWorks is a mechanical CAD software which is used for creating 3d models of the electronics hardware by using the Altium Designer plugin. These 3d models are required for Jamir to complete the mechanical design of the CubeSat and to verify good mounting of the electronic hardware.
After the high-level design was created and components finalised, they are integrated into the design as shown in figure \ref{fig:implementation-workflow} using the design tools from \secref{sec:design-tools}
\subsection{Design verification and schematic capture}
The high-level design shown in \ref{fig:system-block-diagram} is first separated into multiple subsystems, this practice is useful since it allows for design reuse, improves readability of the project and allows different verification strategies to be used for different subsystems.
\item Subsystems based around single ICs, for example power converters, are implemented by using the process outlined in the application note or by copying a reference design or evaluation board design if it exactly matches the application in the DAQ. As an additional precaution, switching converters are simulated in a SPICE simulator to verify the output is stable over the range of input voltages and output currents.
\item Discrete components are used to implement some simple circuits to save on part cost. These are prototyped in circuit.js, then once working in circuit.js are simulated in a SPICE simulator to verify their functionality.
\item Some subsystems have already been verified on other projects and are able to be directly applied to the DAQ system, the schematics for these designs are copied into the project.
Once verified, a subsystem is implemented as a single schematic sheet in an Altium PCB project and connected to other schematics using the hierarchical sheet system as described in \secref{sec:schematic-editor}.
PCB rules are set according to the JLCPCB capabilities page. This is set according to the board manufacturing process used and ensures that the board does not contain geometry which cannot be manufactured or assembled.
The dimensions of the PCB and position of connectors are negotiated with the rest of the design team to ensure it will fit in the CubeSat. The maximum size is negotiated so that there is as much room possible for simpler routing.
Each PCB has a stackup defined, which sets the number of metal layers in the PCB, the thickness of the board, the insulating substrate used and the type of routing on each layer. For example, a standard PCB stackup would have four copper layers (signal-ground-power-signal) with a 7628 substrate. Parameters which determine what stackup to use include cost (since large or less common stackups have additional fees), the number of nets and components required to route and the RF performance of the substrate.
\item Using the manufacturer's sample layout and PCB layout guidelines, if they exist
\item Placing the majority of components on one side to allow cheaper PCB component assembly, and placing large or manually soldered components such as connectors on the opposite side
After the PCB design is finished, an automatic design rule check (DRC) is run to find any errors that would affect manufacturing, based on the PCB rules set in \secref{sec:pcb-layout}. The manufacturing outjob is run to generate artefacts such as Gerber files, the bill of materials for automated and manual manufacturing, and component locations for pick-and-place. These artefacts are sent to JLCPCB to design and partially assemble the PCBs. The manual manufacturing BOM is used to purchase components from component distributors.
After receiving the boards from JLCPCB, some basic tests are conducted (such as ensuring that voltage domains and ground are not short-circuited). After this, additional components are manually assembled either using hot-air or a soldering iron. After soldering, the manual solder joints are inspected to ensure they are not cold joints, and the boards are again tested for short-circuits.
The Raspbian OS is flashed on to an SD card with settings to allow the Pi Zero to connect to a local Wi-Fi network. An external computer is used to connect to the Pi's secure shell (\ssh) server and log into the Pi. After this, the USB Ethernet gadget is configured on the Pi and host computer sides which allows the host computer to \ssh into Pi without needing a Wi-Fi network, which was useful for field debugging.
A combination of Python scripts and C programs were used for different parts of the DAQ. Python is used for tasks which do not require many system calls and do not require high optimisation, such as transferring data from the payload to the radio. The advantage of Python for these tasks is speed of development. Some tasks, such as reading data from the accelerometers, require many system calls and has considerable performance impact when using Python. For these tasks, a program is written in C and compiled on the Pi Zero. C has less overhead compared to Python as it is a compiled directly to ARM assembly, whereas Python is interpreted to an intermediate representation which is executed by the Python virtual machine which adds overhead.
A Samba server was set up on the Pi Zero to allow code developed from a PC to be copied over to the Pi, and to allow an integrated development environment on the PC to access libraries and headers present on the Pi Zero.
To ensure the DAQ performs at a basic level with the camera payload before conducting any large tests, the systems were integrated, and the following tests were performed.
As stated in \ref{sec:environmental-requirements}, it is possible the DAQ will fly with the CubeSat on the PSLV in future work, therefore there are environmental requirements for part of the DAQ system.
These environmental tests will evaluate the resilience of the DAQ in the environment of space and during launch.
IIST recommends a qualification test where the CubeSat placed in a thermal vacuum chamber for $\SI{2.5}{\hour}$ and is heated to $\SI{70}{\degreeCelsius}$. The CubeSat electronics are turned on and tested during the final $\SI{30}{\minute}$ of the test.
Due to time restrictions it was only possible to do a preliminary high-temperature test with a consumer oven on only the electronics section of the payload (the combined camera and DAQ assembly).
IIST recommends a qualification test where the CubeSat is placed into a thermal vacuum chamber for $\SI{2.5}{\hour}$ and is cooled to $\SI{-20}{\degreeCelsius}$. The CubeSat electronics are turned on and tested during the final $\SI{30}{\minute}$ of the test.
Due to time restrictions it was only possible to do a preliminary low-temperature test with a consumer freezer. To prevent condensation from developing on the electronics during the test, which would not occur in the thermal vacuum chamber, the electronics were placed in an airtight bag prior to the test and pressurised with pure nitrogen gas for $\SI{5}{\minute}$ to displace air containing moisture.
IIST recommends that the CubeSat be mechanically qualified using a single-axis electrodynamic shaker table using random vibration, sine-sweep and half-sine shock tests.
\caption{IIST recommended random vibration test profile for qualification of CubeSat for launch on POEM (profile defined in \ref{tabl:random-vibration-profile-iist}).}
Drone tests were used as a qualification platform for the HPR launch since drone tests:
\begin{enumerate}
\item Use the expertise of the UWA Aviation Laboratory, which is participating in the project.
\item Are repeatable, whereas the rocket test can only feasibly be done once per launch season.
\item Have greater control over the position compared to the suborbital rocket launch and will better qualify the machine vision algorithms.
\end{enumerate}
The drone test evaluates the communications between the camera payload and the communications downlink stability in real time. A successful test involves receiving at least one frame from the camera payload at a reasonable quality.
\subsubsection{High-power rocket test flight}
The high-power rocket (HPR) test flight is used as an experimental qualification method for the CubeSat. This DAQ system is used to evaluate the effectiveness of the HPR flight by using accelerometers, but the HPR flight also serves as a milestone for evaluating the effectiveness of this DAQ for this type of application.
One of the objectives of this research is to design a platform for qualification of CubeSats. The first revision of the qualification platform was not used in the final design due to several issues:
\begin{itemize}
\item The STM32L476 did not have enough resources to move data from the sensors and camera payload to the payload at an adequate speed. A benchmark using CrystalDiskMark, in figure \ref{tabl:daq-v1-diskmark} shows that the maximum throughput is $\SI{0.84}{\mega\byte\per\second}$, and while only 60\% of the throughput is being used as shown in \ref{tabl:daq-v1-sensor-datarate}, between reading from the data sources and writing to the storage there is not enough resources in practice to do this at an adequate speed, resulting in the maximum sampling rate of the sensors to be limited.
\item Due to space limitations on the rocket, it was not possible to have two redundant systems. The next revision would use only one DAQ.
\item By the end of this section, it was understood that centimetre level positioning was not required to obtain good results from the camera system.
\item At the end of this revision it was concluded that the STM32 platform was not flexible enough to complete the research objectives in time.
LSM6\-DSOX &$\SI{0.41}{\mega\byte\per\second}$&$16$ byte structs are generated at $\SI{6664}{\hertz}$ for both acceleration and gyroscope data for two sensors. \\
ADXL\-375 &$\SI{0.038}{\mega\byte\per\second}$&$20$ byte structs generated at $\SI{1}{\kilo\hertz}$ for two sensors. \\
Three PCB projects are created for the three boards in the final design:
\begin{itemize}
\item Power and OBDH (referred to as DAQ)
\item GNSS receiver
\item Accelerometers
\end{itemize}
As stated, PCB rules are set according to the JLCPCB capabilities page.
\subsection{PCB sizes}
A PCB size of \SI{80 x 80}{\milli\metre} was used for the boards since this is the maximum size available for the PCB inside the CubeSat. The DAQ board uses this size since the batteries are mounted on the reverse side of the board, and require a lot of board area to mount. The GNSS receiver board uses this size since larger PCB areas result in higher gain for the patch antennas used for receiving GNSS signals. For the MEMS accelerometers, a PCB of \SI{22 x 22}{\milli\metre} was chosen as it is the minimum size possible to fit both accelerometers and mounting holes, and minimising the PCB area maximises the resonant frequency of the accelerometers.
\subsection{PCB stackups}
Four layer stackups are used for the DAQ and GNSS boards since these are cheap to manufacture at JLCPCB for boards under \SI{100x100}{\milli\meter} and they are simpler to route due to more layers. The DAQ uses a standard signal-ground-power-signal stackup. This was chosen since the DAQ board contains many parts and having a dedicated power and ground plane makes routing simpler and improves the EMI performance of the PCB. The GNSS board contains few components and uses a stackup of four ground layers which cover the full PCB to maximise the gain of the patch antennas. The accelerometer board uses a two-layer signal-signal stackup since there are few components to route and to save cost.
All boards use a 7628 substrate since this is the lowest cost substrate and acceptable since the highest frequencies being routed on the PCBs are the L1 signal for GNSS, which is a relatively low \SI{1575.42}{\mega\hertz}. TODO: Loss tangent
All boards use hot air solder levelling (HASL) lead-tin coating due to its low cost, and since this PCB does not require flat surface finishes since relatively large components are being soldered on to the PCB, and since these boards are not commercial boards that are subject to environmental requirements.
\section{Power electronics}
\subsection{Battery selection}
COTS 18650 lithium-ion batteries were chosen as the energy source for the DAQ. Advantages of this battery format include:
\begin{itemize}
\item The 18650 format encases the battery in a rigid metal cylinder which is well-suited for the space environment \cite{knap2020review}. Battery formats which use a flexible pouch, like most lithium-polymer cells, are more prone to outgassing in the vacuum of space \cite{knap2020review}.
\item Compared to other rechargeable battery solutions such as \ce{Ni}-\ce{Cd} and \ce{Ni}-\ce{H2}, \ce{Li}-ion batteries have improved temperature range, energy density and specific energy and cycle life \cite{pathak2023review}.
\item COTS \liion batteries are a mature battery format due to widespread use in consumer products \cite{pathak2023review}.
\item Extensive flight heritage as they have been proven in other CubeSat missions \cite{knap2020review}, and are being used on flagship NASA missions, such as Europa Clipper \cite{krause2021performance}.
\item Low cost as they are COTS grade and are already produced at scale. % TODO: Citation
\end{itemize}
Manufacturers produce a variety of types of \liion batteries with different chemistries, which affect parameters including internal resistance, discharge and charging temperature range and capacity.
The Samsung INR18650-25R \liion battery was chosen for the DAQ platform due to
\begin{itemize}
\item Previous flight heritage on CubeSats \cite{marcelino2021orbit}.
\item Operation over a large temperature range of \SIrange{-20}{75}{\degreeCelsius} and has been proven to be stable at \SI{130}{\degreeCelsius}\cite{samsung2014}.
\item Good capacity of \SI{2500}{\milli\ampere\hour} and high maximum discharge rate of \SI{20}{\ampere}.
\end{itemize}
\subsection{Payload power conditioning design and simulation}
The Texas Instruments TPS61022 boost converter was selected for this 1S system. It has a working voltage range of \SIrange{1.8}{5.5}{\volt} which is suitable for a 1S3P \liion battery pack with a working voltage range of \SIrange{2.5}{4.2}{\volt}, a maximum output current of over \SI{3}{\ampere} and can be configured to have an output voltage of \SI{5}{\volt} which is ideal for the DAQ and the camera payload \cite{ti2021tps61022}. A DC-DC switching converter was selected since it has high efficiency above 90\% throughout the input voltage of a standard 1S battery pack \cite{ti2021tps61022}. The TPS61022 has an operating temperature range of \SIrange{-40}{150}{\degreeCelsius} which is adequate for thermal qualification \cite{ti2021tps61022}. %TODO: This satisfies the requirements in section blah.
The inductor selected for the converter subsystem is the ASPI-0630HI-1R0M-T15, which is a \SI{1}{\micro\henry} coil with a low DC resistance of \SI{10}{\milli\ohm}. The coil has a saturation current of \SI{20.5}{\ampere} which is above the \SI{8}{\ampere} maximum switching current. To make manufacturing simpler, the voltage divider was changed from $R_1=\SI{732}{\kilo\ohm}$ and $R_2=\SI{100}{\kilo\ohm}$ to use $R_1=\SI{750}{\kilo\ohm}$ and $R_2=\SI{100}{\kilo\ohm}+\SI{2}{\kilo\ohm}$ since these resistor values are available on JLCPCB. A \SI{1}{\pico\farad} compensation capacitor was also added. To evaluate the effects of these choices, the converter was simulated in LTspice with a single fully charged lithium-ion battery as the input.
As shown in figure \ref{fig:TPS61022-simulation-ltspice}, during startup there is a small voltage overshoot of \SI{0.06}{\volt} but this is negligible and well within the tolerance of the components. The steady-state output voltage has a mean of \SI{5.006}{\volt} and a peak to peak ripple voltage of \SI{7}{\milli\volt}, or a ripple of 0.15\%, which is more than adequate for this application since it is mainly digital logic.
The simulation also shows that the power system will work up to \SI{3}{\ampere}, which is the output current required to emulate the power capabilities of POEM. Two TPS61022 boost converters will be used in the final design: one will be dedicated for the payload-under-test, and the other will power the DAQ system. This allows the payload to be able to receive the maximum current able to be provided by POEM, and also isolates failures in either system from affecting the other.
As stated in the high-level design (\Secref{sec:hl-power-electronics}), a 1S lithium-ion battery pack was selected. In the final design, three batteries were placed in parallel to form a 1S3P battery pack. Three 18650 cells were selected since this is the maximum number of cells which can fit on a \SI{80 x 80}{\milli\metre} PCB using a 3 cell 18650 battery holder.
The battery pack can be charged by connecting \SI{5}{\volt} power to a charging terminal block. Two \SI{1}{\ampere} TP4056 linear \liion battery chargers were connected in parallel to the battery pack to allow a maximum charging current of \SI{2}{\ampere}.
The DAQ internally uses \SI{3.3}{\volt} for some components, however this system only uses a maximum of \SI{157}{\milli\ampere} therefore efficiency is a lower factor and a linear regulator was chosen. This solution results in a power loss of \SI{270}{\milli\watt}, which was considered insignificant compared to the maximum power output of the \SI{5}{\volt} system (\SI{26}{\watt}). %TODO:
An Advanced Monolithic Systems AMS1117-3.3 linear regulator was chosen due to its cheap pricing on JLCPCB of only \aud 0.20 and since it has been used in past designs with success. It has a high dropout voltage of \SI{1.1}{\volt}, which is acceptable for the \SI{5}{\volt} input \cite{ams2007ams1117}. The AMS1117 has an operating temperature range of \SIrange{-40}{125}{\degreeCelsius} which is adequate for thermal qualification \cite{ams2007ams1117}.
This regulator has been validated in previous boards, therefore implementing this subsystem only required cloning the schematic and layout to the design.
A single schematic was created for the TPS61022 based off the successful LTspice simulation and instantiated twice in the root schematic. A PCB layout based on Figure 10-1 from the datasheet was used and is shown in figure \ref{fig:TPS61022-layout}\cite{ti2021tps61022}.
A first revision of the DAQ used an STM32L476 microcontroller, which had similar peripherals, including UART, {\iic} and SPI, however as a microcontroller it does not have an operating system and does not have significant storage. Storage was provided in the form of a \SI{4}{\giga\byte} embedded multimedia card (eMMC) chip, which was chosen since it is directly soldered to the PCB and should be more resistant to vibration than a micro SD card connector.
% TODO: talk about other cubesats using the stm32 platform?
Due to the issues encountered with the first revision and due to the time constraints, the second revision uses a Raspberry Pi Zero W v1.3 (referred to as "Pi Zero"). This is a development board which integrates the BCM2835 Broadcom system on chip (SoC), \SI{512}{\mega\byte} of RAM and contains a micro-SD card slot, USB interface and peripherals such as UART, {\iic} and SPI \cite{upton2016raspberry}. This board was chosen due to its small form-factor compared to larger Raspberry Pis, simplicity of integration compared to the Raspberry Pi compute modules and low cost. The Raspberry Pi platform has been used in low-cost low Earth orbit (LEO) CubeSat applications \cite{guertin2022raspberry}. It is predicted that in polar orbits that the Pi Zero has a lifespan of 5 years \cite{guertin2022raspberry}, which is adequate for this project since the POEM will cease to maintain its orbit after months. The thermal performance of a Raspberry Pi Zero W is adequate for space applications \cite{guertin2022raspberry}. While a Raspberry Pi Zero 2W was initially selected, which would have had higher performance compared to the Zero v1.3, it was not possible to acquire one due to supply chain issues.
Typically, a Raspberry Pi runs the Raspbian operating system (OS), which is a Debian fork \cite{upton2016raspberry}. Compared to developing for a microcontroller, an OS is easier to develop for as it allows the use of standard Linux utilities for interacting with the system, including \texttt{ssh} for remote control, and having each DAQ task in a separate process with separated memory makes debugging easier. This ease of development was required to meet the time constraint of the project. %TODO: Citation`'
Due to limitations of the BCM2835 SoC only one hardware UART is available \cite{upton2016raspberry}, but more are required for receiving data from GNSS receivers. Receiving data through a software implementation of UART is not ideal since it uses a large amount of CPU resources and is prone to missing bits especially on the non realtime OS of the Raspberry Pi which severely limits its usable baud rate. The XR20M1172 dual hardware UART to add more UART ports to the Raspberry Pi through the SPI bus. This UART has a 64-byte first-in first-out (FIFO) buffer and interrupts, which eliminates the need for expensive polling, and has a maximum data rate of \SI{16}{\mega\bit\per\second}, which is more than adequate for GNSS data \cite{maxlinear2022xr20m1172} which has a maximum baud rate of only \SI{38400}{\baud}. This UART has a temperature range of \SIrange{-40}{85}{\degreeCelsius}, allowing it to survive temperature testing.
\subsection{Camera communications interface and camera data downlink}
\paragraph{Connector}
The final design of the camera payload interface uses a DE-9 connector. This was chosen due to its low cost and ability to be secured using bolts instead of by friction alone. The pinout of the connector is shown in figure \ref{fig:de-9-connector}. The RS-485, power and ground lines are duplicated for redundancy.
A female DE-9 connector is soldered to the DAQ, a male solder cup DE-9 terminal is soldered to 22 AWG copper cable to form the wire harness to the camera payload.
\paragraph{RS-485 receiver}
The SP3485EN-L/TR RS-485 transceiver is used to convert the RS-485 signalling to single-ended UART. The transceiver is fixed in receive mode since there is no way for the POEM to send data to the payload. The transceiver contains a \SI{120}{\ohm} d.c. termination as this is the end of the RS-485 bus, the resistor is sized to handle the power dissipation of a \SI{5}{\volt} RS-485 bus. The receiver contains a PSM712-LF-T7 transient voltage suppressor (TVS) diode array, which is designed to prevent the transceiver from being damaged by electrostatic discharge (ESD) due to manual handling of the connector.
\paragraph{Transport layer}
The camera payload communicates with the DAQ through UART at the transport layer. The settings in table \ref{tabl:uart-settings} were agreed upon and are used on both the payload and DAQ sides:
Speed &\SI{57600}{\baud}& This is significantly more than the \SI{5000}{\baud} connection provided by POEM, but the drone and HPR tests have limited time and require faster transmission to receive any picture during the test. \\
Data bits & 8 &\\
Stop bits & 1 &\\
Parity & None &\\
Flow control & None & This is a one-way connection, so software flow control is not possible. \\
\hline
\end{tabular}
\caption{UART settings.}
\label{tabl:uart-settings}
\end{table}
\paragraph{Presentation layer}
Several simple formats are used in the presentation layer for this project. The data stream may contain either messages or image data.
Messages consist of a header, the message data of a variable length and an end of message marker as shown in figure \ref{fig:message-bytefield}. The header is sufficiently long and unique that it should not appear in the image data stream. Since messages are not often transmitted, this overhead is acceptable.
A chunk protocol is used for the transmission of images in the final design. A preliminary design simply sequentially transmitted each byte of the image, which was adequate when the radio was placed close to each other ($>\SI{10}{\meter}$). However, when testing past this range and at any significant distance, the errors caused the received image to become unrecognisable. The type of error observed consists of contiguous bytes being lost occasionally, rather than frequent corruption of individual bytes as shown in \ref{fig:image-sheds-error-demo}.
\caption{Example of an image received with errors resulting in an unrecognisable image due to dropped bytes causing the image to shift at each error. Only a small proportion of the image is lost ($>2\%$), represented by red pixels at the bottom and contiguous sections of the received image are error-free.}
A chunk protocol was effective against this type of error while being simple to implement. It involves adding a header to signify the start of a chunk, followed by the chunk number and 400 bytes of pixel data as shown in \ref{fig:image-chunk-bytefield}. A \SI{1600x1200}{} pixel image where each pixel is represented by one byte requires 4800 chunks to be sent in this implementation. The number of bytes per chunk is a trade-off between pixels lost per error and effective data rate.
A start of chunk marker of \texttt{\string{0x00, 0xFF, 0x00, 0xFF\string}} was chosen since the image data being transmitted are raw pixels, and it is unlikely for the image to contain pixels that are full or zero intensity, and even more unlikely for them to be adjacent to each other.
When an error occurs causing a string of bytes to be dropped, it causes the next chunk to not be encountered after 400 bytes. When this error occurs, the receiver searches for the next occurrence of the header and restarts receiving from that point. The chunks which were skipped have their pixels filled in the received image to prevent the error causing the image to become unrecognisable.
% TODO: discuss multiple radio options (lora, sik, etc.)
The RFD900x radio transceiver was chosen to emulate the POEM radio service. This transceiver uses the 915 MHz industrial, scientific and medical (ISM) band and transmits with a maximum power of \SI{1}{\watt} using a frequency hopping spread spectrum (FHSS) technique \cite{rfdesign2020rfd900x}. Data rates from \SI{12}{\kilo\bit\per\second} to \SI{224}{\kilo\bit\per\second} are available with the default firmware \cite{rfdesign2020rfd900x}. The RFD900x uses a UART to transmit and receive data \cite{rfdesign2020rfd900x}.
The RFD900x satisfies several constraints. It reduces the time to test since it uses the ISM band, which can be used by anyone provided they follow the Low Interference Potential Devices (LIPD) Class License legislation. The use of the FHSS allows the RFD900x to transmit at the maximum power of \SI{1}{\watt} that is allowable by the class license under the frequency hopping transmitters section \cite{australia2015radiocommunications}.
Distances of \SI{40}{\kilo\metre} line-of-sight is possible using the RFD900x \cite{rfdesign2020rfd900x}, which is far greater than the maximum distance achievable with the rocket and drone tests. The maximum drone test scheduled had an altitude of \SI{500}{\metre}, and the rocket was intended to fly to \SI{10000}{\feet} (\SI{3}{\kilo\metre}).
% This radio transceiver will not be used on the space launch, since POEM will provide a radio downlink, but it must pass environmental testing since wireless communication is the only method of monitoring the system during environmental testing.
The radio transceiver has a temperature range of \SIrange{-40}{85}{\degreeCelsius}, which satisfies the range of temperatures required to pass the temperature testing \cite{rfdesign2020rfd900x}.
The ground station is a laptop running Windows with an RFD900x connected to it over USB serial. There are two Python scripts on the laptop, the first saves the data from the radio to a file and prints any messages to the console.
\begin{enumerate}
\item Opens a serial port to the RFD900x with the UART settings of the ground RFD900x.
\item Create and open a log file to store the received bytes from the radio.
\item Periodically:
\subitem Read the bytes in the serial buffer,
\subitem Save the bytes to the log file,
\subitem Find any start of message mark, and if encountered print all bytes until the end of message mark is received
\end{enumerate}
A second script periodically reads the latest log file and splits it into images.
\subitem Repeat until the image array is full or there are no more bytes left in the image bytearray.
\end{enumerate}
\paragraph{Results}
%TC:ignore
TODO: This might be better placed in the drone test section, but we'll have to deal with this redundant data.
TODO: Should it be hierarchical by system or by test?
%TC:endignore
This system was tested in the drone test at a distance of \SI{120}{\meter} line of sight. The image received is shown in \ref{fig:image-blocking-example} with comparison to an image received without blocking:
\caption{Comparison of two image reception techniques. Image (a) used the blocking protocol with a block size of 400 pixels. Image (b) is an example of what the image would look like without the image blocking.}
\label{fig:image-blocking-example}
\end{figure}
\subsection{Accelerometer data acquisition system}
The STMicroelectronics LSM6\-DSOX is a MEMS inertial measurement unit (IMU) which has a $\pm\SI{16}{\gacc}$ accelerometer and $\pm\SI{2000}{\milli\degree\per\second}$ gyroscope, both with a sampling rate of \SI{6666}{\kilo\hertz}. This accelerometer is used to characterise the random vibration spectrum of launch due to its high sampling rate.
Due to the low full-scale of the LSM6\-DSOX of only $\pm\SI{16}{\gacc}$, the ADXL\-375 was chosen to characterise the shock response of the payload to pyroshock due to its significantly higher full-scale range of $\pm\SI{200}{\gacc}$.
The accelerometer PCB (shown in figure \ref{fig:accelerometers-pcb}) will contain both the ADXL\-375 and LSM6\-DSOX accelerometers and mount them to the chassis.
To maximise the resonant frequency of the assembly, the PCB is mounted on its flat side to the chassis and is fixed with four steel M3 bolts on each corner. Additionally, the dimensions of the PCB was minimised to \SI{22x22}{\milli\metre}, which is the minimum dimension that can fit all components. Maximising the resonant frequency past the sampling frequency will minimise distortion of the response. The PCB is depicted in figure \ref{fig:accelerometers-pcb}.
The accelerometers are connected to the DAQ using the \texttt{SPI0} bus of the Pi Zero. The chip select (\textoverline{CS}) and interrupt (INT) pins are connected to GPIOs and are handled by software, therefore it is not necessary to only use the hardware SPI chip select/chip enable (CE) pins for all chip selects. This is depicted in figure \ref{fig:accelerometers-sch-block}. Although SPI requires more wiring, in the case of these sensors SPI has a higher clock speed compared to {\iic}. For example, the LSM6DSO has supports the {\iic} Fast Mode+ extension (FM+) which gives the {\iic} bus a maximum clock rate of \SI{1}{\mega\hertz}, but the SPI has a maximum clock speed of \SI{10}{\mega\hertz} which is higher. SPI also has higher data rate for a given frequency than {\iic} since {\iic} adds additional flow control bits such as start, stop and acknowledgement bits.
This SPI bus is dedicated for only accelerometers to maximise the number of accelerometers connected, and is not shared with non-accelerometer peripherals.
\caption{Mounting positions of accelerometers on the CubeSat chassis. The top and bottom accelerometer share a common $x$ axis but have opposite $y$ and $z$ axes.}
\paragraph{LSM6\-DSOX software} A C program is written to read values from the LSM6\-DSOX, package them into a binary flat file format and store them to disk. A C program is required since preliminary tests with a Python script with the same functionality resulted in 100\% of the CPU being used to read from one accelerometer.
The \texttt{lsm6dso-pid} platform independent (PID) library was used to provide simple access to the registers of the accelerometer \cite{stmicroelectronics_lsm6dsox_pid}, in combination with WiringPi which provides bindings to GPIO, SPI and other peripherals on the Pi \cite{wiringpi-github}.
To maximise sampling rate and minimise CPU usage, the accelerometer is operated using the FIFO and interrupt features.
Samples from the accelerometer are stored in binary flat file format, consisting of the \texttt{struct datapoint} for each 3-axis sample. A header of \texttt{0x59} is used which signifies the start of a sample. The type field is set to either 1 for accelerometer samples or 2 for gyroscope samples.
To improve reliability, the LSM6\-DSOX program contains a watchdog, implemented using a POSIX timer, which restarts and re-initialises the accelerometer if no samples have been received for \SI{1}{\second}. The process is managed by a \texttt{systemd} service which initialises the program on system boot and restarts the program if it crashes.
\item Set the sampling rate to \SI{6666}{\hertz} and the batching rate to \SI{12.5}{\hertz},
\item Set the FIFO to continuous mode (old samples are automatically discarded),
\item Set FIFO watermark level to 384 samples,
\item Set interrupt pin 1 to pulse on FIFO watermark being reached (this results in a pulse being generated on the INT1 pin when a large amount of data is present in the FIFO to be read, resulting in the interrupt handler being triggered.),
The interrupt handler has the following functionality and is based off sample code from the LSM6\-DSOX PID library \cite{stmicroelectronics_lsm6dsox_pid}:
\item Immediately reset the watchdog timer to 1 second (to prevent it triggering since a read is started)
\item Read the number of samples in the FIFO. For each sample:
\item Read the sample's tag and if it is accelerometer data then read the sample from the FIFO into a 6-byte array
\item Convert each of the three axes (represented by two byte words) to \SI{}{\milli\gacc}.
\item Store the reading in a struct with the correct header and type and write to the log file.
\item For debug reasons, print the current sample every 3332 samples (every half second).
\end{enumerate}
\paragraph{Unpacking data for data analysis}
The accelerometer binary files are read on a PC using a Python script which uses the \texttt{struct} library to unpack the file and convert it to a traditional comma-separated values (CSV) file.
The u-blox NEO-M9N will be used for tracking \cite{ublox2023neo_m9n_datasheet}. This is a multi-GNSS receiver which is able to receive from multiple GNSS constellations simultaneously, which results in a faster acquisition time and greater interference immunity \cite{ublox2023neo_m9n_datasheet}. The receiver can report position with an accuracy of \SI{2.0}{\metre} (circular error probable), which is adequate for a HPR tracking application \cite{ublox2023neo_m9n_datasheet}. The NEO-M9N was used instead of other u-blox receivers due to its high navigation update rate of \SI{25}{\hertz} which is useful due to the high speed of a HPR flight \cite{ublox2023neo_m9n_datasheet}.
Since POEM provides the location of the CubeSat and due to the speed and altitude restriction of the NEO-M9N of \SI{500}{\metre\per\second} and \SI{80}{\kilo\metre} respectively, this receiver will not be present on the space launch and is only required for the HPR launch.
A differential GNSS (DGNSS) solution was considered and tested based on the u-blox ZED-F9P, however the drone test did not require the centimetre level precision of the ZED-F9P, so it was not used in the final design.
% The second revision of the test and POEM emulation electronics (referred to as DAQ v2) contains several improvements and simplifications over DAQ v1.
% \subsection{On-board data handling (OBDH)}
% A Pi Zero is used for the OBDH system instead of an eMMC module and STM32L476 since:
% \begin{itemize}
% \item It reduces the cost of the PCB as the assembly of BGA packages such as eMMC adds significant cost per board,
% \item The Pi Zero W runs an operating system and can be controlled remotely from a PC unlike the STM32, which simplifies development and debugging,
% \item The write speed of the Pi is larger than the STM32 and eMMC combination. % TODO: benchmark write speed.
% \end{itemize}
% While a Raspberry Pi Zero 2W would be preferable due to its multicore design, due to supply chain issues it was only possible to use a Pi Zero.
% DAQ v2 does not have two redundant OBDH due to a lack of room.
A custom rocket named UNO was designed and built by another project member from scratch, it has a height of 290 cm, diameter of $\SI{16.3}{\centi\metre}$ and a dry mass of $\SI{14.42}{\kilo\gram}$ without a motor. It was designed to fly with an M impulse class motor, however due to changes in United States export regulations it was not possible to obtain this motor in the time of this research, and therefore it was only possible to launch with a K impulse class motor which has about 1/10th of the total impulse of the N motor as shown in table \ref{tabl:impulseclasses}.
The rocket was simulated using OpenRocket \cite{openrocket,niskanen2009}, an open-source simulator which can predict parameters such as stability and acceleration based on empirical methods which use the rocket's shape and basic environment parameters such as constant wind \cite{doi:10.1177/0954410017752730,niskanen2009}. OpenRocket is used to ensure the rocket design is stable throughout launch and flight, which is important to ensuring the CubeSat payload does not become damaged by this qualification method. However, as it uses a simple empirical model of the flight, it was not designed to model the effect of the motor and aerodynamic forces on the vibration environment in the rocket. It also does not simulate pyroshock events, instead modelling parachute deployment events as simple changes in the aerodynamics of the rocket \cite{niskanen2009}.
\subsubsection{Flight profile}
% TODO: Add motot thrust curve or something with more detial.
As shown in \ref{fig:openrocket-k-launch} the rocket reaches an apogee of \SI{413}{\metre} at \SI{9.74}{\second} and the total flight time is \SI{30}{\second}.
As shown in figure \ref{fig:openrocket-k-stability}, the stability is above 2.0 calibres for the coast and launch phase, which is a rule of thumb to ensure the rocket is stable and will not veer off course \cite{canepa2005modern}. The short moment of stability below 2.0 occurs when the rocket reaches apogee, which is not an issue since the parachutes are immediately deployed at this point.
As stated, since OpenRocket does not model the vibration environment in the rocket and models the rocket as one solid body, only the acceleration of the whole rocket can be modelled. Pyroshock events are not modelled by OpenRocket. The launch phase lasts only \SI{1.6}{\second} and has a high average acceleration of \SI{5.77}{\gacc}, as shown in \ref{fig:openrocket-k-acceleration}. During the coast phase, the rocket is decelerated by gravity as expected and after parachute deployment the rocket only has a small deceleration force.
The system will be used for the vibration tests on a shaker table, and the rocket test. The data will be recorded as a time series on the OBDH memory. The time series data will be transformed into the frequency domain since existing studies have presented frequency domain plots to present and analyse the response of the system to a test \cite{nasa-pyroshock,nieto2019cubesat}. For the rocket test, the analysis will be split over the several phases of flight - launch, thrust, coast and parachute deployment events, since the forces involved are different in all of these phases.
In the launch and parachute deployments, where pyrotechnics are ignited, an analysis of the shock response spectrum will be performed. This will involve creating the shock response spectrum for the rocket test and shaker table tests, then comparing the slope up to ~1 kHz. If the rocket test SRS slope is on the same order of magnitude as the gradient found in \cite{wang2023numerical} for other low explosives, and it is less than the slope of the SRS from the shaker table tests, then this will show that rocket testing is not an adequate qualification method for shock.
The coast phase, where the rocket motor has burnt out but is still approaching apogee, will be compared to the random vibration test. The random response spectrum will be compared to the spectrum of the rocket test to check how uniformly distributed the rocket test is.
The boost phase will be compared to the quasi-static acceleration tests on the shaker table. It is expected that the acceleration force on the HPR will be greater than those experienced on the launch vehicle, however the key characteristic - a peak in acceleration over a narrow frequency band - should be the same.
\item Machine a heatsink to connect all chips on the Pi Zero module to the chassis or heatsink to ensure better thermal performance \cite{guertin2022raspberry}.
