Aalto-1, multi-payload CubeSat: In-orbit results and lessons learned
M.R. Mughal, J. Praks, R. Vainio, P. Janhunen, J. Envall, A. Näsilä, P.Oleynik, P. Niemelä, A. Slavinskis, J. Gieseler, N. Jovanovic, B. Riwanto, P. Toivanen, H. Leppinen, T. Tikka, A. Punkkinen, R. Punkkinen, H.-P. Hedman, J.-O. Lill, J.M.K. Slotte
AAalto-1, multi-payload CubeSat: In-orbit results andlessons learned
M. Rizwan Mughal , J. Praks a , R. Vainio b , P. Janhunen c , J. Envall c ,A. N¨asil¨a d , P. Oleynik b , P. Niemel¨a a , A. Slavinskis a,e , J. Gieseler b ,N. Jovanovic a , B. Riwanto a , P. Toivanen c , H. Leppinen a , T. Tikka a ,A. Punkkinen b , R. Punkkinen f , H.-P. Hedman f , J.-O. Lill g , J.M.K. Slotte h a Department of Electronics and Nanoengineering, Aalto University School of ElectricalEngineering, 02150 Espoo, Finland b Department of Physics and Astronomy, University of Turku, 20014 Turku, Finland c Finnish Meteorological Institute, Space and Earth Observation Centre, Helsinki, Finland d VTT Technical Research Centre of Finland Ltd, Espoo, Finland e Tartu Observatory, University of Tartu, Observatooriumi 1, 61602 T˜oravere, Estonia f Department of Future Technologies, University of Turku, 20014 Turku, Finland g Accelerator Laboratory, Turku PET Centre, ˚Abo Akademi University, 20500 Turku,Finland h Physics, Faculty of Science and Technology, ˚Abo Akademi University, 20500 Turku,Finland
Abstract
The in-orbit results and lessons learned of the first Finnish satellite Aalto-1 are briefly presented in this paper. Aalto-1, a three-unit CubeSat which waslaunched in June 2017, performed Aalto Spectral Imager (AaSI), RadiationMonitor (RADMON) and Electrostatic Plasma Brake (EPB) missions. Thesatellite partly fulfilled its mission objectives and allowed to either perform orattempt the experiments. Although attitude control was partially functional,AaSI and RADMON were able to acquire valuable measurements. EPB wassuccessfully commissioned but the tether deployment was not successful.In this paper, we present the intended mission, in-orbit experience in op-erating and troubleshooting the satellite, an overview of experiment results,as well as lessons learned that will be used in future missions.
Keywords:
Aalto-1, CubeSat, In-orbit results, Lessons learned, Aalto M. Rizwan Mughal is also associated with Electrical Engineering Department, Insti-tute of Space Technology, Islamabad, Pakistan, Correspondence: [email protected]
Preprint submitted to Acta Astronautica January 20, 2021 a r X i v : . [ a s t r o - ph . I M ] J a n pectral Imager, Radiation Monitor, Electrostatic Plasma Brake
1. Introduction
There has been a significant increase in the design, development, launchand operation of nano and micro satellites since last two decades. A largenumber countries initiated their space activities and a large number of News-pace companies emerged as an outcome. A number of innovative plat-form subsystems, payloads and missions have been proposed, designed andlaunched by universities and small industry thanks to significantly reductionof development and launch costs [1, 2, 3, 4, 5, 6, 7, 8, 9]. This has been madepossible due to availability of Commercial Off The Shelf (COTS), technologyminiaturization and affordable rides. The CubeSat standard, initially per-ceived for educational purposes only, was defined by Stanford and CaliforniaPolytechnic State Universities in 1999 [10]. Since the launch of first CubeSatin 2003, this standard has revolutionized the space industry by playing anincreasingly important role in technology demonstrations, remote sensing,Earth observation and education [11, 12]. More recently, the CubeSats havestarted to increasingly exploit the scientific and commercial use cases [12, 13].Being small in size, they have transformed the traditional design approachof space systems by providing low-cost access to space [14, 15, 16, 17]. Asingle ride of launch vehicle can carry hundreds of CubeSat-class satellites.Many universities are effectively using CubeSats as hands-on tools to teachthe challenging engineering concepts about the design and development ofcomplex interdisciplinary systems. The launch and operation phase providesa unique learning experience to university teams enabling them to learn es-sential skills in mission design and operations [18]. Now a day’s universityCubeSat missions aim at real science and technology demonstration whilealso ensuring the educational objectives. It is important for CubeSat com-munity to share the knowledge, in orbit experiences, lessons learned andmission details which will consequently help other teams to gain valuableexperience and not repeat the same mistakes.The current small satellite literature lacks the whole life cycle: i.e. allaspects relating to mission planning, design, launch, operations and lessonslearned. The teams either report very specific technical information of thedesign, or come up with mission descriptions and in-orbit results. One canbarely find information in the current literature about complete life cycle2overing a wide range of aspects. In order to provide the CubeSat commu-nity with the sufficient details on complete aspects in terms of technologydevelopment, technology demonstration and key experiences, we present thedesign, development and in-orbit experience of Aalto-1, the first satellite ofAalto University, Finland. We present our findings in two papers: the firstone covering the technology development aspects [19] whereas the presentpaper covers the in-orbit results and lessons learned.
2. Mission overview
Aalto-1, shown in Fig. 1, is a 3U CubeSat designed and developed byAalto university and partner organizations. The spacecraft was launched inJune 2017 and hosted three payloads: AaSI, RADMON and EPB.AaSI is the first hyperspectral imaging system compatible with nanosatel-lites, based on a piezo-actuated tunable Fabry–P´erot Interferometer (FPI)which allows for an unprecedented miniaturization [20]. The instrument fitsin a half of CubeSat unit and, within a few seconds, can acquire spectral im-ages in tens of freely programmable channels. The filter works in the spectralrange of 500–900 nm where each channel is 10–20 nm wide. A 512 × ◦ field of view provides a ground resolution better than 200 mper pixel.RADMON, fitting within 0.4 CubeSat units, is one of the smallest par-ticle detectors, which has proven itself capable of taking scientific measure-ments [21, 22]. It measures electron energies in the > >
10 MeV range.EPB is novel deorbiting technology which employs the coulomb drag be-tween the ionospheric plasma and a long charged tether [23, 24]. The tetheris deployed using a centrifugal force and it is estimated that a 100-m tether(such as on-board Aalto-1) could decrease an altitude by 100 km of a three-unit CubeSat within 600 days [25]. A similar experiment was carried on-board ESTCube-1 [26, 27] where tether deployment was not successful [28].While Aalto-1 EPB experiment was improved based on ESTCube-1 groundtest results, yet the deployment of EPB was not successful. This is dueto the fact that Aalto-1 flight hardware had to be delivered soon after theESTCube-1 experiment was carried out and, therefore, the team did not havetime and resources to redesign the EPB module, as it is being done for theFORESAIL-1 mission [25]. 3 igure 1: Overview of Aalto 1 subsystems and photograph of Flight Model (FM).Thehighlighted subsystems are: 1) Radiation Monitor (RADMON), 2) Electrostatic PlasmaBrake (EPB) 3) Global Positioning System’s (GPS’s) antenna and stack interface board,4) Attitude Determination and Control System (ADCS), 5) GPS and S-band radio, 6)Aalto Spectral Imager (AaSI), 7) Electrical Power System (EPS), 8) On-Board Computer(OBC), 9) Ultra High Frequency (UHF) radios, 10) solar panels, 11) electron guns forEPB, 12) S-band antenna, 13) debug connector
In this paper, section 3 briefly introduces mission timeline representingthe launch and operations. Section 4 presents the in-orbit results and lessonslearned of all the payloads. RADMON in-orbit results are introduced brieflybased on previously published results [21, 22]. EPB in-orbit results andlessons learned are discussed in detail, especially the possible reasons of tetherdeployment failure. AaSI detailed in-orbit results are presented here for thefirst time. Furthermore, Section 5 introduces in-orbit experience of platform’ssubsystems. Section 6 discusses the results and concludes the paper.4 . Mission timeline
The spacecraft was launched aboard PSLV-C38 launch vehicle at 05:59Eastern European Time (EET) and the first beacon was recorded by a Soft-ware Defined Radio (SDR) located in South Africa at approximately 08:30EET. The first contact with the Aalto University ground station was estab-lished during the first pass at 10:07 EET.During the consequent passes, several responses were recorded, but werenot decoded due to an unidentified problem in the ground station receptionchain. Later on the problem was troubleshooted to be in mast pre-amplifier.While powering it off provided a directional link with the CubeSat, it cameat a cost – a loss in the signal strength.The mission wise timeline on the commissioning and operations of eachexperiment is presented in Fig. 2. During Launch & Early Operations Phase(LEOP), the first AaSI picture was downloaded and RADMON commis-sioning phase was started. As part of Aalto-1 operations, multiple AaSIcampaigns have been completed. RADMON operations resulted in a usefuldata set during nominal conditions and also during a solar storm. EPB cam-paign resulted in partial success in commissioning phase but failure in tetherdeployment.
Figure 2: Mission timeline
4. Mission payloads
This section describes the in orbit performance of RADMON, EPB andAaSI payloads. The thorough design approach, selection and implementationhas been presented in accompanying paper [19].5 .1. Radiation monitor mission
The RADMON is a small (4 × ×
10 cm , 360 g) low-power (1 W) radia-tion monitor [29, 19]. The monitor detects protons and electrons employinga regular Δ E – E analysis to distinguish between particle species. The de-tectors of the instrument are a 2.1 × × silicon detector and a10 × ×
10 mm CsI(Tl) scintillation detector placed inside a brass envelope(see Figure 3). The envelope of the detector compartment is opaque for pro-tons below 50 MeV and electrons below 8 MeV. The envelope has a 280 µ maluminum entrance window that stops low energy photons and low energycharged particles. A particle must hit both detectors to be registered. There-fore, the thicknesses of the entrance window and the silicon detector set thelower energy threshold for protons to about 10 MeV and electrons to about1.5 MeV. A detailed description of the instrument calibration is presented in[22]. Figure 3: The RADMON radiation monitor cross section. The arrow on the picture showsa particle that is incident within the instrument aperture. The brass case is light-brown.The silicon detector is light blue, surrounded by a blue passive silicon area, which is fixedon a printed circuit board (PCB) shown as dark gray. The CsI(Tl) scintillator is shown ingreen. Under the scintillator there is a photodiode shown in dark blue. White structureon the bottom is an alumina case of the photodiode.
RADMON in-orbit calibration campaign was carried out in September2017. It was discovered that the gain of the scintillator did not match thevalue obtained from ground calibrations, but was about 30% lower. Thereason could not be positively determined, but the deterioration of the op-tical contact between the CsI(Tl) crystal and the photodiode during launchvibrations could potentially be responsible for this decay of performance. Asuccessful in-flight calibration was, however, achieved using data obtained in6 dedicated calibration mode, which allows raw data from detectors to bedown-linked. The in-flight calibration is discussed in detail in [22].The first observational campaign of RADMON started on 10 October2017 and lasted until 2 May 2018. Using these data, it has been demon-strated in [21] that the instrument is able to measure the integral intensitiesof electrons above 1.5 MeV and protons above 10 MeV in Low Earth Orbit(LEO), reflecting the dynamic environment of the radiation belts. Fig. 4shows the temporal evolution of daily electron intensities from October toDecember 2017 with respect to McIlwain L parameter [30] (indicating theequatorial distance of drift shells) together with the Dst (disturbance stormtime) index as a measure for geomagnetic storm intensity [31, 32]. The twoobserved moderate geomagnetic storms result in strong enhancements of theouter radiation belt, while periods following small storms are characterizedby reduced electron intensities in the outer belt.Figure 4 also illustrates the contamination of all electron measurementsby higher energy protons: the constantly increased intensities in the L rangebelow 2 correspond to the proton-dominated inner radiation belt. Furthercomparisons with electron spectra observed in a similar but slightly higherorbit (820 km) by the Energetic Particle Telescope (EPT) onboard the ESAminisatellite (volume < ) PROBA-V (PRoject for OnBoard Autonomy-Vegetation) showed a good agreement for the > RADMON is a successful space experiment and, certainly, it can be im-proved. Minimization of the contamination of electron channels by high-energy protons would be the most valuable improvement for the instrument.The collimator geometry should also be streamlined to achieve an optimalinstrument aperture.The current design is such that particles enter the instrument within a7 igure 4: From top to bottom: Time series of
Dst index and four histograms of inte-gral intensities with respect to L parameter obtained by the different RADMON electronchannels from 10 Oct 2017 to 21 Dec 2017. The z-axis gives color-coded arithmetic dailymean of intensity per bin – note that the color scale is different for all panels in order toenhance the details of all channels which have different sensitivities. Figure adapted from[21] by permission of Elsevier, © ≈ ◦ half-width cone defined by an opening in the brass container. Theopening is manufactured as a right-angle shaft sufficiently larger than thedimension of the silicon detector (see Fig. 3).An incident particle may, therefore, hit a side of the silicon detector ina way that it deposits energy into its active area and its passive area inan arbitrary proportion. Further, it hits the scintillation detector. Thiseffect leads to an underestimation of energy deposited in the Δ E detector.Subsequently, such a particle is misclassified. A silicon detector with twoconcentric active areas would contribute to better particle classification andreduce contamination of electron channels by protons. The detector should8rigger on the central dot and add the energy deposited in the encircling areato its output signal. One of the possible geometries could be a ”sandwich”detector with a thinner layer carrying the central spot and a thicker layerbeneath. In this case, it is even easier to get the correct Δ E signal sincethe energy loss in the thinner layer would not be needed for the pulse heightanalysis. Any signal above the threshold would gate the particle detection bythe Δ E – E detectors below. The thickness of the top detector can be about100–150 µ m and should be optimized for scientific requirements. A thickerdetector would show more edge effects than a thin one, but could have abetter signal-to-noise ratio. The thickness of the entrance window should beadjusted as well during the optimization.Another issue is that the current geometry allows a gradual increase inthe angle of the acceptance cone. The collimator should be designed as aconical opening in the shielding container so that it becomes transparentat sharper energy threshold. It would improve the flatness of the particleresponse at moderate energies.A simulation of the suggested layered design carried out within the Geant4[33, 34] framework is compared to a simulation of the current design in Fig. 5.The ”sandwich” has a thin silicon detector right on top of the Δ E silicondetector. Both detectors are square and of the same size. The instrumentcontainer has a tantalum front wall, which can be optimized further to atantalum lining of the container opening. This reduction is possible sincethe upper thin silicon detector sets the accepted solid angle for a particleto be detected. High energy protons coming within the aperture are stilldetected as electrons. Nevertheless, limiting the solid angle of the instrumentacceptance for such protons improves the quality of the observational data.In a proton-rich environment, such as South Atlantic geomagnetic anomaly,contamination of electron channels could be used as a secondary proxy onthe proton population.As a positive takeaway from the experiment, the successful RADMONre-calibration using in-flight data showed that a dedicated mode allowingthe full pulse height data to be downloaded also from space can render aRADMON-like instrument to a self-calibrating device. Thus, an expensivefull calibration campaign in high-energy beam facilities, reaching hundredsof MeVs in proton energies, can be avoided using this approach.9 igure 5: The contamination of electron channels (e3 and e5 are chosen as examples) byhigh energy protons in comparison to a proposed ”sandwich” design.
The key components of the EPB payload are those of the tether reelingmechanism as shown in Fig 6. These include the tether reel, reel motor (notvisible), tether chamber, tether tip mass, tip mass launch lock (Kaiku), andreel launch lock (Kieku). The reel motor (vacuum qualified piezo motor) isnested inside the reel. The control electronics underneath the tether reel-ing Printed Circuit Board (PCB), separate high voltage PCB, and electronemitters are similar to those of ESTCube-1 as described earlier in the lit-erature [27]. Only changes introduced were related to the revised launchlocks and additional diagnostics. The high voltage converter was changed todouble the voltage from ±
500 kV to ± igure 6: EPB Mechanical parts on the PCB (left) and key tether reeling components:reel, tip mass, tether chamber, and tip mass launch lock (right). The reel lock can be seenon the left panel left side of the tether chamber. The in-orbit tests of the plasma brake started with a commission phase, inwhich the On Board Computer (OBC) sent EPB a number of commands withthe goal of verifying its operational state. This list included essentially all thecommands which were safe to run without any risk of hazard. This restrictionruled out e.g. the commands that would initiate physical changes in thepayload’s status (launch lock burns, motor activation) or the ones not usableat this point of the mission (high voltage or electron gun activation). Thecommands that were run all worked as designed, returning some housekeepingdata for analysis. Most importantly at this point, the data showed that allsystems were at nominal state and that the launch locks had kept the tetherreel and the tip mass intact.The second step in in-orbit tests was to open the two launch locks thathad locked the tether reel and the tip mass during the launch. Each lock wasopened by applying a 150 mA current, which would melt the dyneema stringkeeping the spring loaded lock at closed state. The tether reel lock, namedKieku, had an integrated optical diagnostics system whose state could beread by the OBC at any time. During the burn sequence the system’s stateswitched from “locked” to “deployed” after about 12 seconds of burning asshown in Fig. 7. Similar diagnostics were not available for the tip mass lockKaiku. The duration of the burn current for Kaiku was chosen long enoughto ensure a proper deployment.The final verification of the operational readiness before attempting tether11 igure 7: Flight data showing the deployment of the tether reel lock Kieku. deployment was performed with the help of the photosensor Kyyl¨a. Kyyl¨ais a simple phototransistor placed inside the tether reel chamber. It has thebackside of the tip mass in the center of its field of view. If the tip mass hadbeen ejected from its nest prematurely, the light (e.g. from Earth albedo)entering the chamber could easily be detected in Kyyl¨a’s signal. An exampledata plot from an early Kyyl¨a scan is shown in Fig. 8. The extremely nar-row width of the peaks indicate that even though light is able to enter thechamber, it is able to do so over a very narrow angle only, as the satelliteis spinning. This may be explained as follows. The tip mass, roughly cylin-deric, remains like a plug in the tether opening tube. The tip mass is nottightly in the tube but held to its place by the launch lock. Thus there isa tiny gap between the tip mass and the tube walls that provides the lightwith a passage of the narrow angle. If the tip mass was completely removed,the shape of the peaks would be considerably wider. Another piece of infor-12ation obtained from these tests was the confirmation of the satellite’s spinrate. An approximate seven second periodicity of the peaks coincided pre-cisely to the angular velocity data of the Attitude Determination & ControlSubsystem (ADCS). Simultaneously it provided proof that Kyyl¨a was in-deed measuring real phenomena of its surroundings and not some arbitraryelectrical disturbances.
Figure 8: Flight data from the phototransistor Kyyl¨a. The periodicity of the signal corre-sponds to the satellite’s spin rate at the time.
After the successful initial tests and preparations it was time to attempttether deployment. A controlled spin-up of the satellite could not be per-formed due to the shortcomings of the satellite’s ADCS which are describedin detail in section5. The satellite was nonetheless spinning through naturalcauses and its spin axis and angular velocity ( ≈
50 degrees per second) were,by chance, suitable for taking a shot at deployment.The spin rate for EPB deployment was verified by magnetometer and gy-roscope telemetry data. Figures 9 and 10 present the high resolution measure-13ent data in time and frequency domains respectively and Fig. 11 presentsthe calibrated gyroscope data during the EPB deployment campaign.
Figure 9: Magnetometer high resolution data during EPB deployment campaign
Despite having achieved the desired spin rate around tether deploymentaxis, the deployment attempts all failed, unfortunately. In each attemptthe motor was commanded to make a turn that is relatively small but stilleasily detectable. We couldn’t observe any changes in the tether reel rotaryposition. The vacuum qualified piezo motor has an in-built potentiometerbased rotary encoder. Fig. 12 shows the values measured by this encoderthroughout the tether deployment trials. The peak-to-peak variation of thevalues corresponds to a 1.4 ◦ turn, or 0.4 mm on the perimeter of the reel.The conclusion must be that no detectable motor movement has taken place.If the motor had worked nominally, the turn angle would have been tens ofdegrees. The most noticeable result of the EPB mission is obviously the failure ofthe tether deployment hardware. It is somewhat unclear why this happened,even though several clues exist. Figure 12 shows examples of the measured14 igure 10: Magnetometer high resolution flight data during EPB deployment campaignin frequency domain confirming the spin rate around the spin axis motor voltage during two tether deployment attempts. In normal opera-tion the motor voltage would remain in its nominal value of approximately40 volts. As the plots show, the voltage is cut off and starts a rather rapiddecay as soon as it has been switched on. The motor voltage is generatedwithin the EPB control electronics with the help of a boost converter. InFig. 12 the voltage appears to saturate at the level of the boost converter’sinput voltage. This would indicate that the faulty operation of the boostconverter is the source of all grief.Not all went haywire, though. Several newly developed systems, some in-cluding moving parts, worked exactly as planned. Especially the completelyrenewed design for the tether reel lock Kieku proved to be a reliable workhorse in space. At this point it is important to introduce the reader to thelaunch history of the EPB payload. A very similar payload was first launchedon-board ESTCube-1 [26, 27]. Its fate was identical to that of Aalto-1 EPB.It is important to note that the timelines of the two satellite missions over-lapped in a most unfortunate way. Once the in-orbit results of ESTCube-1were ready and verified, the delivery date of the Aalto-1 flight model hard-15 igure 11: High resolution gyroscope calibrated flight data during EPB deployment cam-paign ware was only four months away. Also, due to the lack of proper on-boarddiagnostics, the reasons for the failure were mostly unknown. Therefore theEPB team was lacking both the proper time and the accurate knowledge ofthe problem in order to make fundamental changes in the motor hardwareand control electronics. Instead, a number of features were added to gatherall the information possible, in order to at least see what is happening incase of repeated failure. All these diagnostics tools described above (Kyyl¨a,Kieku’s optical feedback, motor’s position encoder) worked as planned. Thisallowed the EPB team to have an instant view of the situation in orbit and fi-nally get valuable clues of what happened on-board ESTCube-1 as well. Thelast minute changes could not help the Aalto-1 EPB to complete its mission,but at the very least they helped in compiling a road-map towards moresuccessful missions in the future. A small step for Coulomb drag industry,but a step forward nonetheless. 16 igure 12: Measured values of the rotary encoder of the tether reel motor. These valueswere recorded over several tether deployment attempts. The peak-to-peak variation of thevalues corresponds to a turn of 1.4 degrees. The pre-launch value recorded in the lastground tests was 421.
After establishing communications, the VIS camera was first powered onthe 3 rd of July, 2017. The first housekeeping data from the camera indicatednominal behaviour, and the instrument temperature was ca. − ◦ C. The firstimage, as shown in Fig. 14, was taken on the 5 th of July, while the satellitewas still tumbling. During image acquisition, the satellite was located overNorway with the field of view pointed to the southern direction towardsDenmark. Based on visual analysis, the image quality is good, and no visiblede-focusing or new aberrations are present.The spectral camera was first powered on the 25 th of July, and the instru-ment housekeeping data was nominal. The temperature was around − ◦ C,17 igure 13: Two data sets of the motor voltage during the tether deployment attempts.Notice the saturation at the level of the input voltage ( ≈
11 volts) of the boost converter.In each set the last data point was recorded after the input voltage had been switched off. and the piezo voltages for the FPI were between 26 V and 28 V, which in-dicated perfect health for the FPI unit. When compared to piezo voltagesmeasured on ground prior to launch, there was approximately 10 V differencein one of the channels. This was expected as there is a temperature differ-ence between the measurements (+22 ◦ C at the pre-launch check vs. − ◦ Cin orbit) and the water absorbed by the piezo actuators has evaporated atthe time of taking in-orbit measurements.First images were taken with the spectral camera on the 3 rd of August.From these images, the functionality of the camera optics was verified. Theimaged scene was covered by clouds, and in the false color composite as shownin Fig. 15, one can see spectral variation in the clouds. During this time,the satellite was still tumbling quite rapidly, so the imaged area is movingsignificantly between the spectral frames.18 igure 14: The first image downlinked from Aalto-1. The image is taken with the VIScamera on 5 th of July, 2017 and it shows the coastline of Denmark together with Earth’shorizon. After the performance of optics was verified, the on-board spectral cali-bration method was tested. The calibration is based on measuring a brighttarget (e.g. cloud or desert) and scanning the spectral filter over the cut-off wavelength of the 900 nm short pass filter and taking an average of thepixel values. The sequential images are recorded with very small wavelengthincrement. When the spectral transmission peak passes over the short passfilter, the signal level will drop. When the signal is plotted as a functionof FPI set point voltage, the drop in signal level is visible. The locationwhere the slope is steepest corresponds to the cutoff wavelength of the short-pass filter. Successful calibration measurement was performed on the 5 th of September which is shown in Fig. 16. When compared to measurementsdone on ground, it can be seen that the spectral behaviour is similar, butdue to the cold temperature ( − ◦ C) and different illumination conditionsthe shape of the calibration spectrum is different.The satellite was de-tumbled in June 2018 and the imaging campaign wascontinued immediately after de-tumbling. During this campaign, an image19 igure 15: False color composite of the first spectral image captured by AaSI. The approx-imate wavelengths in the image are R=710 nm, G=535 nm and B=510 nm. The tumblingof the satellite is clearly visible as the frames do not contain much overlap. The bottompart of the image shows as yellow, as the wavelengths 535 nm and 710 nm are extractedfrom the same raw image. mosaic was created from VIS images, and finally on August 6, 2018 thefirst cloud-free images of land targets were acquired. The imaging sequencestarted at the equator above Congo, and continued for about six minuteswhile the satellite was travelling south toward South Africa. The images ofsix different wavelengths were acquired, and, from the resulting spectrum,the red-edge of vegetation is clearly visible, as shown in Fig. 17.An image compression program was uploaded to the satellite during thespring of 2018. This was first tested around the midsummer of 2018, andseveral series of images were taken. In order to downlink the image mo-saics, image compression was required. After compression, the images weresuccessfully downlinked. The stiched mosaic is shown in Fig. 18. The slowtumbling of the satellite is clearly visible in the sequential images.20 igure 16: Calibration measurement comparison. In the top figure, the signal level isplotted as function of FPI set point voltage. Signal derivative is plotted in the bottomfigure. The position with the steepest slope corresponds to the filter cutoff wavelength.The filter cutoff position is visible in both cases, but the measurement performed in orbitis distorted. This is mainly due to the cold temperature, which is outside the instrument’soperation temperature.
This was the first mission ever to demonstrate a hyperspectral camera ona nanosatellite. It was also the first space-borne demonstration of a tunableFPI-based nanosatellite-compatible hyperspectral camera.The main lesson learned was that this technology works in space environ-ment and it can be used for nanosatellite-based hyperspectral imagers. Allof the primary mission objectives were completed, so the AaSI mission canbe considered successful.Not all functionalities of the imager could be verified though. The tum-bling platform prevented imaging of planned targets, and the limited down-link allowed only the use of minimal spectral mode with six wavelengths.However, the tumbling platform showcased the benefits of frame-based spec-tral imaging, as the images in different wavelengths can be overlapped in postprocessing. This is a great benefit in nanosatellite missions, as the imagercan still be used in the case of attitude control malfunction.21 igure 17: The first cloud-free spectral image of a land target (top). The image is centeredon Tshuapa River near Mbandaka. The false color image is constructed from R=752 nm,G=671 nm, B=565 nm. The bottom figure shows the spectrum of the central area of theimage measured at 6 wavelenghts.Figure 18: Mosaic of sequential VIS images . Platform in-orbit performance The in-orbit performance of spacecraft platform which consisted of com-mercial and in house developed subsystems, is briefly presented. While thekey platform subsystems were successfully commissioned, the spacecraft ac-complished its mission with partial success. The design approach of platformsubsystems which consisted of an Electrical Power Subsystem (EPS) [35], anADCS [36], a Global Positioning System (GPS)-based navigation system[37], a Ultra High Frequency (UHF) [38] and S-band [39] radios for Teleme-try, Telecommand & Communication (TT&C), and a Linux-based OnboardData handling (OBDH) [40] is briefly presented in [19].
In order to monitor and keep track of the health of the spacecraft, a num-ber of housekeeping sensors were used. The performance of EPS is presentedin terms of telemetry values of voltage, current and temperature sensors. Theflight data of these sensors confirms that the EPS is functional and providespower to satellite subsystems since its launch. However, there are some is-sues. The telemetry data reveals partially degraded performance of one of thesolar panels as evident by green plot in Fig. 20. This behaviour is likely dueto the un-controlled spin orientation of the satellite. The telemetry data ofsolar panel temperatures, EPS board temperature and battery temperaturesfrom launch date till Aug 2020 is plotted in Fig. 19. The highest temperaturevariation takes place on the satellite surface as evident from central graphsrepresenting panel X and Y, where solar panel temperatures change in ± ◦ Crange. This range remains quite stable throughout the mission representingthat the temperature is at equilibrium. The temperature inside the satellitedepends on operation of payloads and platform subsystems. The telemetrydata of board and battery temperatures, as evident from Fig. 19, representsthat the passive thermal control maintains sufficiently stable temperaturefluctuations.
The commissioning phase of the ADCS functions were met with com-plications, as some of the sensor readings were erroneous. Two of the sunsensors (on the +X and − X directions of the satellite) were malfunctionedand not usable for attitude estimation. In Fig. 21, the gyroscope readingsfrom regular housekeeping data until October 2017 have a low angular rate23 igure 19: Aalto 1 surface and inner temperatures (in ◦ C ) from launch till Aug 2020 resolution. This was because of improper processing of sensor raw data anda problem with the communication channel in the ADCS module. This wasfixed with a small firmware update. Over the course of the mission, some-times the ADCS module is reset and defaulted to idle mode. Such eventwill turn off the ADCS sensors which needs to be manually turned on. Thisshows up as frozen sensor data, visible in Fig. 21 around October–November,2017 and February–May 2018.The first attempt to detumble the satellite was attempted on October2017. The attempt failed because of constant rebooting of the ADCS modulewhen the B-dot control was turned on. The problem was caused by themagnetorquer driver channel which was later fixed with a major firmware andmagnetorquer driver update on June 2018 consequently solving the rebootissue. The detumbling operation was tested again with positive results. Thespin rate of the satellite was reduced close to 0 deg/sec as confirmed bythe telemetry data of Fig. 21. The detumbling control was kept on untilSeptember 2018, after which the satellite started to spin up.The main cause of the uncontrolled spin up of the satellite remains un-known when the B-dot is disabled. Some possible causes are environmentaldisturbance torque, residual dipole moment generated by unknown magneticmaterials or current loop from the solar panels power routing.Although detumbling with the B-dot controller was successful, manyother mission modes, including controlled spin up for tether deployment,24 igure 20: Solar panel current intensities from launch till Aug 2019. The vertical axisrepresents the generated current (in mA) read by each BCR were not successful. The ADCS commissioning modes, including the spinup manoeuvre, has been tried with the magnetorquers only [6, 5]. The re-action wheels showed inconsistencies in their power reading during the earlycommissioning phase and thus have not been thoroughly tested yet.An important lesson learned was to procure the commercial modules atthe early stages of development and test all the functional modes during thequalification phase. Moreover, designing an in house subsystem gives moreflexibility in interfacing and testing.
The OBC − igure 21: Gyroscope data from July 2017 to November 2019. suffered from instability in EPS, resulting in several resets in EPS and thearbiter. A Coronal Mass Ejection (CME) occurred in early September 2017,providing an excellent opportunity for RADMON testing [41]. The satellitewas quickly retasked to collect as much data as possible with RADMON. Aprecious RADMON set of data collection was also interrupted during a CMEdue to OBC reboots. A few unexplained boot events of the OBC, which wereresolved without involvement of the arbiter, occurred during the CME event.It is suspected that these may be related to either radiation or EPS reset.Immediately after the launch, multiple objects launched on the samerocket as Aalto-1 had similar Two Line Element (TLE), and it was unclearwhich TLE set belonged to Aalto-1. The GPS subsystem was one of the firstinstruments successfully operated after contacting the satellite, and navi-gation solutions provided by the receiver allowed determining the correctTLE set. The determined identity was also communicated to the TLE dataprovider [42].It has been observed that the TLE accuracy has been sufficient for mostroutine operations, and the use of GPS has been less frequent than expected.26 igure 22: OBC reboot events during July-Nov 2017 [40] A sub-optimal GPS antenna placement (resulting from a compromise withsolar panel placement) and satellite tumbling have caused delays in obtainingthe first fix after powering the receiver.The commissioning of the UHF transceiver was successful since the firstcontact with the CubeSat was established during first pass over the groundstation. The commissioning phase was met with many challenges which havebeen briefly detailed in [19]. From the telemetry logs, a radio interference inthe Northern direction, close to the horizon, was noted at around 437.22 MHz.Similar kind of interference around the 437.0–437.4 MHz was measured bythe UWE-3 CubeSat mission though the source of interference has not beenconfirmed [43]. The S-band transmitter has not been successfully commis-sioned despite multiple attempts in July 2017 and July 2018.27 . Discussion and conclusions
Although the mission was a partial success in terms of executing the ex-periments, the important lessons learned during this mission have been ap-plied in the design of next variants of payloads and platforms. The RADMONinstrument was successful in commissioning and measurement phases. Itsheritage has been used to design a more complex Particle Telescope (PATE)payload for the upcoming FORESAIL-1 mission [44]. The EPB tether couldnot be deployed due to a failure in tether deployment hardware. The lessonslearned have been taken into consideration in development of the plasmabrake for upcoming FORESAIL-1 and ESTCube-2 missions [45]. The AaSIwas the first nanosatellite-compatible hyper-spectral imager to be flown inspace. Aalto-1 project successfully demonstrated the expertise of VTT inboth visible and hyper-spectral miniature imager designs. The technologyhas many potential future applications to serve CubeSat and/or scientific in-dustry/community. Since Aalto-1, VTT’s hyper-spectral imagers have beendeveloped for Reaktor Hello World, PICASSO, Hera and Comet Intercep-tor missions. The platform has provided successful in-orbit demonstration,although some subsystems lacked the desired performance. An importantlesson learned was to perform a rigorous test campaign while integrating thecommercial and in-house built subsystems.
Acknowledgements
The RADMON team thanks P.-O. Eriksson and S. Johansson at the Ac-celerator Laboratory, ˚Abo Akademi University, for operating the cyclotron.Computations necessary for the presented modeling were conducted on thePleione cluster at the University of Turku.Aalto University and its Multidisciplinary Institute of Digitalisation andEnergy are thanked for Aalto-1 project funding, as are Aalto University,Nokia, SSF, the University of Turku and RUAG Space for supporting thelaunch of Aalto-1.