Professional Documents
Culture Documents
IEEE Aerospace & Electronics Systems Magazine - May 2021
IEEE Aerospace & Electronics Systems Magazine - May 2021
IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE ® (ISSN 0885-8985; USPS 212-660) is published monthly by the Institute of Electrical and Electronics Engineers, Inc.
Responsibility for the contents rests upon the authors and not upon the IEEE, the Society/Council, or its members. IEEE Corporate Office: Three Park Avenue, 17th
Floor, New York, NY 10016, USA. IEEE Operations Center: 445 Hoes Lane, P.O. Box 1331, Piscataway, NJ 08855, USA. NJ Telephone: 732-981-0060. Price/Publication
Information: To order individual copies for members and nonmembers, please email the IEEE Contact Center at contactcenter@ieee.org. (Note: Postage and handling
charges not included.) Member and nonmember subscription prices available upon request. Copyright and reprint permissions: Abstracting is permitted with
credit to the source. Libraries are permitted to photocopy for private use of patrons, provided the per-copy fee indicated in the code at the bottom of the first page
is paid through the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923, USA. For all other copying, reprint, or republication permissions, write
to the Copyrights and Permissions Department, IEEE Publications Administration, 445 Hoes Lane, P.O. Box 1331, Piscataway, NJ 08855. Copyright © 2021 by the
Institute of Electrical and Electronics Engineers, Inc. All rights reserved. Periodicals postage paid at New York, NY, and at additional mailing offices. Postmaster:
Send address changes to IEEE Aerospace and Electronic Systems Magazine, IEEE, 445 Hoes Lane, P.O. Box 1331, Piscataway, NJ 08855. GST Registration No. 125634188.
CPC Sales Agreement #40013087. Return undeliverable Canada addresses to: Pitney Bowes IMEX, P.O. Box 4332, Stanton Road, Toronto, ON M5W 3J4, Canada.
IEEE prohibits discrimination, harassment, and bullying. Printed in the U.S.A.
Editors
Editor-in-Chief–Peter Willett, University of
Connecticut, USA
Associate Editor-in-Chief–Daniel O’Hagan,
University of Cape Town, South Africa
VP Publications–Maria Sabrina Greco,
University of Pisa, Italy
AESS President–Walt Downing, Southwest
Research Institute, USA
Operations Manager, AESS–Judy Scharmann,
Conference Catalysts, LLC, USA
How to Reach Us
We welcome letters to the editor; but, we
reserve the right to edit for space, style, and
clarity. Include address and daytime phone
number with your correspondence.
E-mail: Peter Willett, Publishers: Send books for review to Samuel Shapero, 407 Angier Place NE, Atlanta, GA 30308. If you have questions,
peter.k.willett@gmail.com contact Samuel by e-mail at samuel.shapero@gtri.gatech.edu.
Daniel O’Hagan,
Advertisers
danieloha@googlemail.com
If you are interested in advertising in the AESS SYSTEMS Magazine, please contact Anthony Land at Naylor
Catherine Van Sciver, Associates at aland@naylor.com, or Peter Willett at peter.k.willett@gmail.com
c.vansciver@ieee.org
AESS Senior Members Elevated in 2021
2. Topics of Interest
x
x
x
x
x
x
x
x
x
x
x
x
. Section Organizers
x
x
x
x
x
In This Issue –Technically
we evaluated 4G mobile broadband and machine-type com- time. To the best of our knowledge, this article is the first
munication (MTC) for the aircraft cabin. We studied different analysis of the requirements for massive data transfer from
MTC use cases. While the evaluated use cases do not include and to aircraft. In a first step, a suitable technology has to
massive data transfer, the network architecture within the be selected. Therefore, we study the needed data rates for a
cabin enables the collection of data from the cabin. The per- ground communication system. To achieve this, we ana-
formance of Wi-Fi, 4G, and LiFi inside the aircraft cabin has lyze flight traces to determine the needed data rate on the
been studied in [7]. The authors provide a channel model and ground in terms of the generated data rate per aircraft. We
determine the per-user performance within the cabin. How- also study the challenges and the potential system architec-
ever, they do not investigate any link from the aircraft to ture, including a comparison of existing or proposed com-
the ground network. The case of in-flight entertainment munication systems, such as 5G. Moreover, millimeter
is studied in [8]. The authors propose a flexible 5G wave (mmWave) technology is one candidate for high-
experimental platform for in-flight entertainment serv- speed connectivity and is not yet widely used. Therefore,
ices. Interestingly, they propose to include an on-board we conduct 802.11ad measurements to evaluate the feasi-
data center on the aircraft. While they propose to use bility of mmWave communications for aircraft data trans-
the data center for passenger services, it could poten- fer. The remainder of this article is structured as follows: In
tially also be used to store and process aircraft the section “System Architecture and Challenges,” we
sensor data. analyze different communication system architectures
Several communication systems to transfer sensor and discuss challenges for high-speed connectivity at
data have already been deployed. The history of Gatelink airports. The section “Methodology” introduces the
and the deployments used in the year 2005 are summa- methodology of the flight data analysis. The section
rized in [9]. Although originally designed for communi- “Results of the Flight Data Analysis” shows the results
cation between aircraft and ground stations, Aircraft of the analysis. In the section “Throughput Measure-
Communications Addressing and Reporting System ments of 60 GHz Wi-Fi,” we evaluate the performance
(ACARS) is an option to transmit safety related data. of the 802.11ad standard by measurements. Last, we
Today, most of the ACARS traffic is generated while the present the section “Conclusions.”
aircraft is on ground [9]. To enable higher data rates,
AeroMACS has been introduced for airport surface com-
munications [10]. It is based on 802.16e and uses the spe- SYSTEM ARCHITECTURE AND CHALLENGES
cifically allocated band from 5091 to 5150 MHz. While
the focus is on safety and traffic control at the airport, In this section, we analyze different communication system
data download is one of the mentioned applications [11]. architectures including the performance characteristics of
Additionally, there are deployments specifically for currently available technologies. Furthermore, we high-
transferring data from aircraft based on Wi-Fi or cellular light the challenges of high-volume data transfer for
systems. Some examples are Skywise [12], Gatesync aircraft.
[13], or GroundLink [14]. However, except for Skywise,
they focus on pilot- and safety-related data and not
toward massive transfer of data. Skywise offers an open-
ARCHITECTURE
data platform to analyze the aircraft data. However, due A communication system for aircraft needs to take several
to restrictions in the currently available data links on requirements into account. It needs to offer the required
ground, they are not yet able to offer multi-gigabits per data rate and also be available on airports worldwide. In
second upload speeds as needed in the future. Table 1, we compare the properties of different cellular
The aforementioned existing communication systems and noncellular communication systems in terms of spec-
are not designed to offload a massive amount of data in the trum, data rate, range, operator, and limitations. Spectrum
order of terabits per second during the short turnaround can be shared or licensed. The advantage of shared
Table 1.
Comparison of Different Technology Options for Data Transfer From and to an Aircraft
Technology Spectrum Peak data rate Range Operator Limitations
60-GHz Wi-Fi Shared 6.7 Gb/s (802.11ad) 10 of Airport / local Line of sight
meters operator
Wi-Fi Shared 6.9 Gb/s (802.11ac) 50 m Airport / local Number of
9.6 Gb/s (802.11ax) operator users
AeroMACS Dedicated 50 Mb/s Several Airport / local Number of
kilometers operator users / BS
4G (Cat.21) Dedicated 1.4/0.3 Gb/s (DL/UL) Several Cellular operator Number of
kilometers users / BS
5G Dedicated 20/10 Gb/s (DL/UL) Several cellular operator Number of
kilometers users / BS
5G Dedicated 20/10 Gb/s (DL/UL) 500 m Cellular operator Line of sight
mmWave
spectrum is easy availability; however, it needs to be sight. The advantage of 4G and 5G is the dedicated
shared with other users. For high-capacity communica- frequency and the global availability. On the other
tions, mmWave bands can be used. However, due to hand, the deployment of a specific system on the air-
the high frequency, the range becomes very small. port, such as Wi-Fi or AeroMACS, enables adjusting
Moreover, further parameters need to be taken into to the specific requirements at the airport.
account, such as influence by rain. Additionally, the As a conclusion, several communication systems
antenna size decreases with increasing frequency, exist, which are able to provide high data rates. However,
which is an advantage for mmWave bands. To cover a according to the specific requirements, different options
large area such as an airport, hundreds of base stations might be chosen. Also a combination could be used,
would be needed. The operator can be either a cellular depending on the type of traffic, differing in the amount of
operator that might offer global connectivity or a local data or the time criticality. Nevertheless, the biggest chal-
operator on the airport. The advantage of a global lenge is the global availability across different airports,
operator is that connectivity can be ensured for differ- airlines, and aircraft manufacturers.
ent flights, without the need of a separate contract with
a local operator per airport. Nevertheless, also a global
operator needs to ensure sufficient capacity for the
CHALLENGES
number of aircraft at the airport.
In Table 1, we compare Wi-Fi, 4G, 5G, and Aero- While the capacities of single links are promising in terms
MACS. For Wi-Fi and 5G, we additionally analyze the of the peak data rates, achieving the same high data rates
mmWave bands. Except for AeroMACS, all options for all aircraft at an airport is a challenge. Besides this, fur-
are ready for speeds in the order of multiple gigabits ther challenges exist. In the following, we discuss the
per second. However, in most cases, experienced data main challenges toward providing massive data transfer to
rates are significantly lower than the peak data rates. aircraft.
In particular, for already deployed systems such as
Wi-Fi, it becomes evident that the experienced data
rates are a factor of ten to one hundred below the peak
UPLINK DATA RATE
speeds. This is mainly due to channel conditions The transfer of accumulated sensor data from the aircraft
between user and access point and congestion of the uses the uplink from the aircraft to the base station. How-
channel by multiple users. Additionally, the data trans- ever, cellular networks are designed to provide high
fer from aircraft is uplink traffic, which traditionally capacity in the downlink, not the uplink. In a classical
offers lower data rates than the downlink in cellular setup, this is unproblematic, as data are sent right away.
networks. Hence, the limitation for all applications in However, in the aircraft case, the data is accumulated dur-
the sub-6 GHz frequency bands lies within the number ing the flight and needs to be uploaded during a small
of users with respect to the base stations. For the time. Hence, the uplink data rate is a major limitation for
mmWave frequency, the main limitation is the line of data transfer from aircraft.
OPERATIONS
Figure 1.
In order to transfer data from the aircraft, the transmitter Example for different types of flight, multiple short-haul flights
needs power supply. If an aircraft is parked at a remote per day (aircraft 1) and one long-haul flight per day (aircraft 2).
location or for a longer time during night, this power sup-
ply might not be available. Hence, while the aircraft is
parked, power supply at the transmitter and hard drive on The investigated flight data consists of two one-week
the aircraft needs to be ensured. samples from FlightRadar24 [15], one in summer (23/07/
18–30/07/18) and one in winter (05/02/18–11/02/18).
DATA PROCESSING While we analyze both datasets, we only show the results
of the summer dataset, if not otherwise noted. The data-
The amount of data that needs to be transferred to the sets contain numerous parameters per flight, including
ground depends on the processing location. The data could the position over time, the aircraft, and airline. Particu-
be processed in a data center on the ground. This means larly interesting for this analysis is the flight duration,
that all generated data needs to be transferred, resulting in time on ground, number of aircraft at the same airport
a large amount of data. Alternatively, the data could be and the aircraft size.
processed on board. This can significantly reduce the The flight route and flight duration heavily influence
needed data rate. However, it is limited by weight and the time at the airport. Figure 1 shows two representative
processing constraints on the aircraft. The optimal trade- examples. Aircraft 1 completes multiple short-haul flights
off, where to process which parts of the data is still open. within the same day. The turnaround time between two
One possible driver is the cost for processing power on- flights is short, but the aircraft stays at the airport over-
board versus transfer of the total amount of data. Addition- night. Contrarily, aircraft 2 does a long-haul flight during
ally, different types of data can be treated differently. A the night, but stays at the airport for several hours during
small amount of time critical data could already be sent the day. Hence, the generated amount of data during one
during the flight, whereas the large amount of noncritical day mainly depends on the total duration of flight on that
data would be sent after the flight. day, which is shown in blue in Figure 1. While on ground,
as depicted in green, the aircraft needs to transmit this
WORLDWIDE AVAILABILITY generated data. Hence, for an aircraft to transfer all gener-
ated data, the generated data during the flight(s) needs to
For seamless operation, data transfer needs to be available
be smaller than the amount of data, which can be trans-
worldwide. This is in particular challenging in regions
ferred during the time on ground. With this, we can calcu-
with limited Internet connectivity. For each airport not
late the needed data rate of the ground network Rgnd in
providing sufficient connectivity, other airports need to
terms of the data rate of generating data at the aircraft
support an even higher data rate, additionally resulting in
Rgen as follows:
an increased delay. As we show in the section “Parameter
tf [s] 8 GB
study,” this can be compensated to a certain extent. Never- Rgnd [Gb/s] Rgen (1)
theless, a minimum coverage among airports worldwide tg [s] 3600 h
needs to be available to provide such a service.
with tf being the flight time and tg being the ground time.
We analyze these times in two different ways: 1) per
flight or 2) per day. Per flight means that the generated
METHODOLOGY data during one flight needs to be transferred during the
In order to understand the requirements of transfer mas- subsequent turnaround time. This results in high data rate
sive data from and to aircraft, we studied real aircraft requirements on the ground to transfer a large amount of
traces. These data allow the analysis of the number of data in a short time. On the other hand, the generated data
aircraft at an airport at any time as well as the time are not very time critical. Hence, it might be sufficient to
needed for the turnaround and the duration of the flights. transfer some parts of the data directly and the remaining
part during the night. This means that any data that cannot
be transferred during the turnaround would be saved and
transferred later during the night. We call this approach
per day as all data generated during one day needs to be
transferred at some time within that day. In this way, the
needed data rates are much lower; however, the aircraft
data storage and communication equipment needs to be
powered during night. A more detailed comparison will
be conducted later on.
Several parameters influence the needed data rate. As
already explained earlier, the time within the data needs to
be transferred, per flight or per day, plays an important
role. Furthermore, the aircraft size influences the amount
of generated data. A large aircraft has more sensors and,
thus, collects more data than a small one. Currently, an
A320 generates 10 GB/flight hour [16]. As the envisioned
Figure 2.
Duration of different phases at Frankfurt airport.
kind of data collection framework is not yet used today, a
prediction of future generated data rates is difficult. In
order to relate the calculated numbers of Rgnd and Rgen distribution of the turnaround time during day and night as
and to provide an easier understanding of the influence of well as the time needed for taxiing at Frankfurt airport,
different parameters, we assume Rgen of 500 GB/flight Germany. During the day, aircraft stay at the gate for
hour as a reference. Nevertheless, our analysis is not lim- 1.3 h in the median, whereas during night the median
ited to this and any other generated data rate can be stud- increases to 10.4 h. As Frankfurt airport restricts flights
ied as well. The availability of sufficient connectivity at during the night, the minimum time at the gate during
different airports is another factor. If an airport does not night is 5.6 h. We observe many outliers toward longer
offer connectivity, the data need to be stored until the next times at the gate, as some aircraft stay at the airport for
landing. Then, the combined data of two flights need to be days. However, these cases are not the challenging ones as
transferred. Additionally, the size of the airport and the there is a lot of time to transfer the data. The position of
time of day also influence the needed data rate. At peak the aircraft is based on Automatic Dependent Surveil-
times, many aircraft are present at the airport, leading to a lance—Broadcast (ADS-B) and is not completely accu-
potential bottleneck if the communication system is not rate, as some aircraft switch on the ADS-B very late or
properly designed. The required capacity per airport is switch it off very early. This leads to some aircraft with a
deducted from the single aircraft needs and the number of turnaround time of only a few minutes. As this is not real-
aircraft at the airport. istic, we only take aircraft with a turnaround time of at
least 15 min into account. If not otherwise stated, we eval-
uate the 95th percentile to include almost all aircraft. The
RESULTS OF THE FLIGHT DATA ANALYSIS median times for taxi are 7 min for approach and 12 min
for takeoff. The taxiing times, where the aircraft is already
This section presents the results of the flight data analysis.
on the ground, could be used for additional data transfer.
First, the duration of turnaround and taxiing times is inves-
This would provide an additional 19 min in the median for
tigated, which highly influences the available time for
transferring data. However, in the following analysis, we
transferring the data. Subsequently, we introduce the rela-
focus on the times at the gate only.
tion between the needed data rate on the ground and the
generated data rate on the aircraft. Moreover, we conduct
a parameter study to investigate the influence of different
parameters on the needed data rate. To relate the results to
RELATION BETWEEN NEEDED DATA RATE ON GROUND
actual communication systems, we conduct a case study AND DATA RATE GENERATED ON THE AIRCRAFT
comparing different technologies. Finally, we study the
In this section, we present the relation between the needed
backhaul capacity needed at the airport.
data rate on the ground Rgnd and the data rate Rgen generated
on the aircraft. The calculation is based on the turnaround
times determined in the previous section and Equation (1), as
TURNAROUND AND TAXIING TIMES
introduced in “Methodology.” Figure 3 shows the relation
In this section, we study the duration of turnaround and between Rgen and Rgnd . The case of transferring data per
taxiing at the airport. This is the basis for the calculation flight is depicted in dashed lines. In this case, the data needs
of the needed data rates. Figure 2 exemplarily shows the to be completely offloaded during the turnaround time after
the flight. Due to this, a higher data rate is needed for flights
with turnaround during the day than for flights with turn- CASE STUDY
around during night. However, this only depends on the
In this section, we discuss the generated data rates per
time the flight is scheduled and cannot be influenced.
flight hour that can be supported with 4G and 5G technol-
Independently of day and night, a generated data rate of
ogies. We assume that the data can be transferred per day.
500 GB/h can be supported with 5.1 Gb/s on the ground.
4G supports up to 300 Mb/s uplink data rate with UE cate-
The case of transferring data per day is depicted in solid lines
gory 21 [17], which is currently available with high-end
in Figure 3. In this case, the data that cannot be offloaded
smart phones. With this uplink data rate, Rgen of up to
after one flight can be transferred later on the same day. Con-
65.9 GB/flight hour can be supported. According to the
sequently, the needed data rate for 500 GB/h drops from
IMT-2020 requirements [18], 5G will support between
5.1 Gb/s to 2.3 Gb/s compared to per flight. However, in
50 Mb/s and 10 Gb/s uplink data rate. With the peak
contrary to the first case, if only day or night is used for
uplink data rate of 10 Gb/s, Rgen of up to 2.2 TB/flight
transferring data, the needed data rate increases signifi-
hour becomes feasible. However, experienced data rates
cantly, because the total available time for data offload is
will be lower than peak speeds. Another option to increase
shortened. In particular, the required data rate in the case of
the data rate per aircraft is to use multiple modems to con-
night only is very high, as aircraft on long-haul flights during
nect to possibly more than one network operator. Hence,
night can only transfer data during the day.
with two modems per aircraft, up to 130.8 GB can be sup-
ported with 4G and up to 4.4 TB with 5G. According to
the work in [16], currently 10 GB are generated per flight
PARAMETER STUDY hour of an A320. A data rate of 45.5 Mb/s is sufficient for
this. Hence, this data rate can already be supported by 4G
In this section, we vary several parameters to evaluate their
with a data rate of 15% of the peak uplink rate. To cope
influence on the needed data rate. The varied parameters and
with future demands, 4G can support up to 6 times and 5G
results for an exemplary Rgen of 500 GB/flight hour are
around 200 times of the current rate with one modem.
shown in Table 2. In general, short-haul flights with a duration
As a conclusion, Rgen of multiple TB terabytes per flight
of less than 6 h need a much lower data rate than long-haul
hour become feasible with 5G peak data rates. Therefore, 5G
flights. The reason is that short-haul flights generate less data
is a promising technology to transfer a massive amount of
during the shorter flight duration, whereas the turnaround
data from the aircraft to the ground. Nevertheless, proper
time stays the same. We also analyzed the aircraft size and
deployment and sufficient bandwidth is necessary such that
clustered them in small, medium, and large according to their
these speeds can be supported by the network.
wake turbulence categories. In the per flight analysis, medium
aircraft need the highest data rates as they have short turn-
around times as small aircraft, but longer flight times. Interest-
AIRPORT BACKHAUL CAPACITY
ingly, if we compare different aircraft sizes per day, the
differences vanish. Hence, the relation between ground time While the previous sections focus on the needed data
and flight time during one day is independent of the aircraft rate for one aircraft, this section takes the total
Table 2.
capacity per airport into account. As there are multiple THROUGHPUT MEASUREMENTS OF 60-GHZ WI-FI
aircraft at the airport at any point in time, the total
data rate per airport is a critical parameter, in particu- From the currently available technologies discussed in
lar for large airports. Table 1, 60-GHz Wi-Fi promises one of the highest data
Figure 4 shows the number of aircraft over a week rates. Additionally, it is interesting, as the 60-GHz spec-
at the airport of Frankfurt, Germany. It can be seen trum is not yet as heavily used compared with other Wi-Fi
that the number of aircraft highly varies during the bands. We aim to verify the promised data rates by meas-
day. We can clearly identify the night times as hori- urements. Moreover, the existing hardware can give some
zontal lines, as there is a restriction to fly at night. The first practical insights into other technologies, such as 5G
pattern in Figure 4 looks different for other airports, mmWave. 60-GHz Wi-Fi is standardized as 802.11ad and
as, for example, some smaller airports have the highest
number of aircraft during the night. Other airports do
not have the restriction to fly at night; hence, there is
no time with a constant number of aircraft. For this
reason, the total needed data rate differs per airport
and cannot be generalized. Nevertheless, the used data
transfer systems need to be able to cope with the
expected amount of traffic per airport, depending on
the number of aircraft. For instance, if one aircraft
requires 1 Gb/s, the airport would need to provide up
Figure 4.
to 148 Gb/s concurrently, as up to 148 aircraft can be Number of aircraft at the airport in Frankfurt, Germany, during
present at the Frankfurt airport. one week.
Figure 6.
Measurement results of 802.11ad throughput with different angles
Figure 5. between access point and station.
Measurement setup consisting of an access point rotated with
angle a and a station in in a distance d.
other existing technologies today. However, compared
with the needed data rates as discussed in the section
802.11ay. It is also called WiGig and uses the 60-GHz “Relation Between Needed Data Rate on Ground and
band with a channel bandwidth of 2.16 GHz. 802.11ad Data Rate Generated on the Aircraft,” 1.5 Gb/s is still low.
supports up to 6.7 Gb/s. The standardization of 802.11ay Nevertheless, we believe that with a higher market pene-
is ongoing and with the introduction of MIMO and chan- tration and increasing number of deployments, the hard-
nel bonding, the expected data rate increases to 176 Gb/s ware and implementation limitations vanish and higher
including an extended range. 802.11ad has not been data rates become feasible. Still, there are a number of
widely adapted yet, but a few devices are available. limitations for mmWave links, such as range limitation,
Our measurement hardware consists of two SuperMi- line-of-sight limitation, high number of needed access
cro E300-9D-8CN8TP boxes with a Qualcomm QCA9500 points, or system throughput in a real deployment. Still,
802.11ad chip provided by BISDN GmbH. The manufac- these limitations can be feasible for an airport deployment,
turer states a peak speed of up to 4.6 Gb/s [19]. The as there is a line of sight between the gate and aircraft and
antenna is placed at the outside of the box and uses beam- the distance is small.
forming with different patterns to obtain the best link In conclusion, 60-GHz Wi-Fi is a promising technology
properties as shown in [20]. We measure the throughput for massive data transfer from and to the aircraft if the data
from the station to the access point. As the antennas for rates can be in the promise range or can even be increased in
60 GHz are highly directive, we also investigate different the future. Open questions are related to deployment and
angles between the two boxes. A small impact on the applicability in a real scenario, such as the number of access
angle is beneficial for deployment, as alignment between points needed and how to provide the respective backhaul
stations and access points is uncritical. For the measure- capacity. Also, measurements and tests need to be conducted
ment, two boxes were placed at a distance d ¼ 5 m in an in a realistic environment. These results can also be applied
outdoor environment. One box was acting as an access to 5G mmWave networks. While the frequency is lower, line
point, the other one as a station, as shown in Figure 5. Dur- of sight and range limitations still apply. With the option to
ing the measurement, the access point was rotated with operate a private 4G or 5G network, airports can operate their
different angles a. For the throughput measurement, own cellular network and adapt it to the special environment.
iPerf [21] was used. The throughput of each angle setup We also observe that the measured rates are considerably
was measured by sending UDP traffic for 30 s twice. The below the peak speeds. This has to be taken into account
measurement results are shown in Figure 6. For angles when deciding for a suitable technology.
> 90 , no connection is possible in the outdoor environ-
ment. However, in an indoor environment connections
with larger angles are feasible due to reflections. Figure 6
shows that the throughput is stable at 1.5 Gb/s irrespective
CONCLUSIONS
of the angle. The degraded performance at 90 compared We analyzed real flight traces from FlightRadar24 [15]
with 90 is due to the asymmetric antenna placement. toward the needed data rate for data transfer from the air-
Nevertheless, this throughput is considerably below the craft to the ground. To achieve this, we compare the dura-
advertised peak speeds, although we observe very good tion of the flight with the duration on ground. We study
channel conditions during our measurements. The same two different options: 1) transferring data per flight or
observation has been reported by other papers, who mea- 2) per day. We show that if we transfer data per day, the
sured a throughput of up to 1.6 Gb/s [22] or 2.3 Gb/s [23]. needed data rate decreases up to 47% compared with per
According to them, the low observed throughput is due to flight. We also vary parameters such as flight time or air-
hardware and implementation limitations. The measured craft size. We conclude that long-haul flights are the most
throughput of 1.5 Gb/s is very promising compared with challenging flights. We show that, with today’s 4G
networks, it is feasible to support up to 65.9 GB generated [10] D. Gray, “AeroMACS - Delivering next generation com-
per flight hour, which increases to 4.4 TB with 5G. Hence, munications to the airport surface,” WiMAX Forum, Bea-
5G is a promising technology to transfer a massive amount verton, OR, USA, White Paper, 2015.
of data from the aircraft to the ground. We also highlight [11] B. M. Paolini and S. Fili, “AeroMACS: A common
the challenges of deploying an aircraft data transfer sys- platform for air traffic management applications,” Senza
tem. These are, among others, to provide a worldwide Fili Consulting, Ave, Seattle, WA, USA, White Paper,
availability or to provide a sufficient data rate to cover all pp. 1–14, 2014.
aircraft at one airport. Additionally, an analysis on the [12] Airbus, “Skywise.” Accessed: Nov. 22, 2019. [Online].
data processing needs to be conducted. If a part of the data Available: https://skywise.airbus.com/
could already be processed at the aircraft, the data rate [13] Thales, “Our connectivity solutions.” Accessed: Nov. 22,
needed for transfer on ground can be reduced. 2019. [Online]. Available: https://www.thalesgroup.com/
en/connectivity-solutions
[14] Teledyne, “Wireless ground link system.” Accessed: Nov.
ACKNOWLEDGMENTS 22, 2019. [Online]. Available: https://www.teledynecontrols.
com/en-us/Product
The work presented in this paper has been completed
[15] “FlightRadar24.” Accessed: Dec. 2, 2019. [Online]. Avail-
while S. Hofmann was with Airbus, Central Research and
able: https://www.flightradar24.com/
Technology, Munich, Germany.
[16] Airbus, “Big Data: Airbus is mining the wealth of knowl-
edge for aviation.” Accessed: Jan. 8, 2020. [Online].
evidence that research and development on military cogni- although listed as a topic in the original reference, was not
tive radio systems progress toward an actual system addressed by any of the collected papers. However, spec-
proven in an operational environment. We observe that trum mobility, as a phenomenon, has been touched upon
this formulation already identifies potential data sources in the margins of several papers. The number of articles
as well as alludes to the use of Technology Readiness Lev- by topics is shown below in Table 1.
els (TRL) as a part of the analysis [12]. Third, the papers were coded according to the TRL they
portrayed. TRLs have been used since the 1980s in the U.S.
National Aeronautics and Space Administration. The use
expanded to the U.S. Department of Defense in the 1990s as
METHOD well as globally. TRLs are used to describe the maturity of
This research has been inspired by the findings of [8] as to technology, especially within procurement processes [15].
where do we stand in MCRS research today. Is there going As most of the articles address a single functionality,
to be a “Valley of Death” hindering the evolution of algorithm, or solution, the TRLs are valid and relatively easy
research into products? Or are we already standing on the to apply. For example, TRL 3 was initially described by
brink of that valley? The research method of this survey is NASA as “analytical and experimental critical function and
a systematic literature review (SLR), following the steps characteristic proof-of-concept” and by the European Union
outlined in [13]. The research question for this survey is: as “Experimental proof-of-concept” [12]. Within wireless
Does publicly available technical research data provide communications research, simulation is often used to provide
evidence that R&D on military cognitive radio systems
progresses toward an actual system proven in an opera- Table 1.
tional environment. We observe that this formulation
already identifies potential data sources as well as alludes
Number of Articles by Thematic Categories According
to the use of TRLs as a part of the analysis [12]. to IEEE Topics [15]
The primary sources selected were the IEEE Xplore Digi-
tal Library, Elsevier ScienceDirect, SpringerLink, as well as Applications and services category 14
the Association of Computing Machinery Digital Library. articles
We collected explicitly “military” AND “cognitive radio” New economics 1
related articles. As the first 14 years since the inception of the Machine learning techniques 4
notion of CR have already been covered in [9], we collected
Spectrum sharing and multiple 21
a dataset of military cognitive radio research covering the
access
years 2013 to 2020. The collected papers were then coded in
four dimensions: document type, thematic topic, the technol- Analysis of dynamic spectrum 2
ogy readiness level, as well as the CCMM level. access
First, the collected papers were manually categorized into Routing 23
conference papers, journal articles, doctoral dissertations, and
Security and electronic warfare 41
survey-type articles. The original intention was to use this
grading as an indication of the maturity and quality of the arti- Modeling of spectrum use 21
cle; however, it proved to be far from a useful indicator. Simulation tools and testbeds 11
Second, the dataset was coded into thematic catego-
Waveform design 55
ries, topics, already used in [14]. Spectrum mobility,
Table 2. Table 3.
Within the category CCMM, 3 defense lines of devel- Generally, spectrum sensing can be seen as a component
opment, there is one article that applies defense lines of or subsystem technology within a cognitive radio (as a node
development doctrine, organization, training, materiel, or device) or within a waveform (as a service or a compo-
personnel, leadership, facilities, information/interoperabil- nent thereof). Gupta and Kumar [26] is an example CCMM
ity (DOTMPLFI) framework to assess potential military 4 “system level,” well developed TRL 4, survey. In
capability implications the cognitive radio technology this article, spectrum sensing is brought to a broader
could introduce. context of cognitive radio networking. The article
Category CCMM 1 scope contains five articles—these reviews the spectrum sensing technologies and offers
address a range of topics, for example, from military R&D future research directions for cognitive radio networks.
to general spectrum administration. Nevertheless, the Spectrum sharing, spectrum access, and spectrum regu-
scope or viewpoint of these articles is above the military, lation have been addressed in several papers. The journal
i.e., national, technology wide, or even broader. article of [27] is an example of CCMM 1 “above military”
When the distribution of articles is mapped by CCMM way of addressing spectrum regulation for cognitive radios.
levels and publishing years, it is evident that general inter- At the same time, it stands out also as an example of con-
est is ongoing at the subsystem level of CCMM level 5, ceptual work at a low TRL 1. A comparison can be made
which with 156 articles, is the most populous category. with [28] at TRL 2, where the authors challenge the current
However, mapping CCMM viewpoints by TRLs in Table wireless coexistence paradigm. This article argues for a
4 again portrays heavy concentration to CCMM level 5 more generic solution in terms of frequency reuse and coex-
and to low TRLs 1-3. istence. A new generic medium sharing model that solves
coexistence problems in a simple, efficient, and technol-
ogy-agnostic way is presented. The suggested approach is
compatible with all existing wireless communication tech-
ADDITIONAL OBSERVATIONS nologies and incorporates capabilities for future modifica-
As we have already observed, the most populous catego- tions and additions to support emerging new technologies.
ries across the topics are waveforms, security, modeling As an example of the U.S. Department of Defense
spectrum use, and spectrum sharing. Therefore, it should funded military-oriented R&D, Zeng et al. [29] argues
come as no surprise that these articles concentrate on the that spectrum agility should be complemented with net-
category of CCMM 5 Subsystems (156 articles). work agility to achieve mission success. The authors sug-
Considering paper types across TRLs, we observe that gest the use of a proactive and adaptive cross-layer
a large number of conference papers have been published reconfiguration framework for autonomous network adap-
within the TRL1-4 domain (141 articles). Moreover, tation. As an individual potential solution to a broader
many journal articles, dissertations, and surveys are problem, the suggested approach is well in line with [11].
assigned to low TRL 1. This indicates that significant
research effort remains to clarify military cognitive radios’
functions, requirements as well as security aspects.
DISCUSSION
By far, most of the conference papers have been pub-
lished through IEEE, where IEEE/AFCEA Military The military CR research seems to be prolific on topics
Communications Conference (MILCOM) and Interna- like waveform design and security, not forgetting a steady
tional Conference on Military Communications and interest in CR networking topics. Our data indicate that
Information Systems (ICMCIS) stand out as most pro- the vast majority of papers and topics have been catego-
ductive venues. From the data, less than a dozen prolific rized to CCMM 5 “subsystem” level, indicating that the
authors emerge. However, these authors collaborate with TRLs are still valid tools for assessing maturity, as can be
established research groups we should recognize; inferred from Table 4.
Fraunhofer FKIE, Intelligent Automation Inc., NATO Significantly low numbers of papers address applications,
Science and Technology Organization, Royal Military services, and standardization. If continued, this may harm the
Academy of Belgium, Rockwell Collins Advanced Tech- scale and harmonization of the technology, as well as on
nology Center, Thales Communications, University of interoperability, which is critical for the practical deployment
Nevada, University of Oulu, and Wojskowa Akademia decisions across different communities of interest.
Techniczna of Poland. The annual distribution of military CR research by
In the process of searching, selecting, and coding TRLs in Table 5 indicates continuing interest. This finding
papers, some enticing papers stand out in addition to those is contradictory to general statistics from, e.g., IEEE
already mentioned. Naeem et al. [25] provided a compre- Xplore Digital Library, whereby the numbers of general
hensive survey of network coding techniques in CRNs. (civilian) CR research articles has been on the decline
Our data includes some other conference papers address- since 2014. This may indicate general technology matur-
ing the same topic but from a narrower perspective. ing from singular techniques to address broader system-
Table 4.
CCMM levels by TRL TRL TRL TRL TRL TRL TRL TRL TRL SUM
TRL 1 2 3 4 5 6 7 8 9
CCMM 1 3 2 5
CCMM 2 0
CCMM 3 1 1
CCMM 4 3 5 5 5 1 1 20
CCMM 5 16 60 58 20 1 1 156
CCMM 6 5 5 1 11
SUM 27 73 63 26 2 1 1 0 0 193
Table 5.
Numbers by years to TRL 2013 2014 2015 2016 2017 2018 2019 2020 SUM
TRL 1 5 3 4 2 8 1 4 27
TRL 2 13 9 12 9 8 13 9 73
TRL 3 11 10 9 7 7 3 14 2 63
TRL 4 7 1 8 7 3 26
TRL 5 1 1 2
TRL 6 1 1
TRL 7 1 1
TRL 8 0
TRL 9 0
SUM 36 24 35 18 30 17 31 2 193
Data for year 2020 includes articles published by the time of data collection at the end of the May 2020.
level issues. However, this is not yet evident in military CONCLUDING REMARKS
CR research data.
Machine learning and artificial intelligence seem natu- We collected 193 articles on military CR research from
ral techniques to implement CR, and the interest is on the well-known publishers covering the period of 2013–2020.
rise toward the end of our study epoch. In general, military Our selection of data sources led our dataset to be domi-
CR research seemed sporadic and scattered. Our data sug- nated by subsystem level, low TRL-types of articles.
gest that a transition from individual techniques, node- Therefore, our data indicate that research on military cog-
level subsystem research in the direction of the CRS sys- nitive radio systems remains at the subsystem (compo-
tems engineering approach, and system-level studies has nent) level and at low TRLs. Besides one outlier, there is
not yet happened, and the overall TRLs remain low. no indication that anyone has been able to bridge the gap
The authors recognize that coding, as described, may not from TRL 5 to TRL 7. Toward the end of our study epoch,
be clear-cut and may, at times, be arbitrary. Papers may the increasing interest in machine learning techniques
address topics that overlap more than one category. Although may eventually lead to a shift in this technology’s TRLs
every attempt has been made to ensure that our dataset is in the coming years. Higher system level or high TRL
comprehensive and representative, we cannot claim that all related research questions should be addressed by select-
papers applicable made their way into our dataset. ing different kinds of data sources and remains a potential
future research topic. Nevertheless, research on military [11] NATO Science and Technology Organization, “AC/323
CRSs is nowhere near reaching the “testing validity in a (IST-140)TP/874 cognitive radio networks: Efficient solu-
relevant environment.” The brink of the “Valley of Death” tions for routing, topology control, data transport and net-
has not been reached yet. work management,” 2019.
The absence of publicly articulated comprehensive treat- [12] EARTO, The TRL Scale as a Research & Innovation
ment of the notion of military cognitive radio system (i.e., Policy Tool, EARTO Recommendations. Earto Impact
CCMM levels 1-3) is graphic and has lead to nonintercon- Delivered, 2014.
nected scattered research at the subsystem level (i.e., CCMM [13] S. Keele and others, “Guidelines for performing systematic
levels 4-5). The lack of such treatment also manifests in the literature reviews in software engineering,” Tech. Report,
absence of understanding, which components or subsystems Ver. 2.3 EBSE Technical Rep.. EBSE, 2007.
constitute a MCRS. Therefore, the authors suggest that mili- [14] “IEEE COMSOC best readings—Topics on cognitive
tary operational, acquisition, as well as academic research radio,” Accessed: Mar. 27, 2018. [Online]. Available:
and manufacturing communities, should all come together to www.comsoc.org/best-readings/topic/cognitive-radio
frame a generic, publicly available vision for a MCRS. [15] A. Olechowski, S. D. Eppinger, and N. Joglekar,
“Technology readiness levels at 40: A study of state-
of-the-art use, challenges, and opportunities,” in Proc. Port-
REFERENCES land Int. Conf. Manage. Eng. Technol., 2015, pp. 2084–2094.
[16] L. Rose, R. Massin, L. Vijayandran, M. Debbah, and
[1] J. Mitola and G. Q. Maguire, “Cognitive radio: Making C. J. Le Martret, “CORASMA program on cognitive radio
software radios more personal,” IEEE Pers. Commun., for tactical networks: High fidelity simulator and first
vol. 6, no. 4, pp. 13–18, Aug. 1999. results on dynamic frequency allocation,” in Proc. IEEE
[2] B. Fette, Cognitive Radio Technology, 2nd ed. New York, Mil. Commun. Conf., 2013, pp. 360–368.
NY, USA: Academic Press, 2009. [17] Y. Sagduyu, S. Soltani, T. Erpek, Y. Shi, and J. Li, “A uni-
[3] S. H€am€al€ainen, H. Sanneck, and C. Sartori, LTE Self- fied solution to cognitive radio programming, test and
Organising Networks (SON): Network Management Auto- evaluation for tactical communications,” IEEE Commun.
mation for Operational Efficiency. Hoboken, NJ, USA: Mag., vol. 55, no. 10, pp. 12–20, Oct. 2017.
Wiley, 2012. [18] J. Koivisto and T. Tuukkanen, “Comprehensive capability
[4] “Cognitive radio systems in the land mobile service,” Int. meta model tested by a cognitive radio,” in Proc. IEEE
Telecommun. Union, Tech. Rep. ITU-R M.2330, 2019. Mil. Commun. Conf., 2017, pp. 731–737.
[Online]. Available: https://www.itu.int/dms_pub/itu-r/opb/ [19] T. Riihonen, D. Korpi, O. Rantula, and M. Valkama, “On
rep/R-REP-M.2330-2014-PDF-E.pdf the prospects of full-duplex military radios,” in Proc. Mil.
[5] J. Greitemann, B. Stahl, N. Michels, B. Lohmann, and Commun. Inf. Syst. Int. Conf., 2017, pp. 1–6.
G. Reinhart, “Quantitative model of the technology life- [20] K. Mourougayane and S. Srikanth, “A tri-band full-duplex
cycle for forecasting the maturity of manufacturing tech- cognitive radio transceiver for tactical communications,”
nologies,” in Proc. IEEE Int. Conf. Manage. Innov. IEEE Commun. Mag., vol. 58, no. 2, pp. 61–65, Feb. 2020.
Technol., 2014, pp. 290–294. [21] K. M. Borle, “On robust and secure wireless communica-
[6] R. Gross, R. Hanna, A. Gambhir, P. Heptonstall, and tion system design using software-defined radios,” Ph.D.
J. Speirs, “How long does innovation and commercialisa- dissertation, Syracuse Univ., Syracuse, NY, USA, 2015.
tion in the energy sectors take? Historical case studies of [Online]. Available: https://surface.syr.edu/etd/321
the timescale from invention to widespread commercialisa- [22] A. Robertson, L. H. Tran, J. Colon, T. V. Mai, S. Kompella,
tion in energy supply and end use technology,” Energy E.-H. F. Fu, and J. Molnar, “Network performance and spectral
Policy, vol. 123, pp. 682–699, 2018. behavior of cognitive radios during a coexistence field test,”
[7] S. K. Markham, “Moving technologies from lab to market,” in Proc. IEEE Mil. Commun. Conf., 2014, pp. 1399–1404.
Res. Technol. Manage., vol. 45, no. 6, pp. 31–42, 2002. [23] R. Matyszkiel, P. Kaniewski, R. Polak, and D. Laskowski,
[8] S. Debois, T. Hildebrandt, M. Marquard, and T. Slaats, “Selected methods of protecting wireless communications
“Bridging the valley of death: A success story on danish against interferences,” in Proc. Int. Conf. Mil. Commun.
funding schemes paving a path from technology readiness Inf. Syst., 2019, pp. 1–5.
level 1 to 9,” in Proc. IEEE/ACM 2nd Int. Workshop [24] M. Marwick, C. Kramer, and E. Laprade, Analysis of Sol-
Softw. Eng. Res. Ind. Pract., 2015, pp. 54–57. dier Radio Waveform Performance in Operational Test.
[9] B. Fette, “Fourteen years of cognitive radio devel- Alexandria, VA, USA: Institute for Defence Analyses,
opment,” in Proc. IEEE Mil. Commun. Conf., 2013, Alexandria, VA, USA, 2015.
pp. 1166–1175. [25] A. Naeem, M. H. Rehmani, Y. Saleem, I. Rashid, and
[10] H. Tang and S. Watson, Cognitive radio networks for tac- N. Crespi, “Network coding in cognitive radio networks:
tical wireless communications. DRDC–Ottawa Research A comprehensive survey,” IEEE Commun. Surv.\Tuts.,
Centre, DRDC-RDDC-2014-R185, 2014. vol. 19, no. 3, pp. 1945–1973, Jul.–Sep. 2017.
[26] M. S. Gupta and K. Kumar, “Progression on spectrum [28] I. Jabandzic, S. Giannoulis, and I. Moerman, “Enabling
sensing for cognitive radio networks: A survey, generic wireless coexistence through technology-agnostic
classification, challenges and future research issues,” dynamic spectrum access,” Wireless Pers. Commun.,
J. Netw. Comput. Appl., vol. 143, pp. 47–76, Oct. vol. 106, no. 1, pp. 151–177, 2019.
2019. [29] H. Zeng et al., “From spectrum agility to network agility:
[27] M. Saint and T. X. Brown, “A dynamic policy license for Proactive and adaptive reconfiguration for reliable commu-
flexible spectrum management,” Telecommun. Policy, nication in tactical networks,” in Proc. IEEE Mil.
vol. 43, no. 1, pp. 23–37, 2019. Commun. Conf., 2013, pp. 1663–1668.
semi-active [10], [11] and active hybrid [12], [13] ESS selection details, are described in the “JPL/CSUNSat1
designs attempt to optimize current flow and limitations Development” and “Hybrid ESS Payload Development”
associated with voltage mismatches in the battery or SC sections. Upon completion of the battery payload and the
bank with the addition of power converter electronics. Of CubeSat integration, functional, and environmental testing
these designs, the parallel active hybrid design is shown to were performed to properly qualify all systems for space
have the highest energy efficiency, reliability, and flexibil- operation, as described in the “Hybrid ESS Payload Devel-
ity by incorporating two dc–dc converters between the opment” section. The launch of JPL/CSUNSat1 from the
SCs and the load, and between the battery and the dc link ISS on May 17, 2017 marked the beginning of the primary
[14] to achieve a nearly constant output current from the mission phase to characterize the hybrid ESS payload in
battery with the remaining power supplied by the SCs. space. The performance results for the payload battery are
The addition of power converters can reduce the efficiency described in the “Primary Mission Telemetry” section.
of the overall system through the conversion stages. Pas- Despite operating in a degraded mode after completing the
sive hybrid ESS is by far the most preferred battery-SC primary mission objectives, the spacecraft also successfully
hybrid for research [15]–[18] and commercial [19], [20] completed an extended mission phase to evaluate cycle life
applications, as the battery and SC are connected in paral- performance and model validation. The “Extended Mission
lel directly to the load, thereby providing nearly unim- Telemetry” section describes the results obtained in the
peded current flow as determined only by internal extended mission phase of the mission.
resistances. It reduces the need for power electronics and
control circuitry, which significantly improves the cost,
volume, and most importantly reliability [21].
In this article, we report that JPL and California State JPL/CSUNSAT1 DEVELOPMENT
University Northridge (CSUN) have developed and flight The JPL/CSUNSat1 spacecraft was designed as a 2U
qualified a passive hybrid ESS for demonstration on the “tumbler” CubeSat (i.e., no attitude control capabilities)
low-Earth-orbit (LEO) CubeSat JPL/CSUNSat 1 [22], with the purpose of serving as a demonstration platform
which was successfully deployed from the International for a hybrid battery/SC ESS payload. The avionics con-
Space Station (ISS). The goal of the mission was to dem- sisted of two communication, three power, and one SCP
onstrate a power subsystem that was capable of delivering (single chip package) printed circuit boards (PCBs) con-
high power pulses at low temperatures, and to validate its forming to the CubeSat 10 cm x 10 cm x 10 cm form fac-
performance in an Earth orbiting mission. JPL had the tor. The spacecraft power system provided charging and
responsibility for development of the ESS payload, while load distribution to both the spacecraft bus battery (a sin-
CSUN was responsible for the spacecraft development. gle Li-ion 18650 cell) and the ESS payload. The power
The full chronology of the JPL/CSUNSat1 project, bus supported operation from either the single Li-ion cell
including milestones and significant events, is outlined in or the payload hybrid battery, using a custom-designed
Figure 1. The payload and CubeSat development efforts of power conditioning and control system. To achieve the
the project were completed by May of 2015. The CubeSat coldest possible operating temperatures during flight, the
design and payload cell performance results, as well as cell addition of cold and hot shunt regulator boards were used
Figure 2.
JPL/CSUNSat1 development. A) CAD rendering of JPL/CSUNSat1. B) CAD rendering of JPL/CSUNSat1 avionics PCBs. C) Final fit check
prior to integration. D) Stack assembly. E) Integrated spacecraft with labeled location of the Li-ion and supercapacitor cells in the 3U CubeSat.
Figure 3.
Summary of ground performance testing at JPL. A) Discharge capacity performance of 26650 LFP Li-ion cell at 20 and –40 C. B) Perfor-
mance comparison of COTS supercapacitor cells rated at 100, 150, and 310 F at –40 C. C) Pulse discharge comparison of a mockup hybrid
battery versus Li-ion cell at 15 Amp discharge rate at –40 C. D) Percent of discharge capacity distribution between Li-ion versus supercapa-
citors of the hybrid battery over various temperatures. This data was determined experimentally by separate measurements on current in the
supercapacitor bank and the Li-ion cell during discharge at approximately 2 Hz sampling rate using an external data acquisition (DAQ) cou-
pled with a known constant load source at set temperatures.
Figure 4.
A) Functional block diagram of payload electronics board. B) Prototype engineering model of JPL’s payload electronics board. C) Integrated
hybrid ESS, following thermal vacuum testing.
ion cell at –40 C. This is largely due to the lower imped- payload was designed to be a fully integrated system contain-
ance leading to a faster response time of the SC cells under ing the battery cell, SCs, and a payload electronics board. It
pulse operating conditions [22]. Such power enhance- was constrained to fit within a 1/2U volume including all the
ments at lower temperatures and higher power capability electronics necessary to control the experiment. This
can potentially enable future deep space flight applications approach was chosen since it allowed the payload to be
with high burst power demands, such as communication completely tested at JPL prior to integration into the CubeSat.
events [24], [25]. Another significant benefit of the hybrid The payload electronics as illustrated in Figure 4 con-
ESS is its excellent load sharing capabilities. As tained the battery charger, a bank of discharge resistors,
highlighted in Figure 3D, the SC bank takes a higher per- fault protection (FP) circuitry, switching circuitry, com-
centage of the discharge capacity with decreasing temper- mandable switches, and telemetry collection circuitry.
atures (up to 80% at lowest temperatures of –40 C). Interfaces between the payload and spacecraft were mini-
Consequently, this could also have significant improve- mized in order to facilitate a more independent develop-
ment on cycle life of the Li-ion cell over long cycling peri- ment of the payload and the CubeSat at JPL and CSUN,
ods, due to lower effective depth of discharge and respectively. The only interface connections required
mitigation of known degradation mechanisms [7], [26]. from the payload to the CubeSat were the spacecraft
power bus and the connection to the spacecraft’s I2C com-
munication bus.
PAYLOAD ELECTRONICS Command and control of the payload was executed
The interfacing payload electronics were also designed at entirely through the I2C bus. Several I2C port expanders
JPL as part of the hybrid ESS payload development. The were used to control the various power paths, and to
Table 1.
interface to the telemetry collection analog-to-digital con- environmental testing, including thermal vacuum
verters. The payload electronics also included numerous (TVAC), pyro-shock, and random vibration testing.
power switches which allowed the battery to be directly
charged from the spacecraft power bus and discharged
through a bank of four individually controlled resistors.
This enabled an extended mission mode, allowing the pay-
PAYLOAD ESS
load battery to function as the primary ESS of spacecraft. The Li-ion and SC cells were independently tested using a
The discharge switching circuitry was programmable to small thermal vacuum (TVAC) system as shown in
allow discharge current rates from 1 to 15 A in 1 A incre- Figure 5A-1. The chamber chassis accommodates up to
ments. The payload included I2C based analog-to-digital three test cells, and is packed with aluminum foil to
converters, which captured telemetry including battery improve thermal conduction under vacuum. The TVAC
voltage and current, SC voltage and current, charge cur- chamber featured feedthroughs to allow for voltage, cur-
rent, discharge current, and battery temperature. rent, and temperature monitoring, and to enable functional
The payload electronics also included extensive FP testing of the payload flight cells through its electronics
circuitry which individually isolated the battery and board. The functional block diagram of the TVAC testbed
SCs from the power bus, performed cell balancing of is shown in Figure 5A-2, and further description provided
the two SC cells, and most importantly prevented bat- in a previous work [27]. The nominal base pressure was
tery overcharge (which could lead to catastrophic ther- less than 10-6 Torr which is representative of the space
mal runaway and mission termination). A summary environment. An Arduino microcontroller board was
description of all CSUNSat1 payload FP conditions is employed as part of the testbed to send commands through
provided in Table 1. the payload electronics and retrieve payload telemetry.
Standard functional tests consisted of a short discharge
pulse(s) (<60 s) under vacuum conditions in order to min-
imize heat dissipation. The TVAC test plan included two
ENVIRONMENTAL TESTING AND FINAL INTEGRATION
thermal cycles ranging from 60 to –40 C, with a one to
To achieve flight readiness in a relevant environment two hours dwell period prior to functional testing. The
(Technology Readiness Level 6), both the ESS payload results indicated nominal behavior with the exception of
and the JPL/CSUNSat1 spacecraft were subjected to an observed 200 mV difference between the two SC
Figure 5.
Summary of environmental testing for JPL/CSUNSat1 project. A-1) JPL/CSUNSat1 payload inside JPL’s TVAC chamber, including flight
cell test chassis (picture on right) for TVAC environmental testing. A-2) Functional block diagram of the CubeSat TVAC system. A-3) Rep-
resentative TVAC discharge pulse test data under vacuum (pressure < 10-6 Torr) at –40 C. B-1) Block diagram on the integrated JPL/
CSUNSat1. B-2) Random vibration test profile on the integrated CSUNSat1. The overall root mean square acceleration (grms) in the x, y,
and z directions was determined to be 9.51, 9.49, and 9.53 grms, respectively. B-3) Integrated voltage measurements under TVAC.
cells, most likely attributable to variations in resistance before and after the TVAC testing. As summarized in
between the voltage measuring electronics and contact Table 2, the cell mass differences of all the flight cells
with the cell terminals. Figure 5A-3 illustrates a represen- were negligible after TVAC testing, indicating no electro-
tative pulse discharge test at –40 C of 7 A in hybrid lyte leakage from a cell can breach.
ESS mode through one of the integrated load resistors.
The voltage and current distributions provided good indi-
cations of functionality and performance of the entire pay-
load assembly including optimization of payload
CSUNSAT1 INTEGRATION TESTING
electronics to minimize cell leakage and electrical resis- After successful environment cell testing, the hybrid
tances. Individual cell masses were also measured both ESS payload assembly was integrated into an aluminum
block casing designed specifically to interface to the
JPL/CSUNSat1 spacecraft, as shown in Figure 4C. The
Table 2.
hybrid ESS payload was successfully integrated into the
Summary of Cell Mass Pre- and Post-TVAC spacecraft, and was subsequently subjected to environ-
mental testing.
The pyroshock, random vibration and TVAC testing
Cell Mass (gm)
on the CubeSat with the ESS payload in place were then
Cell ID Pre-Tvac Post-Tvac carried out. The test data (see Figure 5B-2) was used to
calculate the overall root mean square acceleration (grms)
Li-ion 85.120 85.110
in the x, y, and z directions to be 9.51, 9.49, and 9.53
Supercap #1 73.350 73.350 grms, respectively. In addition, no debris or particles were
Supercap #2 75.000 75.000 detected upon inspection after the test up to launch ran-
dom vibration levels.
Figure 6.
Summary of downlinked telemetry data from CSUNSat1 during primary operations. A) Payload battery initial charge cycle. Telemetry cov-
ers over 20 h over multiple downlinks from CSUNSat1. The data provided an initial estimate of the available capacity of the hybrid capacity
at about 2.3 Ah. B) Initial hybrid mode checkout test. This data shows expected load-sharing distribution between the Li-ion cell and super-
capacitor bank at about 20 C temperature. C) Nominal temperature experiment on Jun. 7, 2017 from CSUNSat1 operating in battery mode.
D) Cold temperature experiment on Jun. 14, 2017 from JPL/CSUNSat1 operating in battery mode. E) Supercapacitor performance experi-
ment on Jun. 8, 2017 from CSUNSat1 operating in supercapacitor mode. F) Representative nominal temperature experiment on Jun. 16,
2017 from JPL/CSUNSat1 operating in hybrid mode.
The JPL/CSUNSat1 TVAC test configuration block windows with JPL/CSUNSat1 from the CSUN Tracking
diagram is shown in Figure 5B-1. The TVAC test incor- Stations were 10–20 min for command uplink and/or
porated a solar simulator to provide power generation telemetry downlinks, including the transfer of space-
and command capabilities through the actual ground craft state-of-health and payload battery flight experi-
control equipment. Prior to TVAC testing the spacecraft mental data. The JPL/CSUNSat1 provided telemetry at
solar array was installed. This prevented the CubeSat a resolution of up to 1 s. The initial charge cycle of the
from being directly mounted on the cooling baseplate, Li-ion cell is shown in Figure 6A. The 20þ h of teleme-
which limited thermal conduction within the system. try required multiple downlinks, resulting in a capacity
Coupled with significant thermal dissipation, the space- estimate of 2.3 Ah, in good agreement with ground
craft was unable to achieve sub-zero temperature under test data. During the initial hybrid battery checkout test,
vacuum. Actual test results from JPL/CSUNSat1 under the expected load-sharing performance of the hybrid
TVAC conditions are shown in Figure 5B-3. The 3 V battery was confirmed using a 1 A discharge rate (see
logic supply voltage, bus voltage, Li-ion cell voltage, Figure 6B). At 20 C, the percentage of discharge
and SC voltage behaved as expected under various capacity for the Li-ion cell was 60% versus 40% for
charge and discharge conditions throughout the two to the SC bank, in excellent agreement with ground test
three day TVAC test campaign. The analog data noise data shown in Figure 3D. Overall, the payload battery
resulted from use of a single-ended 8-bit data acquisi- appeared to be in a good state-of-health, based on the
tion (DAQ) system. Upon completing the TVAC test, initial checkout tests during flight.
the JPL/CSUNSat1 successfully passed all environmen- The primary mission phase also included payload
tal testing and was delivered to NanoRacks (CubeSat battery characterization experiments at nominal and
launch services provider) for preparation for launch to cold temperatures in battery, SC, and hybrid ESS oper-
the ISS. ating modes and at pulse discharge rates up to 15 A.
The nominal and cold temperature flight experiments in
battery mode were conducted (see Figure 6C and D).
During the 15 A pulse, the temperature for the nominal
PRIMARY MISSION TELEMETRY test was approximately 4 C. Due to a combination of
The JPL/CSUNSat1 launch from the ISS marked the thermal radiation from LEO and the CubeSat thermal
beginning of the operational phase to characterize the environment, the lowest achievable temperature
hybrid ESS payload. Typical communication time onboard CSUNSat1 was –2 C. Table 3 displays the
Table 3.
internal resistance of the battery observed during both battery assembly. For the hybrid experiment, the opera-
flight experiments, which was 60 mV in battery oper- tion included an observable SC charge cycle. Prior to
ating mode. There was a 5–10 mV decrease from the enabling the hybrid ESS mode, the bank was charged to
nominal versus low temperature experiments, which a voltage difference of <10 mV, to minimize transients
highlights the accuracy of the JPL/CSUNSat1 telemetry and overcharging. As expected, the SC exhibited a
resolution. much lower resistances of about 5.5 mOhm average as
The telemetry from the SC mode is shown in compared to over 60 mOhm for the single Li-ion cell.
Figure 6E. Similar to ground tests, a 150 mV difference in The battery resistances are lower by an average of 2.2
voltage between SC cell 1 versus cell 2 was observed in mOhm during the low temperature hybrid experiments
flight. The corresponding capacitance calculated from as oppose to the battery experiments which suggests
flight data was 318 and 302 F, respectively, with an effec- improved load capabilities from the Li-ion cell in the
tive capacitance of about 155 F for the two cells in series. hybrid ESS. These in-flight results are in good agree-
The internal resistance of SC cell 1 was determined to be ment with ground test results reported previously [22].
over two times higher than that of SC cell 2 (about 20
mOhm total), which is likely due to the resistance differ-
ences in the lead and terminal connectors to the payload
electronics. These differences resulted in no appreciable
EXTENDED MISSION TELEMETRY
impact on overall performance in hybrid mode. The JPL/CSUNSat1 extended mission was carried out to
Figure 6F highlights one of two flight experiments provide life cycle characterization of the hybrid ESS and
on the performance telemetry in hybrid battery mode. to support performance model validation. In addition, the
The temperatures from both experiments, unfortunately, payload ESS was also successfully used as the primary
were similar at about –2 C, resulting in similar values ESS of JPL/CSUNSat1 for a brief period of several hours.
of internal resistances as shown in Table 3. These resis- However, the 2 GB storage device card onboard the space-
tances are significantly higher than values from ground craft failed during flight, prohibiting telemetry data pack-
performance testing, which indicates that non-trivial aging and downlinking. CSUN developed a workaround
ohmic resistances were present in the overall payload that used the conditional sequence execution capability
Figure 7.
A) Discharge capacity profiles from JPL/CSUNSat1 operating in reduced data mode during the extended mission, after its SD card
failure. B) Summary of maximum/minimum temperatures and coulombic efficiency from JPL/CSUNSat1 operating in degrade mode
after its SD card failure. The coulombic efficiencies of the hybrid battery were calculated from estimated charge and discharge
capacities.
along with the response poll feature. The sequence execu- in parallel with the Li-ion impedances containing electro-
tion capability included the feature that each step in the lyte resistance (Ro) in series with three –RC parallel units
sequence was conditionally executed based on one of any known as time constants to account for the total contribu-
of the 44 sensor values on the spacecraft along with abso- tions on voltage polarization during discharge. The –R1C1
lute time and relative time. This capability provided suffi- is the time constant for charge transfer, –R2C2 is the time
cient telemetry data to compete the extended mission constant for the solid electrolyte interphase (SEI) and cath-
successfully. ode electrolyte interphase (CEI), and –R3C3 provides the
estimated time constant corresponding to diffusion [31]. It
is worth noting that the impedance contribution from SEI
PERFORMANCE DATA formation is relatively small compared to charge transfer
and diffusion impedances at the beginning of cycle life
Using the CSUN workaround for telemetry access on the
[32], [33]. Parameterization of the model comes from meas-
payload battery, only 10–12 data points were available for
urements on open-circuit potential along with electrochemi-
each of the 10 charge/discharge cycles performed. Figure 7A
cal impedance spectroscopy (EIS) measurement over
compares the estimated discharge capacities based on the
various temperatures and state of charge (SOC). As illus-
reduced data set from the extended mission phase. No appre-
trated in Figure 8B-1, the impedance data vary significantly
ciable differences were observed in the discharge cycles,
with SOC as expected with the highest impedances at 0%
with an average of 2.6 þ/– 0.1 Ah. The coulombic efficiency
SOC. The electrolyte resistance (Ro) exhibited no signifi-
of the hybrid battery of each cycle was also estimated and
cant change versus SOC but rather varied only with temper-
summarized in Figure 7B. The average value was 96.8 þ/–
atures. The equivalent circuit shown in Figure 8B-2
1.5%. The hybrid ESS cycle life performance is in close
exhibited a sufficiently good fit to the EIS measurement
agreement with ground testing with differences attributable
with changes in both temperature and SOC.
to temperature variations during flight testing.
Despite the lack of telemetry resolution, the extended
mission telemetry data could be used to validate the hybrid
ESS performance model on prediction of useable capacity.
HYBRID BATTERY MODEL DEVELOPMENT AND As the impedance of the SC (Csc) goes to infinite during dc
VALIDATION (continuous) discharge, the capacity contributions of SC
are negligible leaving the Li-ion cell as the sole energy
An impedance-based circuit model of the JPL/CSUNSat1 storage source as described in Figure 8A-2. Solving this
hybrid ESS is shown in Figure 8A-1. The model can be circuit for V(t) results in the following equation:
used to describe both voltage and current distribution by
application of Kirchhoff’s circuit laws. Further mathemati-
cal details on these circuit models are described elsewhere X
3 t
in the literature [28]–[30]. The SC impedance is represented V ðtÞ ¼ Vocv þ Iapp R0 þ Iapp Rn 1 eeRn Cn
n¼1
by a resistor (RESR) in series with a variable capacitor (Csc),
Figure 8.
Summary of hybrid battery model development. A-1) Equivalent circuit model on the operation of the hybrid energy storage system. The
supercapacitor impedance is described by RESR in series with Csc. A-2) The circuit model for performance characterization during extended
mission phase. The Ro is the electrolyte resistance. The charge-transfer time constant is associated with R1 and C1. The SEI time constant is
associated with R2 and C2. The diffusion time constant is associated with R3 and C3. B-1) Example of complex plane plot from EIS measure-
ments for parameterization of impedance-based circuit model for Li-ion cell. A collection of EIS measurements were performed at various
SOC and temperatures. B-2) Each measurement fits sufficiently well to an impedance equivalent circuit containing R0 in series with three
–RC time constants. C-1) Simulation time profile of JPL/CSUNSat1 payload battery on discharge at T ¼ 5, 0, –5 C. C-2) Comparison of
simulation versus the telemetered discharge capacity profile. The simulation current rate was performed at 0.91 A.
with all three time constant variables as parameterized hybrid ESS in flight were in excellent agreement with
by EIS measurements performed on the ground on the ground test results, particularly with respect to the percent
same electrochemical cell chemistries using lookup of capacity contributions between the Li-ion cell and the
tables. The simulation of the discharge profile compar- SC. The large differences in internal resistances from
ing JPL/CSUNSat1 telemetry is shown in Figure 8C-2. flight experiments as compared to ground tests were attrib-
The simulated discharge rate of 0.91 A represents the utable to significant ohmic resistances of the payload bat-
average actual discharge rate through the 3.3 Ohm resis- tery interconnects (i.e., leads, connectors, etc.) and not
tance in the electronic payload. The simulated tempera- due to the inherent cell impedances. Despite operating in
ture of –5 C to achieve minimum residual error is lower LEO, the spacecraft was able to achieve sub-zero tempera-
than the actual payload battery temperature, which again tures critical for flight experiments on JPL’s cold capable
suggests significant ohmic resistances (i.e., leads, con- hybrid ESS.
nectors, etc.) which are mostly present in the payload
assembly. These were unaccounted for from ground-
measured impedances from EIS on individual cells (see ACKNOWLEDGMENT
Figure 8C-1). Overall, these data provided a good agree-
ment with the model and basis for future improvements NASA’s 2013 Small Spacecraft Technology Program
on development of such circuit models, for flight opera- funded the payload development that was carried out at
tions on ESS life, and long-term state-of health the Jet Propulsion Laboratory, California Institute of
characterization. Technology, under a contract with the National Aeronau-
tics and Space Administration and the CubeSat develop-
ment carried out at the California State University,
Northridge. The JPL/CSUNSat1 primary operational
SUMMARY phase was funded by NASA’s Center Innovation Fund.
The JPL/CSUNSat1 project successfully demonstrated a The JPL/CSUNSat1 extended mission phase was funded
hybrid ESS consisting of a Li-ion battery and SC technol- by JPL’s Lunar Flashlight CubeSat Project. This work is
ogies in flight onboard a 2U tumbler CubeSat. Both the dedicated to the memory of Dr. Sharlene Katz, an excep-
primary and extended mission phases were completed suc- tional educator and mentor to many students throughout
cessfully. The performance characteristics of the payload the CSUNSat1 project.
[29] B. Y. Liaw, G. Nagasubramanian, R. G. Jungst, and D. H. [32] M. C. Smart, B. L. Lucht, and B. V. Ratnakumar,
Doughty, “Modeling of lithium ion cells—A simple equiv- “Electrochemical characteristics of MCMB and LiNix-
alent-circuit model approach,” Solid State Ionics, vol. 175, Co1xO2 electrodes in electrolytes with stabilizing
pp. 835–839, 2004. additives,” J. Electrochem. Soc., vol. 155, no. 8,
[30] R. A. Dougal and R. E. White, “Power and life extension pp. A557–A568, 2008.
of battery–ultracapacitor hybrids,” IEEE Trans. Compon. [33] U. Troltzsch, O. Kanoun, and H.-R. Trankler,
Packag. Technol., vol. 25, no. 1, pp. 120–131, Mar. 2002. “Characterizing aging effects of lithium ion batteries by
[31] D. Andre, M. Meiler, K. Steiner, H. Walz, T. Soczka- impedance spectroscopy,” Electrochimica Acta, vol. 51,
Guth, and D. U. Sauer, “Characterization of high-power pp. 1664–1672, 2006.
lithium-ion batteries by electrochemical impedance spec-
troscopy. II: Modelling,” J. Power Sources, vol. 196,
pp. 5349–5356, 2011.
Yaakov Bar-Shalom, University of Connecticut Giuseppe Fabrizio, Defence Science & Athina Petropulu, Rutgers University
Target Tracking & Data Fusion Technology Optimum Co-Design for Spectrum Sharing
Robust Adaptive Array Processing for Radar Between MIMO Radar & MIMO Communication
W. Dale Blair, Georgia Tech Research Institute Systems
Tracking Maneuvering Targets in a World of Over-The-Horizon Radar
On Radar Privacy in Shared Spectrum Scenarios
Netted Sensors Alfonso Farina, Selex (Retired) Multidimensional Sparse Fourier Transform &
Systematic Filter Design for Tracking Radar Adaptivity: Antenna Based Signal Application to Digital Beamforming Automotive
Maneuvering Targets Processing Techniques Radar
Erik P. Blasch, US Air Force Research Lab Demoz Gebre-Egizabher, University of Robert C. Rassa, Raytheon Company
Overview of High-Level Information Fusion Minnesota Business Case for Systems Engineering
Theory, Models, & Representations Design & Validation of Fault-Tolerant Integrated
Multispectral Image Fusion & Night Vision Navigation Systems for Small UAVs Luke Rosenberg, DST Group
Colorization Cooperative & Networked Navigation New Concepts in Maritime Detection
Signal of Opportunity Navigation for Small The Importance of Sea Clutter Modeling
Michael S. Braasch, Ohio University
Characterization & Mitigation of Multipath in Spacecraft in Deep Space Roberto Sabatini, RMIT University
GNSS Felix Govaers, Fraunhofer FKIE Aerospace Cyber-Physical Systems
Fundamentals of Inertial Navigation Multi Sensor Fusion in Distributed Systems
Fundamentals of Inertial Aiding George T. Schmidt, MIT/Draper Lab (Retired)
Maria Sabrina Greco, University of Pisa Inertial System & GPS Technology Trends
Eli Brookner, Raytheon (Retired) Sea & Land Clutter Statistical Analysis & Navigation Sensors & Systems in GNSS Degraded
Radar, Phased-Arrays, Metamaterials (Invisible Modeling & Denied Environments
Man), Stealth, Anti-Stealth, Ultra-Wideband, Advanced Techniques of Radar Detection in Non- Inside Apollo: Heroes, Rules & Lessons Learned
Cognitive Adaptivity, MIMO, 5G: Advances and Gaussian Background in the Guidance, Navigation, & Control (GNC)
Breakthroughs Sensor Selection for Multistatic Radar Networks System Development
MIMO Radar: My Story Taken Out of It
Metamaterial for Low Cost Electronic Scanning, Hugh D. Griffiths, University College London Vince Socci, LHP Software
Wideband Conformal Antennas, Cloaking (The The Challenge of Waveform Diversity How it Works: UAV Technology Overview
Invisible Man), Stealth and Waim. Bistatic & Multistatic Radar How it’s Used: UAV Applications & Business
Around the World in 60 Minutes Opportunities
Cognitive Adaptive Array Processing for Radar
Wolfgang Koch, Fraunhofer FKIE How it’s Managed: UAV Policies & Regulations
Multistatic Exploration: Intro to Modern Passive A Course for New Drone Operators
David Brown, Southwest Research Institute Radar & Multistatic Tracking & Data Fusion
Application of SOSA to Airborne EW Tracking & Sensor Data Fusion Roy Streit, Metron, Inc.
Cognitive DF for Airborne Systems Analytic Combinatorics for Multi-Object Tracking
Kathleen Kramer, University of San Diego & Higher-Level Fusion
Frederick E. Daum, Raytheon Feature Object Extraction
Is There a Royal Road to Robustness? Navigation: The Road to GPS & Getting Hongbo Sun, Nanyang Technological University
MIMO Radar: Snake Oil or Good idea? Beyond It Passive Through-Wall Human Sensing with WiFi
Never Trust a Simulation without a Simple Countering the Drone’s Threat by Radar
Krzysztof Kulpa, Warsaw University of
Back-of-the-Envelope Calculation that Explains It Peter Willett, University of Connecticut
Technology
Nonlinear Filters with Particle Flow A Primer on Various Approaches to Data
Passive Radar Technology: Ground-based &
Real World Data Fusion Association
Moving Platform Challenges
Mark E. Davis, Independent Consultant Noise Radar Technologies Maximum-Likelihood Methods in Target Tracking
Ultra Wideband Surveillance Radar Radar Technology: New Trends & Frontiers & Fundamental Results on Trackability
Passive & Active Radar Distributed Detection & Data Fusion
Antonio De Maio Università "Federico II" di
Napoli Lorenzo Lo Monte, Telephonics Yuanxin Wu, Shanghai Jiao Tong University
Radar Detection, Performance, & CFAR Radar Systems Prototyping Inertial Navigation: Sensing & Computation into
Techniques History & Future of Radar & EW the Future
Optimization Theory in Radar Signal Processing Birsen Yazici, Rensselaer Polytechnic Institute
Uttam Kumar, US Air Force Research Lab
Walter D. Downing, Southwest Research Deep Learning for Radio Frequency Target Machine Learning for Radar Sensing & Imaging
Institute Classification Non-Convex Optimization for Active & Passive
Space Avionics & Scientific Instruments for Radar
Y. Jade Morton, University of Colorado Boulder
Unmanned Space Missions
Satellite Navigation & Sensing
Bridging the Valley of Death
Feature Article: DOI. No. 10.1109/MAES.2020.3040511
an organized framework, which motivates the existence of These are the parts that support the interfaces and pro-
this work. This framework is also compatible with RTCA cesses by which a system can be affected or interacted
DO-326A [20] and RTCA DO-356 [21]. with[20]. Airworthiness security is the necessary protec-
tion that an aircraft must provide to mitigate information
RELATED STANDARDS security threats. The security perimeter includes the por-
tions that support the following:
The development of aeronautical systems is guided by
physical links (e.g., Ethernet ports, wireless
several existing standards. In this section, we identify the
transceivers);
main existing standards with correlation of our work.
The early 1990s were characterized by a rapid increase logical links (e.g., IP stack);
in the extensive software usage in aircraft, engines, and air- network protocols (e.g., DNS, ICMP, gateways,
borne pieces of equipment [22]. This trend has resulted in packet filters);
the industry’s need to create its guide and regulatory mate-
rial to drive system and software development. The SAE network services and clients (e.g., HTML server,
ARP 4754A[23] and RTCA DO-178C [6] were created sat- FTP client/server, IPSEC server);
isfying these needs. It guides the aeronautical community remote applications (e.g., file transfer services,
on system and software development processes [24]. remote monitoring, and web applications).
The SAE ARP 4754A [22] discusses the development
of aircraft systems taking into account the overall aircraft The ARINC 666 [26] defines the airline industry
operating environment and functions. This includes the standards for the electronic distribution of software. It
validation of requirements and verification of the design describes the common principles and rules applied when
implementation for certification and product assurance. transferring FLS from suppliers (SUP) to aircraft mainte-
The RTCA DO-178C [6] has a significant number of nance facilities. The software loading has become a com-
objectives associated with software requirements develop- mon maintenance action. It is therefore desirable to make
ment, using as input system requirements that will be the software readily available to the mechanic. The soft-
implemented by software. There are two levels of software ware may be stored on physical media such as a floppy
requirements on RTCA DO-178C. Software high-level disc, hard drive, CD-ROM, PC Card, or other devices until
requirements generally represent “what” should be needed. To ensure availability, the physical media con-
designed. They include functional, performance, interface, taining the software is either installed on the aircraft or
and safety-related requirements. The software low-level located where aircraft maintenance is performed.
requirements represent the how-to, providing details on The ARINC 667-2 [27] guides the in-service manage-
implementing software in code [25]. They include the fea- ment of the FLS. This standard defines aircraft controlled
tures required for source code development, such as data software, which describes how aircraft software is con-
coupling and control features. trolled regardless of how the software is loaded. Although
The RTCA DO-326A [20] guides by defining activi- ARINC documents provide good information about FLS,
ties for supplementing the aircraft development and certi- they do not specify a framework.
fication process to demonstrate that the effects on the
safety of the aircraft of such unlawful interference are con-
fined within acceptable levels.
METHODOLOGY
The RTCA DO-356 [21] describes guidelines, methods, We initially researched the existing standards in the industry
and tools used in performing an airworthiness security pro- associated with the development of systems. Then, we
cess. Airworthiness security is the protection of the airworthi- looked for related works that eventually describe experience
ness of an aircraft from intentional unauthorized electronic reports on the development of systems with field-loadable
interference. This includes the consequences of malware and capability. It is important to say that there is no other pub-
forged data and access to other systems to aircraft systems. lished paper that describes or evaluate field-loadable capa-
The security perimeter catalogs the parts of the aircraft bility. So, we identified this as a lack of research, which
or system that contact external systems or populations. motivates this work and the framework proposal.
Figure 1.
Results of Survey 1.
With this motivation, we applied two surveys to iden- Finding 2: 95% of systems developers agree that
tify the relevance of creating a reusable framework for the identity is important to ensure completeness and
development of systems with field-loadable capability. correctness (ST2).
Our surveys involved two audiences: systems developers
Finding 3: 80% of systems developers agree that
and aviation maintenance technicians, as these are usually
interchangeability analysis among different redun-
responsible for the loading task. Next, we define the
dancies must be always performed (ST3).
framework itself, by identifying the actors, how an NSW
can become available in the field, its reusable require- Finding 4: 60% of systems developers agree that
ments, and testing procedures. software must be always controlled separately from
the hardware (ST4).
Finding 5: 100% of systems developers agree that a
framework with reusable requirements and tests
SURVEYS
will be useful for System Developers (ST5).
This research conducted two surveys. Survey 1 identified
Finding 6: 60% of systems developers do not spec-
the main needs and practices performed by system devel-
ify the requirements for software loading during a
opers and mapped their knowledge and practices in LASs.
LAS project (ST6).
On the other hand, Survey 2 focused on aviation mainte-
nance technicians, responsible for the task of loading in Finding 7: 60% of systems developers do not con-
aeronautical maintenance facilities. duct tests to ensure different scenarios of software
Survey 1 involved 20 systems developers and included loading (ST7).
eight statements. Out of the 20 system developer, 15 are
considered senior, as they have more than ten years of avi- Survey 2 involved 12 aviation maintenance techni-
ation experience. The other five are considered junior, as cians, responsible for software loading. Out of the 12 avia-
they have less than ten years of experience. The system tion maintenance technicians, eight are considered senior,
developers answered a set of questions with the following as they have more than ten years of aviation experience.
scales: (1) Strongly disagree; (2) Disagree; (3) Neutral; The other four are considered junior, as they have less
(4) Agree; and (5) Strongly agree. Figure 1 presents the than ten years of experience. Survey 2 involved a ques-
results of Survey 1. tionnaire with three questions. The aviation maintenance
We found seven findings during Survey 1. We con- technicians answered the questions with the following
sider responses (3) Neutral, (4) Agree, and (5) Strongly scales: (1) Never (0%); (2) Rarely (0%–5%); (3) Eventu-
agree, to consider that answers provided agree with the ally (5%–20%); and (4) Often (more than 20%). Figure 2
proposed statements (ST1–ST7). The findings identified presents the results of Survey 2.
during Survey 1 are as follows. The findings identified during Survey 2 are as follows.
Finding 1: 100% of systems developers agree that Finding 8: Load capacity—to ensure that the LAS
integrity is important to ensure completeness and has an interface capable of transferring the software
correctness (ST1). from the source (Q1, Q2, and Q3).
Figure 2.
Results of Survey 2.
Figure 3.
F-LAS framework structure.
Figure 4.
Software distribution.
Loading Interface (LI): Physical interface for soft- facilities) to load the NSV inside the LC, and then proceed
ware loading. in aircraft installation. We assume that the professionals
who carry out the software loading have clear and well-
Logbook (LB): Book for aircraft recording mainte-
defined instructions available for the loading task.
nance operations.
Manual Check Procedure (MCP): Procedure for
manual (human) check to ensure the proper loading
of the NSV. REQUIREMENTS
NSV: A newer version of the software. According to IEEE 610.12:1990 [29], a Requirement is a con-
dition or capability that must be met or possessed by a system
Previous Software Version: Software Version
or system component to satisfy a contract, standard, specifi-
before loading the NSV.
cation, or other formally imposed documents. The term
SI: Identification of a software version. stakeholder is essential in requirements engineering. Stake-
holders are the most important sources of requirements. Not
SUP: Entity responsible to supply the NSV. It may
considering a stakeholder often results in fragmentally eli-
be an aircraft manufacturer or a system supplier.
cited requirements, i.e., incomplete requirements[30]. Stake-
holders are those people or organizations that have some
impact on the requirements. In the project of LASs, the stake-
SOFTWARE DISTRIBUTION METHODS holders are system developers, software developers, certifica-
tion authorities, aviation maintenance technicians, suppliers
According to ARINC 667-2 [27], the NSV may be pro-
of supporting equipment for loading, and other interested.
vided using physical or electronic methods. Independently
During the development process, requirements engi-
of the distribution method, it is the responsibility of the neering must elicit the stakeholder’s requirements, docu-
supplier to ensure that the NSV meets specification against ment the requirements suitably, validate and verify the
which it is supplied. The users must ensure that NSV is requirements, and manage the requirements throughout the
properly replicated in any subsequent copies of that entire life cycle of the system [31]. Some of the categories
software. presented in ISO/IEC 25010:2011 [32] are necessary to
The software distribution is different from the soft- allow the field software loading capability, in this work, the
ware load. The software distribution is the process of types of requirements must address: performance, security,
transferring software parts between aviation business part- and fault tolerance. An NSW must achieve the performance
ners or to an aircraft maintenance facility. This includes of the system. The security of the system must ensure the
software parts transferred on physical media or via elec- authenticity and integrity of the loading. Finally, fault tol-
tronic distribution. The software load is the process of erance must ensure when software loading is inappropriate.
transferring the NSV from the LC to the target hardware. Using Actors presented in the section “Actors,” we
Figure 4 presents the possible methods of distribution. create the 23 reusable requirements, and we classified the
When the electronic distribution is used, additional instruc- requirements into four types, depending on the character-
tions must be provided to the users (aircraft maintenance istics of each LAS.
BLR-3: The SUP shall identify NSV. VER-22: The MCP shall contain instructions to ver-
ify that all LAS have the same SI.
BLR-4: The LLC shall be compatible with the LI.
VER-23: The ADS shall provide a message inform-
BLR-5: The LC shall be able to transfer the NSV to ing that different versions of the software exist.
the LAS.
Figure 5 presents the flowchart allowing the identifica-
Manual Check Requirements (MCR)—These require- tion of the types of requirements applicable for each LAS,
ments are for the LAS that need a human inspection to according to loading, checking, redundancy, and inter-
ensure that loading was correctly performed. changeability characteristics. The F-LAS framework con-
MCR-6: The FLSP shall include the MCP. tains a useful flowchart to select which requirements, by
type, apply to a new LAS project.
MCR-7: The MCP shall include the expected SI of
the NSV.
MCR-8: The MCP shall include the expected CRC
of the NSV. TEST PROCEDURES
MCR-9: The MMCP shall contain instructions to Based on the results specified in Surveys 1 and 2, we iden-
verify that all LAS have the same CRC. tified the following three threats.
Automatic Check Requirements (ACR)—These Threat 1: Improper configuration—the aircraft oper-
requirements are for LAS that have a system independent ates with nonvalid software for its configuration, as
to ensure that loading was correctly performed. a result of a loading identity error.
Figure 6.
Test Procedures 1 and 2.
Threat 2: Incomplete loading—the aircraft operates MPC (Step 4a). For the TP2, Steps 1, 2, and 3 are the
with nonfunctional software or with restricted func- same as presented in TP1. The ACCS identifies that the SI
tionalities, as a result of a loading integrity error. and the CRC loaded are the expected (Step 3b). ADS dis-
plays a loading confirmation message (Step 4b). In the
Threat 3: Malfunction with version conflict—the
end, the LAS will be loaded with the NSV (Final State).
aircraft operates with conflicting configurations of
Figure 7 presents Test Procedures 3 and 4 involving
software coexisting.
the scenario of incomplete software loading with the man-
The Test Procedures must exercise the Requirements ual check (TP3) or automatic check (TP4). For TP3 and
identified in this work. The F-LAS contains six Test Pro- TP4, the Initial State and Step 1 are the same as presented
cedures, divided into three scenarios. The scenarios are in TP1 and TP2. Before the complete software loading,
normal software loading, incomplete software loading, the tester abruptly turns OFF the LAS (Step 2). The aviation
and conflict of different software versions. For each sce- maintenance technician checks and confirms the CRC
nario, we have two possibilities: manual (with MCP) or does not match with the MPC (Step 3a). For TP4, the
automated check (with ACCS). ACCS identifies that CRC is different than expected, due
Figure 6 presents Test Procedures 1 and 2 involving to incomplete software loading (Step 3b). The ADS dis-
the scenario of normal software loading with a manual plays a loading error message (Step 4). In the end, the
check (TP1) or automatic check (TP2). Test Procedures 1 LAS will not be loaded with the NSV (Final State).
and 2 guarantee that the loading of an NSV inside of the Figure 8 presents Test Procedures 5 and 6 involving
LAS is correct and complete. TP1 starts with the NSV the verification of different software versions in the
available (Initial State). The connection of the LC and LI units of the LAS. In this hypothetical procedure pre-
allows the transfer of the NSV to the LAS (Step 1). After sented in Figure 8, one of the three units contains the
the complete software loading (Step 2), the tester records NSV. The TP5 and TP6 start with the NSV available
the SI of the NSV in the LB (Step 3a). Finally, the tester (Initial State). The connection of the LC and the LI
checks if the SI of the NSV and the CRC match with the allows the transfer of the NSV to the LAS 1 (Step 1).
Figure 7.
Test Procedures 3 and 4.
Figure 8.
Test Procedures 5 and 6.
After the complete software loading (Step 2), the tester the end, all the LAS units do not have the same software
reboots the aircraft (Step 3). The tester checks that the versions (Final State).
LAS units have different software versions (Step 4a). Test Procedures 1–6 cover all the 23 Requirements
For the TP6, The ACCS identifies a mismatch of the SI identified in the F-LAS framework. Table 1 contains the
and the CRC among units (Step 4b). The ADS informs traceability between Requirements and Test Procedures.
the existence of different software versions (Step 5). In The verification of some Requirements in multiple Test
Procedure is notable.
Table 1.
interruption, modification, or destruction of data and/or testing the inappropriate loading and conflict of software
data interfaces. It includes the consequences of damage versions.
done, false data, and possible access by other systems out- Although the F-LAS framework was presented to the
side the aircraft systems. According to RTCA DO-326A professionals who kindly answered our surveys, we plan
[20], the intended or unintended use of information sys- to conduct a Survey 3 with the respondents of Surveys 1
tems available on the aircraft will impact the airworthiness and 2. This is the first future work that will formally cap-
of future aircraft, such as the potential software update ture their feedback.
error. When external data is transmitted to the aircraft As the second future work, we will explore the use of a
from the external connections, this can add new opportuni- digital certificate signing scheme that could be an optional
ties for the introduction of false or modified data in the verification method instead of using CRC. An extension
security perimeter. of the surveys carried out and synthesized in “Surveys” is
also planned.
As the third future work, we assume that when a soft-
DEVELOPMENT STANDARDS ware version is made available for installation, the supplier
has already made a compatibility analysis between the soft-
We believe that our framework will help to define
ware and other components that coexist in the same system.
parts of the systems architecture, as required by SAE
However, this analysis is of paramount importance and is a
ARP 4754A, identifying the needs of the system so
good future extension for our framework.
that it becomes possible to load new versions of soft-
A fourth future work is a validation of the framework
ware in the field. From the requirements presented and
with ARINC working groups that defined standards
the characteristic of the system design, it will be possi-
ARINC 666 [26] and ARINC 667-2 [27]. We can conduct
ble to refine them to more detailed requirements of
specific surveys with the participants of such working
systems and software.
groups and identify their feedback on our framework.
At the software level, our framework seems to be
suitable for directing the software project to meet the
RTCA DO-178C, especially in section 2.5 of this stan-
ACKNOWLEDGMENTS
dard, which deals with the FLS capacity. Our frame-
work provides specific requirements that address The authors would like to thank: the Instituto Tecnologico
integrity and identity. de Aeronautica (Aeronautics Institute of Technology) for
the support provided in this research; the 20 systems
developers that answered Survey 1; and the 12 aviation
CONCLUSION maintenance technicians that answered Survey 2.
[7] Software Considerations in Airborne Systems and Equip- [18] J. Marques and A. M. da Cunha, “Use of the RTCA DO-
ment Certification, RTCA Standard DO-178B, 1992. 330 in aeronautical databases,” in Proc.IEEE/AIAA 34th
[8] L. Marcil and M. Hawthornthwaite, “Realizing DO-178C’s Digit. Avionics Syst. Conf., 2015, pp. 8D1-1–8D1-6.
value by using new technology: OOT, MBDV, TQC FM,” [19] S. Mahmoodi, U. Durak, S. Hartmann, and S. Jafer, ‘‘DO-
in Proc. IEEE/AIAA 31st Digit. Avionics Syst. Conf., 2012, 330/ED-215 overlay to the IEEE recommended practice for
pp. 6A3-1–6A3-10. distributed simulation engineering and execution process,”
[9] W. Youn, S. Hong, K. Oh, and O. Sung, “Software certifi- J. Aerosp. Inf. Syst., vol. 15, no. 12, pp. 696–705, 2018.
cation of safety-critical avionic systems: DO-178C and its [20] Airworthiness Security Process Specification, RTCA Stan-
impacts,” IEEE Aerosp. Electron. Syst. Mag., vol. 30, no. 4, dard DO-326A, 2014.
pp. 4–13, Apr. 2015. [21] Airworthiness Security Methods and Consideration, RTCA
[10] A. Paz and G. El Bousaidi, “On the exploration of model- Standard DO-356, 2014.
based support for DO-178C-compliant avionics software [22] J. Marques and A. Cunha, “A reference method for air-
development and certification,” in Proc. IEEE Int. Symp. borne software requirements,” in Proc.IEEE/AIAA 32th
Softw. Rel. Eng. Workshops, 2016, pp. 229–236. Digit. Avionics Syst. Conf., 2013, pp. 7A2-1–7A2-14.
[11] U. Eisemann, “Applying model-based techniques for aero- [23] Guidelines for Development of Civil Aircraft and Systems,
space projects in accordance with DO178C, DO-331, and SAE Standard ARP4754A, 2010.
DO-333,” in Proc. 8th Eur. Congr. Embedded Real-Time [24] L. Rierson, Developing Safety-Critical Software: A Practi-
Softw. Syst., 2016, pp. 1–8. cal Guide for Aviation Software and DO-178C Compli-
[12] S. VanderLeest and A. Buter, “Escape the waterfall: Agile ance. Boca Raton, FL, USA: CRC Press, 2013.
for aerospace,” in Proc.IEEE/AIAA 28th Digit. Avionics [25] European Aviation Safety Agency, “Certification memo
Syst. Conf., 2009, pp. 6.D.3-1–6.D.3-16. SW-CEH 002,” 2010.
[13] J. Marques, S. Yelisetty, A. Cunha, and L. Dias, ‘‘CARD- [26] Electronic Distribution of Software, Standard ARINC 666,
RM: A reference model for airborne software,” in Proc.10th 2002.
Int. Conf. Inf. Technol., New Gener., 2013, pp. 273–279. [27] Guidance of the Management of Field Loadable Software,
[14] J. Marsden et al., ‘‘ED-12C/DO-178C vs. agile mani- Standard ARINC 667-2, 2017.
festo—A solution to agile development of certifiable avi- [28] C. Rogers, “Choosing a CRC specifying its requirements
onics systems,” in Proc.9th Eur. Congr. Embedded Real- for field-loadable software,” in Proc.IEEE/AIAA 27th
Time Softw. Syst., 2019, pp. 1–10. Digit. Avionics Syst. Conf., 2008, pp. 5.C.6-1–5.C.6-9.
[15] Y. Moy, E. Ledinot, H. Delseny, V. Wiels, and B. Monate, [29] IEEE Standard Glossary of Software Engineering Termi-
“Testing, or formal verification: DO-178C alternatives and nology, IEEE Standard 610.12-1990, 1990.
industrial experience,” IEEE Softw., vol. 30, no. 3, pp. 50–57, [30] K. Pohl and C. Rupp, Requirements Engineering Fundamen-
May/Jun. 2013. tals: A Study Guide for the Certified Professional for
[16] J. Marques and A. M. da Cunha, “Verification scenarios of Requirements Engineering Exam - Foundation Level - IREB
onboard databases under the RTCA DO-178C and the Compliant. San Rafael, CA, USA: Rocky Nook, 2015.
RTCA DO-200B,” in Proc.IEEE/AIAA 36th Digit. Avion- [31] K. Pohl, Process-Centered Requirements Engineering.
ics Syst. Conf., 2017, pp. 1–8. Boston, MA, USA: Res. Stud. Press, 1996.
[17] J. Marques and A. M. da Cunha, “Cots tool qualifica- [32] Systems and Software Engineering—Systems and
tion using RTCA DO-330: Common pitfalls,” in Proc. Software Quality Requirements and Evaluation
IEEE/AIAA 36th Digit. Avionics Syst. Conf., 2017, (SQuaRE)—System and Software Quality Models,
pp. 1–6. ISO/IEC 25010, 2011.
the Tactical Satellite-4 mission for on-orbit tests [12]. 08C standard [23], section 5.5.3.2.10 specifies the quality
With the help of the Mir Space Station, the solar cells inspection content of interconnectors. In summary, fatigue
were found to suffer severe damage to their structure and of the cell interconnectors is one of the main reasons for
interconnect due to hot-spotting processes caused by SA failure, especially with respect to SA application in
the effect of shadowing from structural members of the extreme environment.
spacecraft [13]. Those space experiments are mainly for However, there are few reports related to on-orbit fail-
the performance of the solar cell itself affected by the ure caused by the fatigue of solar cell interconnectors due
space environment. to the difficulty of on-orbit failure analysis because the
In addition, with demand for increased orbital life and occurrence of on-orbit fatigue is a gradual process. This
with more severe temperature extremes due to various article investigated the output current abnormal drop of a
space missions, the importance of solar cell interconnector LEO satellite SA caused by the fatigue of solar cell
reliability has become more emphasized and the effect of interconnectors. First, the abnormal phenomenon was
thermal or mechanical fatigue of the solar cell interconnect described in detail. Second, all possible failure causes
structure on the SA performance is a subject that has were analyzed and the most likely reason was determined.
received extensive attention in the context of aerospace Then, the failure mechanisms were explained and verified
applications. As early as 1969, Luft et al. [14] investigated by experiments. Finally, the solution to slow down the
the temperature cycling effects on solar panels and found output current drop on the satellite on orbit was given and
that the most common failure mode was the separation of the improvement measures for the development of SAs
interconnections at the cell interface. Some investigations were proposed.
also showed that both contact and fatigue problems are
associated with solar cell interconnects and fatigue of solar
cell interconnectors is one of the key failure modes for
SAs [15], [16]. Luft [17] also studied the design of solar
ABNORMAL CURRENT IN SA OF A LEO SATELLITE
cell interconnectors through fatigue tests under extended The Technology Test Satellite No. 5 (TTS-5) developed
temperature cycling and thought that the most desirable by the China Academy of Space Technology was
interconnector for each specific case must be specially launched at 10:35:12 on December 22, 2017. It is a sun-
considered on all pertinent factors including electrical out- synchronous orbit satellite with an average altitude of
put, space environment, and reliability. Evans et al. [18] 560 Km. It was found that the output current of TTS-5 SA
and Kosseim [19] respectively investigated the experimen- circuits was reduced from 26.8 A 28.2 A in May 2018
tal methods of the fatigue characteristics of interconnectors to 23.6 A 25.3 A in May 2019 by telemetry. The SA of
for different applications. The mechanism of the fatigue TTS-5 consists of four subsolar arrays (SSAs). In order to
behavior of the solar cell interconnector was analyzed by evaluate the power generation capacity of each SSA and
Wiese et al. [20] and a correlation with the relevant micro- the health status of the SA circuit, the output current test
structures of crack growth was given. Meier et al. [21] of each SSA is carried out to locate the health status of
described methods for an estimation of interconnector life- each SSA on May 31, 2019 by sending remote control
time and gave an enhancement method of this lifetime by commands to disconnect each SSA in turn. The test result
microstructural optimization. As the fabrication of space is shown in Table 1, where it can be seen that the current
SAs is a very critical process to ensure the quality and reli- loss is 1 A, 0.5 A, and 1.5 A in the SSA-1, SSA-2, and
ability of SAs and serious power losses can result from SA SSA-3, respectively, and the SSA-4 is normal. Each
rework being done improperly [22], the ECSS-E-ST-20- SSA consists of 13 strings of solar cells in parallel,
Table 1.
Table 2.
Output Current Changes for the SA of the TTS-5 During the Period of a Year and a Half
No. Time of Steady Steady String number change Estimated single string
occurrence current current (bar) current(A)
before falling after falling
(A) (A)
Figure 2. Figure 3.
TTS-5 SA layout. Schematic diagram of single-wing SA circuit connection.
Table 3.
F1: SA circuit short F1-1: Terminal power connector contact point is short-
circuited
F1-2: Inner-panel or interpanel connector contact point is
short-circuited
F1-3: Solar cell positive and negative interconnector is short-
circuited
F1-4: Solar cell is short-circuited
F1-5: SA circuit and substrate is short-circuited
F2: Output current drop due to external F2-1: Solar cell performance degradation
factors F2-2: Solar cell is damaged due to electrostatic discharge
F2-3: Solar cell is impacted by micrometeoroids or space
debris
F2-4: Solar wing fails to orient to the sun after movement
F2-5: SA circuit current telemetry is abnormal
F3: SA circuit open F3-1: Power connector solder joint is open
F3-2: Inner-panel or interpanel power connector is
disconnected
F3-3: Isolate diode body or its solder joint is open
F3-4: Solar cell interconnector is open
F3-5: Solar cell is fractured and open
Table 4.
details), it is found that the stress relief loop of the inter- denudation and corrosion of AO and the actions of other
connector is bent during the solar cell replacement pro- stresses such as high-low alternating temperature, it gradu-
cess. If the adjacent solar cells in a certain string of solar ally breaks and causes the solar cell interconnector to be
cell are replaced twice or more times, the stress relief loop open-circuited.
of the solar cell interconnector will be bent multiple times
and be probably damaged potentially. In addition, the
TTS-5 experienced a severe environment of the LEO AO
when being launched and then entered the work orbit with FAILURE MECHANISM
AO environment [25]. The silver interconnector is suscep-
tible to erosion in the AO environment, resulting in a fur- DAMAGE MECHANISM OF THE INTERCONNECTOR
ther decline in the fatigue resistance of the interconnector.
The damaged interconnector may gradually be broken
DURING SOLAR CELL REPLACEMENT PROCESS
under stresses such as high-low alternating temperature. The solar cell replacement process is shown in Figure 6
Therefore, it cannot be ruled out that the solar cell inter- and the solar cell to be replaced is the B cell between the
connector is open (F3-4). A and C solar cell as shown in Figure 6(a). When replac-
Each solar cell of the TTS-5 is connected in parallel ing, use a blade to cut off the silver interconnector 1#
with a bypass diode. When a certain solar cell is dam- between the A and B cells, slowly tilt it from the left side
aged and its output current is much smaller than that of of the B cell to expose the part where the back of the B
the solar cells in the same string, the bypass diode con- cell are connected with the C cell interconnector, and cut
nected in parallel with the solar cell will turn ON. The off the soldering area and remove the B cell. In this pro-
output current of the string in which the damaged solar cess, the interconnector 3# of the C cell will be bent for
cell is located will only drop slightly, as is not consis- the first time, as shown in Figure 6(b). Then, slowly tilt up
tent with the failure phenomenon. In addition, after the from the right side of the A cell, remove the residual adhe-
solar cell body is completely broken, it cannot be con- sive and silver interconnector on the bottom and put it
nected and its output current cannot be restored, which back in place. During this process, the interconnector 1#
is also inconsistent with the failure phenomenon. There- of the A cell will be bent for the first time, as shown in
fore, it can be excluded that the solar cell is fractured Figure 6(c). Put the replacing D cell into the position of
and open (F3-5). the B cell, use a tape to fix the front of the A and D cell so
that they do not move relative position, turn up the A and
D cells, solder the interconnector on the back of the solar
DETERMINATION OF FAILURE CAUSES
cell, and then paint the NQ-704 silicone rubber on the
According to the above analysis, the most likely reason back of the A cell. During the process, the interconnector
which causes the output current drop of the SA circuit on 1# of the A cell will be bent for the second time, as shown
orbit is the solar cell string open due to the solar cell inter- in Figure 6(d). Put down the A cell and paste it on the sub-
connector open-circuit. In the process of the solar cell strate and temporarily fix it with tape, solder the silver foil
replacement, the stress relief loop of the cell interconnec- of the D cell to the interconnector 3# of the C cell, and
tor was bent and damaged many times during the second- then coat the back of the D cell with the NQ-704 silicone
ary replacement of adjacent solar cell due to unskilled rubber. Finally, put down the D cell and paste it on the
operations, causing microcracks which reduced the fatigue substrate. In this process, the interconnector of the D cell
resistance of the interconnector. Along with the will be bent once, as shown in Figure 6(e).
Figure 10.
Microcrack SEM photograph of the silver interconnector sample Figure 11.
after secondary replacement of the adjacent solar cells. Sample appearance before (left) and after (right) AO test.
Table 5.
Serial number 1# 2# 3# 4# 5# 6# 7# 8#
experiment results show that the output current drop of the solar cell string to be open circuit. And eventually the
the SA of the TTS-5 is due to the silver interconnector output current of the SA of the TTS-5 decreases. Based on
facture, which is caused by two or more times replace- the conclusions, the improvement measures for the devel-
ment of adjacent solar cells. opment of SAs are proposed as follows: 1) Strengthen the
inspection of solar cell fragmentation before loading the
board, and use electroluminescence apparatus to defect
microcracks in order to eliminate early failure cells; 2) It is
SOLUTION AND CONCLUSION prohibited to replace adjacent solar cell fragments twice or
According to the above failure mechanism analysis and more; and 3) Strictly control the replacement process of
experiment results, the temperature alternation on orbit adjacent solar cell fragments and strengthen the training of
may further aggravate the fracture of the interconnector replacement operators. And by adopting those measures,
and thereby reduce the output current of the SA of the subsequent satellites using similar SAs have no longer
TTS-5. In order to slow down the speed of interconnec- experienced a drop in output current of SAs.
tor factures, the temperature gradient of the SA is
reduced by adjusting the on-orbit working state of the
SA oriented to the sun. Through the ground remote ACKNOWLEDGMENTS
control of the additional task program, the TTS-5
This work was supported by the National Natural Science
management execution unit automatically determines
Foundation of China under Grant 61663033.
the geographical latitude of the subsatellite point in real
time. When the satellite turns from the shadow area to
the sunlit area, the satellite attitude will switch from the
REFERENCES
ground orientation mode to the sun orientation mode
187 s ahead of time to ensure that the orientation to the [1] V. A. Shuvalov, G. S. Kochubei, V. V. Gubin, and
sun has been completed before entering the sunlit area. N. A. Tokmak, “Power losses of solar arrays under the
Similarly, when the satellite turns from the sunlit area action of an environment in a geosynchronous orbit,”
to a shadow area, the satellite attitude will switch to the Cosmic Res., vol. 43, no. 4, pp. 259–267, Jul. 2005.
ground orientation mode in advance. And from [2] D. C. Ferguson, R. C. Hoffmann, R. J. Cooper, and
the TTS-5 on-orbit operation in the past few months, J. A. Hughes, “1997–2002 solar array string failures revis-
the output current drop of the SA of TTS-5 is appar- ited,” J. Spacecraft Rockets, vol. 54, no. 3, pp. 542–553,
ently reduced and the mitigation solution is proved to May 2017.
be effective, which also validates the analysis results [3] M. Bodeau, “Observation of sustained arc circuit failure on
for the abnormal output current of SA of TTS-5. solar array backside in low Earth orbit,” IEEE Trans.
From the above analysis and experimental verification, Plasma Sci., vol. 43, no. 9, pp. 2961–2974, Sep. 2015.
the following conclusions can be drawn. The reason for the [4] G. M. Grigorieva et al., “Analysis of geostationary space-
drop in the on-orbit output current of the SA circuit is that craft solar arrays degradation from solar proton flares,”
during the production process of the SA, the operators are in Proc. 6th Eur. Conf. Space Power, May 2002,
unskilled in replacing the adjacent large solar cells of pp. 725–730.
80.0mm40.0mm. When the adjacent solar cells are [5] R. D. Lorenz, “Solar array degradation by dust impacts
replaced twice or more, the stress relief loop of the inter- during cometary encounters,” J. Spacecraft Rockets,
connector is bent and damaged many times to produce vol. 35, no. 4, pp. 579–582, Jul./Aug. 1998.
microcracks, causing the fatigue resistance of the solar cell [6] A. J. Richardson and J. W. Warren, “Solar array degrada-
to be reduced. During flight, the interconnector gradually tion due to meteoroid impacts during extended planetary
fractures under the action of alternating temperature and missions,” J. Spacecraft Rockets, vol. 8, no. 6,
other stresses together with the corrosion of AO, causing pp. 681–683, Jun. 1971.
[7] K. Toyoda, T. Okumura, S. Hosoda, and M. Cho, [20] S. Wiese, R. Meier, and F. Kraemer, “Mechanical behav-
“Degradation of high-voltage solar array due to arcing in iour and fatigue of copper ribbons used as solar cell inter-
plasma environment,” J. Spacecraft Rockets, vol. 42, connectors,” in Proc. 11th Int. Thermal, Mech. Multi-
no. 5, pp. 947–953, Sep./Oct. 2005. Phys. Simul., Exp. Microelectron. Microsyst., Bordeaux,
[8] D. C. Ferguson, “Space solar cell edge, interconnect, 2010, pp. 1–5.
and coverglass designs and their effect on spacecraft [21] R. Meier, M. Pander, S. Großer, and S. Dietrich,
charging and plasma interactions,” in Proc. 35th IEEE “Microstructural optimization approach of solar cell
Photovolt. Specialists Conf., Honolulu, HI, USA, 2010, interconnectors fatigue behavior for enhanced module life-
pp. 002537–002542. time in extreme climates,” Energy Procedia, vol. 92,
[9] D. H. Walker, R. L. Statler, and R. J. Lambert, “Solar cell pp. 560–568, 2016.
experiments on the NTS-2 satellite,” in Proc. Conf. Rec. [22] I. N. Blake, “Lessons learned about fabrication of space
13th IEEE Photovolt. Specialists Conf., Jan. 1978, solar arrays from thermal cycle failures,” in Proc. 32nd
pp. 100–106. Intersoc. Energy Convers. Eng. Conf. (Cat.
[10] R. Hoheisel et al., “Solar cell experiments for space: Past, No.97CH6203), Honolulu, HI, USA, vol. 1, 1997,
present and future,” Proc. SPIE, vol. 8620, 2013, pp. 395–400.
Art. no. 86200U. [23] “Space engineering: Photovoltaic assemblies and
[11] D. M. Wilt, A. M. Pal, S. A. Ringel, E. A. Fitzgerald, components,” European Cooperation for Space Standardi-
P. P. Jenkins, and R. Walters, “Final results from the zation (ECSS): ECSS-E-ST-20-08C Rev.1, Noordwijk,
MISSE5 GaAs on Si solar cell experiment,” in Proc. Conf. The Netherlands, pp. 48–49, Jul. 2012.
Rec. 33rd IEEE Photovolt. Specialists Conf., Jun. 2008, [24] NASA standard electrical, electronic, and electromechani-
pp. 1–4. cal (EEE) parts list, Mil. standard: MIL-STD-975M
[12] P. Jenkins et al., “TACSAT-4 solar cell experiment: (NASA), Washington, DC, USA, Aug. 1994.
Advanced solar cell technologies in a high radiation envi- [25] J. An, W. Zhang, and D. Yang, “Atomic oxygen environ-
ronment,” in Proc. Conf. Rec. 34th IEEE Photovolt. Spe- ment analysis technology for low Earth orbit spacecraft,”
cialists Conf., Philadelphia, PA, USA, Jun. 2009, in Proc. Prognostics Syst. Health Manage. Conf., Harbin,
pp. 000009–000012. China, 2017, pp. 1–5.
[13] V. A. Letin and P. G Babayevsky, “Comparative analysis [26] T. Li, L. Jiang, L. Guo, X. Liu, W. Feng, and R. Zhai,
of solar array fragments degradation at different structure “The effects of atomic oxygen on space solar array,”
scales during and after long-term exposure overboard the Spacecraft Environ. Eng., vol. 27, no. 4, pp. 428–433,
‘Mir’ space station,” High Perform. Polymers, vol. 13, Aug, 2010.
no. 3, pp. S453–S460, Sep. 2001. [27] E. M. Silverman, “Space environmental effects on space-
[14] W. Luft and E. E. Maiden, “Temperature cycling effects craft: LEO materials selection guide,” TRW Space Elec-
on solar panels,” IEEE Trans. Aerosp. Electron. Syst., tron. Group, Redondo Beach, CA, USA, Tech. Rep.
vol. AES-5, no. 6, pp. 943–950, Nov. 1969. NASA Contractor Report 4661, Part 1, Aug. 1995.
[15] A. G. Stanley, “Structural failures in light-weight solar cell [28] SPace ENVironment Information System. (SPENVIS), ver.
arrays under thermal cycling,” IEEE Trans. Aerosp. Elec- 4.6.10, Aug. 19, 2019. [Online]. Available: https://www.
tron. Syst., vol. AES-7, no. 4, pp. 606–612, Jul. 1971. spenvis.oma.be/
[16] M. Pander, S. Dietrich, R. Meier, and M. Ebert, “Fatigue [29] Z. Shen, Z. Wang, X. Zhao, and H. Jin, “Experiment of
analysis of solar cell interconnectors due to cyclic mechani- silver interconnects of photovoltaic solar array in IFM
cal loading,” in Proc. 32nd Eur. Photovolt. Sol. Energy ground-based AO effect simulation facility,” J. Beijing
Conf. Exhib., Munich, Germany, Jun. 2016, pp. 1589–1597. Univ. Aeronaut. Astronaut., vol. 26, no. 6, pp. 676–679,
[17] W. Luft, “Solar cell interconnector design,” IEEE Trans. Dec. 2000.
Aerosp. Electron. Syst., vol. AES-7, no. 5, pp. 781–791, [30] R. E. Neff, W. R. Baron, and H. E. Pollard, “Fatigue simu-
Sep. 1971. lation testing of solar cell interconnectors,” in Proc. Conf.
[18] R. B. Evans, S. Stevens, and P. Campbell, “Fatigue proper- Rec. 18th IEEE Photovolt. Specialists Conf., Las Vegas,
ties of solar cell interconnectors,” International Research NV, USA, 1985, pp. 629–633.
Development Co. Ltd, NASA STI/Recon Tech. Rep. [31] H. W. Boller and J. Koch, “Accelerated fatigue tests of
5027606, Jun. 1977. solar cell interconnectors for simulation of thermal cycles,”
[19] N. M. Kosseim, “Experimental investigation of the fatigue in Proc. 11th Conf. Rec. IEEE Photovolt. Specialists Conf.,
characteristics of silver loop interconnectors,” M.S. thesis, 1975, pp. 153–161.
Montreal, QC, Canada, Dept. Mechanical Engineering,
McGill Univ., 1991.
x
x
x
x
x
x
x
x
IEEE AEROSPACE & ELECTRONIC SYSTEMS SOCIETY ORGANIZATION
2021
OFFICERS VP Member Services Fabiola Colone
President Walter D. Downing VP Publications Lance Kaplan
President-Elect Mark Davis VP Technical Operations Marina Ruggieri
Past President Giuseppe A. Fabrizio
Secretary Francesca Filippini OTHER POSITIONS
Treasurer Robert C. Rassa Undergraduate Student Rep Sarana Tse
VP Conferences Michael Braasch Graduate Student Rep Laila Moreira
VP Education Lorenzo Lo Monte oung Professionals Program Coordinator Philipp Wojaczek
VP Finance Michael A. Cardinale Staff – Judy Scharmann; Amy Krutz
VP Industry Relations Michael Noble
Quantum radar is a hot topic for several reasons, espe- quantum radar [6], is the physicist on this team, and Jeff
cially because of the two recent experiments at X-band Uhlmann (who invented the unscented Kalman filter) is
reported in [2] and [3]. The second reason is the story that the engineer in this collaboration. All three of these talks
microwave quantum radar was used in the real world by were upbeat about the potential for quantum radar. In con-
Chinese to track aircraft and missiles at 100-km range [5]. trast, Jer^
ome Bourassa and Chris Wilson showed that a
The third reason is the assertion that microwave quantum certain important class of quantum radars cannot beat the
radar can defeat stealthy aircraft and missiles [5]. Such optimal classical radar if the transmit signal and the
claims have been published by highly respected peer- so-called “idler” (i.e., the other partner in the entangled
reviewed archival journals, such as Popular Mechanics, photon pair) are amplified equally. This negative result is
but most experts say that these assertions are highly dubi- useful to know. Of course one wonders what happens with
ous [7]. The fourth reason is that the taxpayers of China, asymmetrical amplification of the transmitted signal and
Canada, Sweden, the United Kingdom, the United States, the idler, and this important question was answered at the
Italy, Austria, and other countries have apparently spent a special session on quantum radar at the IEEE Radar Con-
significant amount of money for research on quantum ference in Florence, Italy (September 2020), which we
radars. Consequently, a special session on quantum radar shall review later. Both Wilson and Bourassa are real
was held (virtually) at the IEEE International Radar physicists who have quantum mechanics for breakfast.
Conference on April 30, 2020, which did not occur in The fifth talk, by Daum (not to be confused with the author
Washington, DC. The five talks varied widely in perspec- of this note), was also rather negative. Daum showed that
tive: Three were upbeat, but two were rather negative. the cost of X-band quantum radars is roughly 10 orders of
Marco Frasca and Alfonso Farina presented a new idea to magnitude more than the corresponding classical radar
entangle atoms with photons, rather than the boring old with today’s technology. In the future, with the most opti-
method of entangling pairs of photons; the purpose here is mistic assumptions, the relative cost can be reduced to
to allow a greater number of photons to be usefully only five orders of magnitude. During the chat for this
entangled for microwave quantum radar. Marco Frasca is special session Professor Ravi Advi (the organizer of this
a real physicist who deeply understands quantum mechan- special session) said that he was in tears after learning
ics, and Alfonso Farina is a famous expert in real radar about this cost analysis, and these were not tears of joy.
systems. Han Liu, Amr Helmy, and Bhashyam Balaji But one could put a positive spin on the huge cost of quan-
explained how to use temporal correlation of photons tum radars, because companies typically make a profit of
rather than boring old polarization entanglement to 10%–20% on sales, and hence, quantum radars would put
improve quantum radar performance by roughly 10–20 them in clover. The chat was continuously active through-
dB relative to a classical radar. A crucial point in this com- out the entire 2-h session. This allows for 10 times more ques-
parison is that the classical radar was not optimized, tions from people around the world than is typical at face-to-
which is an important theme throughout all three special face meetings. Many participants said that virtual meetings
sessions. The third talk analyzed the benefits of were much better than face-to-face sessions, because they
using multiple quantum transmit devices in parallel; avoid the cost and time and inconvenience of travel, hotels,
Marco Lanzagorta, who literally wrote the book on and parking. Michael Zatman, the Chairman of the IEEE
Radar Conference, asked some good technical questions, and
he said that this chat was probably the best at the whole con-
Author’s current address: Fred Daum, Department of ference. Zatman asked about the assumptions used to com-
Quantum Information, Raytheon, Woburn MA 01801
pute 10–20 dB of improvement in performance for quantum
USA (e-mail: daum@raytheon.com).
radar relative to classical radar; as we shall see ahead, this is
Manuscript received June 8, 2020; accepted June 11,
2020, and ready for publication January 29, 2021. a recurring crucial question for all three special sessions.
Review handled by Peter Willett. The second special session on quantum radar was held
0885-8985/21/$26.00 ß 2021 IEEE virtually at the IEEE Radar Conference on September 22,
The information listed on this page was valid as of March 31, 2021.
Potential attendees are urged to get up-to-date information directly from the respective conference organizers.
10-14 May 2021 IEEE Radar Conference (RadarConf21) Atlanta, Georgia, https://ewh.ieee.org/conf/radar/2021
2021 USA
VIRTUAL
31 May - 2021 28th Saint Petersburg International Conference on Saint Petersburg, http://www.elektropribor.spb.ru/en/conferences
2 June Integrated Navigation Systems (ICINS) Russia /141
2021 AND VIRTUAL
23-25 June 2021 IEEE International Workshop on Metrology for Naples, http://www.metroaerospace.org
2021 AeroSpace (MetroAeroSpace) Italy
VIRTUAL
8-11 August 2021 IEEE 19th International Symposium on Antenna Winnipeg, https://antem2021.ieee.ca
2021 Technology and Applied Electromagnetics (ANTEM) Manitoba,
Canada
VIRTUAL
14-15 September International Conference in Sensor Signal Processing for Edinburgh, https://sspd.eng.ed.ac.uk
2021 Defence (SSPD 2021) Scotland, UK
AND VIRTUAL
21-23 September 2021 Signal Processing Symposium (SPSympo) Lodz, Poland http://spsympo.p.lodz.pl
2021 VIRTUAL
26-30 September 40th Digital Avionics Systems Conference (DASC) San Antonio, https://2021.dasconline.org
2021 Texas, USA
POSTPONED TO 2021 IEEE 24th International Conference on Information Sun City, https://www.fusion2021.co.za
1-4 November Fusion (FUSION) South Africa
2021
This information is taken from the IEEE Conference Database. For corrections or omissions, contact conference-services@ieee.org. May 2021