Virtualization Of Small-Satellite Downlink Ground Stations by G. Linton and W. Kinsner

From The Space Library

Jump to: navigation, search
Greg Linton1 and Witold Kinsner2

1Canadian Forces School of Aerospace Studies (CFSAS), Winnipeg

2Dept. of Electrical & Computer Engineering, Faculty of Engineering University of Manitoba, Winnipeg, MB, Canada R3T 5V6,



This paper presents the design, discrete event simulation and analysis of a network concept based on the Automatic Packet Reporting System (APRS) to enhance CubeSat downlinks. Established APRS Internet Gateway (I-Gate) nodes, operated by amateur radio volunteers, are aggregated into an example virtualized ground station network. This Systems Tool Kit (STK) simulation includes a set of generalized example satellite models (ExampleSat) for analytical comparisons of link budgets, access times, data rates, antenna designs, orbital altitudes and node locations. These models offset many assumptions associated CubeSat communications design and analysis. The simulation combines ground and space antenna models from 4nec2, node locations extracted from real world I-Gate packets and orbital mechanics into a continental scale RF downlink network. These models take into account amateur radio community surveys and CubeSat literature reviews. The simulation outputs demonstrate increased access time compared to conventional AMSAT/OSCAR ground station designs. Furthermore the aggregated signal to the virtualized ground station network mitigates induced changes in signal due to slant range including Doppler shift. The received signal is compared to white and brown noise to determine the extent of its time invariance for further signal processing. A proposal is given to implement a continental scale Software Defined Cognitive Radio downlink network that provides also utility to the APRS user community. Keywords: Nanosatellites; CubeSats; APRS; Amateur Radio; link budget; downlink; access time; ground station; virtualization; satellite;


One of the limiting factors of CubeSat mission design is the amount of access time the satellite has to downlink its science data from orbit. Using Commercial-Off-The-Shelf (COTS) components [1] [2] a CubeSat payload could can accumulate more science data than it is possible to downlink over the popular amateur radio satellite frequencies [3] within the mission’s expected lifespan. This limitation is induced not only by the orbital mechanics of the CubeSat’s transit past the ground station but also by the transmission capabilities of the small form factor space craft. To solve this problem one approach is to deploy customized ground link stations around the globe such as NASA’s Near Earth Network [4]and the GENSO project [5]. Institutions with limited resources may lack the capability to distribute dedicated ground stations across a wide area to increase a CubeSat mission’s downlink access time. Rather than create a purpose built satellite communications ground station network this paper proposes that CubeSat mission designers could leverage existing, already in use, radio frequency (RF) networks similar to APRS I-Gates into an aggregated system. Although such networks were not designed with satellite communications in mind, their ubiquity represent a distinct advantage. This approach enables improved access time while alleviating issues associated with Doppler induced error and variations in received signal strength.

2. Virtualized Ground Station Network (VGSN)

2.1 System Design

The proposed Virtualized Ground Station Network is a form of overlay network that uses an already deployed and operational network of terrestrial communications packet Radio Frequency (RF) nodes. By aggregating the received signal from a CubeSat to a set of existing heterogeneous terrestrial communications nodes we offset 1) the limited access time available to a single ground station access point 2) the need to deploy infrastructure across a wide area. To investigate the practicality of the VGSN we use the amateur radio Automatic Packet Reporting System (APRS) as a notional example network to build the overlay network on top of. To accomplish this we used a series of models, based on "real world data" within a discrete event simulation in Analytical Graphics Inc.’s Systems Tool Kit (STK) [6]. To create this simulation several key components were investigated and modelled. A generalized model of deployed CubeSats was created based on literature reviews and database queries. Rather than use a generic term we refer to ExampleSat with a suffix to indicate its orbital altitude. Models were created in 4nec2 [7] of the antenna far field radiation pattern for both the ground and space sections of the simulation. An example 4Nec2 far field pattern based on the for the ELFIN CubeSat communications subsystem [8] can be seen in Fig. 2. The selection of ground station antenna models was determined by an informal internet survey of the amateur radio community. The survey responses were used to narrow the selection of antenna types as well as determine proportion of each antenna design used within the simulation. 4nec2 far field models were generated at both 144.39 MHz as well as for 437.385 MHz ground plane (2mGP), whip (GP144-435), J-Pole (J-Ant 144) and an AMSAT/OSCAR compatible Yagi-Uda design (based on the 216CP model to include 16 elements with vertical and horizontal pieces). Each of these models was combined together to create ground and space objects within the overall simulation as shown in Fig. 1. By running the simulation across a 24 hour period we then evaluated the received signal using the energy per bit relative to the channel noise (Eb/N0) as a measure of performance for the VGSN.

Fig. 1: VGSN Model and Simulation Schema.

Fig.2: 4nec2 Far field radiation pattern for ExampleSat-165's dipole antenna.

2.2. APRS and its I-Gates

APRS was originally developed starting in 1982 by Bob Brunigan (WB4APR) to map high frequency US Navy position reports [9]. From its outset, APRS was designed as a robust form of digital communications exchange between geographically dispersed nodes. As such it operates in an unconnected broadcast fashion where packets are not acknowledge nor are re-transmissions requested for lost packets. All of the nodes within a geographic region, such as North America, transmit and receive on a common frequency of 144.39 MHz using the AX.25 protocol [10]. Over time, APRS has been optimized to enable short distance real time crisis operations communications. As GPS enabled radios became widely available a popular use of APRS was for vehicles to beacon out there position updates across the network. APRS can be considered a local tactical information communications infrastructure when operating in a single hop mode. When specialized nodes such as digital repeaters, thus enabling a wider coverage area, the APRS network could be considered regional in nature. At a rudimentary level, APRS nodes are deployed in three different types. Trackers, digipeaters and Internet-Gateways (I-Gates). Trackers can be designed to beacon out packets (on the RF domain) that may include their position, short messages, weather data, sensor telemetry and emergency coordination information for example. Such nodes may also listen for packets from their neighboring nodes. Digipeters are "digital repeaters" that listen for packets on the RF domain and then repeat them (again, on the RF domain) thus extending the coverage of the network. I-Gates perform a similar function to digipeaters but they most often are designed to listen on the RF domain and then transduce those packets to the internet domain to series of servers. These servers provide the contents of the packets to the public whom have created a number of services including a real time mapping solutions such as [11]. Volunteer licensed amateur radio operators whom wish to participate in the APRS network build or buy their own equipment to participate within the network in much the same way as the conventional internet. The APRS network is decentralized with no particular authority controlling its use or operation beyond the established standards for interconnection and accessing its server infrastructure [12]. This makes the network heterogeneous in nature. For the VGSN we focus on the APRS I-Gates to downlink packets from ExampleSat. Each of the I-Gates acted as a ground station within the overlay that is the VGSN inside the STK simulation. Across North America over 842 I-Gates are currently listening to 144.39 MHz for nearby nodes to beacon out packets. A custom Perl script was used to determine the geophysical locations of each of the I-Gates by extracting their GPS position from "real world" packets transmitted across the APRS network to the Tier 2 network servers. These node locations were then used within the WGS84 globe inside the STK simulation. From here the antenna far field radiation patterns created in 4nec2 were attached to the I-Gate locations. This combination of far field models, receiver specifications and WGS84 node locations represents a continental scale RF network simulation as shown in Fig.3.

Fig.3: APRS I-Gate Locations on the WGS84 globe.

In a typical AMSAT / OSCAR [13] ground station a tracking antenna, with a directional main lobe far field antenna pattern, follows the satellite as it travels through its orbit. By contrast APRS nodes, including I-Gates, are typically designed for terrestrial communications and use fixed antennas with omnidirectional far field patterns. None of the infrastructure, nor the frequency used, nor the modulation type, nor the AX.25 protocol [10] used by APRS, nor the APRS packet structure [14] [15] were created with space communications in mind.

2.3 ExampleSat: a "typical" CubeSat

In order to create a representative model of a CubeSat downlink a literature review was conducted to determine the characteristics of the CubeSat mission population. The works of Klofas et al. [16] and the Technische Universität Berlin’s Small Satellite Database (TUB SSD) [17] were instrumental in building a coherent view of the population of CubeSats. A table was created which often describes each spacecraft’s transmitting power, modulation type, data rate, and antenna design. Cross-referencing the satellite mission names from Klofas et al. and the unique NORAD Space Surveillance Network (SSN) identification number we determined the orbital parameters of the CubeSat population. This limits the assumptions associated with modeling ExampleSat and its transmission capabilities. For example, we no longer had to assume an orbital altitude but could select a median value from the population’s multimodal distribution. The operating assumption for the STK simulation was a "worst case scenario for access time" situation, represented by the lowest altitude that is popular for CubeSats. Combined with the literature review data, a picture of a typical CubeSat (ExampleSat-165) emerged as shown in Table 1.

Table 1: ExampleSat-165's Orbital Elements

Both the Right Ascension of the Ascending Node (RAAN) and the Starting True Anomaly were chosen arbitrarily as they do not impact the aggregated simulation results over a 24 hour period for this low satellite altitude. The inclination (between 96° and 104°) and eccentricity (between 0° and 0.0036°) data from the population demonstrated that CubeSat missions are most frequently polar circular in nature. Based on the popularity of the various telecommunications systems designs used by CubeSats, ExampleSat-165’s telecommunications capabilities were determined as shown in Table 2.

Table 2: ExampleSat-165's Telecommunications Capabilities

The above data was derived from creating histograms of the population of 139 CubeSats for each of the pertinent details including; orbital elements, downlink frequency, transmission power, antenna design, modulation type, protocol, data rate, Object ID, satellite name, radio type, mission lifetime, mission status, and more. Admittedly, ExampleSat-165’s downlink frequency of 144.39 MHz was chosen so to be interoperable with the terrestrial APRS network. Additionally only the 144 MHz to 146 MHz band has amateur radio designated as the primary user group. The median downlink frequency for population was 437.385 MHz, which demonstrates the importance of the UHF band for CubeSat development. Dipole antennas designs were the most popular (33) with monopoles next (27), followed by turnstile designs [18].

Table 3: Potential ExampleSat Altitudes

ExampleSat and its variants represent a suite of models that can be reused and contextualized for further CubeSat mission design Table 3 shows the median orbital altitudes populated by CubeSats. These represent the statistical bulk of CubeSat populated orbits and not simply the orbits with the most number of CubeSats at that altitude bin within the histogram. 68 of the 139 CubeSats used the AX.25 protocol, with the second most popular being a proprietary technique. 24 used Audio Frequency Shift Keying AFSK (based on the Bell 202 standard) modulation while 32 used the conventional Frequency Shift Keying (FSK) technique. Interestingly, 1200 baud was the most popular (42) data rate with 9600 baud as second most popular (32).

3. Results

3.1 Access Times

The STK simulation output, shown in Table 4: "Access Times Comparisons over 24 hours" demonstrates there is an order of magnitude improvement in the overall access time when using the VGSN as compared to a conventional AMSAT/OSCAR Yagi-Uda ground station. These access times can be seen in Fig.4 where each green line represents a Line-Of-Sight (LOS) access from ExampleSat-165 to an I-Gate within the VGSN. Beyond the simple increase in access time due to adding additional ground station via the VGSN, we investigated the properties of the received signal from ExamplSat-165 to the VGSN.

Table 4: Access Times Comparisons over 24 hours.

Fig.4: ExampleSat-165 downlinking to APRS IGates across North America.

4. Received Eb/N0 properties

The received signal from ExampleSat-165 was compared in two different scenarios across 24 hours. The first considered a single I-Gate ground station (VE4UM) in Fig.6, while the second considered all 842 North American I-Gates in Fig.5. Within each figure, periods in which no LOS access was possible were removed. Notice in Fig.6 how there are four (4) distinct orbital passes. Stationarity testing using a Phillips-Perron test (PP test) on the signal in Fig.6 shows that it is non-stationary and thus is not Linear Time Invariant (LTI). By contrast, the signal in Fig.5 was shown to have similar stationarity to white noise under the same PP test and thus can be considered an LTI signal.

Fig.5: Received Eb/N0 to VGSN.

Fig.6: Received Eb/N0 to a single I-Gate.

Fig.7: Mitigation of Doppler induced errors via the VGSN.

4.1 AFSK and Doppler shift

One issue associated with satellite communications is Doppler shift induced packet errors. Typically the receiving ground station dynamically tunes its receiver to compensate for the Doppler shift induced by the satellites’ orbital motion relative to the ground. The VGSN concept is an overlay network where the heterogeneous nodes are owned and operated by the community of amateur radio operators. This implies we would not have access to their I-Gates to dynamically tune them. And since we intend to transmit packets to the geographically dispersed LOS accessible I-Gates listening 144.39 MHz at the same time, we cannot change the downlink frequency. This leads to the obvious question, "Does our choice of APRS I-Gates for the VGSN lead to Doppler induced packet errors?" Since the various I-Gate receivers do not experience the same rate of change of slant range for their LOS to ExampleSat-165 different levels of packet errors could be expected. This leads us to a discussion of the AFSK modulation used in APRS that is based on the Bell 202 technique. AFSK uses a 1200 Hz tone to indicate a mark (1) and 2200 Hz tone to indicate a space (0) around a centre frequency of 1700 Hz. Within the STK scenario we determined that the mean Doppler shift created by ExampleSat-165’s transmission was -180 Hz. This Doppler shift is applied 1:1 from the RF domain of the carrier signal into the audio domain of the baseband signal. This may still be within the notch filtering of the receiving I-Gates. The maximum Doppler shift was +3638.151 Hz which would could move mark (1) into the notch filter for a space (0). However, this would cause the packet to fail the Cyclic Redundancy Check (CRC) used by the AX.25 protocol. This Doppler shift issue, shown in Fig.7, illustrates the how AFSK and the AX.25 protocol would act as an initial filter for the VGSN to remove packets with excessive errors. I-Gate nodes that receive packets with any errors, as the protocol is zero tolerant, do not relay them to the Tier 2 network.

5. Conclusions and Recommendations

5.1 APRS I-Gate reception from ExampleSat-165

This work demonstrates that its possible for; a ground station network comprised of nodes using omnidirectional antennas; to receive 1200 baud AFSK packets; from a 165 km high orbiting CubeSat; that uses a dipole antenna on 144.39 MHz. Table 4 shows that when using the base APRS radio systems design for CubeSat communications a Yagi-Uda enabled tracking ground station antenna may not be required to downlink for satellites at 165 km altitudes.

5.2 VGSN received signal

Since the aggregated received Eb/N0 signal from ExampleSat-165 to the VGSN can be considered LTI this implies that the proposed network design abates a number of issues. First, the apparent changes in slant range from ExampleSat-165 to the I-Gates which cause the changes in signal strength shown in Fig.6 are mitigated in Fig.5. Not all of the 842 I-Gate ground station nodes, experience the same spreading loss from ExampleSat-165 at the same time during the orbital passes. This network concept also opens up a number of possibilities for analyzing the combined signal. For example, statistical packet reconstruction or voting methods could be used as Forward Error Correction (FEC). The VGSN would also enable variants of the Viterbi Tree FEC as used within modern communications systems. Digital processing of the aggregate signal is also possible such as filtering.

5.3 APRS, Cognitive Radio and the VGSN

Amateur radio operators may be characterized as early adopters of technology. As such enterprising members of the community have already adapted existing COTS hardware such as the Raspberry Pi and Software Defined Radio SDR USB dongles to create inexpensive APRS I-Gates as shown in Fig.8.

Fig.8: Proposed hardware for Raspberry Pi SDR APRS I-Gate and CubeSat downlink

There is an opportunity to create a continental scale Cognitive Radio (CR) test bed ground station network leveraging this active user community. It is possible for an SDR to sample multiple sections of the RF spectrum in near real time. This would allow a CR system designer to create an APRS I-Gate that would also act as an SDR ground station network simultaneously. This would cost less than $100 CDN per ground station and less than $85,000 CDN purchase the hardware for every enthusiast currently operating and I-Gate. Additionally, since they are licensed operators, their names and addresses are readily available within a public database. Such a CR VGSN could not only adapt to channel changes based on weather and interference noise, but also to the predicted orbits of various spacecraft. This would enable dynamic SDMA and CDMA as practiced by cellular networks today.

5.4 Cautions

We do not recommend that a CubeSat mission designer build and launch a spacecraft that uses APRS I-Gate nodes directly. The wide coverage area from orbit implies that a large number of nodes could be interfered with during the spacecraft's transmissions. It is easy to assume that since ExampleSat transmits at 1W, that APRS nodes which use higher power (45% of 104 surveyed operators use between 1W and 5W for their mobile nodes) and are in close proximity to overcome the satellite's signal. Additionally, since the CubeSat would use packets without a specified address, packets would only be digipeted by I-Gate nodes to the internet, preventing "flooding" of the network. Consultation with the APRS stakeholder community is required before any widespread use of the network for space systems communications. In this context the APRS network is used as an example to explore future space communications designs.


  1. ^  D. Liddle, et al., "STRaND-1 Use of a 500 dollar smartphone as the central avionics of a nanosatellie," in 62nd International Astronautical Congress, 2011.
  2. ^  S. Alexander and R. D. Marlaire, "NASA Successfully Launches Three Smartphone Satellites," April 2013. [Accessed 2016].
  3. ^  ITU, "Amateur-Satellite Service," September 2012. space/AmateurDoc/AmateurSatServiceFreq.pdf.
  4. ^  NASA Space Communications Navigation, "Near Earth Network," July 2014. [Accessed 30 November 2016].
  5. ^  G. Shirville and B. Klofas, "Genso: A global ground station network," in AMSAT Symposium, 2007.
  6. ^  "STK," [Accessed June 2016].
  7. ^  A. Voors, "4nec2," July 2016. [Accessed November 2016].
  8. ^  IGPP, "ELFIN: Communications," [Accessed November 2016].
  9. ^  P. Loveall, How APRS Works, 2005.
  10. ^ W. Beech, D. Nielsen and J. Taylor, "AX.25 Link Access Protocol for Amateur Packet Radio," Tucson Amateur Packet Radio, Tucson, 1998.
  11. ^, " Application Programming Interface • Google Maps APRS,"
  12. ^  [Online]. Available: [Accessed June 2016].
  13. ^ P. Pacier and H. Golding, "The Tier 2 Network," [Accessed June 2016].
  14. ^ AMSAT, "ASMAT," [Accessed July 2016].
  15. ^ M. Lustig, F. Ong and G. Fanti, "Lab 6 Part B AX.25 and APRS," 2015. [Accessed September 2015].
  16. ^ K. W. Finnegan, "Examining Ambiguities in the Automatic Packet Reporting System," 2014.
  17. ^ B. Klofas and K. Leveque, "A Survey of CubeSat Communication Systems 2009 to 2012," in CubeSat Developers Workshop, 2013.
  18. ^ M. Buscher, "TUB Small Satellite Database," [Accessed August 2014].