The goals that the author feels are most important for the ground station communication infrastructure are maximum data throughput, zero downtime, and profitability in both low cost and high potential for exploiting the ground stations as profitable theme parks. This suggests perhaps 10 ground stations in tourism market areas around the globe, each with a data link to a pair of hubs which receive, process, and retransmit information from the spacecraft. Each ground station includes a high-speed laser communication system and a number of independent radio communication systems for tracking multiple spacecraft. The architecture is relatively inexpensive and makes optimal use of the entertainment value of ground stations. It also provides excellent evolution options and is extremely failure soft.
The ground station network will have to move large amounts of data in real time collected by antenna stations across the globe. The use of the existing Internet infrastructure creates significant cost savings and, with some adaptation, meets bandwidth and reliability requirements.
The possibility of a bad DNS list being distributed exists, and in that case the entire ground station network would be disabled for a few hours, although uptime greater than 99.5% is expected of conventional systems. The bad DNS list failure mode can be avoided by using a proprietary set of primary DNS numbers, using external secondary DNS numbers. Therefore, while the Internet gateways may be disabled by a third-party bad DNS list, the main data connections use an internal, and verified, primary DNS list to communicate within the ground station network between receiving stations, hubs, etc.
Each ground station must be able to receive data from others, while sending, in addition to other traffic, the baseline ground station's maximum data throughput with the spacecraft. A redundant pair of hubs meet these requirements in a flexible and reliable structure, with one hub sending data to all remote ground stations through one set of satellite data links, and the other receiving data from ground stations through a different set of satellite data links. Each hub has two fully redundant high-speed connections to the Internet, through which the hubs communicate. A hub's two high-speed connections are purchased from different upstream Internet service providers, which are selected for maximum ability to reroute traffic in the event of individual line or server failures.
Under ordinary circumstances, one high-speed connection at each hub, both in the continental United States, is used to communicate with the other hub and/or communications satellite constellation ground gateways, and the other connection is used to communicate through a firewall with the Internet at large, such as to remote clients and data sources. In the event of a downtime on either line, hub-to-hub communications and hub-to-comsat gateway traffic is routed through the remaining line, with traffic to the rest of the Internet using any available bandwidth remaining.
As each hub has a satellite data link to all other ground stations, either one could support the entire ground station network, although with access to only half of the satellite data links. Nevertheless, a fail-soft mode exists should either hub fail, as links between each ground station and a hub remain, where bandwidth is merely conserved. This is more of an inconvenience than a mission-threatening failure mode.
The mild reduction in capability of a total failure of a hub is the largest single-point failure mode in a centralised approach. Other failure modes involve a link between ground station and hub being disabled, which requires either an expansion of the remaining one or economising with the remaining bandwidth, such as greater compression or limiting the traffic received, if rerouting is not possible. Total failure of a ground station is relatively inconsequential, as rerouting through other visible stations, or through low-Earth orbit (LEO) or geosynchronous Earth orbit (GEO) communications satellite constellations are possible, in addition to storage of the data until an operational ground station enters line-of-sight.
A final fail-safe might be provided by an ordinary dial-in modem over telephone lines. This way, if both data connections were disabled and no other ground stations were in sight, a critical burn would not be missed, as a 56 kbps from Mission Control to the ground station would provide a telemetry and command link with the spacecraft. The dial-in modem would have to be manually activated and use elaborate security precautions. (Exploiting this dial-in modem option, one could wait until the target ground station was the only one visible to an unmanned spacecraft, remotely disable the main TCP/IP links by incapacitating some servers or flooding with forged IP packets, then dial in to the modem with a password and deorbit the spacecraft by command line, if dial-up modem access was permitted.)
The total bandwidth required for each ground station is estimated to be somewhat more than double the throughput of the spacecraft downlink, with the spacecraft downlink bandwidth determining the minimum available on each ground station's "send" link. The average distance from the hub to a ground station is 1/4 of Earth's circumference, or about 10,000 km. The bandwidth would be split between a send connection with one hub and the receive connection to the second, travelling over different land lines and satellites.
This paper deals only with data communications infrastructure, and not the mission control functions or tourist attractions and amenities which may be found at many or all of the ground stations or hubs.
While only three ground stations closer to the equator are required for continuous coverage of spacecraft at lunar distances, as the spacecraft approach LEO, the number of ground stations required to maintain continuous line-of-sight explodes. By 300 km, even assuming very little overlap, about 20-25 ground stations would needed, spread evenly over Earth's equatorial band (most of which being water).
Clearly, while real-time transmission of the large amounts of data collected is very desirable in terms of facilities required in space, it is difficult and expensive to furnish ground stations when LEO spacecraft are over ocean. Further, the theme park spinoff planned for ground communications stations quickly becomes nonviable when ground station sites are determined by geography and orbital mechanics, and not by economic considerations, due to the horizon less than 2,000 km away for a spacecraft in a 300-km orbit.
The number of ground stations is greatly reduced as orbital altitude increases, declining to about 10, even considering about 50% overlap, at 750 km altitude. A network of 10 ground stations ensures that at least three are visible to the Moon at any time, for combined- or parallel-signal data transfer or redundancy, and is an effective minimum number of stations for the reference mission.
At altitudes below this, continuous direct contact with the ground is not possible without a much more extensive ground infrastructure. However, it is possible to communicate directly with all of the broadband satellite constellations, and continuous coverage is available to altitudes of thousands of kilometres. John Montgomery's article, "The Orbiting Internet," in the November 1997 issue of BYTE magazine includes a table listing features and altitudes of various broadband satellite constellations. As these constellations are designed to transmit and receive data from customers on the ground, up to about 100 km from the altitude of the broadband communications satellites, the reference mission spacecraft will be able to transfer data between the satellites.
The fact that in the reference mission as many as four separate spacecraft will be in orbit complicates matters significantly. If only one or two ground stations are visible between four spacecraft, either autonomous, unmonitored operation is required until the number of spacecraft is reduced, or multiple antennas and associated controlling hardware are required at each ground station. However, radio amateurs routinely build private satellite ground stations for several thousand dollars, so rudimentary service to a number of other spacecraft, in addition to the main focus of the primary dish and associated hardware, is possible for an additional cost likely to be less than $100,000.
Small, automated secondary ground stations, without the hardware expense and bandwidth of a laser communications system, could inexpensively add capability to this and other options. A number of small stations could be widely dispersed to provide better LEO and multiple-target coverage. This is practical as only one orbital element is likely to be generating large enough amounts of bandwidth at a time to require a laser communications receiver.
Further, increasing the number of ground stations allow options which combine signals from multiple antennas to increase the effective gain of the ground antenna setup significantly. This leads to an associated reduction in the substantial power, mass, and design requirements for the space-based hardware for a given data transfer rate. Another similar option would be to (eventually) use multiple antennas on the spacecraft to simultaneously beam unique data to multiple ground stations. This provides an excellent upgrade path for lunar base communications. While this is not the case for multiple parallel signals, the supercomputer processing requirements for real-time integration of multiple combined signals is large, so it is very convenient to have a hub-based network layout so that computing resources can be shared or redundant between the hubs, rather than redundant at every ground station.
This option requires three ground stations to be roughly equally spaced around Earth's equatorial region. Latitude constraints are significant, although adequate sites can usually be found in industrialised countries.
As the ground stations must be sited on land evenly around the globe, combinations of sites are relatively inflexible. One combination of three sites may be California (34 N, 118 W), Toulouse (44deg;36' N, 1°26' E), and Hong Kong (22°15' N, 114°10' E). As these are all high-latitude sites, the three ground antennas must be able to point as close as 15° to the horizon. A less technically demanding combination is Kourou in French Guiana (5°9' N, 52°39' W), Bombay (18°58' N, 72°50' E), and either Apia in Western Samoa (171°40' W, 13°54' S) or west of Honolulu (158°10' W, 21°31' N). With ground stations at Kourou, Bombay, and Apia, antennas must point only 25.8° from the horizon.
As no combined- or parallel-signal approaches are practical with only three stations, a decentralised communications structure is one option. Each ground station would have a pair of satellite data links, determined by the spacecraft's downlink throughput, to two other ground stations. As data is processed at each ground station, only the "send" channel of the centralised approach is required in addition to bandwidth for inter-station communications. The average transmission distance is the same in both a centralised and decentralised approach, about 10,000 km.
In a decentralised approach, the raw, unprocessed data is broadcast to other sites, as each site must have the necessary decryption and decompression computing power in any event which would otherwise be idle, so the most secure and highly compressed form of transmission might as well be used, rather than decrypting and decompressing the data, only to recompress and re-encrypt the data for transmission to teams at other ground stations.
This option allows direct contact between ground and LEO spacecraft for only a fraction of any orbit below about 6,000 km altitude. As real-time transmission of data is very desirable, this option uses third-party, geosynchronous satellite constellations as data relays from the spacecraft to the ground, and possibly in the reverse direction as well. Available bandwidth is large, as the following table shows:
Constellation Cyberstar Celestri Astrolink Teledesic Spaceway Skybridge Altitude (km) GEO 1,408 and GEO GEO 700 1466 GEO Maximum Throughput (Mbps) 30 155 both ways 9.6 64 (2.05 on symmetrical links) 6 2 to satellite, 60 to ground Number of Satellites 3 63 LEO, 1-9 GEO 9 288 8 initially 64 Date Operational 1998 2002 late 2000 2002 2000 2001
GEO is 35,786 km altitude (22,300 miles).
Source: The Orbiting Internet by John Montgomery in the November 1997 issue of BYTE
While this option provides continuous coverage for a minimum capital equipment cost and number of ground stations, it is not particularly conducive to the major revenue source of a theme park attraction. This means that while cost is at a minimum, revenues are also low. Single-point failure modes exist with computers and reception systems at each of the ground communication stations.
For the sake of comparison only, a spacecraft-to-ground data rate of 1.2 Mbps and current prices and hardware are assumed. Using a decentralised approach, each station is responsible for a 1.2 Mbps satellite link to one other ground station, in addition to an incoming send/receive link from another station. At early 1998 rates, each station thus requires $32,000 per month for data transmission, plus perhaps $4,000 per month for an Internet gateway's connection. This equates to annual data transfer costs of about $528,000 per ground station continuously active at a peak 1.2 Mbps, including rudimentary control over secondary spacecraft. If each ground station experienced 80% of maximum activity when a spacecraft was in view, 66% of the time, annual costs would be $752,000 at current prices and an arbitrary data throughput.
With the arbitrary data rate and current off-the-shelf hardware, initial setup cost of about $900,000 is estimated for each of three sites, including the requirement to provide rudimentary control over several secondary platforms. This is in addition to the cost of an unknown amount of supercomputer processing power for transmission decompression and decryption, likely to be many millions of dollars for each site, and in addition to an Internet gateway and firewall at some or all sites.
In the centralised approach, average data transfer between each ground station and hub is 1.176 Mbps, assuming the broadcast received by each ground station is 50% the volume of data received from the spacecraft, and a spacecraft is in the sky 66% of the time sending data at 80% maximum data throughput.
With a fault-tolerant centralised approach, a pair of unique satellite links are required at each ground station to both the "send" hub and the "receive" hub. While this makes the system completely fault-tolerant, it doubles the number of satellite links required to eight, for the four ground stations. Assuming 1.176 Mbps of bandwidth, satellite uplink costs are $31,000 per month, and introduces about $10,000 per month of costs for a pair of high-speed Internet connections for interhub communications and Internet gateways. Total data transfer costs of about $836,000 would be incurred.
A less expensive centralised approach would be to eliminate the redundancy in hubs and data lines, increasing the single-point failure modes from just the three hubs to all components of the ground data network and hardware. Annual data transfer costs would be cut to $716,000 at early 1998 prices. Also, the decryption and decompression supercomputers, and Internet gateways and firewalls, are required at only one hub, rather than incurring their maintenance and multimillion dollar capital costs at all three ground stations or two hubs.
The author feels that approximately 10 ground stations is optimal from a technical perspective, with more being welcome depending on the economics of theme park spinoffs. Locating ground stations near the equator has significant advantages in this option, although the availability of third-party comsat relays lends great flexibility site selection with this option.
Provided no two sites are greater than about 100 deg of longitude apart, clustering ground stations in population centres merely increases the amount of an orbit which a low-orbit spacecraft uses a relay to transfer data to the ground to little ill effect. Ground site locations such as Tahiti, Hawaii, San Francisco, Houston, Florida, New York, French Guiana, London, Toulouse, Athens, Cape Town, Moscow, Baikonur, Bombay, Hong Kong, Tokyo, Perth, and Sydney are technically very acceptable and can be made nearly optimal for theme park economics. If economics dictated such, three or four clusters with a few stations each would have little technical effect over the above site combination.
At altitudes greater than about 750 km, direct ground coverage is nearly continuous with a distributed set of ground stations, although at lower altitudes a communications satellite is required to act as a relay as in the 4-station option. However, all but the Teledesic constellation can be used in any part of the volume where indirect communications is required, rather than only the geosynchronous constellations in Option 1. At least three ground stations are visible to a spacecraft at any point, which allows the signal to be combined or for signals to be received in parallel for a much more effective communications system. Further, multiple vehicles can receive the full amenities of at least one ground station, even in rare cases of alignment at a critical data transmission phase, with little necessity to time-critical mission phases to be in the field of view of the most ground stations.
With 1.176 Mbps of average bandwidth to each of 10 ground stations at about $31,000 per month, in addition to $10,000 for a pair of high-speed Internet connections per hub, current total annual costs would be $2.2 million per year.
Ten ground stations at $800,000 each, in addition to a pair of supercomputer setups and Internet gateways, are required. Tourism revenues, however, may be large enough that these capital costs would be lost in the noise.
In this option, enough ground stations are evenly distributed to provide complete coverage to LEO at inclinations between about 30 N and 30 S. This involves roughly 25 ground stations at least, many of which would be only unmanned relay stations in areas of land and sea determined by orbital mechanics. Additional stations can be operated in heavily populated tourism centres, and a large number of receiving antennas are available for multiple spacecraft, or for combined- or parallel-signal approaches.
However, the additional cost of ground stations floating in the Indian Ocean to provide complete coverage is only merited if a space-based relay is not feasible, which is not likely to be the case. It is likely that all LEO coverage required can be supplied through data links directly to comsats, leaving coverage beyond GEO as the main design requirement for the ground station network and reducing the need for complete coverage in LEO. The additional capability does not appear to merit the significantly increased cost.
A large number of ground stations, while involving a greater expenditure than is necessary for communications hardware and data transfer infrastructure, is not only better-performing, more failure-resistant and capable, but also generates a maximum of tourism revenues.
For more than a few ground stations, a centralised structure with redundant hubs is the most failure-resistant approach, although it involves the additional bandwidth two-way traffic for any sites collecting data, rather than a simple broadcast.
While most cost estimates at late 1997 prices are exact quotations with little error, these prices will drop as the launch approaches by an unknown amount. Cost estimates are also highly reflective of the 1.2 Mbps data throughput of the baseline ground communications station, which does not include analysis of mission data transfer requirements.
The following equation describes the minimum angle in degrees between the horizon and a point on the Moon's surface:
In the equation, 6371 km is Earth's spherical radius, and 362,781 km is sum of the Moon's perigee and Earth's radius. Ř (theta) is the maximum difference in longitude between two ground stations around the Earth, with A and B being their respective latitudes.
With ground stations at 30 deg latitude, this is approximated in degrees by:
90 - 0.6 * difference_in_longitude - 1
At 15 deg latitude, the multiplier is 0.53, and at 45 deg latitude, it is 0.71.