As satellites take on those new roles, the technical challenges are myriad. Network architectures must adapt to the new realities of physics in orbit. Handset equipment must be enhanced to reach not just the cellular basestation down the street, but satellites hundreds-or tens of thousands-of miles above the earth. Groundstations must take on new, highly complex control functions. And switching equipment that used to reside safely in air-conditioned racks must be readied to face the stresses of an orbital environment.

Yet to the user, all this must remain transparent. The satellite network must maintain the user-interface metaphors established by the public switched telephone network and the nearly ubiquitous Ethernet.

To meet users' expectations, a satellite-based network must not only offer complete regional or global coverage, but it must also approach the quality of service users expect from an existing terrestrial network. A global voice network such as Iridium, for instance, must provide reasonable-sounding connections and keep transmission latencies short enough to make conversation comfortable.

That requirement has profound implications for the architects of a satellite communications system. It influences a long list of choices that begin with where the satellites will be placed in orbit.

There are basically three possibilities. The most conventional choice for communications satellites has been a geosynchronous orbit. At 35,786 km above the surface of the earth, a satellite's orbital period is exactly one earth day. Hence, if the satellite is in an equatorial orbit and going the right direction, it will appear stationary above the earth. This makes it very easy for a single satellite to serve a single geographic area.

Hughes Network Systems Division has built such a network for the GeoMobile communications program. A single geosynchronous satellite provides two-way voice and data communications for North Africa, Southern and Central Europe, the Middle East and India.

A geosynchronous orbit is a great place for a spacecraft. Atmospheric drag and radiation effects are relatively small, a single satellite covers a very large geographic area, and the technology for design, launching, positioning and switching is well understood. You pop up a satellite and turn on the network.

Network complexity can be minimal as well. There is no need to switch signals between satellites. Ground-station antennas need only aim at one point in the sky, ever. And the path for any connection will be the same: up to the satellite and back down again. This is what is known in the industry as a "bent pipe" architecture. The satellite's only communications function is to receive a signal from earth, shift it over to a different channel-a different frequency and/or a different antenna beam-and send it back to earth. But after decades of communications satellite launches, empty geosynchronous orbits are hard to find.

In addition, the great distance is more demanding of satellite, ground station and-especially-handset radio transceivers. For a voice conversation to work, GEM must make a direct connection from a handset to the satellite and back to another handset, without intermediate hops.

These worries have led architects to consider two alternatives: medium earth orbit (MEO) and low earth orbit (LEO). These orbits-about 10,000 km for MEO and below 1,000 km for LEO-mean that the satellite will appear to move over the surface of the earth. MEO satellites typically have orbital periods of about six hours and LEOs as little as 90 minutes, so from the point of view of an earth-bound observer such satellites are continually zipping across the sky and vanishing. An LEO satellite may only be available to a given ground station for 20 minutes on each orbit.

In the industry, the collection of satellites necessary to provide continuous coverage for a service area is called a constellation. The constellation introduces an entirely new problem into the satellite-based network. The simple bent-pipe approach no longer works. Signals must be picked up by the satellite that happens to be over one handset and relayed either directly or via a ground station to a (possibly different) satellite over the other handset. As the satellites move, the connection may have to be handed off to one or two other satellites.

Latency budgets
Even at lower altitudes the latency budget can be extremely tight, particularly with a series of ground-to-orbit hops. This, projected costs for launching MEO satellites and other considerations led Motorola and Iridium LLC to choose the third approach: a constellation of LEO satellites. The Iridium system, officially turned on late last year, is certainly the most elaborate civilian satellite system yet deployed. Iridium uses 66 active satellites arranged in six polar-orbit planes, with (nominally) one spare satellite in each plane. Each satellite generates 48 L-band beams for communication to handsets, mobile radios, solar-powered phone kiosks, etc.

Articles from designers on the front lines of this new frontier follow, examining some of the key trade-offs in the network architecture. For example, Bill Thomson, lead applications engineer for Advanced Hardware Architectures, explores the key enabling technology that turns the fragile radio link between a handset and a distant satellite into viable data channel-error-correcting TurboCodes.

Joseph Benedetto, principal reliability engineer at United Technologies Microelectronics Center, explores another critical, although far less visible, task. As satellite payloads evolve into sophisticated packet-switching engines, the on-board circuitry must be hardened to withstand radiation. As David Myers, deputy director of defense programs at Sandia National Laboratories, points out in his article, when this hardening is undertaken on something as complex as a Pentium-II CPU, it requires modifications at virtually all levels of chip design.