The Teledesic Network: Using Low-Earth-Orbit Satellites to Provide Broadband, Wireless, Real-Time Internet Access Worldwide

Daniel Kohn
Teledesic Corporation, USA

Teledesic was formed in June of 1990--almost six years ago--with the objective of creating a means of providing affordable access to advanced network connections to all those parts of the world that will never get such advanced capabilities though existing technologies.

Information is becoming increasingly essential to all those things we associate with quality of life: economic opportunity, education, health care, and public services. Yet, most people and places in the world do not now have access even to basic telephone service. Even those who do have access to basic phone service get it through 100-year-old technology--analog copper wire networks--that for the overwhelming part will never be upgraded to an advanced digital capability. Even in the developed countries, there is a risk that whole areas and populations will be denied access to the powerful digital technologies that are changing the world.

The digital revolution is just as fundamental as the industrial revolution and the agricultural revolution before that. It will change all aspects of our societies. Those previous changes took place over many generations--indeed, in parts of the world they are still ongoing today. Driven by advances in microelectronics technologies, where product generations are measured in months, the digital revolution is taking place at a breathtaking pace. The digital technologies that grow more powerful every day in our notebook computers will soon be exploding out through network connections. Yet, outside of the most advanced urban areas, most of the world will never get access to these technologies through conventional wireline means.

While there is a lot of fiber out there in the world--and the number of places is growing--it is used primarily to connect countries and telephone company central offices. Even in a country like the United States, little of that fiber will be extended for local access to individual offices and homes, which represents 80 percent of the cost of a network. In most of the world, fiber deployment likely never will happen.

This is a big problem for all of our societies. If these powerful technologies are available only in advanced urban areas, people will be forced to migrate to those areas in search of economic opportunity and to fulfill other needs and desires. Society now is organized around the economics of infrastructure. With the agricultural revolution, technology--seeds--tied people to the land and brought them together in towns and villages. With the industrial revolution, people came together in increasingly congested urban areas, all organized around the economics of industrial infrastructure--wires, rails, highways, pipes, machinery. To the extent the digital revolution is tied to wires, it is just an extension of the industrial age paradigm. Like the highways and the railways before that, wires are rigidly dedicated to particular locations. If you live along side the main line you prosper. If you live a few miles distant, you are left behind.

It is no longer sound--economically or environmentally--to force people to migrate to increasingly congested urban areas in search of opportunity. The real potential of the information age is to find a means of allowing people to choose where they live and work based on things like family, community and quality of life rather than access to infrastructure. We've done a very good job extending one-to-many communications to most of the world. But having created a means for everyone to see all the benefits of our societies we have also created expectations--legitimate expectations--that will seek fulfillment. We need to provide the means for people to participate fully in the benefits of our societies where they are. We need to create the two-way network links that allow people to participate economically and culturally with the world at large without requiring that they pick up and move to where the infrastructure is.

That is the challenge we set for ourselves with Teledesic. To date, Teledesic has received most of its funding from Bill Gates--the founder of Microsoft, the world's largest computer software company--and Craig McCaw--who founded McCaw Cellular, the world's largest cellular communications service provider before its sale to AT&T in 1994. Their investment is symbolic, as well as financial.

Moore's Law, which says that a microprocessor will do twice as much for the same cost every 18 months, has correctly predicted the exponential growth of the computer industry for over 20 years. However, while computers today are thousands of times faster than those available a decade or two ago, networking has shown only linear growth. Improvements in networking performance, which have required backhoes to dig up streets and replace antiquated copper with modern fiber-optic technology, have not come close to keeping pace. Backhoes do not obey Moore's Law.

The solution we seek to bring into being is wireless access to advanced network connections. Unlike wireline technologies, the cost of wireless access is largely indifferent to location. But to get the bandwidth required for fiber-like service through wireless means, it is necessary to move way up in to the millimeter-wave frequencies--in the 20 to 30 GHz range (the Ka band). But, sending signals horizontally, over the land, in those frequencies is problematic. They are subject to rain attenuation and blocking by terrain, foliage, and buildings. The solution we adopted was simple. Send the signals vertically. This lead us to a satellite-based solution.

The next issue we faced was: what kind of satellite system? Viewed from 1996, it is difficult to predict with certainty all the advanced applications and data protocols that such a network will be called upon to accommodate in the 21st century. But it is reasonable to assume that those applications will be developed for the wireline networks in the advanced urban areas--in other words, the fiber networks.

To ensure seamless compatibility with those fiber networks, it is important that the satellite network have the same essential characteristics as fiber. Those characteristics include: broadband channels, low error rates, and low delay.

The advanced digital broadband networks will be packet-switched networks in which voice, video, and data are all just packets of digitized bits. In these networks you cannot separate out the applications that can tolerate delay from those that can't. People will not want to maintain two networks: one for delay sensitive applications and another for applications that can tolerate delay. Traditional geostationary orbit (GSO) satellites will never be able to provide fiber-like delays.

This leads us to a low-Earth-orbit (LEO) network. To put this in perspective, the space shuttle orbits at about 250 kilometers above Earth's surface. There is only one geostationary orbit, and that is over the equator at 36,000 kilometers--almost 150 times further out than the space shuttle. By contrast, Teledesic's satellites would orbit at about 700 kilometers--50 times closer to Earth than geostationary satellites.

With the combination of a very high minimum vertical angle to the satellite--to overcome the blocking and attenuation problems associated with the Ka band--and the low altitude, geometry takes over, and a constellation of hundreds of satellites is required to cover Earth. The large number of satellites also allows economies of scale in manufacturing and creates a system with very large capacity which allows a low cost of service.

The concept of a network consisting of hundreds of satellites may seem like a radical concept when compared to traditional geostationary satellites but it is less radical when compared with the evolution of networks on the ground. Computer networks have evolved from centralized systems built around a single mainframe computer to distributed networks of interconnected PCs. Similarly, satellite networks (for switched network connections) are evolving from centralized systems built around a single geostationary satellite to distributed networks of interconnected LEO satellites. The evolution in both cases is being driven by some of the same forces.

A decentralized network offers other advantages: A distributed topology provides greater reliability. Redundancy and reliability can be built more economically into the network rather than the individual unit. Also, because a LEO satellite has a smaller footprint within which frequencies can be reused, it is inherently more efficient in its use of spectrum resources. Geostationary satellites will continue to have an important role to play, particularly for broadcast applications where their large footprint is advantageous. But increasingly, geostationary satellites will co-exist with non-geostationary orbit (NGSO) satellite networks.

This evolution toward NGSO systems has resulted in three LEO system types, each focused on a different service segment and using a different portion of the radio frequency spectrum. The best way of distinguishing between these three LEO system types is by reference to their corresponding terrestrial services:

The so-called "little LEOs," like OrbComm, are the satellite equivalent of paging. They operate below 1 GHz, and provide simple store-and-forward messaging. These systems offer low data rates but can provide valuable services in a wide range of settings, such as remote monitoring and vehicle tracking.

The so-called "big LEOs" like Iridium, Globalstar and ICO, have received the most attention. They are the satellite equivalent of cellular phone service, and operate between 1 and 3 GHz.

Teledesic is the first proposed "broadband LEO." It will provide the satellite equivalent to optical fiber. Because it will operate in the Ka band, essentially line-of-sight from the user terminal to the satellite is required, which makes it more appropriate for fixed applications, or mobile applications like maritime and aviation use, where line-of-sight is not an issue. It will provide the advanced, digital broadband network connections to all those parts of the world that are not likely to get those capabilities through wireline means.

When Teledesic was first publicized in early 1994, most people seemed to have difficulty comprehending the services that the Teledesic Network would provide. It is not cellular like hand-held phones, like Iridium and Globalstar, and it is not broadcast video delivery, like Hughes's DirecTV.

Since then, the emergence of the World Wide Web and network-centric computing have provided a compelling model for a different kind of telecommunications: switched, broadband services. Peer-to-peer networking, based on the ubiquity and exponential improvements of personal computing, is transforming the way individuals live and businesses create value. Switched connections communicate from anyone to anyone, and broadband allows the transmission of all forms of digital information--voice, data, videoconferencing, and interactive multimedia.

The Internet today is still at a relatively primitive stage of development, comparable to the first personal computers in the late 1970s. At that time, it was difficult to imagine the pervasiveness and range of applications of personal computing today. By contrast, the World Wide Web already provides a revealing glimpse of the promise of the Internet, with tens of thousands of companies and millions of individuals exploring, publishing and developing on this new medium. Any and all information can and will be digitized, uploaded, and transmitted anywhere.

Well, not quite anywhere. The promise of the information age is constrained by the lack of access to switched, broadband services in most of the developed and virtually all of the developing world. The Teledesic Network will provide a means to help extend these switched, broadband connections on demand anywhere on Earth.

There is an important aspect of these non-geostationary satellite systems that is worth noting. There have been many studies, many of them by the ITU, that show a direct correlation between economic prosperity and teledensity. In the absence of a high level of economic development, however, a country is not likely to attract the investment required for an advanced information infrastructure. NGSO systems like Teledesic can help developing countries overcome this "chicken and egg" problem in telecommunications development.

Once you come out of a geostationary orbit, then by definition, satellites move in relation to Earth. With an NGSO system, continuous coverage of any point requires, in effect, global coverage. In order to provide service to the advanced markets, the same quality and quantity of capacity has to be provided to the developing markets, including those areas to which no one would provide that kind of capacity for its own sake. In this sense, NGSO satellite systems represent an inherently egalitarian technology that promises to radically transform the economics of telecommunications infrastructure. It is a form of cross-subsidy from the advanced markets to the developing world, but one that does not have to be enforced by regulation but rather is inherent in the technology.

I would like to return to the issue of latency in satellite networks that I mentioned earlier as one of our reasons for adopting an NGSO approach. This point is probably worth some elaboration--and please excuse me if I go into too much detail. Even at the speed of light, round-trip communications through a geostationary satellite entail a minimum transmission latency--end-to-end delay--of approximately half a second. This latency causes the annoying delay in many intercontinental phone calls, impeding understanding and distorting the personal nuances of speech. What can be an inconvenience for analog voice transmissions, however, can be untenable for videoconferencing and many data applications.

Applications will be developed for terrestrial networks, not for special networks with non-standard characteristics. Companies that build networks that are not compatible with the predominant data protocols and applications are taking a big business risk that their systems will be usable only for specialized, proprietary applications. History has not looked favorably upon companies that have made big bets on low-quality service. And since telecommunications customers make purchasing decisions based on their most demanding--not their average--application, geostationary satellite systems may not be a feasible choice if even a relative minority of services are latency-sensitive. In fact, most switched data applications are adversely affected by high latency.

Excessive latency causes otherwise high-bandwidth connections to communicate at a fraction of their capacity. And these issues arise not with obscure data protocols or obsolete hardware, but with almost all implementations of the only data protocol with which most people are familiar, TCP/IP, which connects the global Internet and is the standard for corporate networking.

For all "lossless" protocols that guarantee the integrity of the data transmission, latency is a constraining factor on the usable bandwidth. Since a data packet may be lost in transmission, a copy of it must be kept in a buffer on the sending computer until receipt of an acknowledgment from the computer at the other end that the packet arrived successfully. Most common data protocols operate on this principle. The data packet's trip over the geostationary connection takes 250 milliseconds at best, and the acknowledgment packet takes another 250 milliseconds to get back, so the copy of the data packet cannot be removed from the buffer for at least 500 milliseconds. Since packets cannot be transmitted unless they are stored in the buffer, and the buffer can only hold a limited number of packets, no new packets can be transmitted until old ones are removed when their acknowledgments are received.

Specifically, the default buffer size in the reference implementation of TCP/IP is 4 kilobytes, which is 32 kilobits. This means that at any given moment, only 32 kilobits can be in transit and awaiting acknowledgment. No matter how many bits the channel theoretically can transmit, it still takes at least half a second for any 32 bits to be acknowledged. So, the maximum data throughput rate is 32 kilobits per half second, or 64 kilobits per second.

To put this in perspective, if you take off-the-shelf hardware and software, hook up a broadband geostationary link, and order a T1 line (1.544 megabits per second), you expect to be able to transmit about a T1 line worth of data. In fact, any connection via a geostationary satellite is constrained to only 64 kilobits per second, which is 4 percent of the purchased capacity

Changing protocols is not a feasible solution to this situation. The trend in data networking is toward a single "pipe" carrying many types of data (including voice and other real-time data). It is therefore likely to be neither useful nor economical to transmit specific kinds of data using custom, proprietary protocols. In theory, the implementations of standard protocols, such as TCP/IP, can be modified to support higher buffer sizes. But these modifications are rarely simple or convenient, as computers on both sides of any connection need to be upgraded. Moreover, the maximum buffer size possible in TCP/IP is 64 kilobytes, which still only provides 1.024 megabits per second, or 67 percent of a T1 line over a geostationary link.

Even worse, if the geostationary link is not at one of the endpoints of the data transmission but is instead an intermediate connection, there is no method to notify the transmitting computer to use a larger buffer size. Thus, while data packets can seamlessly traverse multiple fiber and fiber-like networks (such as Teledesic), geostationary links are unsuitable for seamless intermediate connections.

The interplay of latency and buffer sizes does not affect all data transmissions, only lossless ones. For real-time data, such as voice and video, where it is not essential that all data be transmitted, "lossy" protocols can transmit higher data rates with less overhead. Unfortunately, real-time applications, such as voice telephony and videoconferencing, are precisely the applications most susceptible to unacceptable quality degradation as a result of high latency.

Instead of attempting to modify the entire installed base of network equipment with which one might want to communicate, receiving seamless compatibility with existing terrestrial networks becomes increasingly attractive. As both bandwidth requirements and the use of real-time data accelerate, the benefits of the fiber-like service that Teledesic offers are only growing in importance.

What all of this discussion makes clear is that no one single technology or satellite system type is going to be appropriate for all communications needs in all settings. The capabilities of fiber cannot be matched for very dense traffic. For basic telephone service, the economics of terrestrial cellular systems are compelling, particularly where no wireline infrastructure exists. Geostationary satellites will continue to play an important role, particularly for video distribution and other broadcast applications, where latency is not an issue and a large footprint is desirable. And each of the LEO system types has an important role to play.

Each of these technologies should be given the opportunity to fulfill its potential without bias from the regulatory structure. But, our international regulatory structures are not evolving as quickly as the technology. The period between World Radio Conferences, for example--two years--is longer than an entire generation of computer chip technology. In the past, the conservative nature of the WRC process actually served a positive function in helping to preserve options for the future. Today, however, there is a serious risk that the failure to take into account new technologies and new approaches could actually foreclose options for the future. It is important that the international regulatory process not be biased toward any particular technology or approach, but rather that it preserve our options for the future.

For the past 30 years of satellite communications, geostationary satellites have been virtually the entire relevant universe and the international satellite spectrum allocations and associated regulations reflect that. Geostationary satellites currently enjoy general priority status in all fixed satellite service frequency bands, by virtue of ITU Radio Regulation 2613. This subjects NGSO satellite systems to unbounded regulatory uncertainty, as their operation would be vulnerable to preemption by any and all geostationary satellites, even those deployed long after the NGSO systems. For someone like Teledesic who proposes a non-geostationary satellite system, special accommodation is required; by contrast, someone proposing a geostationary satellite system need only file the appropriate paperwork with the ITU.

In bands such as the C and Ku bands that already are congested with geostationary satellite systems, it would not be appropriate to change this regime. To allow for the future development of both satellite system types, however, designated sub-bands in which non-geostationary systems would have priority status need to be established in the satellite service expansion bands.

In the Ka-band, where Teledesic proposes to operate, the World Radio Conference last fall made provision for non-geostationary satellite systems. This was an impressive example of the ability of the ITU to adapt to new circumstances and accommodate new technologies. But Teledesic will not be the last iteration of non-geostationary satellite technology and the Ka band is not the final frontier of satellite spectrum. To allow for the future development of both satellite system types, consideration should be given to the long-term spectrum needs for both geostationary and non-geostationary satellite systems. The ITU can play a constructive role in enabling all technologies and system types, preserving each country's ability to make its own choices of systems, services, and service providers.

Recognizing the importance of national sovereignty, Teledesic has put significant effort into designing its system for flexibility of local implementation. For instance, Teledesic is the only NGSO system that employs cells fixed on the ground, rather than moving with the satellite footprint. This Earth-fixed configuration and the small size of these cells enables Teledesic to conform service offerings to national boundaries. Other system features, such as the ability to route a country's traffic through an in-country gateway (up to eight within each 700-kilometer satellite footprint), enhance Teledesic's ability to customize service to conform to each country's regulatory structure.

Also, we should emphasize that Teledesic does not plan to offer service directly, but rather will work in partnership with in-country service providers and, of course, will comply with the regulations of the host countries. Rather than competing with existing infrastructure, Teledesic represents a means by which local service providers can expand the geographic scope and range of services they offer. In this sense, Teledesic is a local service provided through a global network.

Of course, the value of systems like Teledesic--or any technology--ultimately is measured by their ability to enhance the quality and meaning of our lives. The benefits to be derived from the advanced information services they enable are as vast as the areas of need to which they can extend. We hope Teledesic can help play a small role in extending the vision of a Global Information Infrastructure to all the world's citizens.