By Michiel Hegener
Just two or three years after the Internet really took off, it is already commonplace to say that the world is being swept by a telecommunications revolution; so we shan't say it. It is a valid question, though, to ask what is meant by the world in this particular case, because hundreds of millions of people in less well-to-do parts of it have never heard of datacommunications, let alone what it could do for them. Something to worry about? Very much so.
Information and communication technology (ICT or IT)-the Internet in particular-has a huge potential to sweep away poverty and ignorance, which is a far more interesting application than sweeping the world as such or keeping shareholders happy. This article is not about proving that ICT can serve as a powerful engine for economic growth in underdeveloped regions--thousands of examples can be found on the World Wide Web--but it may be worth noting that that idea is now getting widely accepted.
Last June in Toronto, at the world conference called Global Knowledge '97, cohosted by the World Bank and the government of Canada, United Nations secretary-general Kofi Anan said, "What is so thrilling about our time is that the privilege of information is now an instant and globally accessible privilege. It is our duty and responsibility to see that gift bestowed on all the world's people, so that all may live lives of knowledge and understanding." Among the 1,500 attendees listening to those words were many representatives of the very countries where the biggest gains can be made. Several of the donor countries, too, are becoming keenly aware that ICT is not a luxury but a necessity. Leading the pack, the World Bank has really embraced the idea that the way to basic prosperity and justice for all is paved partly with ICT. That doesn't necessarily mean a fiber-optic cable running under the pavement; satellites overhead might do just as well-in some respects even better and, above all, quicker. Although only a very small part of our planet has been wired so far, every place on earth has Internet access via satellite: a PC, a dish antenna and some equipment in between are all that's needed to hook up straight to an Internet backbone, even if you live in a tiny village in the middle of Mali or Myanmar. However, for the individual end user or even for a small business, the price of a very-small-aperture terminal (VSAT) and the tariff per kilobyte sent or received are quite prohibitive, certainly in the poorer regions of the world. Additionally, in many countries, licenses for the use of wireless communications are hard to get, if they can be gotten at all. In spite of that, a VSAT is a viable solution for larger companies and organizations as well as for Internet service providers in developing countries. For instance, about a dozen African ISPs now have their own dedicated satellite links, including a 3- or 4-meter dish antenna on their premises. The dish is trained or a geostationary-earth-orbit satellite (GEO) at 35,786 kilometers (22,187 miles) above the equator, which relays the signals to and from a ground station somewhere in the West that is directly linked to an Internet backbone. The big disadvantage of such a line in the sky of fixed throughput-usually a few hundred kilobits per second-is the waste of bandwidth that occurs at night and the shortage of it during peak hours. The good news amid all of the limitations is that satellite service providers and satellite builders-often closely linked, by the way-are becoming very keen on reaching end users directly, and new technologies enable them to do so in an affordable way. They realize that there is a huge demand for high-speed Internet connections and that only satellites can deliver them quickly to the unwired parts of the world. Where good telephone lines or even ISDN is readily available, the new satellite services have a lot to offer as well: high speed means we are talking hundreds of kilobits per second uplink, and downlinks of a few megabits. Future users of these direct-to-home interactive satellite broadband services owe a debt to the present users of direct-to-home satellite television, a phenomenon that has given tremendous impetus to the entire satellite industry. One of the spin-offs is a huge amount of research into new technologies; another is enough money for the initial, or even the entire, funding of the new, interactive broadband services. Beware: a few years from now you will need to spend only $1,000-$2,000 on equipment to get your own high-speed Internet link. Transmission costs may drop to about a cent per megabyte. And there will be a whole range of dedicated satellite systems to choose from: more than a dozen if all the plans proposed so far become reality. Starting about the year 2000, every square meter of our planet will be showered with connectivity. No strings attached. Or are there?
Before taking a closer look at some of the plans, let's investigate certain basics that apply to them all. There are several ways to get the prices down in order to reach the end user. As no one will be interested in paying thousands of dollars a month for a 64- or 256-kilobit leased channel 24 hours a day, access to the satellite has to be demand assigned. If you want to send out just a short e-mail message, a few kilobits during less than a second is all you need. When you want to search the Web for a while, a 64-kilobit channel would be more suitable. And online viewing of a piece of video might require more than a megabit. When mowing the lawn or hiking in the Adirondacks, most users would select zero bits a second, but they would certainly choose a standby downlink option-if the system offered one-so any incoming e-mail would go straight to them and not to their ISP first. In all cases, you will pay as you go, like you are now doing with your electricity. Apart from such demand-assigned, multiple-access techniques, frequency reuse is another way to bring down the cost of the links, because radiospectrum is scarce and expensive.
Most of the present 250-odd civilian geostationary telecommunications satellites have dish antennae that cover half a continent or more. Now, for broadcasting that's fine: casting a signal out over a broad area is precisely what you want. But if a satellite sends out a signal to just one VSAT, the frequency used is for some time unavailable to others within the entire footprint, like half or the whole of Latin America. With today's level of VSAT use that is tenable, but it isn't when you want to send many, many different high-bandwidth signals to many, many users at the same time-for instance, when every Tom, Dick, and Harry is surfing the Web. Therefore the new, interactive broadband satellite systems will sport many small antenna beams with slightly overlapping footprints of a few hundred kilometers across, together covering very large areas. One and the same frequency can then be used in various spot beams at the same time, provided they don't overlap. The problem here is that you need complex onboard switching techniques between the beams. In spite of frequency reuse techniques like this one, the new, interactive satellite services still need astonishing amounts of radiospectrum--so much that almost all are forced to use the very high frequencies of the Ka-band: between 20 and 30 gigahertz-the spectrum equivalent of the American West in the 19th century. "Go Ka-band, young man," a senior satellite engineer might say these days. Dangers lurking behind the hardly charted hills all have to do with difficult technology, which is of course why the seniors didn't go there themselves. Also, at 30 gigahertz the length of one wave cycle is just 1 millimeter, which makes the whole wave--carrying the Web page you just requested--very vulnerable to raindrops, like a mouse running over a field littered with rat-size rocks. The meteorological hazards can be negotiated to an extent by the Internet transfer control protocol, but when it starts raining real hard, as happens often in tropical regions, you are bound to suffer a complete outage. One possibility is a temporary boosting of the transmission power in the spot beam, which covers bad weather-another advantage of multiple spot beams. For the uplink, though, that doesn't help. A stronger transmitter and a bigger antenna do, but then of course the hardware price will be boosted as well. So, one of the biggest challenges for those who have joined the race is to develop really affordable user kits-VSATs essentially.
A lot of the money for a present-day VSAT is spent on a strong transmitter. It has to be, because geostationary satellites are so far away: lower than 35,786 kilometers they orbit the Earth in less than the 23 hours and 56 minutes the Earth needs to turn on its axis and they'd therefore no longer be geostationary. The huge advantage of these fixed satellite positions is of course that you can use dish antennae that are just as fixed-to your roof, for instance. One advantage of the Ka-band is that to get the same results, smaller and therefore cheaper dishes than for lower frequencies can be used.
Equipping the satellite with more-sensitive receiving facilities also contributes to a cheaper kit for the end user. A much more dramatic move is to lower the satellites: to reach a satellite 800 kilometers above you requires about 1/2000 of the power needed to get the signal to the geostationary arc. At 800 kilometers, a satellite orbits earth in less than 2 hours, so instead of dish antennae you need omnidirectional ones which wastes a lot of transmission power. There are other complexities as well-and other advantages. Phased-array antennae-though technologically still in their infancy-will offer an alternative before too long.
The best-known low-earth-orbit (LEO) broadband satellite system is Teledesic, of the Teledesic Corporation, founded by Bill Gates and Craig McCaw as long ago as 1990. Indeed, there are very few others, as most players intend to use GEOs. Teledesic was publicly announced in March 1994. Although the looks and the design of the satellites still needed some attention, their sheer number-840-was immediately alive and kicking, blazing a broad trail of awe through the telecommunications world and impressing would-be end users around the globe. Those users may have been a little disappointed when it was announced last April that the fleet size would be trimmed down to just 288 satellites plus 36 on orbit spares, even though their orbit altitude went up from 785 to more than 1,300 kilometers. The change followed shortly after the announcement that aircraft manufacturer Boeing had been selected as the main contractor for building the fleet. Both Teledesic and Boeing are based in the Seattle conurbation, as indeed is Microsoft, whose chairman provided part of the seed capital-some tens of millions of dollars. Representatives of Microsoft and Teledesic insist there is no formal link between the two companies. And though $9 billion will be needed to get Teledesic up and running, Gates and McCaw have promised not to pay that amount out of their own deep pockets. As Teledesic president Russell Daggatt says quite significantly, "On the investment side, Bill and Craig will eventually get diluted down to insignificance."
The high number of satellites is needed to make sure that no matter where you are, you always have at least one at more than 40 degrees above the horizon. The reason has a lot to do with the system's availability during adverse weather conditions: the signals have to struggle through only a rather thin slice of the atmosphere if the satellite is more or less above the earth station--your antenna and transceiver, that is. The use of phased-array antennae on the satellites as well as on the ground is another way to boost Teledesic's capacity. Phased array essentially means you can direct the signal by means of electronic steering, without moving parts and instantaneously. So, when online, the parabolic, mushroom-shaped antenna on your roof will always follow one of the satellites as they move across the sky. The satellite, in the meantime, will train its multiple phased-array spot beams on fixed grid cells on the surface, including the one from which you are operating. As soon as another satellite gets closer to you, the signal will be handed over automatically. By using intersatellite links, Teledesic needs in theory only a few large gateway earth stations to interface with terrestrial networks, including Internet backbones. For regulatory and economic reasons, however, there will be close cooperation with local service providers-a policy that will translate into at least one gateway in each country that allows the use of Teledesic.
GEO service providers rely on a lot of proven technology, but Teledesic has chosen not to follow the beaten track. As Daggatt puts it, "Ten years ago LEOs were impossible. Today it's a challenge. Ten years from now, no big deal."
A good overview of the coming Ka-band systems in the March 1997 issue of Via Satellite reported that Teledesic was under "ongoing criticism from its competition regarding the system's feasibility," which is surprising, because you'd expect any infeasibilities to be a source of relief and delight to the other contenders. However, scarcity of available spectrum, of which Teledesic needs quite a slice--1 gigahertz--lends the criticism some justification. If Teledesic flops, the world will have suffered a waste of spectrum, which will of course be the case if any other fleet flops. Some almost certainly will: that is one of the near certainties in this whole endeavor. There are just too many plans.
Part of the Teledesic technology still hasn't been sorted out, but the same is true, to a lesser degree, for the GEO systems. All Ka-band systems still need a lot of research and engineering if they want to be really affordable, which is what they all want. In the meantime, a LEO network has at least two big advantages over GEOs. One is far more even coverage of the globe, especially when the satellites follow near-polar orbits, as is the case with Teledesic. You then get an orbital pattern that resembles the dividing lines between the parts of a peeled orange while the rotation of the Earth is doing the rest. Although this model brings most connectivity to the polar regions and has the lowest satellite density around the equator, it is utterly egalitarian otherwise. Most important, it doesn't favor prosperous areas--a characteristic that should please all who agree with the sentiment, quoted earlier, of Kofi Anan.
Each Teledesic cell--whether on Manhattan or in the depth of the Amazon basin--has a diameter of 80 kilometers and a capacity of 64 Mbps in each direction. It should be added that cells can and sometimes will be shut off. "We certainly would not allow service in a territory where the government doesn't allow it," says Daggatt. That must be bad news for the inhabitants of Tibet, Eastern Timor, and Southern Sudan, to mention a few areas where the population isn't as loyal as the central authorities would like. Worse news even: it looks as if all other systems intend to behave just as obediently. Given the great expectations that the Internet has raised for the cause of freedom and democracy--buzzwords at Global Knowledge '97 really--it is a pity to see how the whims of all sorts of undemocratic regimes willbe nicely catered to by tomorrow's Internet satellite operators. In fairness, it should also be said that their cooperation is rooted in both the noble wish not to bypass local telecom operators without official consent and the sound marketing policy needed to be successful at all. As Ron Maehl, president of CyberStar, the GEO system of Loral Space and Communications, puts it, "We will comply with all the local policies. We are not trying to make a political statement, but we are providing a communications service." For some readers belonging to resistance and rebel movements, this dark cloud has a silver lining: whereas LEO systems can very precisely locate a certain user-by measuring Doppler effects--a GEO system knows only in which spot beam the user is sitting. In the case of CyberStars those beams are about 200 kilometers across, whereas the beams of the GEO system of Hughes Communications, called Spaceway, will measure 650 kilometers on the ground. A beam that covers a part of, say, northern Kenya, may well spill across the border, where it can be used by the Sudan People's Liberation Army if the army pays its bills on time-or has them paid by associates in the West, which is a more likely scenario.
All players in this field like to stress how beneficial their services will be for developing countries, even though their public relations material is usually a bit thin on detail. But in fact only the providers of LEO systems will reach the entire globe-not so much because they are saints, but because that is inherent in LEOs. On the other hand, the multiple spot beams of Ka-band GEOs won't reach every corner of the earth. When asked, both Ron Maehl of CyberStar and Edward Fitzpatrick, vice president of Spaceway, said they would go for the most promising markets first. As Fitzpatrick said, "We will put the capacity where the market is-where it is needed most. The first satellite will be launched in the latter part of 1999. We are focused on North America and Asia. Those will probably be followed not much later by Europe and Latin America. We haven't made a hard decision on that point, but that is the current prespective. The whole U.S. will be covered; Europe too; but in Africa and Latin America it will be more selective rather than ubiquitous." For elaboration of the geography of satellite communications, see "Internet, Satellites, and Economic Development" in the Sept./Oct. 1996 issue of OnTheInternet, which also can be found via the hyperlink version of this article.
A second big advantage of LEOs stems from their proximity to the Earth compared with that of GEOs. When you are online via a GEO and browsing the Web by clicking your mouse, the signal will have to travel about 80,000 kilometers to get to the Web page itself, plus another 80,000 to bring back the result. Given the speed of radio signals-300,000 Kilometers per second-that means you will have to wait half a second after each click to see a change on your monitor--not disastrous, but not very handy either. The same delay may result in disaster, though, when you're in a pinball competition with someone who is online via LEO or cable. Furthermore, the half-second delay is not conducive to smooth videoconferencing or voice conversations-certainly not to smooth interruptions. The defendants of GEO systems maintain that this constitutes about all of the damage the delay of their satellites will cause; the LEO people will of course say there is a lot more to it than that. Whatever the case, all of this matters a great deal. No one wants to bet on the wrong horse. The stakes can be high-for instance, for organizations, companies, or ISPs in developing countries that consider buying a VSAT in order to replace a leased line and get broader Internet access. What to do if they gather that GEOs are no good for high-speed Internet? Wait till 2002, when, hopefully, Teledesic becomes available? Or is the geostationary delay nothing compared to delaying their plans? Who will tell them the truth?
In order to make an attempt to see who is right, we'll have to descend into a muddy mix of propagation delays and default TCP buffer sizes. While keeping your breath, keep thinking of the bottom line: whatever the arguments, consumers will eventually decide for themselves what they like best. And maybe you should also keep your ISOC membership card ready, because this is an issue the society ought to keep an eye on. In essence, it is simple: after a certain number of bytes, TCP wants acknowledgment that they arrived well-and will retransmit damaged packets if necessary. If there is a geostationary satellite between two routers, it takes just over a quarter of a second before the load of bytes arrives on earth again. A similar amount of time passes before confirmation reaches the sender, which will then proceed by sending a new load. Hence, the top per-second speed of a TCP/IP link via GEO is about 1.65 (1:0.6) times the number of bytes sent at a time. The maximum size of this buffer is 65,536 bytes, which means that TCP/IP can reach a top speed of about 865 kilobits per second over a GEO. That much is certain. It is also a fact, though apparently a bit less preordained, that a buffer size of 8 kilobytes is usually used-for instance, in Windows 95. That reduces the top speed to 106 kilobits per second: quite attractive today, but probably not tomorrow. According to a white paper about latency-another word for the round trip delay-which can be found at Teledesic's Web site, "Using a small buffer wasn't just an oversight. Small buffers can improve performance in many common circumstances, such as when one computer serves many users simultaneously (e.g., a popular Web server)." So, is Spaceway creating false hopes by announcing that its system will offer a 384-kilobit TCP/IP uplink? Or is CyberStar, which offers similar speeds? Says Maehl, "We can send at 600 kilobits per second and still be consistent with the TCP/IP protocol." Fitzpatrick is equally adamant: "There are some challenges, and they are easily overcome. TCP/IP can be dealt with effectively." And so is John Stevenson, engineer at Intelsat, which owns the world's largest fleet of GEOs: "The differences between satellite and terrestrial are very minor, and they certainly shouldn't be allowed to turn away people to have no Internet access other than by satellite. They are led to believe that GEOs don't work. It is quite the opposite. They work quite well, and in one hop you get to the Internet backbone."
Intelsat--an international consortium with 141 member countries--prides itself in having carried lots of Internet traffic since the very beginning, when the Internet was still ARPANET. Almost all Internet traffic to and from Africa, for instance, goes via Intelsat: via proprietary VSATs or via lines leased from the local PTT that lead to an Intelsat gateway earth station. Only Djibouti, North Africa, and South Africa have cable links to the rest of the world. The Internet takes up an ever more central place in the Intelsat business plan, whereas Stevenson and other engineers devote a lot of research to the pros and cons of TCP/IP via GEO. One of the best outcomes of this appears to be the pairing of TCP/IP and frame-relay-based multiplexing. "Frame relay is highly complementary to TCP/IP and can support up to 8 Mbps on any single link," says Stevenson. The high speed is achieved by taking some special measures. Increasing the buffer size can be one. But when sending your story about life in Rio all the way to the ISP of your aunt Maud in Liverpool, some other routers using 8-kilobyte buffers stand in the way. A better trick is the use of an overlay protocol, which kind of fools TCP by saying all the time that no packets were damaged, while keeping stock of the ones that were. All retransmissions can then be done at the end of the session.
Both methods reduce the transparancy of the GEO link to the rest of the Internet, but to what extent? And who is paying what price for the speeding-up maneuvers? Daggatt says,"The whole point of the Internet is moving away from application-specific networks and proprietary networks in favor of open, public networks--where even private networks will be virtual private networks operating over the public networks--with common protocols, that is, TCP/IP." Stevenson says, "We are definitely interested in collaborating with people to improve the situation. We don't think it is ideal, but today you can do an awful lot of Internet-based application via GEO with the protocols as they stand. There will be a lot of pressure from users to have TCP/IP with a larger default buffer size. There is nothing preordained in this." But according to Daggatt something is: "The GEO guys say, 'You can modify the protocol.' But if you modify the protocol, the party at the other end has to modify it, too. The whole world has to change. And if you optimize it for a high latency-network, for GEO, it is suboptimal for the ground."
At least two advantages of GEOs over LEOs deserve mention. GEOs may be a lot more expensive than LEOs, but in theory you need only three of them to cover the globe, instead of 60-300 as with LEOs. Note that the equation should include infrastructure on the ground. Either LEO network is very expensive because you need intersatellite links or, not having them, a lot of funding is needed for many gateway earth stations to interface with terrestrial networks because LEOs are short-sighted.
All in all, Teledesic is about two or three times as costly as the average GEO system. Even though all parties are shooting for a hardware price of $1000-$2000, those parties all are cagey--maybe ignorant--about transmission costs. If the GEOs will be cheapest, as you'd expect, that will be something to reckon with. How much are we willing to pay for low latency? That may well be the question. Not for Daggatt, who expects, "We figure our end-user cost will be about one-quarter that of the GEOs. So, the better performance of a low-latency LEO link does not require a cost premium-just the opposite." We'll see. If he's right, the people behind the GEO systems would be damn foolish if they didn't immediately cancel all their plans.
The second advantage of GEOs is that they are good at casting data out over entire continents because they're so high up. As Fitzpatrick sees it, "The whole Internet experience will evolve very dramatically in the next few years. The multimedia machines people have in their homes will be able to store huge amounts on the hard drive: 20 gigabytes in 2000." In other words, magazines, statistics, schoolbooks, pizza recipies, encyclopedias, weather forecasts, andtravel warnings can all be digitized and sent to millions of end users at once and at very high speeds. All of that and more will reach your dish antenna. Courtesy of a preselection option and hard thinking on your part, your receiver will ignore everything except a few magazines about pet rodents, the proceedings of the Perry Como Appreciation Society, and pizza recipies containing oregano. In addition, the user can order specific long files, which should arrive a split second later at high speed. In developed and well-wired regions, these data-casting techniques will even take up a predominant place in the broadband satellite systems. As Maehl sees it, "In developing countries, the satellites will go much closer in the network to the end user. You may have more people with dishes that actually do two-way communication. There it will be a primary means of communication, as opposed to a bandwidth enhancer."
The GEO/LEO debate apparently still needs some time to draw to a conclusion, though everybody seems ready to agree that both have at least some inherent advantages. Such seemed to be the state of affairs when on June 17-presto-Motorola announced Celestri, a $12.9-billion plan for 63 LEOs at an altitude of some 1,500 kilometers, fully integrated with a smaller fleet of GEOs. If that weren't enough, a day later an alliance was announced between SkyBridge, the $3.5-billion 64-LEO constellation of the French company Alcatel, and the $1.6-billion CyberStar GEO system-even though it wasn't immediately clear whether the name of the new venture was going to be CyberStar/SkyBridge or SkyStar or CyberBridge. Says Maehl, "Right now we are not planning any GEO/LEO links because we don't see a product that is more efficiently created by having them. The LEOs won't have links either. It is typically a last-mile system: high-speed, two-way communications for end users and get them into the network."
With so many participants in the race to interactive broadband satellite services, this is not the place to describe all their differences or-an even bigger subject-what they have in common. The overview in Via Satellite has already been mentioned; but don't forget to remove page 58, because AT&T's bid, VoiceSpan, has recently been withdrawn. Also, most plans have a good Web site;see the hyperlink version of this article.
One thing to bear in mind is that almost all important players in the satellite world are aware of the exploding global demand for high-bandwidth Internet connections and of their ability to meet at least some of that demand. Various satellite TV providers, for instance, have serious plans. One is the Luxembourg-based Société Europénne des Satellites, otherwise known as ASTRA, which has chosen the typical scenario of data casting first and becoming interactive a couple of years later. In the first phase, you still need your terrestrial Internet link to order the long files, which should reach your PC fast and cheap an instant later via an ASTRA satellite and your dish antenna. While copycats are crowding the rooftops, the credit for thinking up that model of triple Internet connections-low-speed symmetrical terrestrial access plus a high-speed satellite downlink for long files-goes to Hughes Network Systems, which introduced its 400 Kbps Internet downlink system DirecPC a few years ago in the United States and this year in Europe.
Intelsat is also seeking ways to reach the end user. In fact,
what makes the likes of Teledesic, Spaceway, CyberStar, and Astrolink--the
$3.75-billion GEO plan of satellite builder Lockheed Martin--so
special can be done by others as well-to an extent at least-and
quicker if they have their assets in place, like 24 satellites
in the case of Intelsat. After all, no end user is really interested
in the difference between LEO or GEO, between Ka-band and other
bands, between satellite and cable. What end users are interested
in is simply performance, hardware retail prices, transmission
costs, and, first and foremost, whether a system is available