Mobile Stallite Communications - Part 2 A Technical Critique

Originally published  April, 1997

by Carlo Kopp

© 1997, 2005 Carlo Kopp

In last issue's feature we surveyed the new generation of mobile communications satellites, and briefly reviewed some of the basic technical issues surrounding this new alternative in communications. In this follow-up feature, we will take a closer look at some of the technical issues and fundamental limitations of the existing schemes. The best starting point for any such discussion is to articulate the basic properties which users of long distance computer communications expect of a transmission medium. Using these as a baseline, we will then look at a number of existing schemes and determine what their strengths and weaknesses are in relation to the ideal model. What Would the Ideal Mobile Communications Scheme Offer ?  The trivial answer to this question is infinite zero latency error free bandwidth at zero cost to consumers, with global all weather coverage. As nice as this sounds, fundamental physics and information theory suggest that this cannot be, so the consumer will have to accept some significant compromises. Even so, producing and implementing a scheme which delivers good performance for computer users, as compared to voice users, is not an easy task. To better understand why, it helps to review the most important criteria: Bandwidth is the critical factor in modern computer comms, and the ever increasing demand for multimedia in commercial applications, and imagery transmission in military and other government applications, suggests that bandwidth will remain a critical issue. Whereas traditional circuit switched voice applications can easily merge multiple streams and thus smooth out traffic loads with increasing traffic volumes, computer traffic has the nasty property of being fractal in its statistical properties. If you merge multiple streams of bursty computer traffic, the merged stream will exhibit similar burstiness properties to the component streams (a future feature will discuss the implications of the Bellcore paper in more detail). What this means in turn is that individual users will require multiple Mbit/s class bandwidths to their satellite terminals, if they are to be provided with crisp and responsive interactive performance. Moreover, the shared transmission medium will have to have significant headroom in traffic carrying capacity if it is to accommodate the statistical behaviour of computer traffic without suffering congestion problems. Imagine the consequences of a major congestion collapse on an orbiting global network of routers. Provision of bandwidth is complicated by a number of factors. The foremost of these is spectral congestion, which is a major problem in the US and Europe and is becoming an issue is this country as well. At this time the only bands which are not saturated or approaching saturation with existing traffic are the microwave bands above 15 GHz. Unfortunately, fundamental physics make the exploitation of frequencies above 15 GHz quite difficult. This is because the lower layers of the atmosphere are dense and this density results in significant absorption of microwave signals. For instance water molecules resonate at 22 GHz, and Oxygen at 60 GHz, making even clear air transmission extremely difficult if not impossible in the immediate vicinity of these frequencies. Moisture laden clouds and rain exacerbate this problem, as they are excellent microwave absorbers in their own right. Above 20 GHz heavy rain will knock out transmission with increasing effectiveness with increasing frequency. As a result of these effects, the upper microwave bands are best exploitable in two "windows", one below 22 GHz and the other between 25 to 45 GHz. Above sixty GHz quantum physical absorption makes long haul links impractical. Needless to say infrared or optical frequencies suffer the same problems to an even greater degree. Quality of Service (QoS) is measured by link availability and link error rates. As just noted, the use of the millimetric band is quite problematic primarily as poor weather conditions will compromise QoS very rapidly. Achieving good bit error rates (eg < 10^-9) is generally not a problem with modern receivers, be they microwave or optical, moreover clever use of redundancy in coding schemes can provide excellent resistance to bit errors produced by receiver noise or interference. However, having your link drop out altogether as a cloud passes overhead would be most annoying. An important issue in this context is the ability to cope with interference from natural or man made sources (the latter including intentional jamming in military situations). Conventional modulation schemes do not cope particularly well with interference, and the only defence is to trade bandwidth for redundancy in oder to maintain QoS. Fortuitously, the answer to this problem, as well as the problem of spectral congestion has existed for several decades. It is the technique of Spread Spectrum communications, in which the intended message modulation is spread over a significantly wider bandwidth, by additional modulation with a pseudo-random binary code. The resulting signal appears as noise to a conventional receiver, and is typically ignored by another spread spectrum receiver. A future feature will discuss this in more detail. Latency is the propagation delay incurred by the message as it propagates between sender and receiver. Because radio and light waves travel at the speed of light, which is approximately 3.10^8 m/s in free space, appreciable latencies can be incurred in long distance transmission. This is particularly the case with GEO satellite links. Terrestrial links and LEO satellite schemes therefore have an inherent advantage in the latency contest, simply as they have much lesser distances to cover. However, some latency delay will be incurred with every repeater or router along the way. Therefore schemes which have an advantage in propagation delay latency may lose much ground if many repeaters or routers are used. Coverage is a measure of the footprint of a satellite or constellation of satellites. The footprint of each sat is in turn determined by the type of antennas used, the power transmitted and the sensitivity of the receivers used. In mobile comms or computing applications, the user terminal will require a compact lightweight antenna. This means that good antenna gain and thus terminal sensitivity will require in turn higher power output from the satellite, and better antenna and receiver performance from the satellite, which in turn bites into complexity and thus cost. An LEO constellation will have much smaller footprints per each satellite, compared to say MEO or GEO schemes, with adjacent satellite footprints overlapping one another at their respective boundaries. While the LEO schemes can therefore pack more bandwidth per square kilometre of coverage, they must in turn accommodate a smooth cutover between satellites as one moves out of coverage and another into coverage. This will add complexity to receivers, and thus cost. An issue in the context of coverage is "granularity", or the relative throughput per receiver. Mobile systems mounted on vehicles, ships or aircraft can use a single large high performance antenna/transmitter/receiver which feeds individual users through an onboard LAN. If you wish to feed individual users with mobile laptops or portables, you immediately incur penalties in cost and complexity, particularly at the satellite end. Satellite schemes intended to support individual users will run into a major issue, which is that of population density in the satellites' footprints. Consider an LEO scheme where each satellite has a circular footprint 300 kilometres in diameter. While this satellite is over Tahiti, Bouganville, Baluchistan, the Kalahari or the Simpson desert, it will probably take a handful of connections from geologists, missionaries, the odd tourist and possibly a local government. Consider however the load upon such a satellite over Tokyo, New York, LA, London or Singapore ? If we assume that 5% of the population will each want a 2 Mbit/s connection, then we immediately run up an aggregate bandwidth requirement of the order of hundreds of Gigabits/sec for that satellite alone, and the overheads to manage the state of hundreds of thousands of connections. If we assume that only 0.5% of the population wants a connection, the numbers are still very problematic. This is indeed the Achilles heel of most of the mobile satellite schemes proposed to date. They will indeed provide an unprecedented service for users out in Woop-woop, but are likely to suffer significant difficulties once confronted with the high population densities of the First World. Since most of the world's computer comms users live in the First World, which also produces most of the world's GDP, we must ask a fundamental question - why should shareholders of global communications schemes want to provide a low cost worldwide service when most of the best revenue sources will be unable to extract a truly high quality high speed service from the system and thus are unlikely to subscribe in viable numbers ? Let us however assume that some clever engineering tricks are played and a satellite can be built to carry many Gigabits/s of traffic within its footprint. We are then confronted with the issue of carrying this traffic to adjacent satellite borne routers, and forwarding it to its destination. Should we adopt conventional shortest path routing algorithms, we could simply trace great big lines across the globe between the First World's major population centres, and expect that satellites along these lines will be extremely busy simply carrying traffic between their neighbours. Again, we are likely to confront similar problems in saturation of routers, and thus performance problems due queuing delays. So to avoid saturating satellites along paths between Europe, the US and the Far East, we adopt a routing scheme which channels the traffic through less geometrically advantageous satellites which are lightly loaded with traffic, as they are passing over Third World countries. We then begin to incur latency delays through having to hop across more satellites, and traverse greater distances. In any event, we still end up with ever increasing traffic density as we approach the First World. This raises serious questions about the technical and commercial viability of a number of mobile satellite schemes, particularly in relation to the carriage of high speed computer traffic. Indeed the the only viable near term consumer of high bandwidth digital satellite comms will be the military in the First World. The US DoD Milstar constellation, with four cross-linked GEO orbit satellites, using 60 GHz crosslinks, provides a T1 service for a limited number of channels, and is limited to 2,400 Bits/s for its standard high volume low data rate service. The Milstar I/II is both large and expensive to build and deploy. Conclusions The conclusion we can draw from a basic analysis is that the current generation of proposed mobile satellite communication schemes suffer significant technical limitations in the carriage of computer traffic which will in turn reduce their utility in the highest density and thus best revenue generating parts of the world. They do however provide useful if limited connectivity to parts of the world's geography which are not provided with viable terrestrial links. Whether provision of service to such areas will provide sufficient revenue sources for a follow-on generation of sats remains to be seen.

$Revision: 1.1 $

Last Updated: Sun Apr 24 11:22:45 GMT 2005

Artwork and text © 2005 Carlo Kopp