Satellite Systems Gear Up To Meet The Challenges Of Global Networks

Nothing epitomizes the global nature of everyday life quite like satellite communications. From video and television to voice and basic messaging services, it's the heart and soul of the drive to instantly access information, anytime, anywhere. While many attribute the shrinking world to the Internet, satellites have been steadily shrinking our world for decades. That said, though, it's the Internet that allows us to take full advantage of the terrestrial and satellite networks that span the globe. Furthermore, it's the Internet that will be driving much of the innovation in both networks over the coming years.

While the advances in terrestrial networks will continue exponentially, satellite systems, by their very nature, tend to advance somewhat more slowly. The issue boils down to cost. It's both a lot easier as well as much cheaper to add another fiber-optic trunk line than to put another satellite in space. As a result, bandwidth in space is like gold dust, and it behooves designers to make maximum use of whatever spectrum is available.

This has lead to many advances in higher-order digital modulation schemes, such as 8-level phase-shift keying (8PSK), that lead to greater overall data throughput and spectral efficiency. In addition, techniques like turbo coding, with a promised 50% leap in performance, are about to hit the mainstream. These technologies will go a long way toward lifting the bottleneck that satellite systems are facing as they try to compete with their terrestrial brethren.

In many instances, "compete" may be too strong a word to use, if not downright misleading. Except for the satellite TV (STV) versus cable war, which lately has escalated—thanks to recent FCC rulings—satellite companies are starting to see that more can be gained by complementing terrestrial networks instead of competing with them. This is evident by the recent activation of voice-oriented, bent-pipe, satellite constellations that can connect a caller to virtually every part of the globe, while taking advantage of the current cellular network when appropriate.

Whatever the satellite network, and whether it's for voice, STV, or the more-recent digital radio and data casting, the rising number of subscribers has had the effect of making the semiconductors within the devices almost commodity items. Combined with an overall shortage in RF designers, this has led to a demand for highly integrated, low-cost, off-the-shelf solutions that allow manufacturers to get up and running quickly with minimum overhead.

A perfect example of this cutthroat, low-cost approach to satellite communications is the STV business. With DirecTV (DSS based) and EchoStar (DVB based) now battling both each other as well as the cable companies, cost is everything when it comes to penetrating the home. Beyond that, quality of service and available options are the next big separators, relative to cable.

Success With STV To date, DirecTV and Echostar have both had favorable reviews in the service department, and the recent FCC ruling that allows them to carry local channels has given them appreciable boosts. Current figures put the number of satellite subscribers at about 14 million in the U.S. alone.

The calling card of STV has been the digital quality of its video and audio, as well as the broader content, but this is now being matched by cable companies that are quickly deploying digital set-top boxes (STBs). Running off upgraded lines and headends, these raise the number and the quality of channels available in order to match that of satellite service. This has already slowed satellite penetration in places like Texas, where digital-cable deployment has been particularly aggressive.

Until now, a major disadvantage of STV has been the need for multiple low-noise blocks (LNBs) to facilitate the picture-in-picture and record-and-watch capability. This almost doubles the cost of a satellite system, as everything including the LNB, tuner, demodulator, and finally the MPEG-2 decoder must be duplicated. But, the satellite companies aren't outdone here. Both DirecTV and Echostar have instituted promotions that practically give the basic equipment away—assuming the subscriber signs a contract for at least one year's service. The newer equipment almost invariably comes with dual LNBs, so the subscriber only has to pay for the additional STB and any required cabling.

Even if the cost and quality of service were equal, STV still has two major advantages over cable. The first strikes to the core of what satellites are all about—reach! No matter where a subscriber is, satellite reception is almost guaranteed to be within reach. Cable deployment is still restricted to populated areas. Furthermore, even potential subscribers in populated areas are tired of the monopolistic practices of cable companies, which often don't face competition from other cable companies in their home markets. This monopoly doesn't sit well with subscribers, causing many to opt for satellite in the face of constantly rising costs.

This competition, and the rising numbers of systems being deployed, have affected chip manufacturers. For instance, tuner manufacturers are seeing a rise in demand. For every subscriber there are now at least two STBs, each needing its own tuner. This suits companies like Mitel Semiconductor just fine. Recently, it introduced the SL1935 single-chip, synthesized, zero-IF, quadrature down-conversion tuner. Covering the range from 950 to 2150 MHz, the chip is both DVB (950 to 2150 MHz) and DSS (950 to 1450 MHz) compatible. The device integrates a low-phase-noise PLL frequency synthesizer, a low-noise amplifier, and automatic gain control (AGC). The local oscillator (LO) and crystal reference are off-chip (Fig. 1).

The chip signifies the recent dramatic shift away from the typical "canned" tuners that have taken up so much real estate in both cable and satellite STBs. The high integration is essential if OEMs are to meet the price points and time-to-market requirements that STV providers are demanding.

Key to Mitel's design is the use of a direct-conversion architecture which minimizes the number of discrete components to lower system cost. How Mitel overcame the many hurdles to direct-conversion designs remains under wraps for now. According to Paul Fellows, Mitel's marketing program director for digital TV technologies, "The achievement is the result of the application of good design techniques to minimize LO leakage (the number one cause), and by paying close attention to the block distribution throughout the system." This pithy answer belies the complexity of the technology required, considering the noise margins. But in any event, the design allows the company to package the complete tuner in a 36-pin SSOP for $2.30 each in 100,000-piece lots.

The chip features a dynamic range of −75- to −15-dBm, an input IP3 and IP2 (at an operating sensitivity of −20 dBm) of 5 and 20 dBm, respectively, and a PLL phase noise of −80 dBc/Hz at 10 kHz. The I and Q outputs are optimized for use with the company's VP310 1- to 45-Mbaud demodulator. Supplies required for the demo board include 5 V for the synthesizer, converter, and baseband sections, as well as 30 V for the varactor line.

Mitel isn't alone in the rush to make the tuner a one-component device. The company faces direct competition from the likes of Conexant, which has announced its second-generation, direct-down-conversion tuner, the CX24108. Both companies seek to have the tuner become a commodity item that OEMs can simply pick and place as a unit. Conexant gets a little closer to this ideal by having everything integrated on-chip, including the LO, while running the entire chip off a single 5-V supply.

The Conexant tuner comes in a 48-pin, thermally enhanced TQFP, and feeds the company's CX24110 QPSK demodulator with integrated forward error correction (FEC). The combined chip set weighs in at $10 per set in large quantities. Together, the two chips provide a complete front-end solution, which complements the company's recently announced CX22490/1 STB back-end MPEG-2 audio/video decoder and 2D graphics processor (Fig. 2).

Geographical Differences Conexant's approach highlights some interesting differences between the U.S. market and the rest of the world. Though the company wanted to make the product mainstream, inexpensive, and applicable worldwide, that meant making the feature set less compatible with the U.S. and more targeted toward the larger combined European and Asian STB markets. In the U.S., the cable and satellite STB markets are very distinct, but in Europe and elsewhere, the STB chip-set solutions are very similar for cable, satellite, and even terrestrial broadcast systems. This allows more reuse, a key factor if costs are to come down.

The same concept applies to the software platforms, which also are similar in Europe. As far as DSS or DVB decoding, the CX22490/1 doesn't care because the main difference between the two lies in the multiplexing standard. Otherwise, both are based on MPEG-2. DSS came about in 1994, shortly before the MPEG-2 multiplexing standard was finalized, hence the difference. From an audio/video-decoding point of view, DSS is almost standard MPEG-2 at this point, though the established infrastructure will prevent a full conversion.

In any case, Conexant has implemented the demultiplexing function in firmware, using a proprietary 32-bit microprocessor, thereby allowing the chip to adapt to different packet formats. According to Eric Rayel, senior product-line manager at Conexant, "Overall flexibility is key for the back end, as it is here that many of the differentiators between STBs will be implemented. Key areas include analog video capture, 2D animated-graphics support, multilayer depositing, interactive advertising, flying logos, conditional access methods, and support for the popular personal video recorder (PVR) feature that users are demanding."

PVR allows delivery of the transport stream to a hard-disk drive. Once there, it's recorded in a transport packet format or as a coded elementary stream, possibly scrambled. Tags and markers might also be needed. As no one is certain how the OEMs may want to accomplish the above features, the goal for semiconductor manufacturers is infinite flexibility.

To accomplish this, Conexant uses an embedded 32-bit, 160-MIPS ARM940T RISC processor (Fig. 2, again). The device comes with a software platform that includes fully integrated OpenTV, pSOS, and VxWorks run time and driver-software components. Modem, hard-disk attachment, and USB 1.1 support are provided too. Presently sampling, the CX22490/1 is priced at $25 each (120-MHz CPU) or $28 each (160 MHz) in quantities of 100,000.

Voice Ramps Up While TV remains one of the chief drivers of satellite technology, events have recently unfolded in other areas that altered the skyscape somewhat. Global voice communications has fired the imaginations of investors, service providers, and distributors worldwide—despite the failure of the infamous Iridium project. Right now, leading the pack in this area is Globalstar, with its 48 satellites, orbiting in eight planes, inclined at 52°, approximately 900 miles up in a low earth orbit. Four spare satellites are added for redundancy.

According to Megan Fitzgerald, senior vice president of space operations, "There's a large underserved population that still needs telecom services, with up to three billion lacking basic phone services, and up to 50 million on registered waiting lists. China has 35 million cellular users, and the country isn't even covered with service yet."

The company has taken the wise approach of offering an extension to cellular, instead of trying to replace it. Additionally, it will offer a fixed service to areas that don't have terrestrial, let alone cellular service. Fitzgerald says, "We provide the infrastructure, then partner with regional service providers, which purchase and install the gateway in their territory, and then offer Globalstar as a service."

Globalstar's system is a CDMA-based, bent-pipe system, with all the processing done on the ground. As the CDMA technique is ground based, the company can take advantage of improvements to the CDMA system to enhance its gateways and phones.

So, why is Globalstar still around while Iridium went defunct? The main difference lies in methodology. With Globalstar, when a call is made, it goes from the phone to the satellite, to the gateway on the ground, then through the terrestrial network. With one phone and one satellite, a call can be completed. The system doesn't employ crosslinks from satellite to satellite, as Iridium had attempted. This method was very demanding in terms of people on the ground, the fuel needed on the satellite to maintain the accurate pointing, and the raw processing power necessary to maintain position as well as process the incoming and outgoing calls. Also, as much of that processing was done on board, it was locked into finite capabilities because of that system's TDMA architecture. Globalstar's system features path diversity too, which allows up to three satellites to cater to a call—with soft handoff should one satellite get obscured.

So far, the company has partnered with AirTouch/Vodaphone/Bell Atlantic to serve Canada, the U.S. and parts of Mexico. In Europe, partnerships include France Telecom, with other partnerships covering Northern Europe, Western Europe, China, Russia, and Australia, where Vodaphone has three gateways, giving 100% coverage.

The system uses satellites purchased from Space Systems Loral, Palo Alto, Calif., and has handset partnerships with Qualcomm and Ericsson, with initial tri-mode handsets costing anywhere from $1500 to $2500. "Tri-mode" refers to the option to use the cellular network whenever possible. This saves on call charges, which range from $0.49/min. to $1.50/min., depending on the plan.

Significantly, Motorola has suggested Globalstar as an alternative to its Iridium customers. To this, Globalstar has responded by offering a $295 trade-in rebate toward the purchase of any new Globalstar phone.

Data In The Sky While both the TV and voice markets are, or soon will be, well served by satellite service, that's only the tip of the iceberg as far as development is concerned. Internet access is key and the saving grace for satellites will be a two-way feed allowing them to compete with cable, DSL, and fixed-wireless options.

Getting downstream is easy. In fact, a satellite is perfect for this, with its wide coverage and potentially high data rates. Hughes Network Systems (HNS), which owns DirecTV, has also implemented the DirecPC system which promises download speeds of over 400 kbits/s. Implemented using the company's new DirecDuo dishes, the system simply assigns one or more of the 150 available 6-MHz channels to carry the data. Potentially, with modulation giving two bits per hertz, data rates of over 10 Mbits/s are possible.

As for Globalstar's data pipe, there's much work to be done. Though the system is IS-95 compatible, and already a digital standard, the basic rate is 9600 bits/s—even though lines can be paralleled to achieve over 200 kbits/s. Globalstar's future voice competitor, New Ico—which is only just getting off the ground after being saved from bankruptcy—promises rates of 64 kbits/s by 2002. Neither company can hold a candle to what's available on the ground. In addition, all satellite systems still suffer from one nagging flaw—the return path must be through a phone line. This requires either a regular dial-up modem connection, or digital-subscriber-line (DSL) service.

Of course, the alternative is to have a transmitter provide the upstream to the satellite. But, that places far too much expense at the consumer end, while requiring a complete overhaul of the incumbent satellite constellations. Newer satellites, such as Hughes' Spaceway network, are gearing up for data, but the benefits are a number of years away.

In the interim, the current satellite infrastructure can be employed to facilitate datacasting. This is a point-to-multipoint distribution system that allows ISPs to take advantage of the satellite's wide swath of coverage.

Leading the foray into this technology is Orblynx Inc., which has developed the IDS2000 distribution system to deliver web content, newsfeeds, and streaming media to customers around the globe using satellite networks. The IDS2000 operates at both ends of the network and uses satellites to create an Internet overlay network that uses web caching and point-to-multipoint distribution to deliver content quickly and reliably. IDS2000 will be operated initially using a 2-Mbit/s outbound satellite multicast carrier.

Orblynx is at the cutting edge of a new generation of what are called content distribution networks (CDNs). So far, Orblynx, INTELSAT, Teleglobe, and Swisscom successfully showcased a precommercial service demonstration, employing the prototype IDS platform, at the ITU Telecom 99 exhibition last October.

Digital Radio—Everywhere! As with data casting, digital radio too takes advantage of the potential footprint of a satellite transmission to reach as many receivers as possible. The automotive market is the primary target for this technology, with radio suppliers gearing up to meet the demands of year 2002 car models.

Initial entrants will be costly, at upwards of $500 and, therefore, will be confined to car models in excess of $40,000 for the cost to be absorbed. The subscription fee of approximately $50/month will also have to be factored in.

Despite the high cost of entry, Analog Devices Inc. sees this as a big emerging market and, as a result, is throwing much of its expertise into the technology's realization. ADI's focus is in the lower-IF range of 1 to 2 GHz, and anywhere below that. Therefore, the company is in a prime position to milk the digital-radio phenomenon.

While digital satellite radio has much potential, according to John Greichen, product marketing manager at ADI, a number of hurdles must be overcome. Such issues as the speed of the automobile versus the radio's tracking ability, and the effect of surrounding buildings have to be tackled. Greichen says, "This demands the incorporation of some kind of memory system, but this has to be somewhat more robust than that of a CD player to handle blips of 200 ms or more." The only other option is to hand the problem over to the digital-signal processor (DSP), algorithms for which are still in progress.

For ADI, satellite radio is but one of a slew of wireless applications that have jump-started the market for standard, off-the-shelf (OTS) products. The company has customers clamoring for a good AGC, detector, or any number of parts that the user can quickly design in, instead of starting from scratch. Until digital radio is proven, OTS parts will be the way to go, before finally integrating the solution onto a single chip.

Regardless of the application, almost all satellite technologies have one major requirement—more bandwidth. Unfortunately, the infrastructure is such that bandwidth is gold dust. Plus, the only way to improve the possible data throughput in the near future is to jump to a higher modulation scheme, decrease the effect of noise on the channel, or both. Much of the leading-edge research is being carried out in this area, the results of which will be unfurled over the coming year.

Taking the lead this year is STMicroelectronics, which recently licensed turbo-code, FEC technology from France Telecom. This, combined with a jump to 8PSK modulation, gives the company a major boost in data throughput and noise tolerance.

FEC itself isn't new, and it has allowed communications systems to get within 2 to 3 dB of the theoretical limit (Shannon's limit) of transmission systems. Turbo codes, which double the encoding and then use iterative decoding to get a result closer to the actual transmitted data, allow the systems to reach within 0.5 dB of the theoretical limits of transmission. The end result is that the coding gain (the improvement possible in the bit-error rate for a given signal-to-noise ratio) increases anywhere from 8 to 12 dB depending on the system of turbo codes used, and the data throughput.

The advantages turbo codes give can be applied in various ways. For a given power level (which is often determined by regulatory requirements), the data throughput can be increased, meaning more channels in the U.S. market. For the European market, which doesn't have as great a need for channels, the size of the dish can be shrunk by up to 30%, leading to lower system costs. The permutations are numerous, with implementations that depend on the requirements of the application.

As for the jump to 8PSK modulation, this allows for three bits per symbol, versus two bits for QPSK. When combined with turbo codes, the spectral efficiency of the combination jumps by up to 40% over QPSK with Reed-Solomon/Viterbi FEC. The maximum BER is approximately 10⁻¹².

To be released this summer, STM is gearing up its initial STV0399 8PSK-only offering for the low end. This will be followed by the turbo-code chip to arrive by the third quarter of this year. Both will be done on 0.18-µm CMOS, with the company's STV0299 demodulator and tuner all included. The tuner uses a zero-IF architecture.

The complete solution will eliminate the need for any RF routing on the pc board, and will interface directly to the company's 55xx line of back-end MPEG-2 decoding chips for STBs to give a complete reference platform (Fig. 3). The two-chip, front-end solution will migrate to a one-chip version by early 2001 as volumes ramp up.

STM isn't the only game in town when it comes to turbo codes. Advanced Hardware Architectures has been evangelizing the technology for a number of years, with some success from companies like Comtech, in Canada. In fact, Comtech recently won a Canadian government contract based on its use of AHA's turbo product-code (TPC) technology.

According to Dave Williams, product marketing manager at AHA, "We're doing iterative decoding based on block codes, which doesn't have an error floor, thereby giving us BERs beyond what we can simulate." He continues, "Turbo convolutional codes (TCCs) \[such as those used by STM\] tend to be expensive, but the cost can be hidden within an overall system solution. In addition, TPC technology is not under license."

AHA has recently brought out its Astro-OC3 TPC encoder/decoder, which is capable of 155-Mbit/s data rates (Fig. 4). At the heart of the device is the soft-in-soft-out decoder, which allows the system to say that it reserves the right to change its answer pending further information. The likelihood of whether or not the answer is a one or a zero is determined iteratively.

Key specifications also include an encode latency of less than 10 clocks, flexible coding rates from 0.25 to 0.98 (essential in instances such as poor weather), as well as 32-bit CRC insertion and checking with programmable packet length.

Because turbo codes are relatively new, with respect to availability, silicon is hard to come by. There are, however, a number of things to look out for when evaluating them, including how real it actually is, the data rate, flexibility, latency, and error floors. Many researchers have tried to figure out TCC's error floor, but it's too complex an issue. All they can do is run live data and check for errors. No mathematical models exist. The problem stems from the fact that TCCs are data dependent. Under many circumstances, error floors can just crop up. One advantage TPCs have, though, is that the technologies have varying strengths according to the application. For instance, TCCs show slightly better performance at higher BERs and at lower code rates.

Regardless of whether TPCs or TCCs are used, turbo codes will become an integral part of satellite-system design in the coming years, and they won't stop there. Any application that demands optimal use of bandwidth, such as wireless LANs, will be a prime target for the technology.

The rapid development of satellite systems over the past number of years has not taken place in a vacuum, but in fact reflects and contributes to many cultural changes that also have taken place. The fall of the Berlin Wall, for example, opened the floodgates to the East-West cooperation that's so essential to reducing the overall cost of getting the satellite infrastructure in place to begin with. American companies are launching satellites from Russian-controlled territory, and vice-versa.

The improved communication that the first wave of satellite systems brought about helped cement the bonds of cooperation, as nothing breaks down barriers like open communication. This is a self-propagating phenomenon, with the latest example being the push from within China to join the global community. As companies like Globestar connect the most remote parts of China's regions to the outside world, first by voice and later through data on the Internet, the effects on the world community can barely be imagined. Now could be a good time to begin learning Mandarin.

Companies That Contributed To This Report
Advanced Hardware Architectures (509) 334-1000 www.aha.com Analog Devices Inc. (800) 262-5643 www.analog.com Conexant (800) 981-8703 www.conexant.com DirecTV www.DirecTV.com Echostar www.Echostar.com Globalstar (408) 933-4000 www.globalstar.com	Hughes Space and Communications Company (310) 364-6000 www.hsc.com Ico Global Communications +44 0 20 8600-1000 www.ico.com Mitel Semiconductor (613) 592-2122 www.mitel.com Orblynx Inc. (703) 279-6416 www.orblynx.com STMicroelectronics (781) 861-2650 www.st.com