The bulk of today's integrated circuits are made from one of the most abundant materials on this planet, silicon. Scientists are boosting its physical performance by using do pants. Because silicon is itself too dense for the high speed processing required for some of the higher frequencies used in wireless and mobile-phone transistors, researchers have sought and found a way to add a layer of germanium, which is better suited for high frequencies.
Silicon germanium (SiGe) alloy has the effect of speeding the velocity of electrons flowing through the transistor without generating too much heat. Germanium, found in the hills of Uzbekistan, but also the UK, China, and Belgium, according to U.S. Geological Survey, was used to manufacture the first transistors, but researchers were unable to work with the brittle material, so switched to silicon.
Traditionally used in optical networking systems, SiGe is showing up in some early 3G devices, such as transceivers for base stations from Siemens, NEC handsets, Stanford Microdevices, Anadigics and Texas Instruments. IBM labs, where the SiGe process was invented, sees SiGe as useful for power amplification and frequency filters, as well as high-speed optical network switching.
Judging by the product announcements from the likes of Maxim, Atmel, Infineon, and Texas Instruments, SiGe is not only important in current cellular networks, it being used in Bluetooth circuits, new tuners for digital TV, and Wireless LAN chips.
Despite this apparent trend, many analysts believe that SiGe is not absolutely essential for 3G. "It will be important in 3G but not key," says George Bechtel who tracks the market for Strategies Unlimited Ltd.
"Although SiGe is used selectively in 3G, it is not really widely used," says Malcolm Penn, CEO of Future Horizons, a semiconductor consultancy. "It helps to reduce the number of parts inside devices," says Penn.
SiGe wants some of GaAs market
The most commonly used semiconductors for mobile phones and other high frequency devices is Gallium Arsenide (GaAs). Estimates for 1999 GaAs revenues range from 1.5 billion to 2.4 billion dollars. Analysts forecast it will surpass the 10 billion dollar mark sometime in this year. SiGe could capture a chunk of that market.
While GaAs devices are relatively inexpensive and efficient at gigahertz frequencies and cellular handset power levels, they require a split-voltage power supply, which is a nuisance to cell-phone manufacturers, says Peter Clarke, a journalist for Electronic Engineering Times .
Compared with GaAs, say the experts, SiGe operates on lower voltages, uses less power, has improved input performance, develops less low-frequency noise (good for lower jitter), and offers greater potential for functional integration, that is, the integration of the number of separately functioning parts or subsystems within in a handset or base-station.
SiGe is especially suitable for power conservation, as it requires no negative bias for its power amplifier, so there is no need for an external dc/dc bias source, or a power amplifier with a built-in bias generator.
More significantly, the use of GaAs excludes integration with silicon, say the experts. The economics of integration, enabled by SiGe, may win out in the end. The goal these days is to deliver mobile phones whose bill of materials totals between 30 and 50 dollars and one way to reduce cost is to reduce complexity.
Scaling transistors is one way, but it ultimately adds costs to the chip manufacturers that get passed onto the phone makers. According to the May edition of III-Vs Review , shrinking geometry of silicon devices adds an additional $750 million in new equipment costs, whereas adding SiGe processes to existing lines would cost $2.5 million to achieve the same increase in speed.
SiGe's silicon advantage
Reduction in the number of components is a key way to reduce handset costs. Cell phones include several distinct chips: the basic chip set, an RF device, and a power amplifier. In total, about 350 electronic parts and components are found inside a typical handset. About 12 to 20 of them are integrated circuits, more than 200 are so-called passive components, and the rest are discrete, power management, memory and other bits.
(For a simple overview of the guts of the mobile-phone, have a look here.)
That's far too many components and chips for the cellular phone industry. Designers are working to reduce the number of distinct chipsets. SiGe could well be used to increase functional integration.
It is the norm for handsets to support two or more cellular standards. There are very few, if any, GSM handsets being designed today that operate at a single frequency. Add to the mix the need to support Bluetooth, Global Positioning System, and even wireless LAN in mobile terminals, plus compatibility with digital and analogue networks and you have a design challenge, to say the least. Once again, recent product announcement from multiple vendors suggest SiGe is expediting this type of integration.
Meanwhile, pure silicon solutions are progressing to catch up with SiGe. Phillips and Broadcom (through its acquisition of Newport Communication) are on the cutting edge, working to support higher frequencies and processing out of low-power consuming silicon chips.
Suppliers trying to enter the market quickly will not wait for pure-silicon solutions to become commercial, Behzad Razavi, professor of electrical engineering at UCLA told the audience at a recent industry conference. Ultimately designers will use whatever cost-efficient and power-efficient technology meets their need for next generation products.
In the meantime, the first pure play SiGe foundry in the world is being built in Germany. Called Communicant Semiconductor (Frankfurt am Oder), the planned facility will cost $1.5 billion. It's not just aiming to get a share of the mobile phone chip market, but also wireless LAN, Bluetooth and optical networking components where SiGe use is well established.
The foundry counts Intel as a minority shareholder.
The SiGe market will triple this year, from $100 million in 2000 to about $300 million in 2001, confirms analyst Bechtel. Despite the downward trend in sales projection by mobile phone manufacturers, such as Nokia, Ericsson, Siemens and Motorola, Bechtel confirmed to THEFEATURE his forecasts.
Ask Communicant's CEO, Dr. Klaus Wiemer, if SiGe is really necessary for next generation mobile phones and Wiemer, who has held executive positions at the leading semiconductor foundries, Chartered Semiconductor Manufacturing and Taiwan Semiconductor Corp., as well as Texas Instruments, will give you a "does a fish need water" kind of answer:
"Do you need transistors for a computer... No you don't," says Wiemer, answering his own question. "Given a building with the dimensions of a football stadium you could do it with vacuum tubes."
Valerie Thompson is a freelance business and technology journalist, specialized in emerging networking and computing topics. She lives in Zurich, Switzerland.