Zoran,***OT***
You are correct that SiGe is still in it's infancy stages but you have to admit that it has a lot of promise. I pulled this article out of my files...I don't have the date and the author anymore...I believe it was posted by Semibull in the ANAD thread a while back.
===============================================================
THE CASE FOR SILICON GERMANIUM
The semiconductor industry has become legendary for making chip designs smaller and faster. But the new transistor technology based on silicon germanium offers additional speed for telecommunications applications without further shrinking the circuitry on a bipolar chip. Once a disappointing material that simply fell apart due to internal structural problems, silicon germanium has reached the
stage of being a manufacturable product.
Virtually every major telecommunications company is working on silicon germanium technology. The technology holds great promise for reducing the cost of consumer products (cellular telephones and direct broadcast satellite entertainment services), improving business applications (telephone network transmission), and helping make possible new applications (collision-avoidance automobile radar).
Another reason the technology is important is that it offers additional performance at lower cost than some competing high-speed circuit technologies such as gallium arsenide. "In terms of cost, silicon germanium fits in between gallium arsenide and silicon, but it's closer to silicon, which is its appeal," says John Long, a professor of electrical and computer engineering at the University of Toronto.
Long is doing collaborative circuit design research with NORTEL (Northern Telecom) using silicon germanium for telecommunications applications.
When IBM began looking at silicon germanium in the early 1980s, semiconductor chips used mostly bipolar technology. Each year, those
chips were expected to advance in speed. But IBM scientists foresaw a
time, in about 1990, when scaling conventional ion-implanted base
transistors would become increasingly difficult.
"The rate at which silicon performance in bipolar was progressing was
recognized to be unsustainable," says Dr. Bernard Meyerson, an IBM Fellow at IBM's T. J. Watson Research Center in Yorktown Heights, New York. As the circuitry on chips was scaled down to achieve greater speeds, it became clear that the performance characteristics of bipolar silicon chips would eventually break down at very tiny dimensions.
"The catch is that when you make a device smaller by simply vertically shrinking the profile, the amount of charge in the base of a bipolar transistor must remain fixed," Meyerson says. "So the density of the charge in the layers goes up and up. Unfortunately, the result is that it becomes easier for electrons to leak through that region. The electrical charge leaks like a leaky faucet. That means you encounter a point, as the device gets smaller, where the transistor no longer
works cleanly because there is significant leakage of current. Reliability goes down and power consumption goes up. The resistance in the base also rises, and that causes problems with electronic noise."
An alternative technology was needed if bipolar silicon was to continue to advance in speed. IBM's answer was silicon germanium, a compound which, while compatible with silicon chip manufacturing processes, had some important differences. "Instead of making things go faster by making them smaller, we needed to change the basic physics of the material in the device so we could get additional performance without a reduction in size," says Myerson.
During this time, IBM was also looking for a way to get around the compromises inherent in designing bipolar silicon chips to go even faster. Enhancing speed usually meant compromising some other chip characteristics. Silicon germanium, it was hoped, would permit device designers to simultaneously increase speed, reduce electronic noise, and increase voltage.
It had been known in the industry since the late 1950s that putting small amounts of germanium in silicon would produce design benefits similar to those found in the semiconductor compound gallium arsenide, which provides fast circuit speed but is more expensive to produce than silicon. But producing silicon germanium had remained a challenge. "It was too hard to grow the thin films of silicon germanium," Myerson says. "The germanium atom is 4 percent larger than the silicon atom, so if you put too much germanium in the silicon, the silicon fell apart." As a result, the early silicon germanium devices didn't work. They failed to show either high speed or acceptable operation, owing to the poor quality of materials available at the time.
Myerson's solution was to create a technique for growing silicon germanium thin films that used both ultra-high vacuum (UHV) and chemical vapor deposition (CVD). This clean system, called UHV/CVD, allowed the films to be grown at much lower temperatures than in the past. "While some silicon processes required temperatures of more than 1,000 degrees centigrade, germanium UHV/CVD is done at below 500 degrees centigrade," says Myerson. "That was a key improvement,
because germanium typically separated from silicon at the higher temperatures. The lower-temperature process also prevented other materials, called dopants, from diffusing away from where they were placed. The benefit of that was that you could grow tightly controlled, uniform, and therefore manufacturable films of this alloy. Once you could do that, you could include that alloy in the heart of the transistor to get additional speed from a same-sized bipolar
device."
The first silicon germanium IC ever built at IBM was a digital- to-analog converter completed in 1992. It had 3,000 transistors and 2,000 passive elements, such as resistors and capacitors. In September of 1996, IBM became the first semiconductor company to qualify what is called "a silicon germanium heterojunction bipolar transistor" for manufacturing. While in most telecommunications circuits the effective maximum speed of commercially available silicon transistors is 15 gigahertz of unity current gain frequency (fT) and about 15 gigahertz maximum available power gain (fMAX), silicon germanium transistors being manufactured by IBM have a speed of 46 gigahertz (fT) and 65 gigahertz (fMAX).
"This technology also improves chip yields," says Dave Harame, who manages silicon germanium product development for IBM. "We measure
our yields by using 4,000 transistors in chains-that's 4,000 0.5 x 2.5æ2 emitter area transistors connected in parallels-and we have gotten as much as 78 percent average yield."
The potential economic benefits of such a technology are enormous,
because silicon germanium chips will be made with the same tools as
silicon chips. This means millions of dollars won't have to be invested in new semiconductor tools, as is typically the case when a shift is made from one generation of chip technology to the next. "You're just changing one aspect of bipolar transistor technology," Meyerson says. "Because silicon germanium technology uses the same dimensions as the previous generation of silicon technology, you can reuse existing silicon tools. It is the first time a profound change in the performance of a technology has occurred without a profound investment."
In the early 1990s, IBM converted its mainframes from bipolar to CMOS
technology. As a result, IBM decided to focus the silicon germanium
technology on bipolar applications in the communications industry, for
which IBM felt it was ideally suited. "In computing," Myerson says,
"CMOS has an inherent advantage over bipolar because it uses significantly less power, and thus generates less heat. But bipolar is a lower-power alternative to CMOS at high frequencies (found in
telecommunications). In very high-frequency operations, where the
transistor is switching constantly, the speed advantage of bipolar lets it operate at lower power in continuous mode than CMOS can." The other advantages silicon germanium bipolar brings to high-frequency
applications are that it has higher gain and less noise than its silicon counterparts.
Larry Larson, an engineering professor at the University of California in San Diego, formerly of Hughes Electronics, said he believes silicon
germanium could extend the performance of silicon bipolar technology by anywhere from 50 to 100 percent. Ted Kamins, a scientist at Hewlett-Packard Laboratories in Palo Alto, California, believes "a 50
percent increase in speed is the type of ballpark figure we are talking about, although it could be a factor of two." Larson and Kamins are referring to the cut-off frequency of the bipolar transistor.
Another approach is to take the implanted base silicon bipolar transistor and insert a silicon germanium base in the device without any optimization. "Right now, silicon germanium offers a 20 percent improvement in performance over straight silicon with the same design rules," says John Long, at the University of Toronto. "That means 20 percent faster digital performance, or 20 percent wider bandwidth, in whatever analog circuit you're designing." IBM has gone further by imbedding silicon germanium into its advanced silicon technology, so it can take full advantage of the benefits of silicon geranium.
According to Kamins, the real gains from silicon germanium will be from high-end applications. "The main benefit is going to be extending silicon technology into the frequency range that, up to now, has been covered by more expensive compound technologies like gallium arsenide." Kamins, who was a project leader in HP's silicon germanium development effort from the late 1980s until about two years ago, says that while silicon germanium is fast, its biggest asset is that chips using it can be built using standard silicon processes. "Silicon germanium will not get up to the highest speeds that have been achieved with compound semiconductors using gallium arsenide and indium phosphide, but it can extend silicon technology somewhat into that range. And it will allow you to leverage the overwhelming investment in silicon technology."
An example, according to Kamins, is the difference between the IBM and
HP approaches. IBM's version of silicon germanium was compatible with
CMOS integrated circuits, while the HP design was compatible with
existing HP bipolar processes. "The deciding factor in success is going to be the degree to which the device process is compatible with a company's particular integrated circuit technology. Companies are likely to design devices to be as compatible as possible with existing processes."
Silicon germanium is likely to make many radio frequency applications
more affordable than they are today, says Ronald Finnila, Director,
Corporate Technology at Hughes Research Laboratories in Malibu, California. Among the potential applications for silicon germanium, according to Finnila, is expanding available bandwidth in both wired and wireless local area networks (LANs) for personal computers. "Silicon germanium has the potential for lowering the costs of existing products or integrating more functions on a semiconductor chip, so you can use fewer chips."
For telecommunications applications, IBM combines bipolar silicon
germanium with standard IBM CMOS to miniaturize and thus drive down
the cost of communications components. For example, the technology could be used in cellular phones to combine the functions of several computer chips onto a single chip. That might result in cellular telephones using three to five times fewer individual chips than they do today. This consolidation of chips could result in less expensive handsets for cellular communication and new Personal Communications Services (PCS), an all-digital alternative to cellular.
Long, whose research for NORTEL uses IBM's silicon germanium technology, said the appeal of using silicon germanium in cellular phones is that it would save money on expensive chip packaging by combining the functions of many chips onto just a few. This would also reduce the drain on the cellular phone's battery, because much of a phone's electrical power use today is for communicating between individual chips. Eliminating inter-chip communications would make phone batteries last longer.
Other cellular phone applications for silicon germanium include power
amplifiers (which boost the signal for phone transmission) and down
converters (which handle received phone signals). "Silicon germanium is very high-speed and hence makes a very high-efficiency power amplifier," Larson says. "For the down converter, a silicon germanium device has low noise-which means the device itself doesn't add a lot of noise to the signal-and that's important because you're trying to receive the tiniest signal you can."
Silicon germanium also might be used in products that consumers use, but never see, such as the driver chips that turn lasers on and off inside fiber-optic telephone lines. According to Myerson, lower costs for silicon germanium could result in its displacing other compounds currently used for those driver chips, such as gallium arsenide and indium phosphide. Long also sees a role for silicon germanium devices in fiber-optic telephone lines. "A silicon germanium chip would allow you to run the fiber-optic cable at higher speeds, giving you higher call capacity." Larson is enthused about future uses of silicon germanium. "I think the potential of the technology is really enormous. I once spearheaded an effort to bring that technology in for applications such as collision-avoidance automobile radar and telecommunications."
Silicon germanium could make collision-avoidance radar for automobiles economically feasible for the first time. According to Myerson, "Silicon germanium will allow products that formerly were accessible only to the military to have consumer cost points." Finnila adds, "Car radar is certainly something that's being looked at. For some, but not all, car radar applications, silicon germanium appears to be a good candidate for lower-cost solutions." He says that, at the moment,
silicon germanium is being considered for side- and rear-viewing radar.
There also may be military applications of the technology. Long says that gallium arsenide "has a captive market in the military because it offers many things that are at the leading edge in semiconductors, but basically because it is fast. Silicon germanium is also fast and VLSI, and it might be possible for it to make some inroads in that area."
Another appealing application for silicon germanium may be direct broadcast satellites, which beam data and video programs to small home satellite dishes. Long says that because the receiving dishes are consumer products, any cost reduction made possible by the new transistor technology would be welcome. "If there's one thing silicon germanium needs," says Long, "it is a high-volume product to drive it. Something like direct broadcast satellite TV, which is a mass consumer item in the US and Europe, could provide that kind of volume."
While IBM is the first with a manufacturable silicon germanium technology, there are companies close behind, according to Long. Still other companies are believed to be interested, but are keeping a low profile. So far, according to Long, IBM holds the leading position in silicon germanium.
Myerson estimates that more than 50 companies worldwide are working with silicon germanium, either in connection with circuit design or fabrication equipment. "It has become a substantial effort," he says.
SF |
