1.2 Integrated Circuit Manufacturing
Integrated circuit technology is based on our ability to manufacture huge numbers of very small devices—today, more transistors are manufactured in California each year than raindrops fall on the state. In this section, we briefly survey VLSI manufacturing.
Most manufacturing processes are fairly tightly coupled to the item they are manufacturing. An assembly line built to produce Buicks, for example, would have to undergo moderate reorganization to build Chevys—tools like sheet metal molds would have to be replaced, and even some machines would have to be modified. And either assembly line would be far removed from what is required to produce electric drills.
Integrated circuit manufacturing technology, on the other hand, is remarkably versatile. While there are several manufacturing processes for different circuit types—CMOS, bipolar, etc.—a manufacturing line can make any circuit of that type simply by changing a few basic tools called masks. For example, a single CMOS manufacturing plant can make both microprocessors and microwave oven controllers by changing the masks that form the patterns of wires and transistors on the chips.
Silicon wafers are the raw material of IC manufacturing. The fabrication process forms patterns on the wafer that create wires and transistors. As shown in Figure 1-1, a series of identical chips are patterned onto the wafer (with some space reserved for test circuit structures which allow manufacturing to measure the results of the manufacturing process). The IC manufacturing process is efficient because we can produce many identical chips by processing a single wafer. By changing the masks that determine what patterns are laid down on the chip, we determine the digital circuit that will be created. The IC fabrication line is a generic manufacturing line—we can quickly retool the line to make large quantities of a new kind of chip, using the same processing steps used for the line’s previous product.
Figure 1-1 A wafer divided into chips.
circuits and layouts
Figure 1-2 shows the schematic for a simple digital circuit. From this description alone we could build a breadboard circuit out of standard parts. To build it on an IC fabrication line, we must go one step further and design the layout, or patterns on the masks. The rectangular shapes in the layout (shown here as a sketch called a stick diagram) form transistors and wires which conform to the circuit in the schematic. Creating layouts is very time-consuming and very important—the size of the layout determines the cost to manufacture the circuit, and the shapes of elements in the layout determine the speed of the circuit as well. During manufacturing, a photolithographic (photographic printing) process is used to transfer the layout patterns from the masks to the wafer. The patterns left by the mask are used to selectively change the wafer: impurities are added at selected locations in the wafer; insulating and conducting materials are added on top of the wafer as well. These fabrication steps require high temperatures, small amounts of highly toxic chemicals, and extremely clean environments. At the end of processing, the wafer is divided into a number of chips.
Figure 1-2 An inverter circuit and a sketch for its layout.
Because no manufacturing process is perfect, some of the chips on the wafer may not work. Since at least one defect is almost sure to occur on each wafer, wafers are cut into smaller, working chips; the largest chip that can be reasonably manufactured today is 1.5 to 2 cm on a side, while a wafer is in moving from 30 to 45 cm. Each chip is individually tested; the ones that pass the test are saved after the wafer is diced into chips. The working chips are placed in the packages familiar to digital designers. In some packages, tiny wires connect the chip to the package’s pins while the package body protects the chip from handling and the elements; in others, solder bumps directly connect the chip to the package.
Integrated circuit manufacturing is a powerful technology for two reasons: all circuits can be made out of a few types of transistors and wires; and any combination of wires and transistors can be built on a single fabrication line just by changing the masks that determine the pattern of components on the chip. Integrated circuits run very fast because the circuits are very small. Just as important, we are not stuck building a few standard chip types—we can build any function we want. The flexibility given by IC manufacturing lets us build faster, more complex digital systems in ever greater variety.
Because integrated circuit manufacturing has so much leverage—a great number of parts can be built with a few standard manufacturing procedures—a great deal of effort has gone into improving IC manufacturing. However, as chips become more complex, the cost of designing a chip goes up and becomes a major part of the overall cost of the chip.
In the 1960s Gordon Moore predicted that the number of transistors that could be manufactured on a chip would grow exponentially. His prediction, now known as Moore’s Law, was remarkably prescient. Moore’s ultimate prediction was that transistor count would double every two years, an estimate that has held up remarkably well. Today, an industry group maintains the International Technology Roadmap for Semiconductors (ITRS), that maps out strategies to maintain the pace of Moore’s Law. (The ITRS roadmap can be found at http://www.itrs.net.)
Figure 1-3 shows advances in manufacturing capability by charting the introduction dates of key products that pushed the state of the manufacturing art. The squares show various logic circuits, primarily central processing units (CPUs) and digital signal processors (DSPs), while the black dots show random-access memories, primarily dynamic RAMs or DRAMs. At any given time, memory chips have more transistors per unit area than logic chips, but both have obeyed Moore’s Law.
Figure 1-3 Moore’s Law.
The most basic parameter associated with a manufacturing process is the minimum channel length of a transistor. (In this book, for example, we will use as an example a technology that can manufacture 180 nm transistors.) A manufacturing technology at a particular channel length is called a technology node. We often refer to a family of technologies at similar feature sizes: micron, submicron, deep submicron, and now nanometer technologies. The term nanometer technology is generally used for technologies below 100 nm.
The next example shows how Moore’s Law has held up in one family of microprocessors.
Cost of manufacturing
IC manufacturing plants are extremely expensive. A single plant costs as much as $4 billion. Given that a new, state-of-the-art manufacturing process is developed every three years, that is a sizeable investment. The investment makes sense because a single plant can manufacture so many chips and can easily be switched to manufacture different types of chips. In the early years of the integrated circuits business, companies focused on building large quantities of a few standard parts. These parts are commodities—one 80 ns, 256Mb dynamic RAM is more or less the same as any other, regardless of the manufacturer. Companies concentrated on commodity parts in part because manufacturing processes were less well understood and manufacturing variations are easier to keep track of when the same part is being fabricated day after day. Standard parts also made sense because designing integrated circuits was hard—not only the circuit, but the layout had to be designed, and there were few computer programs to help automate the design process.
Cost of design
One of the less fortunate consequences of Moore’s Law is that the time and money required to design a chip goes up steadily. The cost of designing a chip comes from several factors:
- Skilled designers are required to specify, architect, and implement the chip. A design team may range from a half-dozen people for a very small chip to 500 people for a large, high-performance microprocessor.
- These designers cannot work without access to a wide range of computer-aided design (CAD) tools. These tools synthesize logic, create layouts, simulate, and verify designs. CAD tools are generally licensed and you must pay a yearly fee to maintain the license. A license for a single copy of one tool, such as logic synthesis, may cost as much as $50,000 US.
- The CAD tools require a large compute farm on which to run. During the most intensive part of the design process, the design team will keep dozens of computers running continuously for weeks or months.
A large ASIC, which contains millions of transistors but is not fabricated on the state-of-the-art process, can easily cost $20 million US and as much as $100 million. Designing a large microprocessor costs hundreds of millions of dollars.
Design costs and IP
We can spread these design costs over more chips if we can reuse all or part of the design in other chips. The high cost of design is the primary motivation for the rise of IP-based design, which creates modules that can be reused in many different designs. We will discuss IP-based design in more detail in Section 1.5.
Types of chips
The preponderance of standard parts pushed the problems of building customized systems back to the board-level designers who used the standard parts. Since a function built from standard parts usually requires more components than if the function were built with custom-designed ICs, designers tended to build smaller, simpler systems. The industrial trend, however, is to make available a wider variety of integrated circuits. The greater diversity of chips includes:
- More specialized standard parts. In the 1960s, standard parts were logic gates; in the 1970s they were LSI components. Today, standard parts include fairly specialized components: communication network interfaces, graphics accelerators, floating point processors. All these parts are more specialized than microprocessors but are used in enough volume that designing special-purpose chips is worth the effort. In fact, putting a complex, high-performance function on a single chip often makes other applications possible—for example, single-chip floating point processors make high-speed numeric computation available on even inexpensive personal computers.
- Application-specific integrated circuits (ASICs). Rather than build a system out of standard parts, designers can now create a single chip for their particular application. Because the chip is specialized, the functions of several standard parts can often be squeezed into a single chip, reducing system size, power, heat, and cost. Application-specific ICs are possible because of computer tools that help humans design chips much more quickly.
- Systems-on-chips (SoCs). Fabrication technology has advanced to the point that we can put a complete system on a single chip. For example, a single-chip computer can include a CPU, bus, I/O devices, and memory. SoCs allow systems to be made at much lower cost than the equivalent board-level system. SoCs can also be higher performance and lower power than board-level equivalents because on-chip connections are more efficient than chip-to-chip connections.
A wider variety of chips is now available in part because fabrication methods are better understood and more reliable. More importantly, as the number of transistors per chip grows, it becomes easier and cheaper to design special-purpose ICs. When only a few transistors could be put on a chip, careful design was required to ensure that even modest functions could be put on a single chip. Today’s VLSI manufacturing processes, which can put millions of carefully-designed transistors on a chip, can also be used to put tens of thousands of less-carefully designed transistors on a chip. Even though the chip could be made smaller or faster with more design effort, the advantages of having a single-chip implementation of a function that can be quickly designed often outweighs the lost potential performance. The problem and the challenge of the ability to manufacture such large chips is design—the ability to make effective use of the millions of transistors on a chip to perform a useful function.