When the wafer is completely processed, it will have 100-200 identical chips which perform the same basic functions on it. Only a fraction of these chips will be functional. A top view of a typical wafer would look as follows:

Figure 2.

Each chip is then tested by a computer to determine whether it properly performs the desired electronic functions. If a particular chip is good the tester moves on to the next chip. If a particular chip is bad it drops a spot of ink on the chip indicating that it is to be rejected.

Next, the chips on the wafer are separated from each other. The rejects are thrown away and the good chips are assembled into a package and shipped to the customer. Attachment 1 shows a picture of an unpackaged chip 50 times its actual size and Attachment 2 show a packaged chip which is capable of being used in a customer's system. In this form the chip can now be used in automobiles, computers and the like.

HOW TO DESIGN A CHIP

A chip manufacturer must first conduct a marketing study to determine the functions which its customers would like the chip to perform. Once the functions of a chip are defined, it is the job of a circuit design engineer to develop a circuit to implement these electronic functions. The circuit engineer develops the circuit by making a "schematic" representation of the manner in which transistors must be connected to implement the appropriate electronic function. Often 20 sheets of paper will be used to draw the entire schematic of a complex chip. The schematic would be drawn on paper and look as follows:

The patent laws are available for protection of these electronic circuits provided that the circuit meets the useful, novel, and nonobvious requirements for the patent laws.

The circuit schematic is a paper document and is not useful until it is fabricated on a chip. A layout design engineer must take the circuit schematic and layout patterns which can be imprinted onto a wafer to form a chip. This is a very expensive and time consuming process. Typically, this layout will not rise to the level of invention required by the patent laws. The layout must be done in a timely manner so that the final chip can be available in the market place when it was needed. More importantly, the layout must be very compact to minimize the cost of the chip. The smaller the chip, the more chips which can be put on a single wafer and consequently, the better chance that the wafer will yield more good chips. The layout will be retained on a magnetic tape. Attachment 3 shows the 8 patterns used to manufacture a typical chip having 150,000 transistors on it.

METHODS OF TRANSFERRING THE PATTERN FROM THE DATA BASE TO THE WAFER

The original method for transferring these patterns from the tape to the wafer consisted of converting the tape to glass reticles, converting the glass reticles to glass or chrome masks and then using the mask to imprint the pattern on the wafer. The tape is entered into a computer which converts the information on the tape into a glass reticle. A reticle must be made for each pattern which will be printed on the wafer. A reticle is referred to as a "Mask" in the act. The actual reticle is typically 10 times the actual size of the chip and has a single chip imprinted on it. The pattern which would appear on a reticle are those shown in Attachment 3.

Next, a working mask is made from the reticle. The act includes these objects under the definition of masks. One mask must be made for each pattern. The masks are glass or metal plates and multiple copies of the same chip are contained on the mask. The pattern is now the actual size which must appear on the wafer. The mask are placed in a printer which is basically a camera. The camera prints (i.e., projects light through) the mask and the pattern is then imprinted on the entire wafer. Multiple chips are imprinted at the same time. The set of all patterns successively imprinted is referred to as a "mask work" in the act.

The technology for imprinting these patterns has advanced to the point where the generation of the working mask can be eliminated. This can be done by the use of a "stepper" to imprint the pattern on the wafer. This is typically a more expensive manufacturing step but it is also more accurate. When a stepper is used, the tape is again used to make a reticle for each pattern. As before, the reticle has a pattern for a single chip on it. The reticle is placed in a printer known as a stepper. The pattern is imprinted on the wafer one chip at a time and then it is "stepped" to the adjacent area of the wafer where another chip is imprinted on the wafer.

The newest technology eliminates the reticle. This is a even more expensive manufacturing process but it is even more accurate. The tape is entered into a direct write machine. The direct write machine writes the pattern directly onto the wafer similar to the way a picture is written on a television screen. The machine then steps to the adjacent area of wafer and writes the pattern for another chip. This is covered in Section 4 of the ACT, specifically Subsection (6)(D).

THE COPIES WHICH WE NEED TO PROTECT

Today, many techniques exist to minimize errors in creating the pattern of the circuits. There are computer aided design programs which assist in comparing the circuit schematic to the layout before it is imprinted on the wafer. Nevertheless, it is very rare that a chip having upward of 250,000 transistors on it will work the first time. Inevitably, there will be errors in the circuit design, the layout, or the interreaction between the layout, the circuit design and the wafer processing. It is only after numerous iterations at a cost of millions of dollars that the chip is fully functional and can be sold publicly to customers.

The pirates want to obtain a copy of the pattern only after all of these iterations have been completed. In this manner the pirate can minimize his overall cost. The goal of the pirate is to eventually obtain a copy of the pattern in the form of a tape. The pirate can convert the tape to the various different forms of the pattern needed to manufacture the chip.

The pirate's first problem is that these patterns are considered highly valuable property of the company which originally designed the pattern. Consequently, the paper layout, the tape, the reticles and the working masks are carefully protected by the designing company. They are treated as trade secrets within the company and strict security is used to insure that only employees having a good business need for the patterns may obtain access to them. Subcontractors are often used to convert the tape to the reticles and the masks. Again, there is a strict secrecy agreement between the designing company and the subcontractor. Consequently, the pirate cannot easily get access to the pattern in these formats. Other than stealing the pattern, the only practical way that the pirate can get access to the patterns is from the publicly available semiconductor chip itself.

Since the patterns are imprinted on the wafer (the mask work) to form a semiconductor chip, the job of the pirate is to reverse this process. He starts with a publicly available semiconductor chip which has been assembled in a package. He must remove the lid or plastic covering of the package so that he may get access to the actual chip. Now, he makes a careful photograph of the top pattern of the chip. He carefully blows up this photograph of the chip and draws it on paper or on a computer, just like the original layout design engineer did. The difference is that the pirate has a simple mechanical measuring job as opposed to the original trial and error exercise to minimize the layout which the original designing company had to perform.

Once the top layer has been carefully measured and the information preserved on paper or a tape, this top layer is carefully etched away until the next pattern is exposed. Now this pattern is carefully measured and drawn in the same manner. Each pattern is carefully measured and etched off to exposed the next pattern until every pattern of the chip has been copied. The pirate will now have a tape containing the key patterns which can be converted into the various formats which are necessary to manufacture the chip.

A FAIR REVERSE ENGINERING

Under current copyright law, a copyrighted biography does not prevent a second writer from writing a biography on the same person. The second writer must use different words in the expression of the second biography. The second biography cannot look like the first but the same information could be conveyed. This is analogous to reverse engineering.

The Semiconductor Chip Protection Act of 1983 is intended to protect the photographic copying of the chip but otherwise allows reverse engineering. There is a marked difference between fair reverse engineering and the chip piracy described above. The act of fair reverse engineering could involve the reproduction of the pattern from the semiconductor chip but would not allow this pattern to be substantially copied for use in the production of a semiconductor chip. Instead, the pattern would be used solely for the purpose of teaching, analysis of the chip or evaluation of the circuit concepts or techniques embodied in the chip. A reverse engineering firm should be allowed to analyze the chip, draw a circuit schematic of the chip, and then layout a different pattern. This pattern could be used to fabricate a version of

30-425 0-84--3

the semiconductor chip which is functionally equivalent to the original chip but has different visual patterns on it. The reverse engineering firm could then improve the performance of the chip, reduce the size of the chip and reduce the overall manufacturing cost of the chip. However, this type of cost reduction and performance improvement is also engaged in by the original designing company. Here we have a true cost reduction or advancement in the state of the art.

ECONOMICS OF PIRATING

So far we have been discussing the design and manufacture of a single semiconductor chip. In reality, a complete family of chips are needed so that the customer can develop a complete system. This means a total development would include a main chip, additional chips which are used with the main chip, computers to help the customer develop software to be used with the chip and certain software products to work with the family of chips. The manufacturer must also develop a market for this family of chips. The cost associated with developing this market into a substantial base of customers will often cost nearly as much as the Research and Development Cost. Typical cost of a complete family of chips would be as follows:

Research and Development cost associated with the main chip approximately

Research and Development of additional chips, development tools and soft

ware

Subtotal

Market development cost

Total cost..

Millions

$4

40

44

36

80

Even after a complete family of chips are developed, the Research and Development Cost of upgrading the chips and correcting errors in the chips continue. These costs often run in the area of $10M dollars a year for a complete family.

As discussed earlier, it would be perfectly legal for a company to reverse engineer any part of the chips. Although it may cost $80M dollars to develop the complete family of chips and the main chip cost $4M dollars, it will only cost about $1M dollars to reverse engineer the main chip itself. This is something that the industry must accept.

The typical pirate will simply pick the high volume products in the family of chips and make photographic copies of these. He does not have to copy the entire family, only the main chip. A simple photographic copy of the main chip would only cost about $100,000. The pirate has minimal research and development cost and virtually no market development cost. He enters the market after the original company has fully developed the market. The pirate does not have to recover the research and development cost of the entire family of chips and certainly does not have to recover any market development cost. He is simply interested in making a profit above his manufacturing cost of the chips that he copies. The pirate simply uses price as his weapon.

The abilities of these pirates to copy particular chips within the family of chips dramatically reduces the incentive of the original company to continue to invest in research and development activities. In fact, every chip must be evaluated in light of the risk to chip piracy. As a consequence, many innovated ideas for design of new chips must be cast aside because the return on the investment cannot be justified in light of the threat of chip piracy.

SUMMARY

Under the current copyright law it is not clear whether or not the printing of the pattern on the wafer is a copy. It is even less clear whether or not copying the mask work from the physical/useful chip is a copy under the current law. The bill makes it clear that the valuable masks and mask works are protected even though they may not be copies under the principles of current copyright law. It has taken the SIA 4 years to agree on this extention of copyright law to protect chips. It is our belief that this is the only practical method of protecting our valuable patterns.

The Technology to be protected by the Semiconductor Chip Protection Act of 1983 is the expression of the chip in a particular visual pattern. The masks and mask works would be protected from photographic copying. However, the same electronic functions could be implemented in a chip so long as different patterns were used.