On October 18, 1954, the first transistor radio appeared on the market. Transistors were a big breakthrough in electronics — a new way to amplify signals. They replaced vacuum tubes, which were fragile, slow to warm up, and unreliable. During World War II, there was a big funding push to try to update vacuum tubes, since they were used in radio-controlled bombs but didn’t work very well. A team of scientists at Bell Laboratories invented the first transistor technology in 1947. But the announcement didn’t make much of an impact because transistors had limited use for everyday consumers — they were used mainly in military technology, telephone switching equipment, and hearing aids. (From “The Writer’s Almanac” of October 18, 2013)
In 1964, as a newly-minted electrical engineering graduate looking for an area in which to work and to make a difference, my wife and our new baby moved our pitifully few belongings all the way from the mountains of southern West Virginia to the modern city of Phoenix, Arizona in order for me to start a job in the brand new semiconductor field. In fact, the very word, “semiconductor,” was so new and unknown that in order to follow up on the blank stares about my work, I would add quickly, “you know, transistors,” since everyone had heard about transistor radios by then. The most modern of these newfangled gadgets had at least eight or ten transistors. They were marketed by slogans such as "The new ten transistor radio!" Gadzooks!
Today, the global semiconductor business is larger than $300 billion and is dominated by companies based in the USA, South Korea, Japan, Taiwan, Singapore, and Europe. Intel and Samsung are the two clear-cut leaders. Five of the top ten corporations selling completed products are U.S. based, with two each in Japan and South Korea, and one in Europe. The industry represents the highest levels of capital-intensive ultra-clean manufacturing along with brilliant (and fiercely-protected) intellectual property knowledge in solid-state physics, digital and analog design, software, and applications. It has made possible computers, massive data storage, and communications breakthroughs leading to the present Internet.
I retired in 2002 after 38 years in the industry, mostly with Motorola Semiconductors, which at one time was a leader along with Intel and Texas Instruments. Motorola's business in this field began to fall behind the leaders in the 1990′s and was spun off in 2004 by the corporation, which was experiencing other business difficulties as well, as an independent corporation. The result, Freescale Semiconductors, still is a large semiconductor firm at roughly $5 billion in sales, but is no longer in the international top ten.
It’s interesting to me to look back over nearly 50 years at the industry, since it’s not only a record of dynamic changes for a new and disruptive technology that enabled so many modern products, but also a road map for change for any new industry. In the early stages, there were very few suppliers of any supporting equipment or materials in semiconductors. As a result, if you were in the business, you had to do nearly everything involved. It was the semiconductor equivalent of the River Rouge plant for Ford Motor Company, where iron ingots went into one end on railroad cars, and brand new Fords were driven out the other end of the mammoth factory over a mile long.
While making semiconductors was not nearly so massive and basic, the overall manufacturing concept was similar. In terms of physics, it was even more fundamental, more basic. We did most of the work stages ourselves with equipment we had designed. Starting with purchased raw silicon material, we “pulled” the high purity single-crystal ingots ourselves out of molten raw silicon in our “materials area.” These were sliced and polished, and the “wafers,” as the silicon slices were known to us, were sent to an area known as “wafer fab,” short for wafer fabrication. Wafer fab manufacturing consisted of complex sequences or steps, repeated numerous times in the cleanest environment possible, often in specialized laminar air-flow hoods. Silicon oxide layers were grown in very high-temperature furnaces, followed by a process placing an intricate geometrical outline on the oxide with a “photomask,” a precise high quality glass plate we processed in our own "mask shop." The pattern was developed, like a photographic image, and the result allowed open areas to be etched through the oxide to the base material, while other areas, protected by the “masking layer,” resisted the etch. Various controlled impurities were introduced by “diffusing" them, literally providing the thermal energy to move them slowly at very high temperatures through the open areas, unprotected by oxide, and down into the solid silicon material. This basic process was repeated over and over until the fundamental positive and negative (P-N) silicon junctions and oxide layers formed the various transistors, which were interconnected by pieces of the silicon material itself as well as metallic deposition near the end of the process and a subsequent photo-step to hook up each transistor as needed, and then a final stabilization layer of protective oxide.
The completed “wafer” was tested electrically with a precision set of probes and electrical testing, at a stage called “die probe.” Every wafer had row upon row of these individual patterns, each a potentially valuable semiconductor device. Each one was a “die,” while collectively they were called “dice,” Once they were probed, and the bad ones inked out or otherwise designated, the wafer was sliced with a diamond stylus into individual “chips,” and the good chips were assembled into various shapes and sizes of “packages.” Finally the devices were tested electrically at “final test,” stamped or laser-etched with identifying marks, and available for shipment to customers.
In the beginning, as mentioned above, the manufacturers had to do mostly everything themselves, which included making their own equipment to do the various production steps. However that changed as the market grew and external companies were formed that focused on specific processes: silicon wafer companies, mainly in Germany, the U.S., and Japan, sprang up to produce the ultra-pure silicon and then the wafers themselves; others specialized in the various manufacturing equipment, including the photo-masks and every step of the photo-lithography process, the very high temperature furnaces, every step of depositions, and placement of precise impurities by ion-implantation. Similarly, the final assembly and testing equipment began to be supplied by outside vendors. As a result of these changes, semiconductor companies focused on wafer manufacturing in their own clean rooms, which were considered the mark of a serious player in the chip business. In fact, one industry CEO said “only real men had wafer fabs.” But this business model became outdated for all but the very largest companies, since the capital expense of building clean rooms became untenable for all but the very largest in the business. Intel and Samsung are the two main companies that still produce most all of their products in their own wafer clean rooms, since they have the profitability and financial strength to do so, at least for now, along with the huge volumes of business to utilize the expensive capacity. In addition, their business strategy is predicated on having the most advanced process technology to achieve the fastest performance at the lowest power consumption. A so-called “wafer foundry” industry has sprung up with huge clean rooms able to process wafers for other companies. Thus the clean room processing part of the making and testing a "die" has bifurcated from the design and applications part of the business, and as a result, virtually everyone in the business, other than Intel and Samsung, uses these very large foundries for their mainstream products. In fact, Samsung has a major position in both specialties, not only processing their own wafers in Samsung wafer fabs, but also offering their capacity to certain other companies, including Apple, which designs many of their own custom semiconductor products for use in their own products.
Historically it has taken approximately eighteen months for a new process to be developed and implemented into volume production. A new process generation involves a linear shrinkage of 30% in feature sizes. Saying it another way, the dimension of a feature in the new technology is 70% of the previous version for the same circuit. Thus the area of a new “layout” of the same chip would be 0.7 x 0.7, or about half, and there could be twice the number of “new” transistors in the original area. This phenomenon, first documented by Gordon Moore of Intel, and now known as “Moore’s Law,” predicted the number of transistors would double every one to one-and-a-half years. This assumed the development of new processes would take place every year or two. Doubling the number of possible transistors on a single semiconductor device every year or two has been the primary reason for the inexorable march forward of amazingly complex modern integrated circuits (ICs), along with commensurate improvements in faster speed and lower power consumption, and we have gone from tens or hundreds of transistors in an IC in the mid-1960′s to billions today. Doubling a number over and over for fifty years produces a very large number!
With world-class wafer manufacturing services now available in foundries, most of the current top ten firms have developed leading positions based primarily on their design and applications expertise. They focus on understanding end markets and providing integrated circuits that have not only the computing power but also the software and applications to perform a function. This means the most successful firms have moved beyond products that are simply “components,” but now go beyond this and act as complex sub-systems and even entire systems. This evolution toward more complete solutions and operating at a higher level continues to accelerate.
“The only constant is change itself” is attributed to the Greek writer, Hiraclitus. He had it right. All in all, I feel very fortunate to have gotten into such a breathtaking and earth-changing new industry at its early stages.
If you enjoy this blog, please mention it to others who may be interested.
James Kennedy George, Jr (Jim George)
Author, Reunion, a novel about relationships.
Available in Hard Cover, Soft Cover, and all eBook formats on the Internet from Amazon, Barnes & Noble, and all other Internet retailers, as well as on-order in print format from any book store. In stock at several book stores, including Book People in Austin, Texas, Tamarack on the West Virginia Turnpike, and Hearthside Books in Bluefield, West Virginia.
A number of book clubs in Central Texas have read Reunion in 2013, or have selected the novel for the first half of 2014. The author will be glad to attend your book club for discussions and to answer questions regarding the book as well as the publishing process. Contact him at <email@example.com> for additional information and scheduling.
The official author’s web-site is located on-line at : http://jkgeorgeauthor.com/