This Old Cleanroom
10/01/2001
You're faced with the possibility of a retrofit project: Just how severe does the job have to be?
by Hank Hogan
 |
Jonathan Holder is manager of major construction for the microelectronics division of the IBM Corp. (Armonk, NY) and is supervising IBM's retrofit of a large semiconductor cleanroom in East Fishkill, NY. When the plant closed in 1993, it ran a 125-millimeter wafer semiconductor manufacturing process. After renovation is complete in 2002, the 140,000-square-foot cleanroom will manufacture 300 mm, state-of-the-art product—this is hardly a trivial task.
 |
For instance, in East Fishkill, IBM had to supply services and utilities to 300 mm tools through a slab that had not been built with such an eventuality in mind. "We had to go out and core bore more than 5,000 18-inch-diameter holes in the waffle slab to create these openings to pass all of the services through. We then had to come up with a way to seal that opening off with a cover that was a smoke barrier and was liquid tight to keep spills contained," says Holder.
The IBM case is an example of a complete cleanroom retrofit, with the space gutted down to the walls and floors before being rebuilt. Not all retrofits are as severe.
 |
"The other, more common type of retrofit, is where you have a cleanroom that's not functioning properly and major systems need to be renovated or revitalized, or it's a tool change out," explains Thomas Hansz, president of the cleanroom design and construction firm Facility Planning and Resources Inc. (FPRI; Pittsburgh).
Behind these two different types of retrofits is a single motivation: Cleanrooms must be current. As cleanrooms age, they need to undergo retrofitting and upgrades, be converted to other uses, or be shut down.
 |
Thus, cleanroom owners are faced with many questions: Does it make sense to retrofit? Is it better to build new? What are the problems and pitfalls that might arise with a retrofit?
The answer to these questions depends on the overall economic climate, the technological conditions for a given industry, and the site in question.
Drivers...
As for economics, the fragile nature of the current business climate is pushing end-users to at least consider retrofits. Hansz of FPRI reports that about 50 percent of the company's current projects are some type of retrofit. At the other end of the country, Tim Marrs, a sales engineer with cleanroom manufacturer Clean Rooms West Inc. (Tustin, CA), reports a 25 percent retrofit project rate. Likewise, Ralf Graeber, vice president of United States subsidiary of German cleanroom contractor M+W Zander, notes keen, and recent, customer interest in retrofits.
"A lot of our clients are thinking about upgrading their existing facilities rather than building new facilities just because of the situation in the economy," says Graeber. "So, many clients look into just using office space or warehouse space."
In some industries, technology is also a major driver of retrofits and upgrades. The best example is probably the semiconductor industry. The relentless doubling of transistor count every 18 months—Moore's Law—requires constant innovations in the manufacturing process. Most of the time this involves a process improvement such as a tool change or the substitution of copper for aluminum in metallization traces. Occasionally a change in basic wafer size comes along, such as the current move from 200 mm to 300 mm processing.
Indeed, such changes are anticipated over the useful life of a semiconductor cleanroom. According to Hansz, a feature of one of Intel's standard cleanroom designs allows large cleanrooms to be divided into four sections. Each can be sealed off from the others, enabling retrofitting and renovation to take place on a quarter of the cleanroom at a time.
Disk drive and pharmaceutical manufacturers also face their own version of technological obsolescence. Disk drives double in areal storage capacity annually, and some of that increase comes from innovations in drive manufacturing. Pharmaceutical manufacturers constantly research, develop, market and produce new drugs and compounds.
... and brakes
But the response of these industries to such demands is not the same as the response of the semiconductor industry. Part of this is because the situations are quite different. For disk drives, the areal storage increase is due to improved read/write heads and other factors. These don't necessarily benefit from major cleanroom renovations and retrofits. In pharmaceuticals, governmental regulations act as a deterrent. Once a regulatory body has certified a line, there's great reluctance to change.
"In the pharmaceutical sector where large-scale manufacturing is carried out, flexibility is required in the process and not so much in the cleanroom, as they tend to build again rather than change," comments Conor Murray, technical director for European cleanroom contractor the Ardmac Group.
However, that bioscience preference for new cleanrooms may diminish. For one thing, there are an increasing number of Biocontainment Level-3 cleanrooms being built. These cleanrooms are not designed to work with the most lethal and dangerous biological organisms. That's reserved for Biocontainment Level-4 cleanrooms. However, BL-3 cleanrooms do work with dangerous organisms and do adhere to a fairly high biocontainment level.
Many of these BL-3 cleanrooms are research and development facilities, meaning they could undergo more frequent and far-reaching changes than full production plants. Some of this may involve conversion of office space to cleanrooms and vice versa.
Another factor favoring faster change in bioscience-oriented cleanrooms is the increased pace of pharmaceutical research. There are some 300 new drugs under development by a number of emerging companies. Those drugs are about to move into the clinical phase of trials, which means that production will move from research and development into full manufacturing.
"As new products are introduced, facilities can expect some process equipment changes and so forth, similar in some ways to electronics," predicts Thomas Connolly, manager of advanced technology development for the cleanroom design and contracting firm IDC (Portland, OR).
Bitter medicine
Today, the majority of cleanroom retrofitting and renovation takes place in the semiconductor industry. Even there, with steps taken by some companies to future-proof facilities and plan for upgrades, a retrofit often doesn't make sense.
"That's true, but I also think that's probably a hard pill for people to swallow. What happens is that everyone is optimistic," says Mark Hemingway, vice president in charge of mechanical engineering at the Sterling Engineering Company Inc. (Sturbridge, MA), a full-service mechanical, electrical and plumbing firm. Most of the problems that Hemingway is referring to involve piping and process utilities, but difficulties can crop up in other areas.
Take the IBM case, which is an example where the company decided a retrofit did indeed make sense. According to IBM's Holder, the switch to a 300 mm line meant that the ceiling height within the cleanroom had to move from 10 to 14 feet. The larger size of the equipment, combined with the need for automated overhead transport of wafers, lead to this increase. That, in turn, meant that the company couldn't simply slap up new walls and wheel equipment in. The building had to undergo what amounted to an extensive construction job.
"So over the entire building, over the waffle slab, we had a four inch concrete deck that had to be jack hammered off, and then we had to raise the steel in the building on that level by four feet and rebolt it to the existing columns. It was quite an undertaking," observes Holder.
Part of the process also involved doubling the load capacity of the raised floor, to 2,000 pounds per square inch load rating. The extra weight of the equipment motivated this change.
The height situation was helped somewhat by the use of minienvironments. When the IBM building was constructed in 1985, it came on line as an ISO Class 3 (Class 1) facility, with 90-foot-per-minute velocity out of the air handlers. Thanks to minienvironments, the new cleanroom requirement is a more relaxed ISO Class 6 (Class 1,000), with 22-foot-per-minute air velocity. That enabled IBM to shrink its air handlers and regain some of the lost headroom in the attic over the clean area.
This balancing act allowed IBM to squeeze the entire, taller clean area into the existing building—except for the main exhaust headers. These run the length of the building, and the company couldn't fit those between the top of the air handlers and the roof. So, IBM had to bump up the roof over two bays. This raised area was about 50 feet wide and ran the entire 620-foot length of the building.
Being quiet and saving time
Holder believes that the problem with increased ceiling height is one reason why there will be so few retrofits to 300 mm. Others, such as Connolly of IDC and his co-workers, see it as a tool utilization problem. IDC's contention is that an efficient and economical use of 300 mm tools requires cleanrooms in excess of 100,000 square feet. Very few older semiconductor cleanrooms are this size, and hence very few are candidates for such an upgrade.
In IBM's case, it started with just such a large cleanroom, and the company will end up with a manufacturing floor comprising 106,000 square feet. Some 34,000 additional square feet will be a 300 mm research and development fab. For IBM, another bonus was that the building was vacant, meaning no existing production would be disturbed.
But in the final analysis, neither of these was the deciding factor.
IBM chose to retrofit because the building was quiet. Built at a time when tools were less sophisticated than today, the building had a specially constructed slab capable of active isolation, air handlers supported on bedrock, a separate foundation for the third floor versus the process level, and an isolation joint that surrounds the process floor. During initial construction, the slab took nearly two years to complete. That past investment pays dividends today and will do so tomorrow, according to Holder.
The other advantage from the retrofit for IBM is time. Holder estimates that the company shaved six months off the total schedule and will bring the facility online about half a year sooner due to retrofitting. However, this savings didn't come in the construction phase with its extensive renovation and rebuilding, which was actually slower than it normally would be because this work had to be done within the confines of the existing structure. The time savings arose during the permitting phase.
"We got through the permit stage fairly fast on this building because the site was already designed to handle this building," comments Holder. "The site infrastructure was here. The road system was in place. Much of the permitting work had already been laid down for that building, so it was just a matter of going in and updating it."
Perils and pitfalls
The IBM situation illustrates the work involved in a complete gutting of a plant and subsequent retrofitting. It also shows what some of the challenges may be, and how special circumstances may play a role.
 |
In other cases, the problems associated with retrofitting an older facility are so severe and the constraints so great, that the only solution is a complete changeover. Silicon is the most widely used semiconductor, but it isn't the only one. There have been very successful retrofits in which a silicon facility was converted to one running gallium arsenide or other compound semiconductor. In such cases, the cleanrooms are too large. Thus, some of their square footage is then converted into office space.
 |
However, in many cases, a retrofit isn't as extreme as the one undertaken by IBM. Instead, the retrofit may involve a tool change out or other process improvements. That can stress systems, such as an incident described by Hemingway of Sterling Engineering. A facility upgraded its processing tracks, which dispensed solvent. The old track requirement for exhaust was 80 cubic feet per minute (cfm) per machine. The new requirement was 320 cfm per machine. That increase was much more than the existing exhaust could handle, and so the exhaust system had to be upgraded with the upgrade to the track equipment.
Such unforeseen and unintended retrofit problems can also be external to the cleanroom. For instance, in retrofitting a cleanroom, there may be a requirement to bring it up to code. Over years of use, particularly in research facilities, there may be a tendency for such things as exits to become blocked. With a retrofit or renovation, such code violations will have to be fixed, which can greatly add to the cost and complexity of a retrofit. There also may be requirements from insurance carriers that add a further burden.
Finally, the conversion of space from one use, say as a warehouse, to another, say as a cleanroom, can have other regulation-induced and unintended consequences.
"When you change a warehouse area to a production area, your requirement for parking spaces goes up," says Marrs of Clean Rooms West. Before you were allowed one parking space per 400 square feet if it was a warehouse area, and now it's a production facility and you need two parking spaces per 400 square feet."
Cleanliness is not next to construction
Given all these factors, sometimes a retrofit isn't needed or isn't the most economical way to solve a problem. Andrew Solberg, an engineer in the mechanical department at IDC, uses computational fluid dynamics to model the air flow in a cleanroom. In one particular case, he recalls that a company substantially increased the size of a bin that held bacteria for a pharmaceutical application. The fear was that a change in air handling would be required to maintain the contamination control and cleanliness of the area. After evaluating the air flow, Solberg recommended the use of a single, properly placed diffuser. That alone solved the problem.
In another case, a proposed retrofit was abandoned in favor of a new building. Hansz of FPRI reports that a Canadian semiconductor company hired his firm to add on to an existing building. After studying the situation and running some productivity models, Hansz says he was able to show that the cost per item produced in a new facility would be much less than the cost in a retrofitted plant. So the client decided to forgo the retrofit and expansion in favor of building a new plant at a new site.
If the decision is made to do a retrofit, there are some immediate conflicts that arise.
"Retrofit is a tricky business in cleanrooms because cleanrooms by their nature are clean. They don't lend themselves well to the dirty process of construction," cautions Hemingway of Sterling Engineering.
There have been, of course, some very successful and innovative examples of construction in a functioning cleanroom. In one case, IDC was called upon to strengthen a floor because new, heavier tools required it for seismic reasons. However, the client wanted continuous operations during the retrofit. So IDC developed a scheme to inject concrete from underneath, using existing isolation joints. In this case, retrofitting and construction took place literally under the feet of cleanroom workers.
The need for such ingenuity, however, can be lessened greatly if cleanrooms are designed from the beginning to handle retrofits and upgrades. Such an approach may cost more initially but may actually be less expensive in the long run.
As Hansz of FPRI says, "The best time to reduce renovation costs or upfit costs are during the planning of the facility originally, because it's the changing out of the facility that tends to be the most expensive part of the cleanroom over time."