skip to main content

Should You Fully Automate Your Fab?

By Nanochip Staff

Fully automated fabs offer significantly better yields, output, cycle times, operating costs and flexibility than partially automated ones. Yet some manufacturers have opted to deploy automation solutions on a piecemeal basis.

Can semi-automated fabs achieve meaningfully better results through more extensive automation of their operations? Is doing so worth the cost? And what about the workforce dynamics?

To explore these questions, Nanochip Fab Solutions’ Gary Dagastine sat down with David Hanny and Shekar Krishnaswamy, semiconductor automation director and business development manager, respectively, for Applied Materials.

David HannyShekar Krishnaswamy

NFS: Give us a view of the current automation landscape. Would it be accurate to say that 300mm manufacturers tend to fully automate their fabs, while 200mm manufacturers don’t?

Hanny: That’s somewhat true but it’s misleading to think that wafer size or technology node is the driver. The distinction lies mainly with the experience and technological maturity of the manufacturer. Companies with longer histories of IC manufacturing and with a strong staff of experienced, technically skilled people are more prepared for the use of complete automation solutions to integrate factory-, process- and equipment productivity within a fab and across production sites. These may include systems for AMHS [automated material handling], planning and simulation, manufacturing execution systems—including production dispatching and scheduling— APC [advanced process control] systems, and much more.

The semiconductor business landscape is characterized by a strong push to generate more results from existing assets, thus full automation is a growing competitive necessity. Newer companies, however, may perceive full automation less as an enabling technology and more as an added level of complexity.

Krishnaswamy: Geography plays a role, too. I would add that in the places where most new fabs are being built, such as in Asia, there is more opportunity to automate the decision-making components of fab processing.

NFS: Let’s say I have a fab running well-understood technologies, I’ve already automated a few critical functions like dispatching, and I’m meeting production and cost goals. Why go through the expense and production disruptions involved with fully automating my fab?

Hanny: Even the best-run partially automated fabs can gain substantial benefits from more extensive automation. That’s especially important in the current environment, for two reasons.

First, given today’s highly competitive markets and the short product life-cycles of many consumer electronics, on-time delivery performance at a given quality level is becoming harder to achieve without automated systems. Second, fully automating a fab enables production assets to be used more efficiently, making the fab more competitive and potentially more profitable.

Krishnaswamy: Full automation brings sophisticated decision-making capabilities to production systems. This not only reduces inefficiencies, it increases a fab’s flexibility so that a manufacturer can better capitalize on the opportunities offered by today’s fast-moving markets.

NFS: What are the main points of di”fferentiation between fully and partially automated fabs?

Hanny: Cycle time, higher product quality, the opportunity for yield increases, better on-time delivery performance, better utilization of production assets and, perhaps most importantly since people are often a bottleneck, the use of people in very different ways.

Take quality as an example. Like all fabs, semi-automated ones employ SPC [statistical process control] and use SPC data to maintain a high level of specification conformance. Early steps towards full automation are to implement process control techniques such as run-to-run control. Fully automated factories additionally take advantage of FDC (fault detection and classification) analysis capabilities as wafers move through the line. When a tool’s gas and/or heat levels start to drift, for instance, the system can detect and correct the problem on its own before any wafers are lost. The result is that yield increases, production latency decreases, queue-time violations are reduced, and customers achieve greater product spec conformance.

NFS: How does automation impact on-time delivery performance?

Krishnaswamy: A semiconductor manufacturer’s first objective is to fill a customer order by producing the requested amount of product at a certain quality level by a given date. This sounds simple but in reality it isn’t. Even when everything is going well you need to ensure that you don’t produce parts too late or too early. If you miss a tight delivery date then customer dissatisfaction may increase, and of course you couldn’t very well charge a premium. On the other hand, if you produce parts too early you may be delaying other lots for other customers, or using your resources suboptimally. One result might be that you then must hold the parts in inventory, and incur holding costs.

Here’s a hypothetical, but common, example to illustrate how on-time delivery can quickly become quite complex and difficult to achieve. A customer gives a semiconductor manufacturer a production forecast of one million units. The manufacturer creates a production plan by taking into account all relevant variables and then modeling and verifying optimum production scenarios. However, the customer may change the forecast to an actual production order for two million units. The manufacturer must quickly determine whether they can meet the new requirements. They don’t want to lose the customer but a significant change like that can ripple through the fab and disrupt other scheduled operations.

Meeting these changing or sudden customer needs requires fabs to be nimble, and fully automated fabs are simply more nimble and able to adjust to changing conditions than fabs that are only partially automated. In fact, many of them use advanced planning tools featuring 3D virtual reality animation, or dynamic “virtual factory” capacity planning tools to provide real-time decision-making that maximizes productivity and minimizes costs.

The extensive analytical data collected by statistical process control (SPC), advanced process control (APC) and fault detection and classification (FDC) systems in fully automated fabs is key to the ability to improve overall factory responsiveness and performance.

NFS: What is the impact of full automation on a fab’s workforce?

Hanny: One of the key deliverables of an automation system is consistency: it can drive repetition of the same task again and again, with the same output and within the same amount of time. Humans, by contrast, perform individually. No two people will do tasks exactly the same way, nor will one individual consistently do a task exactly the same way every time. So automation on the factory floor is one way to ensure consistency in timing, task performance and output. It can alleviate bottlenecks and assure process control and predictable outcomes in semiconductor manufacturing.

For example, queue-time violations from WIP [work-in-process] bottlenecks are becoming an increasingly important issue as line-widths on wafers become smaller. If wafers sit and wait too long between process steps they might have to be reprocessed or scrapped. That’s because the next step must take place within a specific time window to avoid wafer oxidization. Or suppose an operator tries to push more lots through a tool. If that is done when a downstream tool happens to be busy or disabled, the wafers may not be able to be further processed in time.

Full automation allows manufacturers to achieve better and more flexible factory performance by performing tasks such as production planning,scheduling, dispatching, and reporting in an integrated fashion.

Also, factory-floor knowledge tends to reside in information silos, such that when a given problem is solved the solution isn’t necessarily shared with others. In a fully automated fab there are centers of excellence, often located in the control room, and problems are solved systematically rather than on an ad hoc basis. Automation makes for fewer people on the factory floor and brings more resources into the control room, where people do higher-level work such as process monitoring and data analytics.

NFS: Earlier you mentioned that full automation of 200mm fabs leads to better utilization of production assets. How so?

Hanny: With full automation, whether at 300mm or 200mm, you can better predict tool availability. You know when WIP will arrive, for example, and can start moving materials to the tool ahead of time, so they are there just when needed, with no delay.

NFS: What are some of the biggest hurdles a semiconductor manufacturer may face in fully automating a fab?

Krishnaswamy: Unanticipated events or exceptions to the norm hugely impact automated systems. You’ve got to design your automation systems such that they will be able to handle exceptions. For example, Applied’s RTD system and Activity Manager can ensure that additional lots do not enter a time-critical process loop if there are too many lots already in the loop and there is a high probability of lots exceeding the allocated cycle time.

Some of the benefits of full automation in semiconductor production.

NFS: You’ve made the case for moving from partially to fully automated facilities. But are there circumstances when moving to full automation isn’t a good idea?

Hanny: One that comes to mind is when a manufacturer isn’t competing in the broad market but instead serves a specific, unique niche where productivity isn’t necessarily the main goal. For example, the low-volume manufacturing of certain products for defense applications is a case where tried-and-true manual or semi-manual manufacturing processes make sense or, in fact, may be required by the customer.

Krishnaswamy: Also some industries, such as medical device manufacturing, must operate according to government regulations that emphasize manual disposition of production processes, which makes it difficult to quickly transition to automated manufacturing.

For additional information, see or contact