The Day the CPK Report Came Out, the Room Fell Silent for Three Seconds
Do you remember? Last year in Q3, for a particular batch, equipment ran as usual, parameters remained unchanged, yet CPK suddenly plummeted from 1.35 to 1.08. In the conference room, as everyone stared at the figures, the air became thick with tension. The production supervisor's face turned ashen as he demanded, "What happened? Why the sudden drop in yield?" A-Zhe from the equipment team scratched his head, confirming the equipment was operating normally. Lili from Quality Control presented data showing abnormal edge values on several wafers. It took us nearly three days to pinpoint that an unnecessary "waiting" step in the process, which exposed the wafers to air for too long, caused edge contamination. Honestly, the production capacity lost from that incident could have covered my overtime pay for half a year.
Where Does the Problem Lie?
Such frustrating incidents are commonplace in semiconductor manufacturing, aren't they? Often, you might assume everything is running smoothly, but in reality, "waste" is occurring unknowingly. What we commonly refer to as "waste" (Muda) is, simply put, any "activity that does not add value." Consider this: Will customers pay for the time you spend waiting for materials? Will they cover your rework costs? Absolutely not! These are all your own losses. Frankly, given the complexity of semiconductor processes, any misstep in a single step can become a financial sinkhole.
How to Actually Implement This?
The core principle of what we often call Lean Production is to identify and eliminate these wastes. The most classic framework is the "7+1 Wastes," which I'll translate into plain engineering terms for you:
- Overproduction: You anticipate issues at the next station, so you produce extra as a buffer. The result? The surplus goes unused and requires storage space.
- Waiting: This refers to the wafer edge contamination case I just mentioned. Machine A waits for Machine B, personnel wait for equipment, or you wait for approvals. DPMO soaring from 300 to 6210 are all consequences of such waiting.
- Transportation: The movement of products or tools within the facility, by itself, does not add value. You might think it's just a few extra steps? Wrong, more movements increase the risk of damage and contamination.
- Over-processing: Investing excessive time and resources to meet a specification that isn't critically important. For instance, if your yield is already 99.9%, does it make sense to expend ten times the effort to achieve 99.99%?
- Inventory: Excess raw materials, work-in-progress (WIP), or finished goods. It consumes space, increases management costs, and can lead to obsolescence or damage.
- Motion: Unnecessary movements by personnel, such as searching for tools, bending, or stretching. These actions consume time and pose potential safety risks.
- Defects: Non-conforming products, rework, and scrap. This is the most direct and painful, directly impacting yield and cost. For example, a single rework can increase your cycle time from 24 hours to 48 hours.
- Non-utilized Talent: The most frequently overlooked waste. Your new hires might have excellent ideas, but you prevent them from contributing due to perceived "lack of experience"; or you simply use senior engineers as glorified operators.
The key takeaway is this: looking back now, isn't it true that "yield abnormalities" and "cost overruns" are often inextricably linked to these forms of waste?
The Most Common Pitfalls
I'll tell you, the most common pitfall is "normalizing the abnormal." Many wastes, because they occur daily, eventually become perceived as "normal." For example, at a certain station, wafers routinely sit on a cart for two hours daily, and everyone thinks, "That's just how it is." Or, to compensate for unstable upstream yield, downstream processes deliberately set stricter parameters, adding an extra wash or bake cycle. This is a classic combination of "over-processing" and "waiting." I recall a new assistant engineer once asking why we turned a particular screw 10 times daily when 5 turns were clearly sufficient. Upon asking, he discovered it was a habit stemming from a five-year-old oversight that no one had ever bothered to correct. Simply put, no one had truly considered, "Why are we doing it this way?"
One Action You Can Take Today
Today, choose one step in your workflow that "frustrates" you, and ask yourself: "Does this truly add value?"