That day, when the CPK report came out, the room was silent for three seconds before I thought of a countermeasure.
I remember one time, our machine switched to a new batch of raw material, and there shouldn't have been any major issues, theoretically. However, at the morning meeting the next day, Quality Control suddenly presented a report showing that the Cpk of a critical parameter had dropped to 1.08, and DPMO had skyrocketed to 6210! The entire conference room fell silent instantly, and everyone's faces turned grim. The boss frowned and asked, "Wasn't it well-controlled last time? How did we mess up again?" Frankly, I believe you and I are absolutely no strangers to this "good then bad" situation. To put it simply, we didn't thoroughly implement the Control stage and failed to establish poka-yoke mechanisms.
Where Did the Problem Occur?
In fact, many times we solve problems in the Improve stage, optimize processes, and then feel like the job is done. But did you know? What's most easily overlooked is failing to "lock in" these improvement results. This is like spending immense effort to fix a leaky roof, but forgetting to inspect it regularly, only for a heavy rain to bring it back to square one. The core of the Control stage is to ensure that those problems that cause you headaches won't resurface. Simply put, it means updating your "control plan" and establishing a "monitoring mechanism" so that problems are detected before they even begin to sprout.
How Is It Done in Practice?
A control plan sounds academic, but in essence, it's a table clearly outlining the "key parameters" for each process step, their "target values," "tolerance ranges," what "measurement tools" to use, "how often to measure," and "how to handle abnormal measurements."
- Update the Control Plan: Incorporate all the critical variables identified in the Measure and Improve stages, along with your optimized target values and tolerances. For example, if a certain chemical concentration was originally monitored every eight hours, but you found it significantly impacts yield, you might need to change it to every four hours, or even install an online monitoring instrument.
- Establish Monitoring Mechanisms: This part "activates" the control plan. The most common tool is the Control Chart. You set the upper and lower limits, and an alarm sounds as soon as a measurement point goes outside these boundaries. For example, if we set the average for a certain thickness parameter at 100um, with upper and lower control limits at 95um and 105um. When you find seven consecutive points falling below 95um, even though they haven't yet gone out of tolerance, the trend is already very clear, at which point you can intervene early, rather than waiting until products are scrapped to deal with the aftermath.
- Poka-Yoke Design: To be honest, this is the ultimate form of Control. If errors can be made impossible from the source, that's the ideal. For instance, designing different sized components so they cannot be interchanged, or making sure the machine cannot start if its settings are outside the acceptable range.
The Most Common Pitfalls
The most common pitfall I've encountered is when control plans are merely written on paper, without actual "execution." Whenever asked, people would say, "Oh, yes, it's in the SOP." But upon inspection, no one was following the measurements, or the measurement data wasn't being analyzed. Another time, we improved the temperature control of a machine, boosting Cpk from 0.8 to 1.5, and everyone was thrilled. However, because the new temperature setpoint wasn't "locked in" the machine's program, after some time, a new OP accidentally reset it to the old parameters, and the Cpk gradually declined again, causing us to lose a large batch of wafers. Therefore, a control plan must not only be written down but also ensured to be diligently executed, monitored, and, when necessary, accompanied by poka-yoke designs.
One Thing You Can Do Today
Check if you have a control plan that clearly outlines "key parameters."