That Day, the Surface Roughness Gauge's CPK Report Infuriated the Plant Manager
"Damn it, why is the roughness of this batch of goods problematic again? Didn't we say MSA passed?" That day, the Plant Manager looked at the report, and his anger boiled over. I silently looked at that report: CPK 1.08, DPMO 6210. The numbers looked decent, but the problem was, we were using the "wrong" gauge to measure the "right" product. Surface roughness, a critical indicator in semiconductor processes, often causes widespread chaos. Have you ever encountered similar frustrating situations? The gauge says it passed, but as soon as the product goes out, customers come back to you with complaints?
Where's the Problem? Why is CPK Visible Yet Unattainable?
To put it plainly, many times when we perform MSA, we only focus on the gauge's repeatability and reproducibility, i.e., EV (Equipment Variation) and AV (Appraiser Variation). However, surface roughness measures "microscopic" dimensions, making it extremely sensitive to the environment, measurement points, and even the sample's placement angle. Do you think a calibrated gauge is foolproof? Frankly, many times we overlook two critical details: "gauge resolution" and "standardization of the measurement method." Especially when measuring very small Ra or Rz values, if the gauge's resolution is similar to your process variation, then no matter how high the CPK, it's merely self-deception.
So the key is, you must know how precisely your gauge can "see," and whether your measurement process ensures consistency in every measurement.
How to Do It in Practice? Validate with a "Microscopic World" Mindset
To tackle surface roughness MSA, in addition to traditional EV/AV analysis, I recommend two additional steps:
- Resolution Check: Take several standard samples with extremely small known roughness differences (e.g., an Ra value difference of 0.005um) and have the gauge measure them. If the gauge cannot differentiate between them, even if MSA passed, it means it cannot effectively monitor your process. In this case, your gauge's resolution might be the culprit for process instability.
- Measurement Point Sensitivity Analysis: For the same sample, perform multiple measurements at different locations to see the extent of the variation. If the variation is significant, your problem might not be the gauge, but rather that the sample itself has non-uniform roughness, or your measurement method does not clearly specify the measurement points.
In other words, you must ensure that your gauge and measurement process genuinely reflect the product's "true variation," rather than gauge or human-induced "noise."
The Most Common Pitfall: Ambiguous Measurement Point Specifications, Everyone Has Their Own Way
I remember one time, while investigating a customer complaint, we found that for the same wafer, Engineer A measured an Ra of 0.5nm, while Engineer B measured 0.7nm. Both claimed they followed the SOP, and MSA had passed for both. What was the result? After some investigation, we discovered that our SOP only stated "measure the central area of the wafer" but did not specify "which point" within the central area. Engineer A preferred measuring a little to the left, and Engineer B preferred a little to the right. Although both were within the "central area," the wafer itself, due to process effects, had subtle roughness differences at different points. This is a typical pitfall of "ambiguous measurement point specifications"; everyone has their own approach, rendering MSA ineffective even if passed.
One Thing You Can Do Today
Re-examine your surface roughness SOP and clearly specify the measurement points.