When the Cpk Report Came Out, the Entire Room Fell Silent for Three Seconds
I recall several years ago, we introduced a batch of advanced measurement instruments, expecting our production line quality to soar. What was the result? The following month's Cpk report plunged directly from the original 1.3 to 1.08, and the DPMO figures climbed to 6210. In the meeting room, the production line manager's face was paler than a lab's cleanroom suit, and the only sound was the air conditioning. Everyone exchanged glances; with upgraded instruments, why had things gotten worse? Finally, our gaze turned to measurement engineer Xiao Li, who looked more helpless than anyone that day.
What Was the Problem?
Simply put, the culprit was 'temperature,' an invisible killer. Our new batch of instruments indeed offered high precision, but they were also more sensitive to changes in ambient temperature. Consider this: thermal expansion and contraction of metallic materials is common physics knowledge. Your gauge itself is metal, and the wafer being measured is also metal. When the factory ambient temperature rises from 23°C in the morning to 25°C at noon, and then drops back to 23°C in the evening, the gauge's dimensions will change accordingly. Although imperceptible to the naked eye, for nanometer-level measurements, even these tiny variations are enough to derail your Cpk.
The key point, therefore, is that your gauge will yield different results at different temperatures. This doesn't mean your product actually has an issue, but rather that your 'measurement system' itself is unstable. This is why we need 'temperature compensation.'
How Is It Done in Practice?
The most direct method is 'calibration.' But not just an annual external calibration like before. You need to calibrate more frequently and strategically.
- Establish a Temperature Calibration Curve: In the lab, use constant temperature equipment to simulate different temperatures (e.g., 20°C, 23°C, 25°C), measure a standard part, and record the gauge's reading deviation at each temperature. You will obtain a 'temperature vs. deviation' curve.
- Online Temperature Sensing: Install high-precision temperature sensors near the gauge to monitor on-site temperature in real-time.
- Automated Compensation: Input the gauge reading, real-time temperature, and calibration curve into the system, allowing it to automatically calculate and compensate for temperature-induced errors. For example, when the temperature rises from 23°C during calibration to 25°C, the system will automatically adjust the reading according to the curve, ensuring it remains consistent with the 'true value' at 23°C.
For instance, if your gauge measures a 100µm standard part as 100.000µm at 23°C, but measures it as 100.002µm at 25°C. Then, when the on-site temperature is 25°C, your system should automatically subtract 0.002µm from the gauge reading, thereby yielding an accurate result.
The Most Common Pitfall
We fell into a major pitfall back then: we thought placing the gauge into a 'constant temperature and humidity' measurement station would solve everything. It turned out that even if the temperature inside the measurement station was stable, the gauge's 'body temperature' still needed time to reach equilibrium with the station's environment after being moved in from outside or influenced by the operator's hand temperature. Data measured during this 'adaptation period' is garbage.
Frankly, oftentimes we only focus on the precision of the equipment itself, overlooking variables like 'people' and 'environment'. Initially, we didn't wait for the gauge to fully adapt to the environment, eagerly wanting to see the data, which resulted in a lot of false signals, causing Cpk to drop and leading to finger-pointing. Only later did we discover that after every shift change or instrument relocation, we needed to wait at least 30 minutes for the equipment's temperature to stabilize before starting measurements; otherwise, it was all for naught.
One Thing You Can Do Today
Check your gauge's ambient temperature monitoring and compensation mechanism.