cp and cpk
Cp and Cpk are the two most important process capability indices in Statistical Process Control (SPC), measuring how well a process's natural variation fits within specification limits. Together, they provide a complete picture of process precision and accuracy.
Cp — Process Potential
C_p = \frac{USL - LSL}{6\sigma}
- Compares the specification width (USL − LSL) to the process spread (6σ, which contains 99.73% of output).
- Ignores process centering — it asks "if the process were perfectly centered, would it fit within spec?"
- Measures the potential capability of the process.
Cpk — Process Performance
C_{pk} = \min\left(\frac{USL - \bar{X}}{3\sigma}, \frac{\bar{X} - LSL}{3\sigma}\right)
- Measures the distance from the process mean to the nearest specification limit, in units of 3σ.
- Accounts for centering — a process that is precise but off-center will have high Cp but low Cpk.
- Always Cpk ≤ Cp. They are equal only when the process is perfectly centered.
Interpreting Cp and Cpk Together
| Situation | Cp | Cpk | Meaning |
|---|---|---|---|
| Good process | 1.5 | 1.5 | Precise AND centered — excellent |
| Off-center | 1.5 | 0.8 | Precise but not centered — shift the mean |
| Wide spread | 0.8 | 0.7 | Too much variation — reduce σ |
| Both bad | 0.7 | 0.4 | Needs major improvement |
Key Insight: If Cp is high but Cpk is low, the fix is simple — re-center the process (adjust the target). If Cp itself is low, the process needs fundamental improvement (reduce variation).
Semiconductor Industry Standards
- Critical Process Steps (gate CD, thin film thickness, overlay): Cpk ≥ 1.67 typically required.
- Standard Process Steps: Cpk ≥ 1.33 is the minimum acceptable level.
- New Process Introduction: Often start at Cpk ~1.0 and improve toward 1.33+ during ramp.
- World-Class Processes: Cpk > 2.0 (six-sigma quality).
Calculating Cp/Cpk: Important Notes
- Sample Size: Need at least 30+ measurements for reliable estimates. Small samples can give misleading values.
- Normal Distribution: Cp/Cpk assume the data follows a normal distribution. Non-normal data requires transformation or alternative indices.
- Process Stability: Only calculate Cp/Cpk on in-control data (no special causes present). An out-of-control process's capability indices are meaningless.
Cp and Cpk are the universal language of process quality in semiconductor manufacturing — they enable quantitative comparison of process performance across tools, fabs, and companies.