Scenario
During a yield meeting at a wafer fabrication plant, you observe the daily total number of particles on a report and notice that the C-chart occasionally goes out of control. Your supervisor asks if the process has drifted again, but you clearly recall that equipment utilization and wafer batch counts haven't changed, so how could this be? You ponder internally, "Is the cleanliness of the metrology equipment unstable? Or are there other factors not considered?"
Plain Language Explanation
In process monitoring, we often need to track the number of defects on products, such as particles on wafers, solder joint defects on PCBs, or scratches on panels. In such cases, many instinctively think of using a C-chart. However, a C-chart is not a panacea; the key to choosing between it and a U-chart lies in whether your 'unit of inspection' is consistently the same size each time.
Imagine you are a Quality Control engineer, and your daily task is to inspect products for flaws.
- C-chart: It's like you inspect a fixed size of 100 wafers every day and record the total number of particles on these 100 wafers.
* Underlying assumption: The 'opportunity' for inspection is exactly the same each time.
* Applicable scenario: For example, inspecting 50 PCB boards of the same size daily and recording the total number of solder joint defects on them.
- U-chart: On the other hand, the number of wafers you inspect daily might vary (e.g., 80 today, 120 tomorrow), or the wafer sizes might differ. In this situation, you cannot just look at the total number of particles, because the 'opportunity' for inspection has changed. You need to calculate the average number of particles per wafer (total count divided by the number of wafers) and then monitor this defects per unit.
* Underlying assumption: The probability of defect occurrence is uniform, but the 'opportunity' for inspection is not fixed.
* Applicable scenario: For example, inspecting wafers from different batches and varying quantities daily, recording the average number of particles per wafer.
Simply put, a C-chart looks at the 'total score,' while a U-chart looks at the 'average score.' When to look at the total score versus the average score depends on whether your 'denominator'—the sample size—is constant.
Practical Judgment
The most crucial criterion for choosing between a C-chart and a U-chart is whether your 'sample size' is constant.
| Trait/Chart | C-chart (Number of Defects) | U-chart (Defects Per Unit) |
|---|---|---|
| Applicable Scenario | Fixed sample size, monitoring total number of defects | Variable sample size, monitoring defects per unit |
| Sample Size | Must be fixed (e.g., inspecting 100 wafers each time, 50 products of fixed size) | Can vary (e.g., different number of wafers inspected each time, different product sizes, different inspection areas) |
| Monitoring Metric | Total number of defects per sample | Defects per unit per sample (Total number of defects / Sample size) |
| Control Limits | Calculated based on the total number of defects | Calculated based on defects per unit, and adjusted with varying sample sizes |
| Common Examples | Total number of solder joints on fixed-size PCBs, total number of particles on fixed-batch-size wafers | Average number of particles on wafers of varying batch sizes, scratch density on panels of different sizes (number of scratches per square centimeter) |
Practical Operation Recommendations:
- First clarify the definition of 'sample': What exactly constitutes your 'one sample'? Is it 100 wafers? Or one wafer? One batch?
- Determine if the sample size is fixed：If the number of products, area, volume, time, etc., that constitute the 'opportunity for defect occurrence' is completely fixed and unchanging during each monitoring, then use a C-chart.
- If the sample size varies：If these 'opportunities for defect occurrence' change over time or across batches, such as inspecting 100 wafers today and 80 tomorrow, or if product sizes or areas differ, then you must use a U-chart. Otherwise, the control limits of a C-chart will generate false out-of-control alarms due to variations in sample size, leading engineers to misinterpret the process state.
- Provide accurate data：When using a U-chart, in addition to the defect count, you must provide the actual size of each sample so that the system can correctly calculate defects per unit and dynamically adjust the control limits.
How InsightFab Does It
InsightFab features a built-in intelligent judgment module that recommends the most suitable control chart type based on your uploaded defect data and sample size information, automatically calculating control limits. This eliminates the hassle of manual judgment and complex calculations.
Golden Quote
"To control defects, don't just look at the quantity; it's crucial to also consider whether the underlying 'opportunity' is consistent to avoid misinterpreting the true state of the process."