- FDC
- Fault Detection and Classification, Often Using Equipment Process Data
- MES
- A System Recording Production Execution and Histories
- Setpoint
- The Setpoint the Equipment Is Intended to Reach
- Trace
- Actual Readings Collected Over Time
Six Common Data Types and the Questions They Answer
| Data | Examples | Purpose |
|---|---|---|
| Images and Defect Records | Optical images, SEM review micrographs, Klarf defect files | Locations, Morphology, Distributions, and Inspection Signals |
| Metrology | CD, Film Thickness, Overlay, and Stress | Dimensional and thin-film thickness deviations |
| Tool Telemetry | Pressure, Flow, Power, and Temperature | What Happened During Processing |
| Recipes and Configurations | Recipe Versions, Offsets, Components, and Maintenance Records | Planned Conditions and Actual States |
| Process History | Wafer ID, step, tool, chamber, and track-in/out timestamps | Which Objects Actually Underwent Which Processes |
| Electrical & Test | Process-Monitor Electrical Tests, Wafer Probe / Final Test, STDF Files, or Bin Maps | Process parameters, functional die test, and bin classifications |
KLARF is a common inspection-data exchange format, but versions, fields, and image references must be confirmed. SEM images are often grayscale; color may be added in post-processing, and a layout is not an inspection image.
- 01OutcomesObserving Deviations on Wafers
Images, defect coordinates, CD metrology, and test bins describe post-process outcomes.
- 02ConditionsQuerying Equipment Processing Conditions
Recipes, FDC sensor traces, maintenance logs, and materials provide process context.
- 03AlignmentAligning Across Identity, Time, and Version
Confirm identical wafer IDs, correct route steps, and actual execution conditions.
- 04DeliverClarifying What Can Be Compared and What Remains Missing
Address missing data and definition differences first, then select an appropriate analysis.
Recipe Setpoints and Actual Telemetry Are Distinct Evidence
A recipe sets gas flow, but actual chamber flow varies with MFC response. Sensors detect drift, but readings depend on calibration, port position, and sampling rate. A sensor alarm does not automatically convict a specific component.
Four Data Alignment Conditions That Must Be Recorded
- Identity:Wafer ID, Lot, actual tool and chamber; slot numbers require carrier and timestamp binding.
- Time:Timezone, clock source, drift, and recipe phase; track-in/out does not equal in-plasma time.
- Space:Coordinate origin, units, notch orientation, die/stepper layout, and inspection care areas.
- Version:Active recipe version, inspection recipe, analysis method, and dataset snapshot.
Advanced: Communication Protocols Do Not Guarantee High Sampling Rates
SECS-II, GEM, and HSMS govern messaging, tool state models, and transport. Trace sampling depends on tool controller capability; EDA/Interface-A provides dedicated data collection. Using SECS/GEM does not guarantee 10 Hz raw waveforms or microsecond time synchronization.
Clarify What Data Can Answer Before Selecting Algorithms
If only step averages are available, raw transient spikes cannot be recovered. If chamber history is absent, tool-level correlation cannot be attributed to a specific chamber. Reports must document these gaps rather than guessing.
WAFERPILOT / APPLICATION TO THIS TOPIC
Structuring Disparate Data into an Investigation Context
WaferPilot organizes checks and source references according to connected data and agreed methods, delivering comparable results, candidate explanations, missing information, and next-step recommendations. Cross-source connectors require field mapping and data checks for each system.
Deliverables for Engineers:An inventory of available data, cross-source mappings, comparable conditions, and missing items, ensuring subsequent analyses use validated entities and baselines.
View the Current Demo and Capability Scope →Related Reading & References
Below are verifiable industry references; specific processes, permissions, and acceptance criteria are governed by fab site specifications.