2880 casesVM replay rowsUNKNOWNInspect metric contract
5 contextsModule contextsUNKNOWNInspect metric contract
4 familiesModel familiesUNKNOWNInspect metric contract
0Live server runsUNKNOWNInspect metric contract
SIMULATION / SA-VMET-SMALLDATA-001
Small-data virtual metrology replay
VM trust gate
A browser CSV lab for sparse semiconductor metrology: compare module context, sample plan, data age, sensor health, physical metrology fraction, and model family before trusting a prediction.
EXPLORE
Start with the observable.
Use this page when the simulation is the primary artifact. The linked lesson explains the concept; the linked lab keeps the visual replay surface available.
- Filter module context, sample plan, physical metrology fraction, and model family to see when uncertainty replaces confidence.
- Compare prediction error, guardband, Cpk proxy, escape risk, false-hold risk, and sampling cost instead of optimizing only model error.
- Use decision gate and next safe check to decide whether the row supports monitoring, sample-more, hold-for-review, or bounded release language.
- Open the model card before reusing the table in a lecture, reading-room note, or process-control backlog.
BROWSER INTERACTIVE LAB
Filter the saved CSV. Inspect matching replay rows.
No server solver is running here. The browser filters a precomputed table and redraws the plot locally.
Loading replay rowsReading the saved CSV in this browser.
prediction error nm proxyguardband nm proxyescape risk proxy
Each metric is normalized independently for shape inspection. Vertical positions do not compare magnitude across metrics.
n/aprediction error nm proxy
n/aguardband nm proxy
n/aescape risk proxy
Change one or more controls to find a saved row. No unrelated CSV value is substituted.
| case id | module context | sample plan | data age | sensor health | physical metrology fraction | model family | measured sample count | uncertainty sigma proxy | prediction error nm proxy | guardband nm proxy | cpk proxy | escape risk proxy | false hold risk proxy | sampling cost index | decision gate | next safe check |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Loading saved replay rows... | ||||||||||||||||
RUN CARD
Conditions before conclusions.
The browser filters saved CSV rows and redraws local plots. It does not train a model, run a live solver, connect to equipment, ingest uploaded wafer data, or execute APC/dispatch logic.
Primary knobs
- module_context: CMP thickness, lithography CD/overlay, plasma etch CD bias, implant sheet resistance, ALD film thickness
- sample_plan: dense_baseline, periodic_skip, adaptive_uncertainty, excursion_confirm
- data_age: fresh, aging, drifted
- sensor_health: stable, noisy, shifted
- physical_metrology_fraction: 0.10, 0.25, 0.50, 1.00
- model_family: mean_shift_baseline, ridge_surrogate, gaussian_process_proxy, conformal_guardband_proxy
Saved outputs
- measured sample count
- training rows
- drift score
- uncertainty sigma proxy
- prediction error proxy
- guardband proxy
- Cpk proxy
- escape-risk proxy
- false-hold-risk proxy
- sampling cost index
- decision gate
- next safe check
EXPORT
Download the public artifacts.
These exports are the supported public files for review. PDK files and restricted third-party material are not redistributed.
Small-data VM replay CSV
/data/small-data-virtual-metrology-replay-v80.csv
Open file ->Web result JSON
/data/small-data-virtual-metrology-replay-web-v1.json
Open file ->Model card
/data/small-data-virtual-metrology-replay-model-card-v1.md
Open file ->METHODS
How the number was made.
Methods are written for citation discipline: the extraction rule matters as much as the plotted value.
Generate a 2880-row factorial teaching table over module context, sample plan, data age, sensor health, physical metrology fraction, and model family.
Keep this method attached when reusing the figure or metric.
Compute measured sample count, training rows, drift score, uncertainty proxy, prediction error proxy, guardband proxy, Cpk proxy, escape-risk proxy, false-hold-risk proxy, sampling-cost index, decision gate, evidence state, and next safe check.
Keep this method attached when reusing the figure or metric.
Use official public source anchors for vocabulary and problem framing only; all visible rows, formulas, labels, and limitations are original SemiAgora teaching content.
Keep this method attached when reusing the figure or metric.
LIMITS
What this page does not claim.
The MVP is useful because the limits are visible. These statements prevent a teaching simulation from being cited as silicon evidence.
The virtual-metrology values are not calibrated to CMP, lithography, etch, implant, ALD, or any other real process module.
Carry this boundary into any derivative note, paper draft, or slide.
Model-family labels are educational proxies; no machine-learning library, Gaussian-process fit, conformal-prediction package, APC system, or commercial VM tool is run on the public page.
Carry this boundary into any derivative note, paper draft, or slide.
A real VM program needs owned sensor streams, matched physical metrology, drift monitoring, data lineage, uncertainty validation, process-owner review, and quality governance.
Carry this boundary into any derivative note, paper draft, or slide.
No live server model training, uploaded wafer data, equipment-control action, or APC execution is enabled.
Carry this boundary into any derivative note, paper draft, or slide.
This lab teaches evidence discipline; it cannot approve sample reduction, release a lot, replace metrology, qualify a recipe, or set a manufacturing control policy.
Carry this boundary into any derivative note, paper draft, or slide.
EVIDENCE
Trace the claim back to artifacts.
The evidence route remains the catalog-level browser; this page is the simulation-level reading card.
The CSV preserves every module context, sample plan, data-age state, sensor-health state, metrology fraction, model-family label, and decision metric used by the browser lab.
SA-VMET-SMALLDATA-001
The JSON stores row counts, role views, public boundaries, source anchors, methods, limitations, and high-risk examples.
SA-VMET-SMALLDATA-001
The Metrology/Yield lane now has a virtual-metrology surface beside SPC, DOE, CD/overlay, ellipsometry, TLM, and feedforward-control evidence.
SA-VMET-SMALLDATA-001
The public page teaches measurement judgment only: no wafer-data upload, no model training service, no process-control action, and no release decision.
SA-VMET-SMALLDATA-001
Source spine
These links anchor the public teaching model. They do not convert the replay into a qualified process recipe or signoff result.
NIST virtual metrology dynamic samplingOfficial 2025 NIST publication anchor for virtual metrology, drift/shift, uncertainty quantification, adaptive sampling, and public CMP-dataset validation framing.IRDS / NIST virtual metrology white paperOfficial white-paper anchor for virtual-metrology benefits, data-quality issues, confidence metrics, correlation with real metrology, and model-maintenance concerns.NIST metrology and process-control uncertaintyOfficial anchor for why measurement uncertainty can create false pass/fail decisions in process-control loops.NIST CHIPS Metrology ProgramOfficial program anchor for fit-for-purpose semiconductor measurements as devices and manufacturing stacks become harder to monitor and predict.NIST/SEMATECH process monitoring handbookPublic handbook anchor for SPC/SQC vocabulary and checking whether a process has degraded.NIST process capability handbookPublic handbook anchor for Cp/Cpk capability vocabulary and assumptions.
Public execution boundary
No server-side live ngspice execution, uploads, accounts, payments, comments, newsletter signup, or job/event submissions are active on this page.