Monitoring

Fab Overview

Live status of every lot on the line: defect rate, dominant pattern and drift alarms.

Lots

Pattern sequence in test order · click a lot for details
LotToolStatusDefect rateDominant patternSequence · 25 wafers
Loading lots…

Active drift alarms

requires engineering attention

Defect pattern distribution

all lots
Monitoring

Lot Monitor

Replays a lot in test order. A drift alarm fires when the same defect appears on at least 3 of the last 5 wafers.

Select a lot to replay

Each wafer is classified in the order it was tested; the first wafer that triggers a drift alarm receives a full root-cause analysis and a cost recommendation.

Analysis

Wafer Inspection

CNN classification, out-of-distribution check, similar historical cases, drift context and AI root-cause analysis.

Select a wafer to inspect

Pick a wafer from a lot, an index from the 30k-wafer history, or upload your own wafer map.

Analysis

Upload Wafers

Bring your own wafer maps. Files are validated on the server and previewed here; then inspect a single wafer or replay them all as a lot with drift detection and a cost decision.

Supported formats

Encoding: 0 = no die (background) · 1 = pass die · 2 = fail die. Any grid size; maps are resized to 64 × 64.
JSON

A 2-D array, {"wafer": [[…]]}, a list of those, or {"wafers": [{"wafer_id": "W1", "wafer": [[…]]}, …]}.

CSV / TXT

One row per line, values separated by commas, spaces or tabs. Several wafers: separate blocks with a blank line; a line like # W1 names the next wafer.

NPY / NPZ

A 2-D array (one wafer) or a 3-D stack (N, H, W). Pickled object arrays are rejected.

Images

A rendered wafer map in three colours. Background is taken from the image border, the warmer colour is read as fail. Always check the preview.

Download a sample with one real wafer per defect class: JSONCSV
Analysis

Review Log

Engineer confirmations and corrections of model predictions.

IDTimestampWaferPredictedReviewCorrect classComment