Analyse your first wafer lot¶

For: anyone new to tsmap. No wafer files, no installation, and no prior knowledge of the app needed. Goal: load a real multi-wafer lot, find its worst wafer, and discover why it is the worst. You will need: a browser. Nothing else. Time: about 10 minutes.

At the end you will have found a genuine process-corner effect in a 13-wafer lot, and you will know which parts of the interface to reach for when you do it with your own data.

This is a guided walk-through with a known answer — follow it in order. For the full reference on any screen it touches, see the user guide.

1. Open tsmap¶

Go to wafertools.github.io/tsmap/app/.

The browser version needs no download and parses files entirely on your own machine — nothing is uploaded. Everything below works identically in the desktop app if you have it installed.

You should see an empty toolbar with Open files, Add files, and — in the middle of the window — Load sample data.

2. Load the sample lot¶

Click Load sample data.

This loads PVT-LOT-05, a synthetic 13-wafer lot generated for demonstration — no real device data. It is a process-corner lot: each wafer was run at a deliberately different corner of the process window, and tsmap labels each one automatically.

3. Import the tests¶

A test selector appears before parsing. This happens for every STDF and ATDF file — it is the import step, not a warning.

The lot has 7 tests (six parametric, one functional), and all of them are already ticked. Small lots arrive pre-selected; a large one starts empty instead, so that "import everything" can never be the accidental default. See Memory advisory for where that line falls.

Click Import 7 tests →.

You should now see a grid of 13 wafer maps, each card titled like W01 · TT, with a Summary panel down the right-hand side, headed Lot PVT-LOT-05 · 13 wafers — every wafer comes from that one lot, so the panel names it. The · TT suffix is the process corner — those were applied automatically along with the sample.

4. Read the lot summary¶

Before looking at any single wafer, read the panel on the right. It reports:

• 13 wafers, 79.0 % mean per-wafer yield, 2,873 dies analysed.
• A Findings section — 4 findings, 1 Unusual and 3 Notable.

The top finding reads "Edge-local failure pattern detected. Ring 4 (edge) shows reduced yield." tsmap found that on its own: nothing was configured to look for it.

Mean per-wafer yield is not lot yield. It is the arithmetic mean of 13 per-wafer percentages, so every wafer counts equally regardless of how many dies it has. The label says which one it is — worth reading carefully whenever you quote a number from here.

5. Find the worst wafer¶

Scroll to Wafer Yield at the bottom of the panel. It lists every wafer in slot order by default. Click Yield on the small Slot | Yield toggle to re-sort worst-first.

The order is now:

This is the moment the lot explains itself. The two worst wafers are both SS. The two best are both TT. Yield is tracking the process corner, not the slot number — and slot order, the default view, hides that completely.

6. Look at the worst wafer¶

Click the ⤢ expand icon on the W11 · SS card to open it on its own.

Use the Plot mode button in the toolbar to switch between Hard Bin and Soft Bin. Hard bin tells you whether a die failed; soft bin tells you how — the same failures, resolved into finer categories. Hover any die for its coordinates, bin, and test values.

Close the expanded view when you are done.

7. Compare the corners¶

Click Insights (the chart icon at the right of the gallery toolbar). You get three sub-tabs: Overview, Distributions, and Correlation.

On Overview, find the Group by: picker and choose Split.

Every panel now reports per corner instead of per wafer. Yield by Split ranks the corners directly, and the bin pareto shows which bins each corner is losing dies to — SS failing differently from FF is a different problem from SS simply failing more.

Click a bar in Yield by Split to drill into that corner's individual wafers, then ← Back to return.

8. Export one result¶

Any of these work from the lot view:

• Summary report (top of the Summary panel) — a self-contained HTML report of the whole lot.
• Test values CSV / Functional CSV — the underlying numbers.
• Download gallery PNG (camera icon in the toolbar) — the wafer grid as an image.

What you have done¶

• Loaded a 13-wafer lot and imported its tests.
• Read a lot-level summary and let tsmap surface an edge-related pattern unprompted.
• Re-sorted by yield and found that the worst wafers share a process corner.
• Switched between hard and soft bin to separate whether from how.
• Grouped the whole analysis by split to compare corners rather than wafers.
• Exported a result.

Where to go next¶

• Your own data — Supported file formats and, for CSV/JSON/Parquet, Column mapping.
• Splits on your own lots — the sample's corners were pre-assigned; assign your own in Wafer splits.
• Something not working — Troubleshooting.
• The rest of the interface — the user guide.