Report Analyzer — Documentation

← Documentation

Report Analyzer — user guide

This guide follows the same order you'll use the software in: getting TestStand to write the right report format, importing the reports, organizing them, analyzing them, and exporting the result. The screenshots use synthetic demo data (product BDB-100), with Report Analyzer 4.0.1.

1. Set up TestStand to write ATML reports

Report Analyzer reads ATML XML IEEE-1636.1:2011, the native, lightweight report TestStand already knows how to write: you don't need a results database or a server. That isn't always what a station is producing: many sequences are set to write HTML or plain text, which this tool can't read. Check this first.

In TestStand, open Configure → Result Processing and set the report format to ATML 5.00. From then on, every completed run writes a file Report Analyzer can import.

TestStand's Result Processing entry, with the report format set to ATML 5.00 and the output folder where the files will land.

2. First run

The application opens on the Import module with an empty dataset. Nothing is loaded yet: this is where you point it at your reports folder. The tabs across the top are the modules: Import, Grouping, History & Statistics, Gage R&R (Beta), and Error Distribution.

3. Import your reports

Drag the folder where your station writes reports onto the window, or browse to it, and bulk-import. The reader processes every file, expands multiple-numeric-limit steps into individual measurements, and discards the setup and cleanup SessionActions that would otherwise flood the dataset with noise.

A folder of 1,543 reports, imported: one row per run, with its station and socket.

If your sequences always produce the same kind of noise, save the import filter as a filter set and reuse it on every import instead of configuring it each time.

After importing you get one row per run, with date, serial number, product, overall result, duration, station, and socket. That's your dataset; every other module works from it.

Files with several runs (append mode)

If TestStand writes several runs to the same file, each run comes in as its own execution, and the import tells you so: "N files contained more than one run — each run was imported as its own execution".

Importing the same folder again

Next week you don't start over: import the same folder again and only the new reports come in. The ones already in the dataset are skipped, and the app tells you how many.

On re-import: 51 new executions from 51 files, and "508 runs were already in this dataset and were skipped".

Incomplete or damaged reports

A report TestStand was still writing when you copied it is never half-imported: the import counts it as failed and, when you open the details, says why ("The file is incomplete: it ends before the report is closed…"). The same goes for an empty file or one that isn't an ATML report. The rest of the folder imports normally.

4. Group your runs

Analyzing everything together is almost never what you want. Groups let you split a dataset by whatever dimension matters — shift, batch, fixture, station, before and after a process change — and then run any analysis on one group instead of the whole pile.

Create a group, assign runs to it, and it becomes selectable throughout the application.

5. Read the statistics

History & Statistics is where most of the work happens. The table gives you, per measurement: how many values, how many valid, how many failed, mean, minimum, maximum, the limits, tolerance, nominal, sigma, bias, and the indices — Cg, Cgk, Pp, Ppk — plus %EV, %K, %RE, and the verdict. Pp and Ppk are performance indices, calculated with the overall standard deviation of every run.

Click a measurement's row and the two panels below update: the history chart plots every value in run order against its limits, and the histogram shows the distribution against the normal curve. Cells outside the active standard's acceptance limits appear shaded, so a long table gets scanned in seconds instead of read row by row.

The history of 5V_OUT Voltage: the measurement moves toward the upper limit from one day to the next, before a unit fails.

The history chart can be drawn as lines, points, bars, spline, or steps, with markers, limits, and fill area toggled separately, useful when several measurements are overlaid and the chart gets crowded.

Flags instead of a false number

When an index makes no sense, the table flags it next to the measurement name instead of giving you a misleading number: "no variation" if every reading is the same, "two-state" if the measurement only takes two values (a relay, for example), and ⚠ when the readings don't match the pass/fail TestStand wrote, which usually points to a misread limit.

Type 1 study: Cg and Cgk

Cg and Cgk tell you whether the measurement system is capable: they come from measuring the same master unit many times. Import those runs into a separate dataset; mixed with production, the indices lose their meaning. With the standard your customer requires, the table gives Cg, Cgk, and the verdict per measurement, and shows you which one has bias.

A Type 1 study: 5V_OUT Voltage repeats well (Cg 4.51) but has a bias that brings its Cgk down to 0.94.

6. Gage R&R (Beta)

The Gage R&R module, still in Beta, analyzes a crossed study: several parts, measured by several operators, with repeats. It also goes in a separate dataset. The flow is New → Select all → Auto-detect grid → Save grid → Results: the grid builds itself, with the serial number as the part and the report's operator as the operator.

The results give, per measurement, the %GRR, the ndc, and the AIAG verdict, with the variance components and the ANOVA table, so you can see whether the variation comes from the equipment (EV) or the operator (AV).

A study of 10 parts × 3 operators × 3 repeats: 3V3 Rail comes out Acceptable (1.48%), ADC Channels · CH3 Conditional, and PWM Frequency Unacceptable, because of one operator's bias.

7. Analyze the failures

Error Distribution answers a different question: not "is this measurement capable" but "what's costing us yield." It gives you the overall pass rate, the status breakdown by group, and the ranking of the most frequently failing measurements: the short list of what to fix first.

Failures by station and socket

The Failures by station and socket table has one row per station and socket, with the worst fail rate at the top in red, following the selected group. It tells you whether the problem is the unit or the fixture. The socket is read from each report starting with version 4.0.1; anything you imported earlier has no socket until you import it again.

8. Export a report

The Report Wizard walks you through what goes into the document in three steps. The first step is measurement selection: include everything, or narrow it down to the ones the report is about. The following steps cover which groups to include, the output format, and the cover-page data — company, author, product, variant, and analysis type — which are printed on the first page along with your logo. At the end you choose Export PDF or Export Excel.

Failure summaries have their own export, separate from the wizard:

Exporting the failure and error summary, with the same cover-page fields.

9. What the exported reports look like

The Type 1 MSA report opens with a cover page carrying your logo and the metadata you entered, followed by a summary table of every measurement with its n, mean, limits, indices, and verdict. Then, one page per measurement: statistics and indices side by side, the histogram, and the run history against the limits.

The failure summary report is shorter: cover page, totals, and the frequency of each failure, ready to attach to a corrective action.

The Excel version carries the same analysis, plus the raw values, so whoever receives it can verify every number instead of taking the PDF on faith: a summary sheet with the verdict per measurement, and one sheet per measurement with its raw data. That's typically what a customer quality engineer asks for.

10. Settings reference

General

Significant digits used throughout the application, default number of histogram bins, and whether long group names are shortened for display.

Analysis: the MSA standard

This is the setting that changes your numbers. The application ships with four standards — AIAG MSA 4 (2010), AIAG MSA 3 (2002), Ford Global, and VDA 5 / Q-DAS — and each uses different factors and a different Cg acceptance limit, so the same data can pass with one and fail with another. Choose the one your customer requires, and the preview shows you the formula that will be applied.

If a customer imposes factors that aren't among the four, define your own standard with its own numerator factors, sigma denominators, and acceptance limit.

View

How many groups are shown at once, the default history chart mode, and the series style.

Export

Default output folders for PDF and Excel, and the company logo printed on every report's cover page. Set the logo once and every report you generate carries it.

Licensing and activation

The download includes a 14-day trial with every feature, with no key or account. When you're ready, buy a license; checkout doesn't ask for any machine details. Then copy the Device ID from the activation screen, send it with your order number through Activate your license, and get your key on the next business day, to paste into that same screen. One license activates one computer, forever.

If you change computers, contact us with the new Device ID and we'll reissue the key at no cost.