In Circuit Test Program Release: 6 Checks Before Handover

An in circuit test program is released once and then trusted for years, which is why the release review matters more than any single test run. A program that passes on the developer’s desk can still miss a net, hold a limit that no board can meet, or point at a fixture that was built for a different panel. Six checks catch those faults before production does.

Why In Circuit Test Release Is a Reviewed Step

The release review is the point where the program stops being a file and becomes the measurement that decides whether boards ship. Everything after that moment is production pressure, and changes made under pressure are the ones that reduce coverage without anyone noticing until a field return appears.

A defined release therefore has a signature, a date and a set of evidence. The in circuit test program version, the fixture revision, the limit file and the correlation result are recorded together, so a later question about a specific board can be answered from documents rather than from memory.

Coverage Review: What the Program Actually Tests

Coverage is checked net by net against the schematic and the bill of materials, not by counting test points. Opens, shorts, missing parts, wrong values and polarity each need their own verification, and a net that is accessible but untested is a gap that no pass rate will reveal.

The review also has to confirm the opposite case: that the program does not test nets that are not brought out to a test point. Where a design has a net with no access, the coverage report should say so, and that gap is what drives the in circuit test coverage figure the customer is given. Where a net is accessible only through a connector or a socket, the document should say so, because the coverage figure changes once that connector is populated.

Fixture Verification Before the First Run

The fixture is checked mechanically before the program is judged. Probe alignment against the panel, planarity across the test area, support posts under heavy components and the vacuum seal all affect whether contacts are made at the specified pressure rather than at whatever pressure the board allows. Support posts are also checked under the heaviest components, since a board that flexes under probe pressure changes the contact force on every probe across the panel.

Probe wear, spring force and cleanliness are recorded at the same time, since a fixture that has been in storage may have corroded tips. Confirm the fabrication notes used to build the panel match the fixture drawing, because a panel revision that moved a test point invalidates the release even when the netlist is unchanged.

Limits, Tolerances and Guard Bands

A limit that is set exactly at the nominal value will fail good hardware, while a limit that is opened to stop failures will pass bad hardware. The guard band between the two is what makes the test useful, and it has to come from the component tolerance plus the measurement uncertainty of the tester itself.

Resistance, capacitance and diode drop each need their own treatment. Where a reading sits close to a limit on every board, the limit is wrong rather than the boards, and the fix belongs in the program rather than in a re test policy that quietly accepts out of limit results. The test method itself, including continuity and isolation limits for bare and loaded boards, is described in IPC-9252, which is a useful reference when setting guard bands.

First Article and Correlation With a Known Board

Correlation uses a board that has been measured on another instrument or a board with a known injected fault. The same board is run on the released program, and the two sets of readings are compared; a measurement path that is offset by a fixed amount is a fixture problem, while a random spread is a contact problem.

The correlation board should also carry a deliberate fault on each test type, because a program that detects nothing will happily report a full pass. Record the readings, the board identity and the date, and keep the board with the program so that the next release can be compared against it.

Handling False Failures and Retry Rules

False failures are inevitable at some level, and the way they are handled decides whether the test remains meaningful. A retry limit, a defined re clean step and a rule for scrapping a board that fails repeatedly are part of the release, not an operator decision made at the station.

Any board that passes only after several retries is a candidate for review, because the same marginal contact that produced the failure may have hidden a real one. Read the retry data with the solder joint acceptance criteria so that a process problem is not dismissed as a fixture quirk. Where a station reports the same failure at the same net on every board, stop the line and check the fixture before the program is adjusted.

Documentation for the In Circuit Test Release

The release file holds the program revision, the fixture drawing revision, the netlist, the coverage report, the limit file, the correlation results and the name of the person who approved them. Each document is dated, because the value of the file comes from being able to reconstruct the state of the test on a given day.

Store the file where production can reach it without a request, since the first question after a failure is always which version was running. The in circuit test record also belongs with the wider quality records for the same lot, so that test data and inspection data can be read together.

Operator Training and Shift Handover

A released program still depends on the operator loading the fixture correctly, verifying the panel identity and reacting to the first failures of a shift in the right order. Training covers those three points and the specific alarm messages the station can produce. The release note should list the alarms the operator is expected to act on and the ones that require an engineer.

Shift handover should pass on the state of the fixture, any retries that were seen and any program change that is pending. An unrecorded change from the night shift is the most common way a released program becomes an unreleased one in practice.

Change Control After Release

Any change to the program, the fixture or the limit file is a new release, even when the change looks trivial. A revised limit or an added test step invalidates the earlier correlation, and the release evidence has to be produced again for the new state.

Keep the superseded versions rather than overwriting them. When a customer asks why a board shipped in a given month, the answer depends on which version was loaded at the time, and a file that only holds the current revision cannot answer that question.

FAQ

What should an in circuit test release package contain? The program revision, the fixture drawing revision, the netlist, the coverage report, the limit file, the correlation results and the approver’s name, each dated. The package is what allows any board to be traced back to the test state it was measured under.

How is test coverage verified? Net by net against the schematic and the bill of materials, covering opens, shorts, value errors and polarity. Counting test points is not coverage verification, because an accessible net that is never measured still counts as a gap.

Why use a correlation board? Because it proves the released program measures a known board the same way another instrument or an injected fault does. A fixed offset points to the fixture and a random spread points to contact quality, and both are found before production starts.