How Much Test Coverage Do You Actually Have?

Test coverage is easy to assume and hard to prove, and the gap between the two is where escaped defects live. Building a defensible coverage argument requires understanding what each test can and cannot see, and where the boundaries between them fall.

What Continuity Testing Proves

A continuity test drives a current through each net and measures the voltage that results, which proves that the net exists and that it is not shorted to a neighbour. It proves nothing about the value, orientation or presence of the components attached to it.

That distinction is often blurred in discussion, which leads to the belief that a board that passes test has been verified. It has been verified for one class of defect, and the other classes remain open. Our flying probe notes describe the method.

In-Circuit Test and Guarding

In-circuit test goes further by measuring the impedance at each node while the surrounding components are guarded, which requires access to more nodes and a more complex fixture. The measurement can detect a missing component, a wrong value and, in many cases, a reversed one.

The guarding is the difficult part. A parallel path through another component can mask the measurement, and the tester’s ability to isolate it depends on the network topology rather than on the tester’s specification.

Optical Inspection and Its Limits

Automated optical inspection checks the presence, position and orientation of components and the appearance of the solder fillets. It is fast and it runs in line, which makes it the default for volume production.

Its limit is that it sees only what is visible. A joint under an area array package has no visible fillet, so optical inspection can confirm that the component is present and correctly placed and say nothing about whether it is connected. Our optical inspection notes set out the limits.

X-Ray and the Area Array Blind Spot

X-ray inspection sees inside the assembly and can identify voids, bridging and missing balls under an area array package. What it cannot do reliably is detect a joint in which the two surfaces are in contact without having merged, because there is no gas gap for the X-ray to resolve.

The technique therefore proves the presence of solder rather than the quality of the connection, and a design that depends on X-ray alone has a residual risk of the specific defect that X-ray cannot see. Our BGA inspection notes compare the methods.

Boundary Scan and Its Coverage Question

Boundary scan moves the observation point inside the device, which removes the need for physical access to the nodes between two scan-capable parts. Its coverage is limited by the scan chain itself: nets that do not connect two scan-capable devices are outside the chain’s reach.

The chain also has to be intact. A design error in the scan path disables the capability for the whole board, so the path should be treated as a critical net with its own routing rules and its own continuity check.

Functional Test and the Definition of Passing

A functional test exercises the product under realistic conditions and confirms that it behaves. It is the only test that validates the design rather than the construction, and it is also the test with the weakest coverage guarantee, because a functional test tells you that the unit works but not that it will continue to.

The value of the test depends entirely on the definition of passing. A test that checks a subset of the specification gives a subset of the assurance, and the gap should be known and recorded rather than assumed to be small.

Combining the Tests

The usual arrangement runs an assembly-defect test first, to remove the units that are wrong, and a functional test afterwards, to remove the units that do not work. The first is cheap per board and covers many defects; the second is expensive and covers the remainder.

Where the volume is high, a sample-based functional test is sometimes substituted for a full one, with the assurance coming from the process rather than from the test. That is a legitimate decision, provided it is made explicitly and the escape risk is understood.

Documenting the Coverage

A coverage statement should name the defect classes tested, the method used for each and the classes that are not covered. It should also record the access required, so that a future change to the board does not silently remove a test point.

The statement is what allows the test strategy to be defended when a defect escapes. A programme with a documented gap is a programme with a known risk, while a programme with an undocumented gap is one that will be judged on the quality of its assumptions after the fact. Our design for testability notes describe the provisions that make the statement possible.

Process Control and Verification

On a design of this kind, in-circuit test is the item that decides how the rest of the board is arranged. The process window is set by the narrowest step in the flow, so an improvement anywhere else shows up as margin rather than as yield until that step is addressed. A short note on the drawing about handling, storage or packaging is often worth more than an extra decimal place on a tolerance.

Where a value sits close to a process limit, the drawing should say so, since the shop can then open the process window rather than working to a nominal figure that carries no tolerance. A stack-up that is drawn rather than described removes most of the ambiguity from a quotation, and it lets the fabricator price the board against the dielectric and copper weights that will actually be used.

Process Control and Verification

On a design of this kind, in-circuit test is the item that decides how the rest of the board is arranged. The process window is set by the narrowest step in the flow, so an improvement anywhere else shows up as margin rather than as yield until that step is addressed. A short note on the drawing about handling, storage or packaging is often worth more than an extra decimal place on a tolerance.

Process Control and Verification

On a design of this kind, in-circuit test is the item that decides how the rest of the board is arranged. The process window is set by the narrowest step in the flow, so an improvement anywhere else shows up as margin rather than as yield until that step is addressed. A short note on the drawing about handling, storage or packaging is often worth more than an extra decimal place on a tolerance.

FAQ

Is one hundred per cent test coverage achievable? No, because no single method detects every defect class and no combination is exhaustive. The practical goal is to cover the classes that the process can produce, with the residual risk documented and accepted.

Does more test always mean better quality? It means more defects are caught at test, which is not the same thing. Quality improves when the process produces fewer defects, and a test programme that compensates for a weak process becomes a permanent cost rather than a solution.

What does gopcb provide as test evidence? We provide the test method used, the programme revision, the pass and fail counts and, where requested, the record of an individual board by its identifier. For prototypes we also provide the first article measurements against the drawing.