Test Coverage: Design for Testability: Test Access and Coverage Rules

Design for test is one of the few engineering disciplines that costs nothing to apply and saves a great deal when it is missing. Its whole purpose is to make a fault visible at the stage where it can be corrected cheaply, rather than at the stage where the product is already in a customer’s hands.

What Test Coverage Means in Practice

Coverage is the proportion of the possible faults that a test programme can detect, and it is never a hundred per cent. A continuity test finds opens and shorts between nets but says nothing about a component that is fitted in the wrong orientation or a resistor of the wrong value.

In-circuit test reaches further, because it measures the impedance at each node and can therefore detect a missing or incorrect component. The limit is access: every node the tester must reach needs a probe point, and the probe point has to be reachable from the fixture side. Our flying probe notes describe the alternative where a fixture is not economic.

Providing Test Access

A test pad is a piece of copper that exists only to be probed, and it should be large enough for the probe to land on without slipping, typically a square of about a millimetre, with a clearance of a similar amount around it. Where the pad is on a dense board, the fixture may need to approach it from both sides, which doubles the mechanical complexity.

The pads need a grid, because the fixture is drilled on a standard pitch and a pad that falls between grid positions cannot be reached. The common pitch is around a tenth of an inch, and maintaining it is a constraint on placement as much as on routing. Our test coupon notes describe the related verification structures.

Test Points and the Bill of Materials

A test point placed under a component is worse than no test point, because it exists in the netlist and cannot be used. Placement rules should keep test pads clear of component courtyards, especially where the component has a body that sits close to the board.

Where a node is inaccessible, the designer has three options: bring it out to a test pad on another layer through a via, accept reduced coverage, or rely on a functional test that exercises the node indirectly. The third option is often the right answer for a node that is probed only to confirm a connection.

Clustering and Sequencing

A test programme is faster when it is organised well. Grouping the tests by fixture access rather than by net name reduces the number of probe movements, and sequencing them so that the most likely failures are found first means the programme can stop early on a bad board.

The same reasoning applies to the order in which power rails are checked. Testing a rail before its own supply is confirmed produces a failure that is real but misleading, and the operator spends time on the wrong fault. Our short circuit notes describe how a fault is localised.

Boundary Scan and Its Advantages

Boundary scan moves the test inside the device. A device with a boundary scan chain can drive its own pins and read the state of its inputs, so the tester can verify connections between two scan-capable devices without physical access to the nodes between them.

The requirement is that the devices support the standard and that the chain is routed correctly on the board. A chain that is broken by a design error disables the whole capability, so the scan path should be treated as a critical net and given its own routing discipline.

Functional Test and Its Cost

A functional test exercises the assembled product under conditions that resemble its use, and it is the only test that confirms the product works as a system. It is also the most expensive to develop, because it needs a test rig, stimulus and a definition of passing.

Most programmes combine the two: an in-circuit or boundary scan test to catch assembly defects, followed by a functional test to confirm behaviour. The first is fast and finds most defects; the second is slow and finds the rest.

Design Rules for Testability

The rules are short. Keep a test pad on every net that matters, place those pads on a standard grid and clear of components, provide a way to isolate the power rails so that a short can be localised, and give the tester a ground reference close to the measurement points.

Where a design has a programmable device, provide a header for the programming and debug interface, and make sure it is accessible after assembly rather than only before. A product that can be programmed in the fixture is far easier to support than one that needs a separate operation.

Test Coverage as a Design Review Item

Coverage should be reviewed with the same formality as the layout rules, and the review should happen before the test programme is written. An inaccessible node found at that point costs a layout change; the same node found later costs a fixture that is already drilled.

The report to the customer should state the coverage achieved and the nets that are not covered, with the reason. That record is what allows the test strategy to be defended later, when a defect escapes and the question of why it was not detected is asked. Our design release checklist lists the items to confirm.

Treating testability as a layout constraint rather than as an afterthought keeps the cost of quality low, because the cheapest defect to find is the one that a well-placed pad has already made visible.

Process Control and Verification

Reviewing the design before the data is released is far cheaper than correcting it after the panel is in the tank, because every step downstream inherits the decision made at the front end. 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.

The measurements that matter are the repeatable ones: conductor width and spacing, annular ring, finished hole size, plating thickness and surface finish are all verifiable on a coupon that travels with the panel. Documenting the assumption is part of the design work, and a short note on the drawing prevents a question that would otherwise arrive a day later and cost a day of schedule.

FAQ

How many test points should a board have? Enough to reach every net that a fault could affect, which in practice means most of them. The exceptions are nets that are inaccessible by construction, such as those under a ball grid array, and those are covered by boundary scan or by a functional test instead.

Does adding test pads increase board size? It can on a dense board, although the pads are often placed in the space that the routing already leaves. Where the design has no room, moving the test access to the bottom side or reducing the coverage on non-critical nets are both acceptable answers.

What does gopcb check for testability? We check the probe access of every test pad, the grid alignment of the fixtures we would use, the clearance between test pads and components, and the isolation of the power rails. Where a net cannot be reached we report it before the board is made rather than after the fixture is drilled.