Electrical Test in PCB Production: Opens, Shorts and Fixtures

A bare board that passes visual inspection can still contain a broken track inside a via, a short between two inner layers, or a connection that is marginal enough to fail in the field. Electrical test is the only step in the fabrication flow that verifies the connectivity the designer intended, and it is applied more than once because each stage carries its own risk of introducing a defect.

Where Testing Happens in the Flow

Three points in the process normally carry a full electrical test. The first is after inner-layer etching, where a fault can still be repaired or the layer scrapped before value is added. The second is after outer-layer etching, which is the last chance to correct a defect on a panel that has not yet been coated. The third is on the finished board, before shipment.

Each test is a filter, and the cost of the defect rises with every step it survives.

Why Early Detection Matters

The economics of testing are captured by the rule of ten: the cost of finding and correcting a defect grows by roughly an order of magnitude at each stage of assembly. A broken track found at bare-board test costs a patch wire or a scrapped panel, measured in minutes and a few dollars. The same track found after components have been placed, reflowed and the assembly tested costs the parts, the rework labour and the inspection, and if it escapes to the end customer it can cost the account.

<img src="https://www.gopcba.com/wp-content/uploads/2024/10/op1.jpg" alt="electrical test of a bare PCB panel with flying probe heads” />

Testing is therefore not an overhead added to protect the customer. It is the mechanism that keeps defects from accumulating cost as they travel down the process.

What the Test Actually Checks

Bare board electrical test verifies continuity and isolation. Continuity confirms that every net on the netlist is connected from end to end, through every via in the path. Isolation confirms that no two nets that should be separate are connected, including through the inner layers where a short cannot be seen. The test does not verify impedance, solderability or mechanical quality; those need their own checks.

The test programme is derived from the netlist, and the netlist has to match the artwork. Where a repair has been made, the test points and the repair record should be reconciled before the board is shipped, or the test will confirm the wrong design.

Fixture Test and Flying Probe

The traditional approach uses a dedicated fixture with spring-loaded probes, one per test point. A fixture is fast and cheap per board once it exists, which suits volume production, but it has to be built for each design and it costs money before the first board is tested. It also requires test points on the design, and on a dense board those points compete with components for space.

A flying probe moves a small number of heads over the board and measures each net in turn. It needs no fixture, so it suits prototypes and low volumes, and it can reach points that a fixture cannot. It is slower per board, and the trade is normally decided by volume and by how often the design changes. Multilayer PCB prototype requirements covers the test access that a prototype should include for either method.

Parameters to Agree With the Customer

Test conditions are a specification, and they should be written down. The data source and format define what is being tested. The electrical conditions define the test voltage, the current limit and the resistance threshold for continuity, and the insulation resistance and voltage for isolation. The fixture method and the probe selection define which points are reached. The test stamp or marking identifies the boards that passed.

The repair specification is the part most often left vague. It should state what kinds of repair are permitted, who approves them, and how a repaired board is re-tested and marked. A repair that is not documented is indistinguishable from an escape.

Using Test Data for Process Control

A test station is a data collection point, and the value of the data is in the pattern. Faults that cluster in one area of a panel usually point at a process condition, such as contamination on a press plate or a poor vacuum seal during exposure. Faults that cluster on one part number often indicate a conflict between the customer’s design rules and the shop’s process capability, and they need a conversation rather than a process tweak. Faults that appear at random are the hardest to resolve and usually need equipment or tooling investment.

<img src="https://www.gopcba.com/wp-content/uploads/2023/05/Delkin-Devices-Engineers-Working-in-Factory.jpg" alt="test fixture probing a finished printed circuit board” />

Summarising the failure percentages by type, tracking them over time, and feeding the results back to the process owners is what turns testing from a screening step into an improvement loop.

Repair, Re-test and Concession

Some defects can be repaired. An open track on an outer layer can be patched, a short can be cut, and a missing feature can sometimes be added. Repairs have limits: the number permitted per board, the area in which they may be made, and the maximum length of a patch. Every repair changes the electrical characteristics locally, and a patch on a controlled-impedance line changes the impedance at that point.

Boards that fail outside the permitted repair window should be scrapped or shipped only against a documented concession agreed with the customer. Shipping a board with an undeclared repair is the fastest way to lose a qualification.

Documentation and Traceability

Test Point Design

A test programme is only as good as the access it has. Test points should be placed on nets that are otherwise unreachable, on the power rails, and on any signal that is difficult to probe once the product is assembled. A round pad of about 1 mm with a 0.8 mm solder mask opening is a workable standard, and it should be given its own footprint rather than shared with a component pad, because probing wears the surface and a worn pad becomes a soldering problem later.

Spacing matters as much as size. Fixture probes need clearance between adjacent test points, and flying probe heads need room to approach from above without colliding with tall components. Keep test points away from the board edge, where the probe may not reach, and away from fine-pitch component pads, where the probe risk of damage is highest. PCB design guidelines for manufacturability covers these clearances alongside the rest of the layout rules.

Finally, record which test points are functional and which are for manufacturing only. A test point that a customer expects to use for a firmware interface has to be documented, and one that exists only for fabrication has to be listed so that nobody designs a product function around a pad that will not be present on the next revision.

Test records should identify the panel, the design revision, the test programme revision and the result, and they should be retained for the period the customer requires. When a field failure occurs months later, the test record is what establishes whether the board was tested, with which programme, and whether it passed. PCB design quality characteristics describes the wider set of evidence that a fabricator is expected to keep.

FAQ

Does every board need 100 percent electrical test? Most customers specify it, and it is the only way to guarantee connectivity on each unit. Sampling is cheaper but leaves the untested boards unverified.

Can a flying probe replace a fixture? For prototypes and small batches, yes. For volume production, a fixture is faster and usually cheaper per board, provided the design has the test points to support it.

What do I need to provide for test? A netlist that matches the artwork, a defined list of test points where the design is dense, and a written specification for the electrical conditions and the permitted repairs.