Nx Gen Inventors
EngineeringTechnical paper

Designing for test before you design the board

2 min read
Test bench instruments and probes laid out on a workbench
Key takeaways
  • Coverage is fixed the moment placement is locked.
  • A test point costs millimetres; a missing one costs a spin.
  • Write the test plan beside the specification, before the schematic.

Test coverage is a layout decision. Deciding it late is what turns a two-spin project into a five-spin one.

The question that arrives too late

A board reaches the bench and the first question is how much of it can actually be reached. If that answer is a surprise, the design is already finished. The panel is cut, the fixture is quoted, and every remedy is a revision.

This is not a test engineering failure. By the time a board exists, coverage has already been decided — silently, by placement. The set of nets a bed-of-nails can touch was fixed when the last component moved.

Three decisions that belong in the specification

Before the schematic, three things need writing down. Which functions must be provable in-circuit. Which can be proven functionally, at the edge connector. And which will only ever be proven in the field, by telemetry.

That split is not a testing preference; it is a cost model. It determines how many test points the layout carries, where the fixture datum sits, whether boundary scan earns its licence on this design, and how long a functional test takes per unit at volume.

The cheapest test point is the one you placed before you needed it.

What it changes on the layout

Once the split exists, the layout has a requirement rather than an afterthought. Mandatory nets get test points placed with the component, on the same side, at a pitch the fixture house can actually build. Probe-side keep-outs are drawn before the copper pour, not negotiated after it.

The cost is real but small: a few millimetres of board area and some placement discipline. The alternative cost is a revision, and revisions are measured in weeks.

In practice

On a recent industrial controller, the test plan named twenty-two nets as mandatory in-circuit before a single line was drawn. Placement worked around them from the start. The first fixture reached all twenty-two, and the pilot build ran without a coverage exception.

Nothing about that outcome was clever. It was ordinary work, done in the right order.

Written by the engineering delivery team. Published articles carry a named author once attribution is confirmed.

Have a question the article did not answer?

We would rather answer it than have you guess.