Test fixture design for prototypes and pilot builds: how we do it, and where AI helps

A test fixture is a device that locates and holds a product or circuit board (the device under test, or DUT) and makes repeatable mechanical and electrical connections to it, so the same test can be run on every unit. It can be as simple as a 3D printed nest with a few spring-loaded probes, or as large as a production in-circuit tester.

Formlabs' guide to jigs and fixtures puts the difference simply: a jig holds the workpiece and guides the tool, while a fixture holds the workpiece (Formlabs). A test fixture is a fixture that also carries the electrical, optical, or mechanical interfaces the test needs.

Keysight, a test equipment maker, describes in-circuit test (ICT) systems as using a bed-of-nails fixture to contact specific PCB test points, checking each component, while functional testing validates the overall behavior of the device.

It also notes that ICT is less effective for low-volume boards, where functional testing, boundary scan, or flying probe testing may be more suitable (Keysight).

Flying probe maker Takaya explains the tradeoff: a dedicated jig must be created for each board, which costs money, time, and storage space, while a flying probe needs no dedicated jig, takes longer per board, and suits prototype and high-mix, low-volume boards (Takaya).

Boundary scan is defined by IEEE 1149.1, which standardizes test logic built into chips for testing the connections between chips on an assembled board (IEEE SA, P1149.1).

On most of our prototype and pilot projects, the fixture that matters is a functional test fixture, often with pogo pins. Bed-of-nails ICT is usually the contract manufacturer's tool at production volume.

For one or two prototypes, bench testing with hand-held probes, clips, and a short written procedure is usually the right call. A custom test fixture earns its cost when hand testing becomes slow, inconsistent, or risky. Our rules of thumb:

The question we ask is simple: will this exact test be run more times than it takes to build a fixture for it, or will a bad result be expensive? If yes, design a fixture.

Every fixture answers a list of tests. If you ran a design FMEA, its detection controls are your starting list (see our DFMEA for prototypes walkthrough). For each test, write the measurement, the pass limits tied to a requirement, the instruments, and how long it may take.

If that list does not exist yet, write it before any CAD.

For where these tests sit in a program, see stages of product testing and their purposes and design verification and validation.

A rigid part has six degrees of freedom: three translations and three rotations.

Good fixture design constrains each one that matters, without over-constraining the part, because redundant constraints add stress and accuracy problems (Formlabs).

In practice we locate from the same datum features the drawing uses (mounting holes, a board edge, a housing face), add a clamp or lid that cannot be closed on a misplaced part, and make wrong orientation physically impossible.

A spring-loaded contact, often called a spring probe or pogo pin, is made of a moving plunger, a coil spring, and a conductive barrel. The plunger presses against a target such as a pad on a circuit board (Smiths Connectors, via ESA ESCIES).

For a fixture, we decide which nets need contact (power rails, ground, programming and debug pins, key signals, outputs), choose probe tips for the pad type, and route wiring so it can be traced and repaired. Where the board has a connector, mating the connector is often more reliable than probing it.

The board has to be designed for this. Ask for test pads on the nets the fixture needs, on one side if possible, with spacing your fixture builder can probe, and a programming and debug header. If the chips support it, boundary scan can reach nets that probes cannot. Our board bring-up guide covers the design-for-test items to add before layout is frozen.

A functional test fixture is only as good as its script. The script sets up power in steps, runs each measurement, compares it to limits, and saves a record per unit with a serial number.

Open-source frameworks exist for this. For example, OpenHTF is an open-source Python hardware testing framework originally created at Google that structures tests as phases with measurements and limits and produces a test record for each run (OpenHTF on GitHub).

Commercial test executives do the same job. Whatever the tool, keep limits in one place and version the script with the firmware and hardware revisions.

A new fixture can be wrong in ways that look like product failures.

Keysight describes validating a test program on a known-good board, often called a golden board, to tune tolerances and eliminate false failures (Keysight).

We do the same on functional fixtures, then check repeatability by measuring the same units several times, and with more than one operator if more than one person will use it.

The NIST/SEMATECH Engineering Statistics Handbook explains how gauge repeatability and reproducibility studies quantify that measurement variation (NIST/SEMATECH, Gauge R&R studies).

If the fixture handles mains voltage, large batteries, stored energy, or moving parts, add guards, interlocks, and current limits, and treat the fixture itself as electrical lab equipment: IEC 61010-1 sets general safety requirements for electrical test and measurement equipment and laboratory equipment (IEC 61010-1:2010).

For electronics, the fixture is also part of your ESD control.

ANSI/ESD S20.20 is the ESD Association's standard for an ESD control program that protects electrical and electronic parts, assemblies, and equipment (ESD Association).

Ground the fixture and the operator, and use ESD-safe materials where the board touches the fixture.

A pogo pin test fixture is the usual first PCB test fixture on a prototype or pilot project: a base plate holds spring probes under the board, a nest locates the board on alignment pins, and a lid or lever presses the board down onto the probes. It can program the microcontroller, power the board, and run a functional test in one step.

A bed of nails test fixture is the same idea scaled up: many probes, often one or more per net, wired to an in-circuit tester that checks components one by one. It is fast per board, and it is built for one specific board revision. That is why, for prototypes and frequent revisions, we usually prefer a smaller functional fixture, plus flying probe testing at the board house if component-level checks are needed.

Points to settle early:

3D printing is a natural fit for prototype fixtures because the nest can match the exact geometry of a part and be reprinted when the design changes.

Formlabs' guide notes that 3D printed fixtures can include features that are hard to machine, recommends threaded inserts or captive nuts instead of tapped holes in plastic, suggests building flat datum faces into printed fixtures so they can be inspected, and warns that some SLA resins creep under continuous load, recommending SLS for parts that stay loaded (Formlabs).

It also notes that ESD-safe printing materials exist (Formlabs).

How we use it: printed nests, lids, and probe plates for early builds. Machined aluminum or other plastic plates where probe alignment, wear, or heat matter. And metal hardware (pins, bushings, inserts) wherever something slides or is clamped. More on materials and processes in our guide to 3D printing in prototype development.

AI tools are useful for the writing-heavy parts of fixture work and weak at the parts that need physical judgment.

NIST's Generative AI Profile names confabulation, confidently stated but false content, as a core risk of generative AI, and it also warns about over-reliance on AI output and about exposing sensitive data through AI tools (NIST AI 600-1).

For fixtures that means three rules on our projects: never let AI-generated limits go into a script unchecked, run every AI-drafted script against a known-good unit before using it on real units, and check a tool's data terms before sharing schematics, CAD, or test data.

For a broader tool view, see AI tools for hardware prototyping in 2026.

LA NPDT does design, engineering, and prototyping in-house at our facility in Ruston, Louisiana: 3D printing, CNC machining, urethane casting, and bridge tooling, plus electronics and firmware.

That means the same team can design the product, the fixture, and the test script, and change all three together when a build reveals a problem. We use AI to speed up drafting procedures, scripts, and log reviews, and we keep the limits, the fixture design, and the sign-off with our engineers.

Our view is practical: a fixture is worth building when it makes results repeatable and trustworthy, or when it will travel to a manufacturer with the product. If it does neither, test on the bench and spend the time on the design. For what changes when you move from prototypes to a pilot run, see what breaks from prototype to pilot production, and for SBIR teams, why measurement fixtures matter in SBIR Phase I to Phase II.

If you want the plan first, our Stage 1: Review/Analysis + Plan is a fixed $800 and is not credited toward later stages. Or see our electronic design services and rapid prototyping, or book a free 20-minute initial consultation. Phone: +1 318-200-0526.