Back to directory
Reference

In-circuit test questions, answered

Fixture design, program development, platform selection, used equipment, and cost — the questions we field from engineers and buyers every week.

Platform Selection

What is the difference between in-circuit test and flying probe test?

Open

ICT uses a bed-of-nails fixture with a dedicated probe for every net, so all nodes are contacted at once. That makes it fast — seconds per board — and it can power the board and test powered nets, but it needs a mechanical fixture per design. Flying probe touches nets sequentially with a few moving probes and needs no fixture at all, so it is cheaper for prototypes and low volume but far slower per board. Most programs use ICT for volume and flying probe for prototypes, high-mix low-volume builds, or boards with no room for a fixture.

When should I use flying probe instead of ICT?

Open

Use flying probe when there is no fixture and a fixture would not amortise: prototypes, engineering builds, high-mix low-volume production, or boards where probe access is so limited that a bed-of-nails fixture is impractical. It is also the sensible choice when a design is still changing, because a fixture built to a moving design becomes obsolete. The trade-off is test time — flying probe contacts nets sequentially, so it is much slower per board. As a rule of thumb, below a few thousand boards on a specific design flying probe is usually cheaper overall, above that ICT wins.

Can boundary scan and ICT run in a single test pass?

Open

Yes, and it is increasingly common. Several boundary scan toolchains integrate with major ICT platforms so that the JTAG chain is exercised through the same tester and, where supported, the same fixture and test pass. This matters for cycle time — running both as one operation is faster than two — and for fixture design, because the boundary scan access requirements need to be planned alongside the probe field. Ask prospective vendors specifically whether their tool integrates with your platform, as this materially changes both the fixture requirements and the total cost.

What is IEEE 1149.1 boundary scan and when should I use it instead of ICT?

Open

IEEE 1149.1 boundary scan tests interconnect through the JTAG cells built into compliant devices, so it needs no physical probe access. Use it for BGA and fine-pitch packages where probe access is impossible, for boards with very few test points, and to cover nets a fixture cannot reach. It cannot test analog nets or devices without boundary scan, so in practice it complements ICT rather than replacing it.

What is the difference between JTAG and ICT for BGA coverage?

Open

On a BGA, a bed-of-nails fixture can only reach the outer balls a probe can physically touch — interior balls are effectively invisible. Boundary scan covers interconnect between BGA pins and the rest of the net through the devices' JTAG cells, which makes it the practical answer for BGA and fine-pitch coverage. Most production strategies run both: ICT for the rest of the board and boundary scan for the inaccessible nets.

What is the difference between ICT and functional test?

Open

ICT is a structural test: it verifies that components are present, correctly oriented, the right value, and that nets are not open or shorted. It generally does not power the board up in a full sense or exercise device firmware. Functional test (FCT) powers the assembly and checks that it behaves as designed — running firmware, communicating over interfaces, producing correct outputs. ICT catches manufacturing defects; FCT catches design and integration problems as well. Most production lines need both, and they are complementary rather than alternatives.

Should I use AOI or AXI instead of ICT?

Open

They test different things and do not substitute for one another. AOI inspects solder joints and component placement optically; AXI inspects hidden joints such as BGA balls using X-ray. Neither verifies electrical continuity or component value. ICT measures electrical behaviour but cannot see a joint it cannot probe. A common and effective combination is AXI for BGA and hidden joints, AOI for visible assembly quality, and ICT for electrical structural coverage — with boundary scan added where electrical access is impossible.

Fixture Design

What board data does a fixture builder need to quote a bed-of-nails fixture?

Open

At minimum: the fabrication and assembly data (Gerbers and drill file), the netlist, and a test-point report identifying which nets are intended as probe targets with their coordinates and side. A CAD file or IPC-D-356 netlist is ideal because it defines probe locations directly. Without a defined test-point list the builder has to infer targets, which is both slower and a frequent source of quoting error. Also useful: board thickness and material, panel or array format if the board is processed in a frame, and any mechanical constraints such as tall components or keep-out areas on either face.

Vacuum or press-fit actuation — how do I choose a fixture type?

Open

Vacuum actuation uses a sealed chamber to pull the board onto the probe field and is generally the simpler and lower-cost option for small and medium boards. Press-fit and mechanical actuation apply force through a lever or press and are usually preferred for large boards, high probe counts, boards needing high probe force, or where vacuum routing is impractical. Consider board size and rigidity, the total probe force required, available shop air, and how much deflection the assembly tolerates. Large boards under vacuum can bow enough to cause unreliable contact on edge probes, which is where mechanical actuation tends to win.

How many actuations will a fixture last before probes need replacing?

Open

It depends heavily on probe type, spring force, board surface finish, and contact cleanliness rather than on a single fixture-wide number. Probes are consumables and are expected to be replaced periodically, usually individually as contact resistance drifts or failures appear rather than as a full set. Boards with HASL finishes and any contamination wear probes faster than clean ENIG surfaces. The practical approach is to monitor contact resistance trends and false-fail rates rather than counting actuations — rising resistance on specific probes tells you which to replace before yield degrades.

How long does it take to build a new ICT fixture?

Open

Typical lead time from released board data to a delivered fixture is several weeks, commonly in the four to eight week range for a standard single-sided assembly, with double-sided and high-complexity designs running longer. The variables that move it most are probe count and density, mechanical complexity, whether non-standard probes or connectors must be procured, and the queue at the fixture house. If program development is in scope, add that schedule separately. Ask for the fixture build and the program development as two dates rather than one, so a slip in either is visible early.

What is the difference between a wired and a wireless ICT fixture?

Open

A wired fixture routes each probe through individual wiring to the tester interface, which is straightforward and repairable but becomes unwieldy at high probe counts where the bundle is large and prone to damage. Wireless fixtures use a spring-loaded interface block or translator that mates the probe plate directly to the tester's receiver, eliminating the wire bundle. Wireless is generally preferred at high probe counts and for repeated fixture changes, because it is faster to swap and less fragile. It requires the tester to have the matching receiver hardware, which is worth confirming early.

Used Equipment

What are the biggest risks when buying a used in-circuit tester?

Open

The three that cause most problems are software licensing, calibration status, and spares availability. Confirm in writing that the machine's software licence transfers with the hardware — an unlicensed tester can be unusable regardless of condition. Verify that calibration can be traced and is current, and get the certificate. Then check that the specific pin cards, probe assemblies, and consumables you will need are actually obtainable for that model, because discontinued platforms can leave you with a working machine you cannot maintain. Also establish whether the system was decommissioned working, and ask for a demonstration or a witnessed power-on before payment.

Can a used ICT fixture be adapted to a different board?

Open

Sometimes, but treat it as a re-engineering project rather than a swap. The mechanical frame, actuation system, and sometimes the probe field can be reused if the new board is similar in size and outline. What almost always needs rework is the probe plate: new probe locations drilled, probes re-loaded for the new netlist, and the wiring or wireless interface re-mapped. Budget for a new probe plate, probe replacement, and re-correlation against a known-good board. It is most viable when the new board is a revision of the original or shares a similar test-point pattern.

Cost & Turnaround

How much does a bed-of-nails ICT fixture cost?

Open

A single-sided bed-of-nails fixture typically runs roughly $8,000 to $25,000, driven by probe count, actuation type (vacuum vs press-fit), mechanical complexity, and whether test program development is included. High pin-count, double-sided, or wireless fixtures for fine-pitch and high-current boards sit at the top of that range or above. Used fixtures cost a fraction of new, but need rework and re-probing for a new board layout.

How much does in-circuit test program development cost?

Open

Program development is usually quoted separately from the fixture and typically ranges from a few thousand dollars for a straightforward board to tens of thousands for a complex one. The drivers are the number of devices requiring library work, node count, whether boundary scan must be integrated, whether powered tests are in scope, and how much debug and correlation the board needs. Boards with clean DFT and an accurate netlist sit at the low end because the work is largely translation. Boards with poor test-point access, many custom devices, or an unreliable first article sit at the high end — and the debug phase, not the CAD translation, is almost always what sets the cost.

What is the total cost of bringing up ICT on a new board?

Open

The full first-time cost of bringing up ICT on a new board has four parts: fixture build, program development, first-article correlation, and the internal engineering time to review and approve it. Quoting usually covers the first two. First-article correlation — proving the fixture and program against a known-good board and reconciling discrepancies — is frequently additional. Budget also for probe spares, documentation, and at least one round of ECO-driven rework if the design is still moving. Ask any vendor for these as separate line items rather than one lump sum; it makes both the comparison and the internal approval easier.

How do ICT and flying probe compare on turnaround time?

Open

Per-board, ICT is far faster — typically seconds against minutes for flying probe — because every net is contacted simultaneously rather than sequentially. But turnaround depends on what you are measuring. If it means time to first tested board on a new design, flying probe is usually quicker to deploy because there is no fixture to build: you can often test a prototype within days. ICT requires fixture and program lead time, commonly several weeks. So flying probe wins on time-to-first-test and ICT wins on time-per-board once running.

What causes the gap between an ICT quote and the final invoice?

Open

The common ones are scope creep after the quote, additional engineering-change iterations, library development for devices that turned out to need it, unexpected fixture rework after first-article correlation fails, expedite fees, and freight. Most arise because the original scope was underspecified rather than because the vendor acted badly. The practical defence is a written scope that names probe count, actuation type, sidedness, whether program development and first-article correlation are included, and how many ECO iterations are covered before additional charges apply.

Program Development

How long does Keysight 3070 test program development take?

Open

Most new-board programs take three to six weeks from CAD data and a first article, including device library work, debug, and correlation against a known-good board. Boards with clean DFT coverage and an accurate netlist move faster. Boards that need significant library development, or that have poor test-point access, run longer — the debug phase, not the CAD translation, is usually what sets the schedule.

What does an ICT test program development project actually involve?

Open

A typical project runs through five stages: importing CAD and netlist data and building the board model; developing or importing the device library; generating the initial test program including shorts, opens, and analog tests; running it on a first-article board and debugging the failures; then correlating against a known-good board and documenting coverage. Most of the schedule sits in stages four and five rather than in data import. Ask a vendor to quote and schedule these stages separately so a slip is visible early and attributable.

Can a test program be reused on a different ICT platform?

Open

Not directly. Test programs are written in platform-specific languages and reference platform-specific device libraries, so moving between Keysight, Teradyne, SPEA, or other platforms is a translation and re-verification effort rather than a copy. Some tooling can assist with conversion of test statements and device models, but the output must be re-verified on the target platform against a known-good board. Budget for library gaps and debug. If you anticipate running on more than one platform, raise it at quoting stage because it changes scope.

Can you convert a GenRad 228x program to a modern platform?

Open

Yes, but conversion is a redevelopment effort rather than a one-click translation. The source program's test statements and device library are translated where tooling allows, then re-verified on the new platform against a known-good board. Expect partly manual work and coverage gaps that must be re-covered, so budget debug and correlation time alongside the translation itself.

How do I validate a new ICT program before it goes into production?

Open

Validate on a known-good board first: every test should pass, with margin. Then validate on boards containing known defects — deliberately created or previously diagnosed — and confirm each defect is caught by the intended test. Check that the program does not produce false failures across a run of good boards, since a program that fails good boards gets disabled by the line and stops protecting anything. Record the coverage achieved and the defects deliberately not covered. Review all three before releasing to production.