In-Circuit Test: Test Points, Fixtures and Limits

In-circuit test is the workhorse of assembled board testing. It checks the board as a collection of components and connections rather than as a working product, which means it can find a wrong value or a missing part before anyone powers the design up. It also makes demands on the layout that have to be met at the design stage, because a board that cannot be probed cannot be tested this way.

What In-Circuit Test Does

The test uses a fixture with a bed of spring loaded probes, one for each node the test needs to reach. The board is pressed down onto the probes, and the tester measures between them.

What it measures is deliberately simple. A continuity check between two nodes confirms that a track exists and no short connects them. A resistance measurement across a component checks its value and confirms it is present. A capacitor measurement detects a missing part or a wrong value, though measuring capacitance in circuit is less reliable than resistance because of the surrounding network. Power rails can be checked for shorts before power is applied, which protects the assembly from damage.

The strength of the method is coverage of structural faults. It finds missing components, wrong components, reversed polarised parts, open joints, bridges and damaged devices in seconds, and it does so without the board being functional or the firmware being loaded. That combination makes it the standard test for volume production of boards where a functional test alone would be slow to develop and slow to run.

The second strength is diagnosis. When the tester reports a failure, it reports a specific node or component, which points the technician at the problem. A functional test tells you the product does not work; an in-circuit test usually tells you why.

Test Point Design Rules

The fixture is the design constraint, and the rules below are what make it buildable.

• One test point per node. Every net that the test needs should have at least one accessible pad, ideally on the side that faces the fixture.
• Pad size. A round pad of about 1.0 mm to 1.5 mm gives the probe a reliable target without crowding the board. Below 0.8 mm the fixture becomes fragile.
• Spacing. Adjacent probes need clearance for the spring housing, and a 2.54 mm grid is the classic arrangement. A 1.27 mm grid is available and costs more.
• Clearance from tall components. A probe has to travel vertically, so the area around a test point must be free of components above a low height.
• Keep away from the board edge and mounting holes, where the fixture frame and the board support posts sit.
• Do not share a test point with a component pad unless the pad is large and the process allows it, since the probe damages paste and solder.
• Add a dedicated ground and power test point where the measurement benefits from a short reference path.

Two additional rules are less obvious. Test points should be on one side of the board where possible, since a double sided fixture costs more and is harder to maintain. And the copper under a test point should be solid rather than a thin trace, because the probe pressure and the measurement both benefit from a low impedance contact.

What In-Circuit Test Cannot Reach

The method depends on physical access, and access is the first thing a dense design removes.

A ball grid array has no accessible node on either side of the joint, so the nets that connect only through the balls cannot be probed. A design can recover some of those nets by routing them out to a test point elsewhere on the board, but that may require more layers or longer traces. Where it is not possible, the coverage of those nets has to come from somewhere else, usually boundary scan or a functional test.

Fine pitch parts present a second problem. A test point on every node may simply not fit, and the trade is to probe the nets that matter most: power rails, ground, and the signals that are hard to reach from the edge connector. That is a judgement call, and it should be recorded so that the test coverage is a known quantity rather than an assumption.

In-circuit measurement also has electrical limits. A capacitor measured in circuit is in parallel with everything else connected to its two nodes, so a wrong value may be masked by the surrounding network. Large capacitors and low value resistors are measured with more confidence than small ones. Diodes and semiconductors can be checked for a forward drop, but a device with a parallel path may look correct even when it is faulty.

Finally there is the fixture itself. A bed of nails becomes impractical above a certain node count, and a board with a few thousand nodes may need a mechanically complex fixture that costs more than the test program saves. At that point the alternative approaches start to win.

Alternatives and How They Combine

A flying probe tester moves a small number of probes from point to point under program control. There is no fixture, which makes it ideal for prototypes and small batches, and it can probe finer features than a bed of nails. It is much slower per board, so it is rarely used for volume.

Boundary scan, or JTAG, uses the internal shift register of compliant devices to drive and read pins without physical probes. It covers the nets between devices with boundary scan support, which is often a large part of a dense board, and it requires no test points. The limitation is device support: parts without the capability are invisible to it.

Optical inspection and X-ray cover the solder joints rather than the electrical function, and they complement rather than replace in-circuit test because they do not verify that a correct component was fitted or that a track is intact.

A functional test applies power and exercises the product, which is the only test that proves the design works. It is usually slower to develop and to run than in-circuit test, and its diagnostics are coarser, but it catches interactions that no structural test can find.

The practical arrangement is a combination. In-circuit test or flying probe covers structural faults at the beginning of the line, boundary scan covers the dense interconnections, and a functional test at the end confirms the product works. Which share of the coverage each one takes is a design decision, and the total is what matters rather than any single method.

Fixture Cost, Programming and Test Time

In-circuit test has a cost structure that is unlike any other part of production, and it is worth understanding before the decision is made.

The fixture is a one time cost that scales with node count, probe type and mechanical complexity. A simple single sided fixture for a few hundred nodes is a modest investment. A double sided fixture with fine pitch probes and a vacuum hold down for a large board can cost an order of magnitude more, and it is specific to one board design, so a revision that moves the test points may require a new fixture.

Programming is the second cost. The test program has to be generated, debugged against real boards, and tuned so that measurement tolerances do not produce false failures. On a board with many components this is a significant engineering effort, and it interacts with the design: a component with a wide tolerance or a strong parallel path generates more false calls.

Test time is the third and it is recurring. Every probe contact takes time, and a program with thousands of measurements can take minutes per board. Reducing test time usually means reducing coverage or grouping measurements, which is why a test engineer may ask for a design change that makes several nets measurable at once.

The economics are therefore volume dependent in a specific way. A fixture and a program are worth paying for when the production quantity is large enough to amortise them, which for many products means a few thousand boards. Below that, a flying probe or a well designed functional test is usually cheaper overall, even though the per board test time is longer.

Design for Test Checklist

• Allocate test points on the fixture side of the board, one per node where coverage is required.
• Use 1.0 mm to 1.5 mm test pads on a consistent grid, and confirm the grid with the test house.
• Keep the area around each test point clear of tall components and of the fixture frame.
• Route the nets that are buried under area array packages out to test points where possible.
• Provide dedicated ground and power test points close to the devices being measured.
• Leave the test points free of solder mask and free of paste, and do not place a test point under a component that will be fitted.
• Confirm the mechanical interface: board size, tooling holes, edge clearance and support positions all belong to the fixture.
• Involve the test house before the layout is released, because test point placement is far cheaper to change on screen than in a fixture.

FAQ

• Does every net need a test point? No. Cover as many as the space allows, and use boundary scan or functional test for the rest. The important thing is knowing what the coverage is.
• Can a component pad be used as a test point? It can, but the probe damages the pad and the paste, so a dedicated pad is preferred.
• How many test points fit on a board? It depends on the grid and the component density. A 2.54 mm grid with 1 mm pads is the most economical arrangement.
• Is in-circuit test still relevant? Yes, for volume production of boards with accessible nodes. On very dense boards its share of the coverage drops, and flying probe, boundary scan and functional test take over more of the work.

Summary

In-circuit test turns a board into a set of measurable nodes and checks them in seconds. It is fast, it diagnoses clearly, and it finds the structural faults that a functional test would take far longer to isolate. In return it demands physical access, which is why test point provision belongs in the layout rules rather than in a late review.

A 1.0 mm to 1.5 mm pad per node, a consistent grid, clearance from tall parts and a defined side for probing are enough to make a board testable. Where the density makes that impossible, the coverage has to be supplied by boundary scan, by a flying probe on the test strategy, or by a functional test, and the total coverage is what should be tracked. That planning is cheapest when it happens alongside the assembly and capability discussions, before the design is committed to a production fixture that cannot be changed cheaply.