Easing PCB stress from bed-of-nails: model board stress prior to building the fixture.

All bed-of-nails fixtures--whether vacuum, pneumatic or mechanical--exert forces on boards to ensure adequate electrical contact at test nodes. When board logic was relatively sparse--because components were larger and their heat dissipation prohibited spacing them close together--maintaining a constant force distribution across the board during test was not difficult.

The picture has changed considerably. Because of the logic density and capability of today's components, they often occupy clusters of high concentration on the boards. Thus, force density on the boards varies widely. Such clusters on complex boards often include SOICs and BGAs that have hundreds of pins on very tight centers. A "simple" fixture design that merely places nails on those pins subjects some areas of the board to much higher loading than others, which creates bulges, bending moments and other deformation of the substrate that threaten to induce solder-joint and other stress-related failures. And whereas through-hole device legs could at least partially compensate for such stresses to minimize any damage to the board under test, SMT components are much less forgiving. Pb-free solders compound this. Pb-free solders are more brittle, and manufacturers already report increases in solder-joint defects.

The only solution is to meticulously design boards and fixtures that distribute stresses as evenly as possible and thereby minimize stress-induced damage. Doing so effectively requires a stress model--a "map" that identifies potential trouble spots.

PCB designers must distribute logic nodes and board components as much as possible to avoid "hot spots" that increase the likelihood of board damage during test. Fixture designers must carefully manage the force distribution as well. Their efforts may include reducing the probe force, applying backup blocks and placing adequate numbers of push-fingers. Accomplishing these goals requires manufacturers to accurately measure stresses on the board during test.

[FIGURE 1 OMITTED]

Here we discuss issues of fixture design and two specific methods for creating a stress model. The more common technique uses strain-gauge measurements at discrete board locations. We will also look at an alternative that incorporates CAD, finite-element analysis (FEA) and design-for-test (DfT) rules to model the whole board before the fixture is built.

Proper fixture design. A test fixture screens PCBs for defects; it should not create them. Preventing damage requires treating the fixture itself as a designed part of the process. The days of merely connecting board I/O points to tester pins by randomly sprinkling push fingers and supports around the fixture have passed. Probes, push fingers and supports exert forces on the board during test and therefore must be considered during fixture design. Fixtures must distribute forces carefully to eliminate the damage threat. In addition, clearances between fingers and board components must conform to known DfT guidelines, especially around deflection-sensitive components like BGAs.

Fixture construction must not delay production schedules. The challenge is to design and build a safe fixture while maintaining typical turnaround times of three to 15 days, depending on the fixture's complexity. The most accurate and ideal approach would automatically incorporate critical DfT rules, and become standard practice for designing and constructing all test fixtures. As much as possible, a design for a properly performing fixture should be finalized before building commences. Adding fingers and supports after the fact, such as in response to fixture damage in pre-production or full production, is difficult and expensive, as well as unacceptable in principle. But how do you calculate the stresses that the board will undergo precisely enough to design and build such a fixture?

Strain gauges. One historical approach to measuring board stresses uses strain gauges. Developed in the 1930s, these gauges can measure the unit strain at a specific board location. Their electrical resistance varies in proportion to the amount of strain. This method is reactive because it requires both a completed fixture and the board under test to perform the measurement. Strain gauges were developed and are primarily used to measure structural metals. PCBs are typically made from FR-4/G-10--a nonmetallic insulating material that is orthotropic and exhibits material properties very different from those of metals. Also, strain gauges cannot easily predict local component stiffening effects and induced moment couples. And, of course, correcting any fixture-design problems discovered by the strain gauge will require engineering changes that incur high costs in both time and money.

The strain gauge method can investigate only a few small areas on the board at a time. As a result, these measurements can provide only a snapshot of only those areas. Other parts of the board and areas underneath components that are inaccessible to the strain gauge may experience high stress as well, so results from this method cannot adequately reveal problems.

The location of strain gauges is critical. Many manufacturers place them around high-cost components such as BGAs and ceramic pin-grid arrays (CPGAs) where component damage is the major concern. Yet the board may fail because of a cracked solder joint around a capacitor worth only pennies.

Moving a strain gauge even slightly on the board may yield drastically different results. And the reliability of the results will depend a great deal on the human factor. Two operators may have very different views of where to place the gauges to measure the strain on the board under test.

Furthermore, most areas of highest stress experience steep strain gradients. A strain gauge averages strain in the area that it covers, so it will return a value somewhat less than the true maximum.

Another problem is that the strain gauge is an external device with respect to overall test dynamics. Results will not necessarily represent actual board stress during test. The board's sensitivity to temperature and humidity may affect the results as well, as will the material from which the gauge is made. Such factors must be considered when selecting an appropriate gauge for a particular application.

[FIGURE 2 OMITTED]

The board-stress model. An alternative approach is the board-stress analysis (BSA) method to "gauge" the fixture before building it. This approach incorporates CAD, FEA and DfT rules, integrating the packages with custom software. A BSA details stresses that the board will experience during test. The analysis provides an overview of the entire board--rather than only individual areas--identifying potential high deflection points to ensure the best possible fixture design. It can be run quickly in parallel with other tasks during fixture preparation.

The accuracy of BSA has been verified on numerous fixtures by comparing its results with those from strain gauge measurements. Because of the risk of board damage during bed-of-nails test, we now perform BSA during the design of every test fixture that we build. Through extensive stress testing on real boards (and after adding an appropriate safety factor), we determined that a maximum deflection of 0.010" will avoid board damage. Our experience has shown that properly designed fixtures will not exceed the 0.010" limit.

The figures show an example of a BSA performed on a fixture prior to its release for manufacturing. The colors in Figure 1 represent the magnitude of the deflection, while the blue contour lines show the rate of change in deflection in a manner similar to that of a topographical map. The map also shows component outlines relative to the predicted deflection.

This figure demonstrates the power of the stress-analysis tool and the comprehensive picture that it presents of the entire board. Note that changes to critical fixture features such as probe placement, probe force and finger/support locations will invalidate the analysis. Eliminating even a single push finger could damage boards on high point count fixtures.

This example predicts a maximum board deflection of 0.008". Because the analysis provides a view of a fixture that does not yet exist, if the model had predicted a deflection greater than the standard 0.010", the fixture's designers could have made the necessary modifications before building it. Re-running the analysis after incorporating the design changes would ensure compliance, again before fixture construction.

The map in Figure 2 details the locations of push fingers and board supports used in the analysis and their relative location to components.

Manufacturers should view strain-gauge measurement as a complement to the BSA. BSA should be performed on all fixtures during design. Strain-gauge measurements will validate the finished product and ensure that both the fixture and board have been constructed correctly.

Gary St. Onge is vice president of the Test Fixture Group, Everett-Charles Technologies (ectinfo.com); [e-mail]. Jesse Carpenter is also with ECT.