Hector D. Amador
Senior Mechanical Design Engineer
Robert J. Knapik
Mechanical Designer Intel Corp. Chandler, Ariz.
R & D efforts that involve the testing of physical-hardware generally fall into one of two categories: proof-of-concept or precision data collection.
Mechanical test fixtures used in proof-of-concept experiments are typically built without undergoing rigorous design analyses. Here, the approach works fine but falls short for precision data-collection experiments because it ignores important metrics such as structural flexure, manufacturing variation, and tolerance accumulation, which taken together, could hurt data integrity.
An example of the latter is a fixture for qualification and reliability testing of electronic circuit boards designed by engineers with Intel's Assembly Technology Development organization. In many applications circuit boards must endure repeated bending as from users pressing connected key contactsin cell phones. This subjects boards to both out-of-plane and in-plane lateral loads, where lateral loads depend on how boards are retained in a housing.
A main goal of the fixture is to collect board cyclesto-failure data and use it to calibrate finite-element models. The models predict fatigue-damage rates. One particularly challenging aspect of the fixture design is the interfaces that touch boards. Interfaces cannot exert lateral loads on boards because sliding introduces unknown frictional loads that would corrupt the FEA calibration. To get around the problem the fixture uses pairs of opposing, long, narrow rollers to apply bending loads to boards without stretching them as they flex. Rollers attach to a moveable support and touch boards on a narrow strip across their width over a range of locations and in between board components. The rollers themselves are supported by staggered rows of shaftmounted roller bearings that control deflection under load.
A center subassembly lets operators adjust the vertical gap between upper and lower roller subassembly pairs to accommodate different board thickness' and control the amount of initial pressure roller rods apply to a board when securing it before testing. In addition, the center subassembly positions rollers symmetrically about a board center over a wide range of spans, and supports and locates the two roller subassembly pairs during tests. Here, a center-mounted, threaded shaft operated by a thumb wheel drives threaded upper roller carriers, moving them symmetrically inward or outward by virtue of righthanded threads on one rod end and left-handed threads on the other. The other two shafts support and guide the roller carriers' embedded shaft bearings.
Applied loads react through a base subassembly that is sized both to fit the largest board under test and stay within about one-tenth of the master assembly volume. The base subassembly also uses a threaded shaft and a pair of smooth shafts for adjusting the distance between pairs of roller subassemblies. It includes four adjustable corner guards that prevent boards from working loose and rotating in plane during tests.
BEFORE THE CHIPS FLY
Before machining actual hardware, components were first evaluated with GD&T ( Geometric Dimensioning & Tolerancing per the ASME Y14.5M 1994 specification), RSS (root-sum square) tolerance analysis, and FEA.
Each dimension of each component has associated with it a GD&T tolerance that describes the spatial relationships between part features and surfaces. In essence, GD&T looks at how tolerances of individual parts affect the function of an assembly. As such, it helps prevent the overtolerancing of individual parts, which in turn, can lower manufacturing costs.
RSS tolerance analyses quantify fit between components, assuring they can be assembled properly. RSS also quantifies the range of possible size or feature-to-feature-dimensions within or between subassemblies, including those characterizing an entire fixture or master assembly. This allows the prediction of maximum positional variability (before loading) between the desired and possible positions of any roller rod within a fixture or between corresponding roller rods of different fixtures in the master assembly. In other words, it represents the deviation of a board's imposed deflection, at a particular roller rod, from the deflection desired for a test. Results of both tolerance studies drove iterative adjustments of final manufacturing dimensions.
Finite-element analysis identified appropriate part geometries and materials that kept stresses and deformations within spec. In fact, FEA predictions of maximum component stresses prompted design changes that will let the fixture work continuously for about five years without replacement of any machine components; an order of magnitude longer than the six-month PM interval allowed by the customer.
Components subjected to high loads and wear or those difficult to protect from normal corrosion are made from 300-Series stainless steels. Most largersized parts and those required in large quantities are made of a 6000-Series aluminum, chosen for its relative stiffness, low weight, and ease of machining.
All aluminum parts are anodized to prevent the formation of aluminum oxides that could contaminate test boards or damage bearings, threaded joints, and electrical connections. Rollers made from the material prevent electrical shorts should loads damage or penetrate board solder resist. Anodizing also discourages galvanic-corrosion reactions between steel/aluminum galvanic pairs. For this same reason the fixture is spec'd to operate only in conditions typical of indoor laboratories, or about 25C and less than 12% humidity. As of this writing the fixture has been assembled successfully with no unexpected interferences and holds tolerance on deflection during circuit-board tests.
A FEW LESSONS LEARNED
The difficulties encountered were minor and had mainly to do with unspecified or incorrectly specified dimensions for purchased hardware (bearings, snap rings, etc.). For example, don't finalize component or assembly designs involving purchased parts until verifying key dimensions. It has been our experience that catalogs often include misleading, vague, or incorrect dimensions. Obtain purchased parts early and check them thoroughly.
Finally, when doing structural analyses at the component level, try to empirically validate analyses as you finish the design of a subassembly, before proceeding. Small errors introduced when approximating component-level structural boundary conditions invariably accumulate, causing unacceptable deviations at the subassembly level.Key design specs
3 1.5 in. to 8
3 4 in.
3 23 in. to fit within the work area of the material testing machine.
Intel Corp., www.intel.com