|
|
|
|
It’s the Analysis That Counts Evaluation Engineering, October 2002 by John Starr, CirVibe, and Wayne Tustin, Equipment Reliability Institute Many manufacturers have found optimized vibration test to be the most efficient means of finding production flaws, crediting it with discovering 60% to 80% of the problems. In fact, many companies have concluded that the value of thermal screening is overrated.1 Tests during the development of electronic products can provide mounds of information, but little is gained without detailed analysis. With detailed analysis, substantial immediate cost savings can be realized during production test and later from higher reliability. Electronic products, such as commercial off-the-shelf (COTS) products used by the military, have many physical dimensions and material properties that cannot be tightly controlled, yet are very critical to life. As a result, understanding the product is a critical element in vibration testing.2 The most common use of vibration test in developing reliable electronic products has been a combination of testing and evaluation with empirical relationships. This approach is adequate for most circuit cards. But, empirical relationships provide guidelines, not a real numerical definition of life at the point of failure. Since the empirical guidelines do not address design details, many unexpected failures on printed-circuit cards can and too often do occur. Failures can be extremely costly, depending on when in development or service they occur. Replacing empirical equations with effective analysis leads to substantial savings. Does Analysis Work? You may question the capability of mechanical analysis to aid in developing electronic systems. But remember that all systems, including electronic systems, are subject to numerous physical laws and mechanisms. Analysis always works. When detailed analysis disagrees with tests, the fault usually is our lack of understanding of the physical product. Vibration-Caused Failures For most components on modern electronic circuit cards, the most severe stresses result from card deformations defined by mode shapes at natural resonances. Random vibration commonly is used in product testing, with various resonances excited simultaneously, much as they are in service. Whenever a failure takes place during highly accelerated life testing (HALT), environmental stress screening (ESS), highly accelerated stress screening (HASS), or other testing, you need to identify the root cause of that failure.3 When vibration is understood at the root cause level, design changes can be implemented with the greatest probability of success and at the lowest cost. Vibration of
electronics can be quite complex. Physics of failure (PoF) analysis
translates test data into data defining exposure at the
point-of-failure level. Detailed analysis of designs can show why
life expectations of identical components can vary significantly
with location. Substitute Hardware The design of reliable products requires further knowledge. If a component has multiple suppliers, will substitution change life capabilities? Since components are designed for electronic function, substitute components can differ in various structural properties. Circuit boards can vary similarly in thickness and bending modulii. Differences result in natural frequencies, responses, and life capabilities. All mechanical properties affect stress, and stress affects life capability exponentially. The designer must evaluate whether proposed changes will affect reliability. With all the expected variations in circuit and component parts, how do we interpret our test results? If one prototype unit passed one test, what can we predict for other units? Electronics products are difficult to control mechanically because there is little or no control on important physical parameters. Testing of all variations is not practical. Analysis allows extrapolation of test experience to cover critical mechanical parameter variations. Circuit-Card Complexity Common expressions illustrate our inability to understand our test results and the difficulty of defining an effective stress screen for electronics, such as “each electronic product is unique.” Other expressions, following life tests or field returns, are “cannot duplicate failures,” “no fault identified,” and “re-test OK.” Such phrases are common because of the statistical complexity of test control and test items. Detailed Analysis Three large contributors to statistical variations in test results are the following:
Circuit-card components fail under vibration as a result of fatigue from cyclic stresses, i.e., inertial forces and mode-shape-caused component bending. Unfortunately, these stresses cannot be quantified by measurements during a test. Empirical formulas have attempted to define life capabilities through simple curvature approximations. These methods often fail because they can’t properly cover all variations in circuit-card details. Component life is affected by curvature in both directions. At best, a simple formula provides a crude approximation. Since the stress/life relationship is exponential, large life-capability errors result. A test provides response measurements and pass/fail information, with the amount of detail available determined by the allocated funds. However, when you add PoF analysis, the available information expands dramatically. The CirVibe software program is an example of PoF analysis useful in developing reliable electronics.4 It converts a geometric description into a mathematical model, then solves this model, extracting detailed stress cycling data for every component on the test item. The software-program methods were based on decades of experience in applying numerical analysis to design and development of structures. This experience included design, development, and test of numerous electronic products. This automated program develops finite element analysis (FEA) models from simple geometric descriptions of a circuit card and its components. The finite element detail is generated internally, so you don’t need FEA expertise. Interfaces to computer-aided-design (CAD) programs speed the development by translating CAD data to circuit-card analysis models. FEA applies laws of mechanics and determines product information beyond the capability of any test program. Analysis can optimize accelerometer positioning. Taking advantage of new tools to use current PC power, the detailed calculations can turn a few accelerometer measurements into:
Adding modern FEA to the test process turns accelerometer readings into definitions of life capability of every component. Extending the analysis to include any design variations is very simple: repeat the analysis with new parameters. Design changes such as component details can be evaluated in minutes. More complex, ruggedizing changes such as layout, support conditions, stiffener additions, or similar changes can be performed in a few hours. Design changes can be qualified virtually, without the time and expense of building and testing a prototype. Many options also can be considered. Product understanding gained from detailed analysis is valuable in the design of test fixtures used to attach circuit card(s) to the shaker’s vibrating table. Too often, decisions are based on results obtained with faulty fixtures in an attempt to match the geometry, but unfortunately not the dynamics, of in-service usage conditions. Detailed pretest PoF analysis identifies which components are to be driven at each natural frequency and to what level each is to be driven. It also predicts the change in damage that will result from a change in drive level over a frequency range. This product understanding is used to optimize the stress screen. Screens can be tailored by excitation control in frequency bands to properly excite critical parts of the test article without using excessive test-article life. Figure 1 shows stress-screen effectiveness at a component level. It also illustrates life capability under one set of requirements. Conclusions Since a test cannot provide any measurements descriptive of point of failure, test alone can be a hit-or-miss approach to gaining knowledge about a product. The industry phrase “the ESS process is unique for each electronic product” demonstrates this fact. It’s the combination of test and analysis that provides real knowledge. Tests create real failure data. Subsequent analysis provides numerical definition of the failures experienced. By numerical definitions, we mean definitions of exposure to fatigue damage at the component level. These numbers are transferable from one design configuration to another. When numbers can be transferred from one design to another through analysis, each test program benefits from all past experience. Test programs become more efficient. For life testing, definitions of life capabilities relative to requirements are more accurately defined, reducing the risk of failures. For stress screening, screen effectiveness can be defined at a component position level. Screens can be optimized much earlier in the process. The benefits of analysis include cost savings of stress-screen programs and savings from producing a more reliable product. References 1. Hobbs, G.K., “Reliability—Past and Present,” Sound and Vibration, April 1997. 2. Starr and Abner, “Understanding Vibration of Electronic Systems,” www.vibrationandshock.com/news/news5/nl5.htm. 3. Gray and Tustin, “Electronics Testing into the 21st Century: Success in Test Is in Capabilities, Not Specifications,” www.equipment-reliability.com/articles/art2.htm. 4. CirVibe Circuit Card Vibration Software, Users Manual Version 3.0, http://www.cirvibe.com/. About the Authors John Starr is a consulting engineer at CirVibe. He has an M.S. in engineering mechanics from Michigan State University and extensive experience in structural analysis and test development. CirVibe, P.O. Box 47394, Plymouth, MN 55447, e-mail: jstarr@cirvibe.com Wayne Tustin is president of Equipment Reliability Institute and a graduate of the University of Washington with a B.S.E.E. He has broad experience in shock and vibration testing and teaching of these test methods. Equipment Reliability Institute, 1520 Santa Rosa Ave., Santa Barbara, CA 93109, 805-564-1260 CirVibe
Vibration Published by
EE-Evaluation Engineering |
